Home

Product User Manual - H-SAF

1. Radar data mm h PR lt 0 25 0 25 lt PR lt 1 00 1 00 lt PR lt 10 00 10 00 lt PR PR lt 0 25 63 45 26 0 25 lt PR lt 1 00 8 14 16 1 00 lt PR lt 10 00 5 21 8 10 00 lt PR 0 0 Table 44 The contingency table for the three precipitation classes defined in fig 11 of Chapter 3 evaluated by comparing H01 with radar data The averages of POD 0 43 FAR 0 81 and CSI 0 15 have been obtained using radar data on one year of data 1 December 2009 30 November 2010 In Table 44 it is possible to see that 87 of no rain is correctly classified by H01 There is a general precipitation underestimation by H01 Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 131 183 HSAF Product Validation Report PVR 01 6 3 2 Rain gauge validation Dec 09 Jan 10 Feb 10 Mar 10 Apr 10 May 10 Jun 10 Jul 10 Aug 10 Sep 10 Oct 10 Nov 10 TOT ith RRzozs mm h 0 14 0 30 0 28 0 17 0 28 0 35 0 48 0 60 0 63 0 41 0 26 0 24 ORB Rreo2smm n 0 76 0 87 0 92 0 72 0 75 0 63 0 69 0 74 0 68 0 73 0 70 0 55 Bra RRv025mm n 0 10 0 10 0 07 0 12 0 14 0 22 0 23 0 22 0 27 0 20 0 16 0 18 0MB POD with RR2imm h_ 9 19 0 25 0 32 0 21 0 25 0 31 0 34 0 41 0 46 0 35 0 32 0 31
2. Country Type of interpolation Quality control Y N Belgium Barnes over 5x5 km grid Y Bulgaria Co kriging Y Germany Inverse square distance Y Italy Barnes over 5x5 km grid N Poland No Y except cold months Turkey No Y Table 9 Data pre processing strategies H01 H02 Country Spatial matching Temporal matching Spatial matching Temporal matching Belgium N A N A N A N A Bulgaria N A N A N A N A Germany matching gauges are each overpass matching gauges are each overpass is searched on a radius compared to searched on a radius compared to the of 2 5 km from the next hourly of 2 5 km from the next hourly rain IFOV centre amount IFOV centre amount Italy mean gauges value each overpass Gaussian weighted each overpass is over 15x15 km area compared to mean gauges value compared to the centred on satellite next hourly centred on satellite next hourly rain IFOV amount IFOV amount Poland mean gauges value each overpass mean gauges value each overpass is over the IFOV area compared to over the IFOV area compared to the rectangular next 10 minutes rain rectangular next 10 minutes rain amount amount Turkey weighted mean of each overpass weighted mean of each overpass is the gauge values compared to estimated at the minute averaged 3kmX3km grid rain for Temporal structure within matching satellite IFOV by using semi variogram the g
3. PR OBS 1 BE DE sL Fom r PO TU winter 2010 Version 1 4 radar radar radar radar radar gauge gauge gauge gauge gauge gauge NS h 283 1538 102 36 1959 20190 0 7389 249 27828 NS NR 6958 38044 14064 1914 60980 63546 1662 57869 5689 128766 NR ME 0 17 0 08 0 21 109 0 0 34 0 07 0 29 0 57 0 27 ME sD 0 18 4 51 1 64 2 29 1 41 0 73 1 42 0 86 2 36 0 87 sD MAE 0 58 0 95 0 97 160 03 0 59 0 92 0 63 1 27 0 64 MAE MB 0 59 1 19 1 45 3 81 1 26 0 36 1 15 0 44 2 22 0 49 MB cc 0 07 0 01 0 05 0 18 0 03 0 07 0 04 0 10 0 02 0 08 cc RMSE 0 83 1 59 167 2 58 1 55 0 81 1 43 0 91 2 43 0 93 RMSE RMSE h 238 388 427 700 890 166 360 178 628 195 RMSE NS 110mm h 13916 105474 44429 18722 182541 24675 424 27425 20998 73522 NS NR 110mm h 1131 13432 3079 124 17766 43386 397 20876 1138 65797 NR ME 1 10mm h 1 63 E 1 00 0 84 1 19 1 74 0 84 1 03 0 54 1 49 ME sD 1 10mm h 1 05 1 66 1 80 1 61 1 65 477 1 48 2 43 2 37 sD MAE 1 10mm h 1 84 ma 1 76 1 60 se 2 02 147 1 92 176 MAE MB 1 10mm h 0 16 0 32 0 44 0 47 0 33 0 25 0 47 0 45 0 64 0 32 MB cc 1 10mm h 008 0 11 0 04 001 0 08 0 17 0 08 0 21 0 04 0 18 cc RMSE 1 10mm h 2 40 2 12 2 10 1 85 2 12 2 48 1 70 2 67 2 43 2 53 RMSE RMSE m h_ 99 124 117 137 Hae 97 106 137 186 111 RMSE NS 11566 28 0 20771 117 0 0 380 2827 NS NR mm 9 4 o 17 423 1 o 4 428 NR ME 210mm h 4 51
4. Precipitotion in mm sys gouge error corrected jor 5 9 August 2010 mm sau 7 IH 100 oem ee BS o ia 2o en so aw 20 an te nl A 10 3 s mH N La MI AAN r o 7 an al EE W i 1E 20E E E Figure 60 two day totals ending at 9th August 0 UTC interpolated on a 1 x1 evaluation grid as derived from SYNOP messages Global Precipitation Climatology Centre GPCC operated by DWD Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 91 183 Product Validation Report PVR 01 Figure 61 Hourly precipitation sum mm for H01 satellite data crosses time stamp 2010 08 07 05 43 UTC and for RADOLAN RW left filled raster 2010 08 07 05 50 UTC and station data right dots 2010 08 07 06 00 UTC Data used PR OBS1 data for eastern part of Germany in the given period were available for 5 43 UTC 5 57 UTC 7 03 UTC and 14 11 UTC Only these data are analysed in this case study Statistical score evaluation A first look to the results Fig 15 shows that rain rates detected by satellite product are in the same area of Germany as those indicated by the ground data In Table 1 2 the result of the categorical statistic of the validation with both RADOLAN and rain gauge data are listed The Probability Of Detection POD of precipitation gt 0 25 mmh gain for validation with RADOLA
5. pA E Figure 48 H01 image of August 14th 2010 at 6 06 left compared with upscaled radar at 6 05 right The scale corresponds to thresholds of 0 1 1 and 10 mm h 1 Scores evaluation The scores obtained for the present case study Table 1 are very good especially if compared with the long period scores In particular the product appears just very slightly overestimating while in the long period it is heavily underestimating and probability of detection is high with low false alarm ratio Sample 19 Mean error 0 31 Standard deviation 1 41 Mean absolute error 0 94 Multiplicative bias 1 18 Correlation coefficient 0 60 Root mean square error 1 37 URD RMSE 1 27 POD 0 91 FAR 0 42 CSI 0 55 Table 20 Scores obtained with the comparison with radar data in mm h 1 The time evolution of the fraction area with rain measured by radar gt 0 25 mm h and the Equitable Threat Score ETS is reported in next figures Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 i Product H01 PR OBS 1 Date 30 09 2011 Page 82 183 HSAF Product Validation Report PVR 01 m os a Lrotiititrotiil Fraction area 20 25 mm h 14 15 16 17 z 12E osE oe o4 ma oon E tot m gt Ra Lahirsbontinbuiten 14 15 16 17 18 Figure 49 Time evolution of fraction area with rain measured by radar gt 0 25 mm h and Equitable Threat Sc
6. Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 118 183 Product Validation Report PVR 01 Figure 94 Instantaneous precipitation field on 15 August 2010 at 07 05 UTC top row and 15 00 UTC second row derived by SHMU radar network left column and H01 product right column In the figure below the precipitation values are shown as satellite IFOVs projected over the radar composite domain White contoured circles represent 120 km rain effective range of the radars inside which data are included in the statistical scores computation In case of precipitation fields observed at 15 00 UTC prev figure second row very similar features can be observed as in the case above Compared to the radar field the intensities are overestimated by HO1 in each of the precipitation cells The maximum value observed by radars is 15 mm h while by HO1 it is 66 mm h Moreover these maxima were not detected within the same cell unlike in case of the 07 04 UTC passage It should be noted that lower radar intensities or undetected precipitation by radars especially near the Slovakia Poland border could be also caused by the radar beam blockage and or attenuation in the precipitation Scores evaluation Since only a small fraction of validation area white contoured circles in Fig 2 is covered by precipitation in case of the satellite passage at 07 04 UTC common statistical s
7. Radar rainfall estimation is influenced by several error sources that should be carefully handled and characterized before using it as a reference for ground validation of any satellite based precipitation products The main issues to deal with are e system calibration e contamination by non meteorological echoes i e ground clutter sea clutter clear air echoes birds insects W LAN interferences partial or total beam shielding rain path attenuation wet radome attenuation range dependent errors beam broadening interception of melting snow contamination by dry or melting hail hot spots variability of the Raindrop Size Distribution RSD and its impact on the adopted inversion techniques Some of them are typically handled by resorting to standard procedures some others requires the availability of dual polarized observations Generally speaking there are not correction methodologies applicable worldwide The knowledge of the radar system and the environmental conditions makes the difference when approaching such problems During the Precipitation Product and Hydrological Validation workshop held in Bratislava the 20 22 of October 2010 the Precipitation Product Validation Group PPVG has decided to set up a working group on the radar data use in the validation procedures This WG is not aimed at promoting the acceptance of shared data processing chain What really matter for us is the characterization of t
8. HSAF Product Validation Report PVR 01 The gauges density ranges between 7 for Bulgaria where only 3 river basins are considered to 27 km for Turkey These numbers should be compared with the decorrelation distance for precipitation patterns at mid latitude Usually the decorrelation distance is defined as the minimum distance between two measures to get the correlation coefficient Pearson Coefficient reduced to e A recent study on the H SAF hourly data for Italy shows this decorrelation distance varies from about 10 km in warm months where small scale convection dominates to 50 km in cold months when stratified and long lasting precipitation mostly occur In Figure 13 the value of the linear correlation coefficient is computed between each raingauge pair in the Italian hourly 2009 dataset as function of the distance between the two gauges aow 20w 10w B PPV Countries Rain gauges 10E 20E Figure 12 Rain gauge networks in PPVG Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 34 183 Product Validation Report PVR 01 200812 200901 200902 200903 200904 08 RO oo E 0 6 200910 200911 e OR 200905 E 200906 02 200907 E 200908 joe 200909 PEPE ee rE 20 40 60 80 km Figure 13 Correlation coefficient between raingauge pairs as function of the distances between the gauges Col
9. jeteorological stations precipiaton stations ciate statons Figure 20 Distribution of the raingauge stations of Iskar River Basin Goudenhoofdt E and L Delobbe 2009 Evaluation of radar gauge merging methods for quantitative precipitation estimates Hydrol Earth Syst Sci 13 195 203 d lidati Doc No SAF HSAF PVR 01 1 1 son iati HSAF Product Validation Report PVR 01 Issue Revision Index 1 1 Rn Product H01 PR OBS 1 Date 30 09 2011 Page 50 183 2 conventional fnal automati a Watershed of r Varbitsa up to sp Djebel Figure 22 Distribution of the raingauge stations of Varbica River Basin The instrument The following information should be provided in this section Tipping bucket with heating measures the precipitation with increments of 0 1 mm quality index of the measurements between 1 and 10 7 8 Product Validation Report PVR 01 Product H01 PR OBS 1 Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Date 30 09 2011 Je usar Page 51 183 Weighing type measurement with heating rim measures the precipitation with increments of 0 1 mm quality index of the measurements between 1 and 10 8 9 Conventional precipitation gauges type Wild measuring 24 hourly totals of precipitation The rainrate is given only by the automatic stations for a 60 minutes interval Those
10. 45 0 E 275 300E 325E 350E 375E 400E 425E Figure 42 Automated Weather Observation System AWOS station distribution in western part of Turkey The instruments The gauge type of the network is tipping bucket where each has a heated funnel The minimum detection capability of the gauge is 0 2mm per tip In the maximum capacity of the instrument is 720 mm h at most The operational accumulation interval is 1 minute so that alternative cumulation intervals such as 5 10 20 30 minutes are possible Data processing Quality control High quality of the ground data is critical for performing the validation of the precipitation products The validation results or statistics can provide meaningful feedbacks for the product developers and additionally the products can be used reliably only if there is a confidence present about the ground data at a certain level For this reason some predefined quality assurance QA tests are considered for the precipitation data in order to define the confidence level First of all a flagging procedure is defined as described in next table QA Flag Value QA Status Brief Description 0 Good Datum has passed all QA Test 1 Suspect There is concern about accuracy of datum 2 Failure Datum is unstable Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 74 183 HSAF Product Validation Report PVR 01 Ta
11. g wE we WE a Figure 6 The network of 3500 rain gauges used for H SAF precipitation products v tion The rain gauge networks of PPVG is composed of approximately 3500 stations across 6 Countries Figure 6 A key characteristic of such networks is the distance between each raingauge and the closest one averaged over all the instruments considered in the network and it is a measure of the raingauge density Instruments number and density are summarized in Table 3 Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 22 183 HSAF Product Validation Report PVR 01 Country Total number of gauges Average minimum distance km Belgium 89 11 2 Bulgaria arm 7 Germany 1300 17 Italy 1800 9 5 Poland 330 475 13 3 Turkey 193 27 Table 3 Number and density of raingauges within H SAF validation Group the number of raingauges could vary from day to day due to operational efficiency within a maximum range of 10 15 only in the Wallonia Region only in 3 river basins only covering the western part of Anatolia Most of the gauges used in the National networks by the PPVG Partners are of the tipping bucket type and hourly cumulated see Table 4 The rain gauge inventory Chapter 4 proposed by rain gauge WG annex 2 on the instruments the operational network and the approach to match gauge data
12. Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 13 183 Product Validation Report PVR 01 ORR OSI SAF PDF PEHRPP Pixel PMW PP PR PUM PVR RMI RR RU SAF SEVIRI SHM SSM I SSMIS SYKE Tes TKK TMI TRMM TSMS TU Wien U MARF UniFe URD UTC vis ZAMG Operations Readiness Review SAF on Ocean and Sea Ice Probability Density Function Pilot Evaluation of High Resolution Precipitation Products Picture element Passive Micro Wave Project Plan Precipitation Radar on TRMM Product User Manual Product Validation Report Royal Meteorological Institute of Belgium alternative of IRM Rain Rate Rapid Update Satellite Application Facility Spinning Enhanced Visible and Infra Red Imager on Meteosat from 8 onwards Slovak Hydro Meteorological Institute Special Sensor Microwave Imager on DMSP up to F 15 Special Sensor Microwave Imager Sounder on DMSP starting with S 16 Suomen ymp rist keskus Finnish Environment Institute Equivalent Blackbody Temperature used for IR Teknillinen korkeakoulu Helsinki University of Technology TRMM Microwave Imager on TRMM Tropical Rainfall Measuring Mission UKMO Turkish State Meteorological Service Technische Universitat Wien in Austria Unified Meteorological Archive and Retrieval Facility University of Ferrara in Italy User Requirements Document Universal Coordinated Time Visible
13. Levizzani V Mugnai A Laviola S Petracca M Sano P F Zauli CNR ISAC CNMCA VS EUMETSAT The results of WGs said that is not possible to consider radar and raingauge fields like the true and the accuracy indicated in the Table 57 RMSD is the degree of closeness of measurements of a quantity to its actual reference value The reference value of precipitation fields is not available and the measurement available is a limited picture of the reference Then it is important to evaluate which are the limits of available reference and then to understand the sources of errors of data used to evaluate the satellite outputs Taking in account this consideration a direct comparison of the requirements with the result of validation is not correct since they have different meanings the requirements indicate what error is allowed by the user to the satellite product to be significantly useful threshold or to produce a step improvement in the application target or to produce the maximum improvement before entering saturation optimal it is the RMSE of satellite v s reference the result of validation activities indicate the difference between the satellite measurement and the ground measurement utilized as a reference it is the RMSD of satellite v s reference Requirements Result of PR OBS1 threshold q target optimal validation Accuracy RMS gt 10 mm h 30 20 10 Accuracy RMS 1
14. HSAF Product Validation Report PVR 01 Table 40 The main statistical scores evaluated by PPVG for H01 during the summer period Rain rates lower than 0 25 mm h have been considered as no rain m 128 Table 41 The main statistical scores evaluated by PPVG for H01 during the autumn period Ra lower than 0 25 mm h have been considered as no rain Table 42 The main statistical scores evaluated by PPVG for H01 during one year of data 1st December 2009 30th November 2010 Rain rates lower than 0 25 mm h have been considered as no rain 129 Table 43 The averages POD FAR and CSI deduced comparing HO1 with radar data Table 44 The contingency table for the three precipitation classes defined in evaluated by comparing H01 with radar data Table 45 The averages POD FAR and CSI deduced comparing H01 with rain gauge dat 131 Table 46 The contingency table for the three precipitation classes defined in Section 3 evaluated by comparing H01 with rain gauge data Table 47 User requirement and compliance analysis for product H01 Table 48 Summary of the raingauge characteristics Table 49 Number and density of raingauges within H SAF validation Group Table 50 Data pre processing strategie Table 51 Matching strategies for comparison with H01 and H02 Table 52 Matching strategies for comparison with H03 and HOS Table 53 List of contact persons Table 54 Questionnaire Table 55 List of precipitation events selected for statistical analy
15. In this report only the Wideumont radar has been used The data of this radar are controlled in three steps Data processing First a long term verification is performed as the mean ratio between 1 month radar and gauge accumulation for all gauge stations at less than 120 km from the radar The second method consists in fitting a second order polynomial to the mean 24 h 8 to 8 h local time radar gauge ratio in dB and the range only the stations within 120 km and where both radar and gauge values exceed 1 mm are taken into account The third method is the same as the second but is performed on line using the 90 telemetric stations of the SETHY Ministry of the Walloon Region Corrected 24 h images are then Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 49 183 Product Validation Report PVR 01 ensar calculated New methods for the merging of radar and raingauge data have been recently evaluated Goudenhoofdt and Delobbe 2009 In this report only instantaneous radar images are used 4 6 Ground data in Bulgaria NIMH 4 6 1 Rain gauge The network The maximum number of available raingauges for this project is 37 distributed over 3 basins The average distance between stations is about 7 km with a very high variance Generally in the plain area distance is lower than in the mountainous areas Watershed of r Iskar up to station Novi Iskar
16. PR OBS 1 Date 30 09 2011 Page 3 183 Index 1 The EUMETSAT Satellite Application Facilities and H SAF 2 Introduction to product PR OBS 1 2 1 Sensing principle 2 2 Algorithm principle 2 3 Main operational characteristics 3 Validation strategy methods and tools 3 1 Validation team and work plan 3 2 Validation objects and problems 3 3 Validation methodology 3 4 Ground data and tools used for validation 3 5 Spatial interpolation for rain gauges 3 6 Techniques to make observation comparable up scaling technique for radar data 3 7 Temporal comparison of precipitation intensity 3 8 Large statistic Continuous and multi categorical 3 9 Case study analysis 4 Ground data used for validation activities 4 1 Introduction 4 2 Rain Gauge in PPVG 4 2 1 The network 4 2 2 The instruments 4 2 3 Data processing 4 2 4 Some conclusions 4 3 Radar data in PPVG 4 3 1 The networks 4 3 2 The instruments 4 3 3 Data processing 4 3 4 Some conclusions 4 4 Rain gauge and radar data integrated products in PPVG 4 4 1 INCA system 4 4 2 RADOLAN system 4 4 3 Some conclusions 4 5 Ground data in Belgium IRM 4 5 1 Radar data 4 6 Ground data in Bulgaria NIMH 4 6 1 Rain gauge 4 7 Ground data in Germany BfG 4 7 1 Rain gauge 4 7 2 Radar data 4 8 Ground data in Hungary OMSZ 4 8 1 Radar data 4 9 Ground data in Italy DPC Uni Fe 4 9 1 Rain gauge 4 9 2 Radar data Doc No SAF
17. For the later overpass however the opposite effect can be observed The precipitating area observed by ground stations is overestimated by PR OBS 1 leaving unrecognized only one spot of rain in the NE Poland The maxima of rainfall are generally properly located however its spatial distributions are more fuzzy right panel One of the reasons for that might be time shift of 7 and 4 minutes respectively in those cases Statistical scores The ability of PR OBS 01 product to recognize the precipitation was analysed using dichotomous statistics parameters performed for all overpasses available for the 17 of May 2010 The 0 25mm h threshold was used to discriminate rain and no rain cases In the next table the values of Probability of Detection POD False Alarm Rate FAR and Critical Success Ratio CSI are presented Par ter Scores Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 115 183 HSAF Product Validation Report PVR 01 POD 0 36 FAR 0 44 CsI 0 28 Table 33 Results of the categorical statistics obtained for PR OBS 1 Low value of POD than the high value of FAR indicate that the product ability to recognize the stratiform precipitation is not satisfactory The quality of PR OBS 1 in estimating the convective precipitation is presented on the next figure One can easily notice that PR OBS 1 underestimates th
18. Table 56 Mean Residual and Mean Absolute Residual values obtained for three algorithms for spatial interpolation using cross validation approach Table 57 Simplified compliance analysis for product PR OBS 1 2 3 Table 58 Errors of the ground reference provided by all validation groups Table 59 RMSE and standard deviation of interpolation algorithms for 3 different regular grids 182 List of figures Figure 1 Conceptual scheme of the EUMETSAT application ground segment Figure 2 Current composition of the EUMETSAT SAF network in order of establishment Figure 3 Geometry of conical scanning for SSMIS Figure 4 Flow chart of the precipitation rate processing chain from SSM I and SSMIS Figure 5 Structure of the Precipitation products validation team Figure 6 The network of 3500 rain gauges used for H SAF precipitation products validation Figure 7 The networks of 54 C band radars available in ther H SAF PPVG Figure 8 Geometry of conical scanning left and IFOV right of SSMI Figure 9 Left Gaussian filter Right section of gaussian filter a Figure 10 Left Original Gaussian matrix Right Reduced matrix to dimensions M xK Figure 11 Main steps of the validation procedure in the PPVG Figure 12 Rain gauge networks in PPVG Figure 13 Correlation coefficient between raingauge pairs as function of the distances between the gauges Colours refer to the months of the year 2009 Figure 14 Radar netwo
19. The main aims of INCA precipitation for PPV group are to identify the INCA precipitation products which can be considered as precipitation ground reference and used for validation of H SAF products both instantaneous and cumulated precipitation fields to identify the techniques of comparison the INCA precipitation products with satellite precipitation products to develop the code to be used in the PPVG for a correct verification of satellite precipitation product performances with INCA to produce a well referenced documentation on the methodology defined to perform H SAF products validation based on these techniques and INCA precipitation products Activities First step e identify experts contact persons inside INCA community which can provide information on INCA system like methods of precipitation data integration product formats data coverage products availability and quality e collect and study INCA methods and products and to consider how these methods meet requirements of H SAF precipitation products validation e compare precipitation field reconstructed using radar data raingauges data and INCA products for some case studies Start Time End time December 2010 March 2011 First Report 31 of March 2011 Second step e develop common upscaling software tools for proper upscaling of identified INCA products into native H SAF product s grids Doc No SAF HSAF PVR 01 1 1 Issue Rev
20. 03 11 64 S 13 09 8 58 Bms B ME Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 126 183 Product Validation Report PVR 01 sD 0 82 156 0 79 1 20 5 04 0 00 36 531 5 32 sD MAE 11 39 164 11 95 13 09 8 58 37 085 13 31 MAE MB 0 00 0 09 0 00 0 05 0 09 0 25 0 013 0 09 MB cc 0 77 0 77 0 00 00 0 109 0 77 cc RMSE 1141 1247 1167 12 03 14 06 8 58 52 056 14 40 RMSE RMSE 100 84 100 95 92 30 75 30 _ Sm Bae RMSE Table 38 The main statistical scores evaluated by PPVG for H01 during the winter period Rain rates lower than 0 25 mm h have been considered no rain In Table 38 it can be seen that the scores obtained by radar data are quite different from the scores obtained by rain gauge data for light precipitation rain rate lt 1 mm h Besides the RMSE of this precipitation class has the highest value This is due probably to the difficult not only of the satellite product but also of rain gauge and radar instruments to measure small precipitation intensities This aspect has been highlighted also in Section 4 2 and 4 3 on ground data description Germany is the only country which has performed the validation using both radar and rain gauge data The results reported in Table 38 appear quite different in particular for precipitation with rain rate lt 1 mm h
21. 2040 Figure 80 Precipitation rate from the Hungarian radar network at its original resolution at UTC right panel HO1 product left panel at 6 45 UTC Conclusions Note that the same blue colours in the radar and the HO1does not correspond to the same rain rate H01 dark blue 1 2 mm light blue 4 5 mm radar dark blue 0 1 mm light green 5mm The HO1 well detects the precipitation area but HO1 overestimates the precipitation values Product Validation Report PVR 01 HSAF Product H01 PR OBS 1 Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Date 30 09 2011 Page 106 183 5 5 Case study analysis in Italy Uni Fe It is here reported the case study analysis of 06 July 2010 on Italian territory performed by University of Ferrara 5 5 1 Case study 6 of July 2010 Description On July 06 the Azores anticyclone avvected very warm and moist air on the Tyrrhenian coasts where a weak trough induced cyclonic circulation and instability in the early morning After 10 00 UTC deep convections initiated in the Po Valley and in central Italy along the Apennines chain causing waterspouts along the northern Adriatic coasts hailfalls and supercells storms in Central Italy O6JUL2010 00Z 500 hPa Geopotential gpdm und Bodeniar uch hPa SEVIRI HR VIS image at 12 00 on July 06 shows a well developed convective cluster over central Italy while small scale scattered convec
22. 6 04 6 04 sD MAE 12 02 12 70 13 143 12 61 14 24 14 36 14 25 14 25 MAE MB 0 15 0 07 0 112 0 09 0 05 0 16 0 06 0 06 MB cc 0 27 0 11 0 274 0 04 0 05 0 04 0 04 0 04 cc RMSE 12 70 13 58 14 328 13 48 15 38 16 74 15 59 15 59 RMSE RMSE 89 66 92 99 9180 92 95 45 83 28 93 93 82 RMSE Table 40 The main statistical scores evaluated by PPVG for H01 during the summer period Rain rates lower than 0 25 mm h have been considered as no rain In Table 40 it can be see that the scores obtained by radar data are quite different from the scores obtained by rain gauge data for light precipitation rain rate lt 1 mm h Besides the RMSE of this precipitation class has the highest value As it has been said in Section 6 2 this is due probably to the difficult not only of the satellite product but also of rain gauge and radar instruments to measure small precipitation intensities The statistical scores evaluated for precipitation classes 2 and 3 using both rain gauge and radar data are very similar The best scores have been calculated on coastal areas by Turkey A general precipitation underestimation by H01 is reported in Table 38 using both rain gauge and radar data for rain rate greater than 1 mm h Besides a precipitation overestimation by H01 has been found for light precipitation rain rate lt 1mm h The Slovakian team has obtained the worst results also during summer period An investigation on this result is in pro
23. Figure 23 Network of rain gauges in Germany Figure 24 Pluvio with Remote Monitoring Module 4 7 2 Radar data Radar based real time analyses of hourly precipitation amounts for Germany RADOLAN is a quantitative radar composite product provided in near real time by DWD Spatial and temporal high resolution quantitative precipitation data are derived from online adjusted radar measurements in Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 53 183 HSAF Product Validation Report PVR 01 real time production for Germany Radar data are calibrated with hourly precipitation data from automatic surface precipitation stations The combination of hourly point measurements at the precipitation stations with the five minute interval radar signals of the 16 weather radars C Band Doppler provides gauge adjusted hourly precipitation sums for a 1km x 1km raster for Germany in a polar stereographic projection Radar site Longitude E Radar site Longitude E M nchen 48 20 14 11 36 46 Rostock 54 10 35 12 03 33 Frankfurt 50 01 25 08 33 34 Ummendorf 52 09 39 11 10 38 Hamburg 53 37 19 09 59 52 Feldberg 47 52 28 08 00 18 Berlin 52 28 43 13 2317 Eisberg 49 32 29 12 24 15 Tempelhof Essen 51 24 22 06 58
24. It seems to confirm the difficulty of the ground networks to measure light rain rate The statistical scores evaluated for precipitation classes 2 and 3 using both rain gauge and radar data are very similar A general precipitation underestimation by H01 is reported in Table 38 using both rain gauge and radar data Only for light precipitation rain rate lt 1mm h there is an overestimation by H01 compared with radar data The best scores using radar data have been evaluated in Belgium RMSE Cl1 238 Cl2 99 CI3 100 and using rain gauge in Italy RMSE Cl1 195 Cl2 111 Cl3 92 The Slovakian team in general obtained the worst results An investigation on this result is in progress In the frame of this investigation Slovakian team tested the conformity of validation software with common validation methodology Moreover the meteorological situation was analysed in the first half of February and it was shown that data from H01 shows high precipitation intensities LO0mm h in some regions of western Slovakia and central Europe One of these regions was measured also by western Slovak radar Maly Javornik This radar captured precipitation intensities of 1 mm h in average Using MSG RGB imagery only low level clouds or clear atmosphere was detected in relevant timeframes To ensure about the meteorological situation also rain gauge measurements map was checked and it is in agreement with radar measurements After these tests prelimina
25. Page 104 183 Product Validation Report PVR 01 TAN IRS a ler UN B y tonnen S SS Figure 77 Synoptic chart at 00 UTC on 18th of July 2010 Data used A m CPR N gt N T Figure 78 H01 product left panel Cloud type from NWC SAF right panel Precipitation rate from the H ungarian radar network at its original resolution in middle Comparison In this cold front weather situation during the whole day H01 did not detected the middle size thunderstorms Conclusions Note that the same blue colours in the radar and the HO1does not correspond to the same rain rate H01 dark blue 1 2 mm light blue 4 5 mm radar dark blue 0 1 mm light green 5mm The H01 in most cases well detects the precipitation area but the middle size thunderstorms were not detected Improvement of the H01 spatial resolution would help the detection 5 4 3 Case study 10 of September 2010 Description Acyclone over Mediterranean causes precipitation in Central and South Europe Lot of precipitation was measured mainly in the central part in Hungary Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 105 183 HSAF Product Validation Report PVR 01 ID J R SI HELYZET 2010 09 10 00 WC erator con ones Wits Data used uae konpazit Esi Kompozit Canh
26. URD POD FAR CsI 399 436 0 88 1 48 1 28 1 55 0 70 1 71 142 0 81 0 12 0 73 Table 36 Statistic scores for HO1 Conclusions H01 product is successful to catch rainy area In other words frontal system is well described generally by this product algorithm in terms of areal matching For quantitative estimate there is an overestimation Performance of h01 algorithm is acceptable 5 9 Conclusions Twelve case study analysis of HO1 have been here reported for 2010 Stratiform and convective precipitations during summer and winter periods have been analysed in different countries Rain gauges with 10 minutes refresh time radar data and nowcasting tools have been used to highlight different characteristics of the satellite product The case studies here proposed have pointed out that different statistical score values are obtained during summer and winter period In summer when more convective events occur all the countries have observed that HO1 reproduces the rainfall patterns and amounts with quite good confidence About the convective systems it has been observed that H01 did not well detect the small medium size thunderstorms This effect is due to a typical size of these convective cells which does not exceed the H01 SSMI and SSMI S IFOV Capturing of convective cores by satellite IFOV or in upscaled radar image is strongly dependent on the mutual position of convective core and sa
27. and E Roulin 2008 and A Rinollo An optimization of this code to be used by all the partners of the PPVG represent one of the next step Anyway different approaches over different Countries are leading to very similar values in the considered skill scores indicating probably two things 1 none of the considered approaches can be considered as inadequate and more important 2 the differences between ground fields and satellite estimates are so large that different views in the data processing do not results in different numbers 4 3 Radar data in PPVG In this section the complete inventory of the radar data used in the PPVG with some considerations are reported as first results of the Radar WG Annex 3 4 3 1 The networks In the HSAF project satellite based precipitation estimations are compared regularly with the radar derived precipitation fields However radar rainfall products are influenced by several error sources that should be carefully analyzed and possibly characterized before using it as a reference for validation purposes However we have to emphasize that the radar data used for validation purposes is not developed by the validation groups themselves They are developed within specialized radar working teams in many of the countries It is not the aim of the PPVG to improve the radar data used however it is specifically expected from the current activities to characterize radar data and error sources of the ground data
28. coming from the radar networks of the PPVG Main error sources of radar rainfall estimations are listed in the Radar Working Group description document Annex 3 1 system calibration 2 contamination by non meteorological echoes i e ground clutter sea clutter clear air echoes birds insects W LAN interferences partial or total beam shielding rain path attenuation wet radome attenuation range dependent errors beam broadening interception of melting snow contamination by dry or melting hail hot spots variability of the Raindrop Size Distribution RSD and its impact on the adopted inversion techniques Oo Sey Or Moreover several studies have been on radar quality assessments like S alek M Cheze J L Handwerker J Delobbe L Uijlenhoet R 2004 Radar techniques for identifying precipitation type and estimating quantity of precipitation COST Action 717 Working Group 1 A review Luxembourg Germany or Holleman I D Michelson G Galli U Germann and M Peura Quality information for radars and radar data Technical rapport 2005 EUMETNET OPERA OPERA_2005_19 77p Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 38 183 Product Validation Report PVR 01 500 L oos MB PPV Countries Weather radar units Horizontal beam extent of 100 km O Horizontal beam extent of 200 km Figure 14 Radar networks i
29. composite production Uinmendort Rostock Preprocessing II 2 x in 60 min summation to hourly composits statistical clutter suppression interpolation 2 Preprocessing III of Radar data with station data every 60 min smoothing precalibration Calibration of Radar data with station data every 60 min calculation of calibration params and interpolation calibrations a a intersection of different calibration ite m TET A D f en A procedures for best result Tene Fecuanetl SOONG eee a Figure 26 Flowchart of online calibration RADOLAN DWD 2004 4 8 Ground data in Hungary OMSZ 4 8 1 Radar data The network The main data used for validation in Hungary would be the data of meteorological radars There are three C band dual polarized Doppler weather radars operated routinely by the OMSZ Hungarian Meteorological Service The location and coverage of the three Hungarian radars are shown here below the measurement characteristics are listed in Table 16 All three radars are calibrated periodically with an external calibrated TSG the periodicity is kept every 3 months Figure 27 location and coverage of the three Hungarian radars d lidati Doc No SAF HSAF PVR 01 1 1 e HSAF Product Validation Report PVR 01 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 55 183 Year of installation Location Radar type Paramet
30. distribution of 10 minute precipitation obtained from RG data measured at closest to the given time slot is presented The RG derived precipitation map was prepared using Near Neighbor method PG rainat 15082010 1800 UTC 2 a a ese aA ahii S e F 0 a bia 36 ape 5 u 7 Me ota rate 1 08200 1459 UTC g ge 2 TERREN aa M ga t th k A Figure 85 PR OBS 1 at 1459 UTC on the 15th of August 2010 right panel and 10 minute precipitation interpolated from RG data from 1500 UTC left panel On both maps the precipitating areas reveal the lightning activity seen on the previous figure however the PR OBS 1 tends to overestimate the preci ting area Yet the precipitation measured Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 111 183 Product Validation Report PVR 01 Je usar in the central Poland was missed by H01 On the other hand the precipitating area seen on the satellite derived rainfall map in the North Poland right panel correspond to lightning activity observed in this region previous figure Fact that this rainfall is not present on the RG map may be explained by the ground network density that is rather sparse in this region It should be also stressed that the maximum of convective precipitation seen on the satellite derived map is shifted westward and more fuzzy than the one on the ground based precipitati
31. radiances at all frequencies and polarisations of the SSM I SSMIS channels e convoluting the monochromatic radiances with the instrument model so to simulate brightness temperatures comparable with those that would be measured from the satellite e finally collect the simulated Tg s in the Cloud Radiation Database Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 17 183 Product Validation Report PVR 01 ensar When the satellite passes the acquired data are pre processed by the instrument processor and made available for the precipitation generation chain that includes e an initial preparation of the dataset to be processed sea land mask emissivity preventive classification of cloud nature e the retrieval algorithm that searches for the maximum likelihood solution in the hydrometeor profiles available in the CRD also using the error structure available in a database e the uncertainty estimator that appends the retrieved precipitation rate with information on likely error this information is also used for updating the error structure database Q OFF LINE ACTIVITY gt SE REAL TIME ACTIVITY gt PRECIPITATION RATE ERROR ESTIMATE Figure 4 Flow chart of the precipitation rate processing chain from SSM I and SSMIS 2 3 Main operational characteristics The operational characteristics of PR OBS 1 are discussed
32. 1 Date 30 09 2011 Page 45 183 Product Validation Report PVR 01 Je usar 1997 Since June 2005 areal spatial and temporal high resolution quantitative precipitation data are derived from online adjusted radar measurements in real time production for Germany The data base for the radar online adjustment is the operational weather radar network of DWD with 16 C band sites on the one hand and the joined precipitation network of DWD and the federal states with automatically downloadable ombrometer data on the other hand In the course of this the precipitation scan with five minute radar precipitation data and a maximum range of 125 km radius around the respective site is used for the quantitative precipitation analyses Currently from more than 1000 ombrometer station approx 450 synoptic stations AMDA 1 ll and AMDA III S of DWD approx 400 automatic precipitation stations AMDA III N of DWD approx 150 stations of the densification measurement network of the federal states the hourly measured precipitation amount is used for the adjustment procedure In advance of the actual adjustment different preprocessing steps of the quantitative radar precipitation data are performed These steps partly already integrated in the offline adjustment procedure contain the orographic shading correction the refined Z R relation the quantitative composite generation for Germany the statistical suppression of clutter the gradient
33. 1 Date 30 09 2011 Page 57 183 HSAF Product Validation Report PVR 01 4 9 Ground data in Italy DPC Uni Fe 4 9 1 Rain gauge The network The maximum number of available raingauges is about 1800 irregularly distributed over the surface On the average however a number of stations have low quality data failure or data transmission problems and their data are missing 9999 recorded This number of no data stations is highly varying on hourly daily basis and ranges from few units to a hundred In case of data acquired but not transmitted recorded the first transmitted measure is the cumulated value over the time when the data were not transmitted The average minimum distance between closest stations is about 9 5 km with a very high variance in some regions such as Tuscany in central Italy it is below 5 km while in Emilia Romagna Po Valley it is more than 20 km A study of the decorrelation distance between stations as function of the mutual distance has been carried out for the 2009 dataset The decorrelation distance is defined as the minimum distance between two observations that makes the Pearson correlation coefficient between the two measures decrease below e Results are shown in Figure 28 where the decorrelation distance is plotted as function of the distance between stations It appears that there is a large variability of this parameter from higher values around 60 km for cold months when large p
34. 2010 Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 96 183 HSAF Product Validation Report PVR 01 POD 0 65 0 48 0 11 0 55 0 43 0 07 FAR 0 79 0 53 0 90 0 77 0 70 0 96 CsI 0 19 0 31 0 05 0 20 0 21 0 03 Table 27 Results of the categorical statistic of the validation for whole month June 2010 The contingency tables next two figures for both kinds of validation data show that in the lowest three classes more than 50 of PR OBS 1 data fall in the same class better results are with radar data RRsatin class1 MRRsatinclass2 M RRsatinclass3 MH RRsat in class 4 100 0 0 0 w n n 60 a Z Zo 0 0 x0 a 20 2 1 1 o o class 1 class 2 class 3 class 4 ss 1 class 2 class 3 class 4 Oe RR lt 0 25 025 RR lt 1 1 lt RR lt 10 AR gt 10 O lt cRACO 25 O25e RRct 1 lt RA lt 10 RR gt 10 RR Radar RR Radar Figure 67 Contingency table statistic of Rain Rate mmh 1 for PR OBS1 vs radar data 00 0 w 0 w n w 0 w Zo Zo a r 0 x 2 2 1 1 o o clase 1 class 2 class 3 class 4 clase 1 class 2 class 3 class 4 Oc RRCO25 0 25 lt RR lt 1 1 lt RR lt 10 RR gt 10 Oe RR lt 0 25 0 25 lt RA lt 1 lt RR lt 10 RR gt 10 RR Rain Gauge RR Rain Gauge Figure 68
35. 5 km from the next hourly rain of 2 5 km from the hours are compared IFOV centre amount IFOV centre 3 6 12 and 24 hours Italy Nearest neighbour the average rainrate Nearest neighbour rain amounts in the over a given hour Is same number of compared to next hours are compared hourly rain amount 3 6 12 and 24 hours Poland mean gauges value each overpass is mean gauges value rain amounts in the over the pixel area compared to the over the pixel area same number of next 10 minutes rain hours are amount compared 3 6 12 and 24 hours Turkey weighted mean each overpass is weighted mean rain amounts in the semi variogram compared to the semi variogram same number of gauges value corresponding 1 gauges value over hours are compared centred on satellite IFOV minute rain rate centred on satellite IFOV 3 6 12 and 24 hours Belgium and Bulgaria use raingauges only for cumulated precipitation validation Table 52 Matching strategies for comparison with H03 and HOS Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 i Product H01 PR OBS 1 Date 30 09 2011 Page 142 183 HSAF Product Validation Report PVR 01 Conclusions After this inventory some conclusion can be drawn First it seems the raingauge networks used in this validation activities are surely appropriated for the validation of cumulated products 1 hour and hi
36. Contingency table statistic of rain rate mmh 1 for PR OBS1 vs rain gauge data Results of the continuous statistic show positive Mean Error ME in both periods with both kind of ground data in the first class which means that H SAF product overestimates small precipitation amounts The opposite is for the other classes Standard deviation SD with 2 4 mmh for the class RR gt 0 25 mmh is the highest for validation with rain gauge for 3 June the correlation coefficient CC with 0 43 for radar data is the best analogue to the results for POD see above For detection of precipitation RR gt 0 25 mmh there are nearly the same results for both kind of ground data and for both periods which means the chosen period is representative for June 2010 RRimmh 3 June 2010 June 2010 3 June 2010 June 2010 rain rain rain gauge radar gauge radar gauge radar rain gauge radar Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 97 183 Product Validation Report PVR 01 0 57 0 94 30 80 1 82 1 48 217 2 22 0 00 23 02 6 10 1 50 1 23 1 56 1 68 30 80 2 18 18 10 72 0 64 0 68 0 72 0 62 0 00 a 0 14 0 26 0 12 0 36 0 22 0 29 0 04 0 07 1 92 1 64 2 24 2 41 30 80 gt 29 33 12 03 Table 28 Continuous statistic Conclusions The resu
37. Date 30 09 2011 Page 80 183 Product Validation Report PVR 01 Je usar Here are two examples of H01 files compared with radar data upscaled to the same grid The first example is of the afternoon of August 15 and the second in the early morning of August 16 next two figures T ji TEP T f 1 f fi Figure 46 H01 image of August 15th 2010 at 16 41 left compared with upscaled radar at 16 40 right The scale corresponds to thresholds of 0 1 1 and 10 mm h 1 The product matches the rainfall pattern quite good E 2 fe a SS ain Figure 47 H01 image of August 16th 2010 at 4 56 left compared with upscaled radar at 4 55 right The scale corresponds to thresholds of 0 1 1 and 10 mm h 1 Also in this case the matching is quite good It is possible to see that in both cases the matching is quite good with correct reconstruction and estimation of rainfall patterns and amounts and in particular of the delineation between light precipitation and precipitation greater than 1 mm h 1 In analyzing the H01 files it is noted that many of them have not the standard resolution 128 pixels per line but a lower one Here is an example with 60 pixel per line early morning of August 14th Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 81 183 Product Validation Report PVR 01
38. Italy uses 3 6 12 24h accumulations without gauge correction in Hungary 3 6 12 24h data is used but only the 12h and 24 hourly accumulations are corrected by rain gauges in Poland and Slovakia no rain gauge correction is applied Poland has only 6 and 24 hourly data Turkey has 3 6 12 24h data and applies rain gauge correction for 1 hourly data It is important to note that techniques used for accumulation are numerous even within the same country the can differ from one accumulation period to another E g in Hungary the 3 6h accumulations are derived from summing up the interpolation of the 15minute frequent measurements into 1 minute intervals whereas the 12 and 24 h accumulations are summed up from 15 minute measurements but corrected with rain gauge data All above implies that more probably the quality and error of rainfall and rain rate accumulations is differing from one country to another and cannot be homogeneously characterized Conclusion of the questionnaire Maintenance All the contributors declared the system are kept in a relatively good status Correction factors for error elimination These correction factors are diverse in the countries not homogeneous distribution of correction methods all contributors compensate for non meteorological echoes Clutter RLAN interferences implemented in Hungary Slovakia in development Poland and Slovakia correct attenuation In other countries it is not accounted for So
39. PVR 01 Results of the horizon model for Mal Javorn k and Koj ovsk ho a radar sites are shown on Figure 103 To evaluate the radar visibility over the whole radar network composite picture of minimum visible height above the surface was created and is shown on Figure 104 Layer visibility 500 1000 Layer visibility 500 1000m Terrain elevation Above the surface Above the surface Minimum visible height Minimum visible height Minimum visible height Minimum visible height above the sea level above the surface above the sea level above the surface igure 103 Radar horizon model output for Maly Javornik left and KojSovska hofa right radar sites Colour scale on left corresponds to the products showing heights above the sea level scale on right corresponds to the products showing heights above the surface Figure 104 Composite picture of minimum visible height above the surface over the whole radar network Compositing algorithm selects the minimum value from both radar sites In next step minimum visible heights above the rain gauge stations were derived from the composite picture Distribution of rain gauges according to the minimum visible height of radar beam is shown on Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 153 183 HSAF Product Validation Report PVR 01 next figure It should be noted that while radar beam
40. Report PVR 01 Je usar Use of spatialization technique Due to the time and space structure of precipitation and to the sampling characteristics of both the precipitation products and observations used for validation care has to be taken to bring data into comparable and acceptable range At a given place precipitation occurs intermittently and at highly fluctuating rates Various maps time series analysis statistical and probabilistic methodologies are employed in the validation procedure classically but some additional new aspects such as the spatial coverage verification model of point cumulative semivariogram PCSV approach Sen and Habib 1998 are proposed for usage in this work Each precipitation product within the H SAF project represents a foot print geometry Among these HO1 and H02 products represent an elliptical geometry while HO3 and HOS have a rectangular geometry On the other hand the ground observation rain gauge network consists of point observations The main problem in the precipitation product cal val activities occurs in the dimension disagreement between the product space area and the ground observation space point To be able to compare both cases either area to point product to site or point to area site to product procedure has to be defined However the first alternative seems easier The basic assumption in such an approach is that the product value is homogenous within the product fo
41. Turkey has 3 6 12 24h data and applies rain gauge correction for 1 hourly data It is important to note that techniques used for accumulation are numerous even within the same country the can differ from one accumulation period to another E g in Hungary the 3 6h accumulations are derived from summing up the interpolation of the 15minute frequent measurements into 1 minute intervals whereas the 12 and 24 h accumulations are summed up from 15 minute measurements but corrected with rain gauge data All above implies that more probably the quality and error of rainfall and rain rate accumulations is differing from one country to another and cannot be homogeneously characterized 4 3 4 Some conclusions Maintenance All the contributors declared the system are kept in a relatively good status Correction factors for error elimination These correction factors are diverse in the countries not homogeneous distribution of correction methods all contributors compensate for non meteorological echoes Clutter gt RLAN interferences implemented in Hungary Slovakia in development gt Poland and Slovakia correct attenuation In other countries it is not accounted for gt Some of the countries are testing new procedures for dealing with VPR Italy and Partial Beam Blockage PBB effects VPR Vertical Profile of Reflectivity used in Turkey v This means that the corresponding rainfall estimates are diverse and the estimation of thei
42. U T C by the polarimetric radar located in Gattatico Emilia Romagna Italy a Doc No SAF HSAF PVR 01 1 1 Product Validation Report PVR 01 a Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 66 183 va 25 03 2007 09 30 GMT gat 25 03 2007 09 30 GMT gat oz Gone maer Idro Meteo Figure 36 Measured upper panel and VPR corrected lower panel PPI of reflectivity observed on 03 25 07 at 0930 U T C by the polarimetric radar located in Gattatico Emilia Romagna Italy Rainfall estimation Quantitative rainfall estimation is one of the first application of the radar network The estimation of rainfall at the ground takes advantage of the dense network of raingauges spread all over Italy This network is one of the most dense in the world with more than 1700 gages and it is used for tuning and correcting the operational Z R algorithms of non polarimetric radars In order to evaluate the benefits of upgrading the new radar installations to full polarimetric radars and for considering the benefit of existing polarimetric radars many studies have been carried on by Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 67 183 HSAF Product Validation Report PVR 01 Research Centres and Regional Authorities belonging to the network e g Silvestro et al 2008 As an example in
43. ZHAIL RTR CTR 3h 6h 24h acc WRN precip 1h acc SRI 1km 2km agl Processing Doppler method clutter Clutter filtering Clutter Removal VPR chain removal atenuation frequency domain IIR Correction Z R correction yes filter A 200 b 1 6 VPR gt No Atmospheric attenuation Z R a 200 b 1 6 correction Z R a 200 b 1 6 RLAN filtering in development Is any quality NO in development NO NO map available Description of National composite National composite CAPPI Projection instanteneous SRI Projection CAPPI 2 km Azimuthal Equidistant radar product azimutal equidistant Projection Mercator Resolution 250 m used in HSAF standard elipsoid Resolution 1 km Threshold Rain Validation Resolution 1 km Threshold 31 5 dBZ Gauge Correction with Threshold 5 dBZ No No rain gauge correction limited number of rain gauge correction gauges Description of Acc Periods 1 6 24h Acc periods 3 6 12 Acc periods accumulated National composite 24h 1 3 6 12 24h radar product PAC Projection National composite Projection Azimuthal used in HSAF azimuthal equidistant CAPPI 2 km Equidistant Validation standard elipsoid Projection Mercator Resolution 250 m Resolution 1 km Resolution 1 km Threshold Treshold 0 1 mm No Threshold 31 5 dBZ Rain gauge correction rain gauge correction No rain gauge correction applied for 1h Rain Acc 11 Annex 4 Study
44. analysis Both components large statistic and case study analysis are considered complementary in assessing the accuracy of the implemented algorithms Large statistics helps in identifying existence of pathological behaviour selected case studies are useful in identifying the roots of such behaviour when present This Chapter collects the case study analysis performed by PPVG on H01 for the year 2010 The Chapter is structured by Country Team one section each The analysis has been conducted to provide information to the User of the product on the variability of the performances with climatological and morphological conditions as well as with seasonal effects Each section presents the case studies analysed giving the following information description of the meteorological event comparison of ground data and satellite products visualization of ancillary data deduced by nowcasting products or lightning network discussion of the satellite product performances indications to satellite product developers indication on the ground data if requested availability into the H SAF project In the future the PPVG will test the possibility to present case study analysis in the test sites indicated by the hydrological validation team in order to provide a complete product accuracy and hydrological validation analysis to the users Product Validation Report PVR 01 HSAF teen Product H01 PR OBS 1 Doc No
45. are d lidati Doc No SAF HSAF PVR 01 1 1 e HSAF Product Validation Report PVR 01 Issue Revision Index 1 1 pnd Product H01 PR OBS 1 Date 30 09 2011 Page 21 183 e ground data error analysis radar and rain gauge e point measurements rain gauge spatial interpolation e up scaling of radar data versus SSMI grid temporal comparison of precipitation products satellite and ground e statistical scores continuous and multi categorical evaluation e case study analysis 3 4 Ground data and tools used for validation Both rain gauge and radar data have been used until now for H01 validation As said in the previous section during the last Precipitation Product Validation Workshop held in Bratislava 20 22 October 2010 it has been decided to set up Working Groups to solve specific items of the validation procedure and to develop software used by all members of the validation cluster A complete knowledge of the ground data characteristics used inside the PPVG has been the first item of the working groups this is necessary to understand the validation results and to define the procedure to select the most reliable data to represent a ground reference A complete report on the results obtained by the Working Group on rain gauge radar and ground data integration are reported in the Chapter 4 with a complete inventory of the ground data used within the PPVG 500 1000km PPV Countries Rain gauges
46. but have also been used as a ground reference for statistical analysis of the PR OBS 2 product during selected precipitation events Case study PR OBS 1 vs INCA 15 August 2010 15 00 UTC This is the first case study elaborated at SHMU which compares the PR OBS 1 product with precipitation fields produced by the INCA system In order to make precipitation fields from the microwave instruments and ground observations at 1 km resolution comparable the INCA Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 163 183 HSAF Product Validation Report PVR 01 precipitation fields have been upscaled into the PR OBS 1 native grid using the Gaussian averaging method Ellipses in next figure represent the satellite instrument IFOVs with colour corresponding to the upscaled radar rain gauge and INCA analysis rain rate value in case of next figure from a to c respectively or the satellite rain rate value in case of next figure part d As can be seen in next figure part b the rain gauge network captured intense precipitation near the High Tatras mountain in the northern part of Slovakia where only low precipitation rates were observed by radars next figure part a The resulting INCA analysis is shown in next figure part c The corresponding PR OBS 1 field next figure part d shows overestimation even when compared with the rain gauge adjusted field o
47. by Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 177 183 HSAF Product Validation Report PVR 01 spatialisation method Therefore this method seems to be the best for creating the geographical distribution of H SAF products error for countries characterized by terrain geographical configuration similar to the Polish one Conclusions The analysis performed for ME of H 05 3 h cumulated product obtained using data from Polish network of rain gauges showed that Natural Neighbour interpolation method seems to be the best one for creating maps of H SAF products error However application of Natural Neighbour method does not allow for extrapolating the distribution beyond the area defined by stations what is a disadvantage of this methods As the maps are to be created for the whole H SAF domain presented above results should be verified over other countries Therefore in the next step of WG5 activities the study will be performed for other countries and for the errors calculated using radar data Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 178 183 Product Validation Report PVR 01 e usar 15 Annex 8 Comments on the Validation Results for Products PR OBS 1 PR OBS 2 And PR OBS 3 Please note following paper is an historical record Casella F Dietrich S
48. civil protection purposes National Weather Radar Coverage 125km Figure 30 Italian radar network coverage The existing sixteen C band weather radars that belong to Regional Authorities ENAV and AMI are listed below Bric della Croce Owner Regione Piemonte Polarization on going upgrade to polarimetry Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 60 183 Product Validation Report PVR 01 Je usar Settepani Owner Regione Piemonte and Regione Liguria Polarization dual San Pietro Capofiume and Gattatico Owner Regione Emilia Romagna Polarization dual Monte Macaion Owner Regione Trentino Alto Adige and Provincia autonoma Trento Polarization single Teolo and Loncon Owner Regione Veneto Polarization single Monte Midia Owner Regione Abruzzo Polarization single Monte Rasu Owner Regione Sardegna Polarization single Fossolon di Grado Owner Regione Friuli Venezia Giulia Polarization single Linate and 12 Fiumicino Owner ENAV Polarization single Brindisi Owner Italian Air Force Polarization single Grazzanise Owner Italian Air Force Polarization single Pisa Owner Italian Air Force Polarization single Istrana Owner Italian Air Force Polarization single The first C band radar of new generation directly managed by DPC located in Tuscany Italy is operational si
49. data in the PPVG In this chapter the first example of precipitation fields integration has been provided Section 4 4 3 INCA and RADOLAN products The INCA system a tool for the precipitation products validation is available in Slovakia and Poland in both countries being run in pre operational mode In Germany similar precipitation analysis system called RADOLAN is being run operationally This tool is already used for validation of the H SAF precipitation products in Germany The study performed in the PPVG INCA WG showed that the accuracy and reliability of the raingauge stations significantly affect final precipitation analysis of the INCA or INCA like systems In order to solve this problem an automated blacklisting technique is going to be developed at SHMU currently blacklisting is used in manual mode The next step will be to develop the software for up scaling the INCA precipitation field into the satellite product grid The grids of INCA and RADOLAN have similar horizontal resolution to the common radar grid The up scaling software will allow to provide case study analysis and statistical score evaluation for future considerations on the opportunity to use these precipitation integration products in the H SAF validation programme 5 Validation results case study analysis 5 1 Introduction As reported in the Chapter 3 the common validation methodology is composed of large statistic multi categorical and continuous and case study
50. grid is a 5x5 km regular grid with 240 columns and 288 lines Moreover a Digital elevation model is used to provide a mask of Italy in order to 1 screen out sea pixels too far from the coastlines and 2 process the pixels with the elevation above sea level 4 9 2 Radar data The network The Italian radar data have been not used for the validation of the current version of H01 but the verification of the satellite product with those data is in progress The results will be presented at the next review of the project The Italian Department of Civil Protection DPC is the authority leading the national radar coverage project in order to integrate the pre existent regional systems made of ten C band fixed regional installations five of them polarimetric and one transportable X band polarimetric radar two systems owned by the Italian company for air navigation services ENAV and three managed by the Meteorological Department of the Italian Air Force AMI After its completion the Italian radar network will include twenty five C band radars including seven polarimetric systems and five transportable dual polarized X band radars see next figure The Italian Department of Civil Protection is developing the radar network in Southern Italy and thanks also to the fruitful collaborations with Regional Authorities ENAV and AMI integrated all the existing radars in one national network with a clear advantage for both severe weather monitoring and
51. ground station representativeness have been applied for instance by applying Gaussian filters but the statistics of residual errors are not available this problem affects radar to a minor extent than raingauge that may explain why comparisons with radar finally are not worse than with rain gauge the raingauge s representativeness of IFOV pixel geolocation is retrieved by using the information made available by satellite owners and it is not perfect it is necessary to evaluate how much mislocations impact on the accuracy of the comparison The effect is clearly larger for convective precipitation This may explain why product PR OBS 3 is apparently performing better than PR OBS 1 and PR OBS 2 the high resolution minimizes mislocation errors Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 180 183 HSAF Product Validation Report PVR 01 similarly time mismatching is a source of error more effective for convective precipitation hence the advantage of PR OBS 3 and also of radar contributing to reduce the effect of intrinsic lower accuracy parallax errors introduce mislocation of satellite precipitation with associated comparison errors larger for convective precipitation because of deeper penetration in the upper troposphere These and maybe other error sources need to be analyzed in detail in order to determine their contribution t
52. gt 0 25 mmh in comparison with both radar and rain gauge data was quite good for higher rain rates the probability of detection is lower although lower false alarms The quantitative precipitation amounts were overestimated for small amounts and underestimated generally for rain rates greater 1mmh 0 80 1 18 14 78 1 96 1 84 1 75 1 92 0 00 5 79 4 86 4 88 1 55 1 67 1 45 1 68 14 78 11 70 11 28 9 88 0 51 0 56 0 62 0 54 0 04 0 22 0 22 0 28 0 17 0 44 0 27 0 28 0 07 0 19 0 13 2 28 2 25 1 93 2 25 14 78 13 05 12 28 10 89 Table 25 Continuous statistic Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 94 183 HSAF Product Validation Report PVR 01 5 3 2 Case study 3 of June 2010 River Danube catchment Description On the beginning of June 2010 the weather was determined by a low pressure area over eastern part of Central Europe Wet hot air out from Mediterranean Sea was directed around the low pressure vortex Bergthora contraclockwise out from north to Bavaria and arrived overhead the near ground cold area By this air advection on 3 of June fell lon lasting rain in the catchments of the rivers Regen and Danube and caused a Danube river flood Precipitation amounts over 24 hours reached between 80 mm and 155 mm Gew sserkundlicher Monatsbe
53. main stati mm h have been considered as no rain ical scores evaluated by PPVG for H01 during the spring period Rain rates lower than 0 25 In Table 39 it is possible to see that the scores obtained on land areas comparing HO1 with radar data are similar to the scores obtained with rain gauge for all the precipitation classes The best scores have been calculated on coastal areas by Turkey A general precipitation underestimation by H01 is reported in Table 39 using both rain gauge and radar data for all precipitation classes The best scores using radar data have been evaluated in Germany RMSE Cl1 192 Cl2 96 C13 93 and using rain gauge in Italy RMSE Cl1 156 Cl2 89 Cl3 83 The statistical scores obtained during the spring period Table 39 are very similar to the ones obtained during the winter period for rain rates greater than 1 mm h Table 38 6 2 3 The summer period PR OBS 1 BE DE HU TU DE spring spring spring spring spring spring spring spring spring spring spring 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 Version 1 4 radar radar radar radar radar gauge gauge gauge gauge gauge gauge NS lt imm h 4950 36210 8667 5224 55051 13988 506 17327 41407 41407 NS NR lt 1mm h 3793 29711 12905 3890 50299 24207 2165 12434 49208 49208 NR ME lt 1mm h 0 19 0 23 0 16 0 91 0 04 0 18 0 13 0 10 0 15 0 15 ME SD 0 20 0 84 1 37 2 06 1
54. next figure is shown the cumulated rainfall estimates versus gage measurements obtained for the event observed on 06 01 2006 by the dualpolarized C Band radar of Mt Settepani The figure shows the comparison between a multi parameter algorithm that uses polarimetric data Silvesto et al 2008 and a simple ZR relationship Marshall Palmer bis ZR200 MultiParametric 200 Cumulated rain radar mm 100 150 200 250 Figure 37 Cumulated radar rainfall estimates versus gage measurements for the event observed on 06 01 2006 by the dualpolarized radar located in Settepani Liguria Italy 4 10 Ground data in Poland IMWM 4 10 1 Rain gauge The network The maximum number of rain gauges in the Polish ATS Automatic Telemetric Station national network is 950 Each ATS post is equipped with two independent rain gauges of the same sort One of them is heated during the winter period and the other one is not Therefore precipitation information is derived from 475 points at the time Fact that rainfall is measured by two equally sensitive instruments two meters away from each other at the same post enables to apply simple in situ data quality control during summertime During winter non heated rain gauge is covered with a cup to prevent it from being clogged by the ice and damaged Because of that the precipitation information derived from ATS network in winter cannot be verified using this method It can be stated that dur
55. not sufficient and they do not reflect real ground reference of precipitation fields obtained results can be considered as a ceiling guess of radar measurements quality indicator with regards to terrain visibility This result includes also the error of rain gauge network itself Also averaged mean error root mean square error and relative root mean square error values were computed for 68 rain gauge stations located in radar horizons Averaged mean error 0 184 mm h for instantaneous or 4 42 mm for 24 hours cumulated precipitation Averaged RMSE 0 395 mm h for instantaneous or 9 48 mm for 24 hours cumulated precipitation Averaged URD_RMSE 69 3 for 24 hours cumulated precipitation It should be noted that all computations in this study were based on 24 hour cumulated precipitation and only re calculated into instantaneous precipitation Values of errors in case of instantaneous precipitation can be significantly higher because of short time spacing Therefore it is planned in the future to calculate errors of radar measurements separately for instantaneous and for cumulated precipitation 12 Annex 5 Working Group 3 INCA Precipitation for PPV PROPOSAL The precipitation ground reference can be only based on certain conceptual models The validation activity inside H saf project is composed by hydrological and product validations Precipitation captured by river basin is transformed by set of processes into the river discharge T
56. of March 2011 Final Report 31 of July 2011 Third step code possible Matlab realization for e ground data quality check e comparison between rain gauge and satellite products Start Time End time June 2011 November 2011 Codes delivery and related documentation 30 of November 2011 Doc No SAF HSAF PVR 01 1 1 H SAF Issue Revision Index 1 1 haapaa Product H01 PR OBS 1 Date 30 09 2011 Page 137 183 Product Validation Report PVR 01 Composition of the working group Coordinator Federico Porc University of Ferrara supported by Silvia Puca DPC Italy Participants from Belgium Bulgaria Germany Poland Italy Turkey FIRST REPORT Coordinator Federico Porc University of Ferrara supported by Silvia Puca DPC Italy Participants Emmanuel Roulin and Angelo Rinollo Belgium Gergana Kozinarova Bulgaria Claudia Rachimow and Peter Krahe Germany Rafal lwanski and Bozena Lapeta Poland Ibrahim Sonmez and Ahmet Oztopal Turkey Emanuela Campione Italy 0 Introduction This document reports on the outcomes of the inventory completed about the raingauges used as ground reference within the validation groups Moreover some general conclusion is drawn based on the raingauges validation activities carried on in the last years by the Validation Group of H SAF The inventory was structured in three sections dealing with the instruments used the operational networ
57. pe 50 H 08 0 A O 25 lt PRe1 1 lt PR lt 10 PR210 PR20 25 0 25sPR lt 1 1sPR lt 10 PR210 PR20 25 Class mm h Class mm h Probability of detection False alarm rate 1 1 09 09 08 08 07 07 a 08 Braingauge 0 6 raingauge g 05 mradar Z 05 radar 04 INCA 0 4 inca 03 03 02 0 2 a 0 1 01 E o o RR 2 0 25 mm h RR 21 mm h RR gt 0 25 mm h RR21mmh Threshold Threshold Critical success index 1 09 08 07 _ 06 Braingauge Bos radar 04 INCA 03 o2 ot o RR 0 25 mm h Threshold RR 21 mmh Figure 117 As in previous figure except for event 4 stratiform Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 i Product H01 PR OBS 1 Date 30 09 2011 Page 169 183 HSAF Product Validation Report PVR 01 Conclusion The INCA system as a potential tool for the precipitation products validation is available in Slovakia and Poland in both countries being run in pre operational mode It is still relatively new system undergoing continuous development More sophisticated algorithms of the precipitation analysis e g assimilation of the 3 D radar data can be expected from its development in frame of the ongoing INCA CE project In Germany similar precipitation analysis system called RADOLAN is being run operationally This tool is already used for validation of the H SAF precipitation products in this country
58. precipitation showed that PR OBS 1 has difficulties with proper recognition and estimation of this type of precipitation 5 7 Case study analysis in Slovakia 5 7 1 Case study 15 of August 2010 Description During the day a cold front was moving over Slovakia territory towards North East next figure The cold front was accompanied by thunderstorms and occasional torrential rainfall causing severe floods in some river catchments in the western half of Slovakia lela 15 08 2010 00 00 UTC Figure 93 Synoptic situation on 15 August 2010 at 0 00 UTC Data used Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 i Product H01 PR OBS 1 Date 30 09 2011 Page 117 183 HSAF Product Validation Report PVR 01 The HOa data from two satellite passages over the SHMU validation area on 15 August 2010 have been selected the DMSP16 observation at 07 04 UTC average observation time of the SHMU validation area and the DMSP15 observation at 15 00 UTC As ground data the instantaneous precipitation field derived by the SHMU radar network is used The closest coincident fields 5 min time frequency to the satellite passages have been selected from 07 05 UTC and 15 00 UTC The radar composites used consist of data from two radars one is situated at Maly Javornik and the second at Kojsovska hola The rule of maximum value selection is applied in the composition The original spatial resolution of the radar fi
59. radar at Kojsovska hola right Figure 41 Map of relative RMSE left and Mean Error right over the SHM radar composite 72 Figure 42 Automated Weather Observation System AWOS station distribution in western part of Figure 43 HO1 and H02 products footprint centers with a sample footprint area as well as the Awos ground observation sites Figure 44 Meshed structure of the sample H01 and H02 products footprint Figure 45 Synoptic situation on 15 August 2010 at 6 UTC zoom in the surface map Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 9 183 HSAF Product Validation Report PVR 01 Figure 46 HO1 image of August 15th 2010 at 16 41 left compared with upscaled radar at 16 40 right The scale corresponds to thresholds of 0 1 1 and 10 mm h 1 The product matches the rainfall pattern quite good 80 Figure 47 H01 image of August 16th 2010 at 4 56 left compared with upscaled radar at 4 55 right The scale corresponds to thresholds of 0 1 1 and 10 mm h 1 Also in this case the matching is quite Figure 48 H01 image of August 14th 2010 at 6 06 left compared with upscaled radar at 6 05 right The scale corresponds to thresholds of 0 1 1 and 10 mm h 1 Figure 50 Surface map on 22 August 2010 at 06 UTC MSLP and synoptic observations c 82 Figure 49 Time
60. rain gauges according their altitude above the sea level To simulate terrain visibility by meteorological radars Shuttle Radar Topography Mission SRTM data were used as an input into radar horizon modeling software developed in SHM Details about SRTM can be found at http en wikipedia org wiki Shuttle_Radar_Topography_Mission or directly at http www2 jpl nasa gov srtm SRTM model provides specific data set of terrain elevations in 90 m horizontal resolution in the whole HSAF area where HSAF validation by radars is performed Modelling software parameters were adjusted for single radar according real scanning strategy Radar Site Mal Javorn k Koj ovsk ho a Tower height 25m 25m Range 1200pixels 240km 1200pixels 200km Resulted resolution 200m pixel 166 67m pixel Min elevation 0 1 deg 0 8 deg Refraction 1 3 standard atmosphere 1 3 standard atmosphere Elevation step 0 01 deg 0 01 deg Azimuth step 1 40 deg 1 40 deg Layer minimum 500m 500m Layer maximum 1000 m 1000 m Max displayed height 5000 m 5000 m Radar horizon model provides the following outputs maps of radar range e terrain elevation e minimum visible height above the sea level e minimum visible height above the surface Layer visibility defined by minimum and maximum levels Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 152 183 HSAF Product Validation Report
61. sD MAE 0 62 0 61 0 83 1 28 0 72 067 oet 0 70 067 0 67 MAE MB 0 63 0 56 1 30 2 88 094 jose o75 0 81 071 071 MB cc 0 08 0 06 0 13 0 21 009 007 0 09 Jo08 oo7 o07 cc RMSE 0 95 0 90 1 39 2 25 1 14 1 02 o78 117 11 05 1 05 RMSE RMSE __ 200 192 12 313 80 487 12 218 64 156 35 254 79 229 RMSE NS 2656 16684 10585 3737 35216 15170 1067 7273 29505 29505 NS NR 1563 15893 5219 929 23604 18721 1593 4688 27839 27839 NR ME 1 38 1 46 0 15 2 57 1 01 1 55 1 33 0 91 1 36 1 36 ME sD 1 46 1 67 2 76 4 30 2 00 212 1 87 222 209 209 sD MAE 2 01 1 81 2 01 3 21 1 92 208 4 74 194 198 1 98 MAE MB 0 40 0 30 0 91 2 49 053 034 foss Joas joss 0 39 MB cc 0 15 0 09 0 19 0 26 012 J009 0 24 foie Jori foii jcc RMSE 2 59 2 23 2 78 5 01 249 263 229 245 252 252 RMSE RMSE 105 39 95 83 156 85 318 04 MS 102 12 88 94 123 92 106 RMSE NS 8 168 275 0 627 470 2 0 795 795 NS NR 26 96 12 0 134 297 37 0 341 341 NR ME 11 79 12 67 12 819 12 51 14 23 14 36 14 25 14 25 ME sD 3 68 4 69 6 401 464 559 854 604 604 sD MAE 12 02 12 70 13 143 1261 1424 1436 14 25 14 25 MAE MB 0 15 0 07 0 112 009 005 foie f 0 06 006 mMB cc 0 27 0 11 0 274 0 04 005 0 04 f 0 04 J0 04 cc RMSE 12 70 13 58 14 328 13 48 1538 1674 15 59 15 59 RMSE RMSE 210mm h_ 89 66 92 99 91 80 95 45 83 28 93 93 82 RMSE Table 39 The
62. satellite pixel the ground rain rate value was calculated as a mean of all rain gauges measurements recorded within that pixel e Temporal matching satellite derived product is combined with the next corresponding ground measurement As the ground measurements are made with 10 minute time resolution the maximum interval between satellite and ground precipitation is 5 minutes 4 11 Ground data in Slovakia SHMU 4 11 1 Rain gauge The network In Slovakia there are overall 98 automatic rain gauge stations potentially available for the H SAF project The real number of usable gauges varies with time because on average about 20 of them are out of operation Mean minimum distance between rain gauges in the complete network is 7 74 km Map of the rain gauge network in Slovakia containing also climatological and selected hydrological stations is shown in next figure Figure 39 Map of SHM rain gauge stations green automatic 98 blue climatological 586 red hydrological stations in H SAF selected test basins 37 The instruments Type of all the automatic rain gauges is tipping bucket without heating of the funnel The gauges are able to measure precipitation rates ranging from 0 1 to 200 mm h at 10 min operational accumulation interval Shorter accumulation interval of 1 min is also possible which makes the instruments suitable for case studies in the H SAF project The data processing The rain gauge data are not used at S
63. sio in po 00900 Tanesi sonor 11900051 tongue asane 12E 1599201 e gasaveaoua ax ongtude 15712435E asa 22098620E 25 9998067916 E mmn attude 695719 szan arses 4so0a7019292 ax attude sa 70660N 5 029 so ss0a1 591000570006 N space resolution 1m hm hon 1m wan ceca Compania or T naiona Composite ot national Compose ot Znatna Compoate ofS eat dete umber of radars ie neterodt radars radars radars international radars lumber of precipitation stations oland ont ST HU CHEN ZAMG Number ot precipitation stan 1200 475 Poland oniy wnt mp Btcklst for precipitation stations 2 ves Yes Yes eso Densty ot raingauge stations ep of dona 9 preciptiaton stations 2 1o To mo insanianeous precipan basea OT output aata on raingaugenetwor imo resoiution smn No Yes 15min Yes 15 minute imenn instantaneous precipitation basea oniy en radar network ime resolution smn no Yes S minute Yes 5 minuto timatines Instantaneous preciptation based on combined raingauge and radar smn Yes 10 minutes Yes 5 minutes Yes 5minutes network tmo resolution tines naat pao besed eniyan Yes min 5 min available Yes min 5 min available Faingauge network tmo tervals S min 1 3 8 12 1824 hours no piii maina timelines 1 3 6 12 24 hours 1 3 6 12 24 hours Cumutavepreciptation based oniy on Yes min min avalale ves min S min avaable radar network time intervals timelines 5 1 3 6 12 18 24 hours No 1 3 6 12 24 hours 13 612 24 hours r prectpietion based on Yes min 1
64. stations are located in Varbica and Chepelarska river basins There are no automatic stations in Iskar river basin Data processing There is quality control on the data In this Project the point like gauges data are interpolated for using Co kriging interpolation of the ground measurements taking into account orography 4 7 Ground data in Germany BfG The H SAF products are validated for the territory of Germany by use of two observational ground data sets SYNOP precipitation data based on the network of synoptical stations provided by the German Weather Service DWD and RADOLAN RW calibrated precipitation data based on the radar network of DWD and calibrated by DWD by use of measurements at precipitation stations Time interval Annotation 6h 12h hourly Number Resolution Synoptical stations 200 Precipitation 1100 stations RADOLAN RW Near real time Near real time Automatic precipitation stations 16 German radar 1 hour Near real time Quantitative radar composite sites product RADOLAN RW Radar data 1 kmx 1 km after adjustment with the weighted mean of two standard procedures Table 14 Precipitation data used at BfG for validation of H SAF products 4 7 1 Rain gauge The network The data used are compiled from 1300 rain gauges About 1000 are operated by DWD while about 300 are operated by other German authorities The average minimum distance between stations is 17 km The
65. the same two areas of Germany as those indicated by the ground data In Table 7 8 the result of the categorical statistic of the validation with both radar and rain gauge data are listed Statistical scores The results for validation with radar data for 5 6 December are better than for the whole month December Probability Of Detection of precipitation RR gt 0 25 mm was 0 32 with less False Alarm Rate of 0 73 and Critical Success Index is 0 17 more worse than summer results Since there were not detected hourly precipitation data in both radar and PR OBS1 this class has no amounts and for rain gauge we have got false alarm rate of 100 3087 3087 661 0 32 0 30 0 27 0 23 0 73 0 94 0 68 0 92 0 17 0 05 0 17 0 06 x Table 29 Results of the categorical validation statistic of case study5 6th December 2010 L csi ii 0 11 0 05 0 00 0 12 0 07 Table 30 Results of the categorical statistic of the validation for whole month December 2010 The contingency tables next two figures for both kinds of validation data show that only in the lowest class more than 50 of PR OBS 1 data fall in the same class Generally in winter we have an underestimation by satellite data B RRsatin class1 MIRRsatinclass2 RRsatinclass3 W RRsat in class 4 1w 1o 0 so 2 w n 0 eo 0 Za Eo 0 x a 2 2 1 w o gt class 4 clase 2 clase 3 clase 4 class 1 class 2 class 3 class 4 O
66. the classification algorithm starting from the radar variables observed on 09 14 08 at 0500 U T C by the polarimetric radar operated by Piemonte and Liguria regions Reconstruction of vertical profile of reflectivity Rainfall estimation might be heavily perturbed by the presence of melting snow due to the enhancement of reflectivity factor caused by the increase in size and dielectric constant without a corresponding increase of rain rate This well known problem is usually handled by retrieving the Vertical Profile of Reflectivity VPR and correcting the observed measures Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 65 183 HSAF Product Validation Report PVR 01 The algorithm developed by ARPA SIM for VPR retrieval and correction is currently under test in order to be implemented within the DATAMET system It is based on the computation of mean VPR shape Germann and Joss 2002 and assuming it to be uniform in the whole radar domain on the retrieval of the reflectivity at the desired level by the simple adding of a constant quantity in dBZ units The original algorithm is modified and integrated with a VPR diagnosis and analysis phase to handle different operative problems Fornasiero et al 2008 As an example next figure shows the measured upper panel and VPR corrected lower panel PPI of reflectivity observed on 03 25 07 at 0930
67. with the satellite estimates in the PPVG has pointed out that the rain gauge networks available in the PPVG are surely appropriated for the validation of cumulated products 1 hour and higher but probably not for instantaneous estimates The comparison of satellite rain rate with hourly cumulated ground measurements surely introduces intrinsic errors in the matching scores that can be estimated as very large The validation of instantaneous estimates should be carried on only when gauges cumulation interval is 10 to 15 minutes as in Poland Values cumulated over shorter intervals 5 or even one minute as it is done in Turkey are affected by large relative errors in cases of low moderate rain rates Studies are undertaken in order to quantitatively estimate the errors introduced in the validation procedure comparing the instantaneous satellite precipitation estimation with the rain gauge precipitation cumulated on different intervals Moreover the revisiting time 3 4 hours of HO1 makes impossible or not reasonable to validate the product for 1 24 hours cumulated interval The WG has also pointed out that different approaches for the estimates matching are considered in the PPVG One of the next step of the WG will be to define in collaboration with the GeoMap WG Annex 7 the spatial interpolation technique and to develop the related software to be used in side the PPVG Country imum detectable Maximum detectable H
68. year 6 2 5 The annual average Table 42 The main statistical scores evaluated by PPVG for H01 during one year of data 1st December 2009 30th DIC 09 DIC 09 DIC 09 DIC 09 DIC 09 DIC 09 DIC 09 DIC 09 DIC 09 DIC 09 NOV 10 _ NOV 10 NOV 10 NOV 10 NOV 10 NOV 10 NOV 10 NOV 10 NOV 10 NOV 10 Version 1 4 radar radar radar radar radar gauge gauge gauge gauge gauge NS 29288 227633 95385 48615 400921 81665 2557 69632 56782 210636 NR 22513 135195 54308 17061 229077 126858 9283 98986 41485 276612 ME 0 10 0 04 0 27 0 96 0 10 0 24 0 05 0 19 0 05 O sD 0 20 1 05 1 50 1 94 1 14 0 90 1 20 1 02 1 25 1 01 MAE 0 60 0 71 0 84 1 27 0 77 0 63 0 73 0 67 0 73 0 66 MB 0 77 0 93 1 51 3 00 1 20 0 55 1 09 0 63 1 11 0 68 cc 0 09 0 07 0 12 0 22 0 09 0 07 0 08 0 10 0 07 0 08 RMSE 0 91 1 09 1 53 217 1 26 0 95 1 22 1 06 1 28 1 05 URD 216 82 244 55 345 20 490 69 25W02 202 35 263 55 216 35 291 99 NS 25907 174323 83357 35510 320651 70563 5015 51881 42540 169999 NR 8715 80036 23716 3492 115959 100734 6157 42917 15994 165802 ME 1 10mmh 1 18 1 20 0 01 2 33 0 84 1 64 0 80 1 00 0 78 1 36 sD 1 10mm h 1 39 1 82 2 89 4 51 2 09 2 16 RaT 2 50 1 90 2 23 MAE 1 10mm h 1 74 1 72 1 94 3 09 1 81 2 14 1 71 2 00 1 56 2 03 MB 1 10mm h 0 47 0 44 0 98 2 30 0 61 0 34 0 65 0 50 0 58 0 42 cc 1 10mm h 0 22 0 20 0 21 0 24 0 20 0 17 0 22 0 22 0 18 0 19 RMSE 1 10mm h 2 31 2 21 2 96 5 11 2 46 2 73 2 47 2 72 2 08
69. 0 1 1 and 10 mm h In this case the satellite reconstructs correctly the higher rainfall zones missing the lower rainfall ones Scores evaluation The scores obtained for this case study are given in next table Sample 18 Mean error 0 45 Standard deviation 0 64 Mean absolute error 0 74 Multiplicative bias 0 48 Correlation coefficient 0 42 Root mean square error 0 90 URD RMSE 1 14 POD 0 24 FAR 0 27 CsI 0 22 Table 22 Scores obtained with the comparison with radar data in mm h 1 Unlike in the summer case here the product shows the same underestimation pattern as in the long period analysis and a low probability of detection It can be added that the radar data cumulated over 24h revealed to be underestimated compared with interpolated rain gauge data This of course only worsens the conclusion about the product The time evolution of the fraction area with rain measured by radar gt 0 25 mm h and the Equitable Threat Score ETS is reported in next figure Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 89 183 Product Validation Report PVR 01 w gt ov Frection area 20 25 mm h t ES Figure 58 Time evolution of fraction area with rain measured by radar gt 0 25 mm h and Equitable Threat Score ETS during the present case study Conclusions From visual and statistics
70. 0 1 mm 2000 Y 120 1440 Germany 0 05 mmh 3000 Y 60 Italy 0 2 mm N A Y N 60 Poland 0 1mm N A y 10 Turkey 0 2 mm 720 Y 1 Table 8 Summary of the raingauge characteristics only 300 out of 1800 gauges are heated information not available at the moment a value about 300 mmh can be assumed for tipping bucket raingauges Most of these shortcomings could be avoided by using instruments based on different principle or mechanisms The German network and a part of the Bulgarian network as an example are equipped by precipitation weighting gauges that allow continuous precipitation both solid and liquid measurements with higher accuracy Other option could be the use of disdrometers that give more information about the precipitation structure and a more accurate rain rate measure In table 5 relevant characteristics of the raingauges used in the different countries are reported 4 2 3 Data processing The partners of the Validation Group have been using a variety of different strategies to treat gauge data and to compare them with satellite estimates Some are using interpolation algorithms to get spatially continuous rainfall maps while others process directly the measurements of individual gauges All the data in the network except for cold months in Poland are quality controlled there is no information about the techniques used but usually quality control rejects data larger than a given threshold and in c
71. 0 31 ae 0 87 0 97 0 98 0 83 0 76 0 65 0 62 0 62 0 55 0 58 0 65 0 66 0 75 Sa 0 09 0 02 0 02 0 10 0 14 0 20 0 22 0 25 0 29 0 24 0 20 0 19 0 16 Table 45 The averages POD FAR and CSI deduced comparing H01 with rain gauge data Radar data PR lt 0 25 0 25 lt PR lt 1 00 1 00 lt PR lt 10 00 74 58 mm h 10 00 lt PR PR lt 0 25 0 25 lt PR lt 1 00 1 00 lt PR lt 10 00 Satellite data 10 00 lt PR Table 46 The contingency table for the three precipitation classes defined in Section 3 evaluated by comparing H01 with rain gauge data The averages of POD 0 32 FAR 0 72 and CSI 0 18 have been obtained using rain gauge data on one year of data 1 December 2009 30 November 2010 In Table 46 it is possible to see that 90 of no rain is correctly classified by H01 There is a general precipitation underestimation by the satellite product H01 6 4 User requirement compliance In the next table the statistical scores obtained by the yearly validation of HO1 with radar and rain gauge data are reported The statistical scores reach the thresholds stated in the User Requirements in all cases using rain gauge data as ground reference and in all cases except for precipitation lower than 1 mm h using radar data as ground reference table 6 9 This result might be explained by considering the hi
72. 0 m for the Italian radars 500 m for one of the Hungarian radars and 250m for the other two Polish radars can work with 125 m and 250 m resolution and in Turkey it is 250 m for all the radars The scan strategies within the PPVG countries are well balanced and similar to each other though they vary from one radar to the other even within countries All radars are regularly maintained and calibrated which is a good indicator of the continuous supervision of quality of radar data and the important element to sustain radar data quality 4 3 3 Data processing The Tab 08 is provided to summarize the available products generated from radar measurements and the processing chain used to generate them Finally the list of the radar products used for the validation work is included in the last row Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 40 183 Product Validation Report PVR 01 Je usar Radar rainfall products are obtained after processing the measured radar reflectivity at different elevations of the radar scan strategy After each elevation the PPI Plan Position Indicator products and the CAPPI Constant Altitude PPI products are calculated PPI is the measurement of the radar antenna rotating 360 degrees around the radar site at a fixed elevation angle CAPPI products are derived from this by taking into account the radar displays which give
73. 0 minutes Yes min 5 min available Yes min 5 min available combined raingauge and radar 5 min 1 36 12 18 24 hours Salabe in future e paeas k oor comme raingaugo aaa sill nt aeza h 136224 cases wile sot no 2932009 case2 No 1282010 eases No 2082010 eases no 15 1682010 cases no avanabiny af own sofware Tor upscaling INCA data into native satellite grid baa hid ae ne nie oe yes No No No m yes no wo no os no no o no 0s yes No o No 08 yes no no no Table 13 INCA Questionnaire It is also here presented the first case study elaborated at SHM Annex 5 which compares the H01 product with precipitation fields produced by the INCA system In order to make precipitation fields from the microwave instruments and ground observations at 1 km resolution comparable the INCA precipitation fields have been upscaled into the PR OBS 1 native grid using the Gaussian averaging method see Chapter 3 Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 47 183 HSAF Product Validation Report PVR 01 Ellipses in Figure 18 represent the satellite instrument IFOVs with colour corresponding to the upscaled radar rain gauge and INCA analysis rain rate value in case of Fig 1 a b and c respectively or the satellite rain rate value in case of Fig 1 d As can be seen in Fig 1 b the rain gauge network captured intense precipitatio
74. 01 PR OBS 1 Date 30 09 2011 Page 42 183 Validation 600m resolution Threshold No S60 Activities No rain gauge correction Resolution 2 km Threshold 7dBZ Rain gauge correction applied for 12 24 hourly data Table 11 Inventory of the main radar data and products characteristics in Belgium Italy and Hungary POLAND SLOVAKIA TURKEY List of Available PPI PCAPPI RHI MAX CAPPI 2 km MAX Products EHT SRI PAC VIL VVP Etops PPI HWIND VSHEAR HSHEAR PPI 0 2 CAPPI LTB SWI MESO WRN Base VIL List of non operational Cmax ETOPS products LMR CMAX Hmax EBASE UWT VAD SHEAR SWI VIL RAIN Acumulation MESO ZHAIL RTR CTR Precip Intensity 1h 1 3 6 12 24h WRN 3h 6h 24h acc precip 1h acc SRI 1km 2km agl Processing chain Doppler method clutter Clutter filtering Clutter Removal VPR removal atenuation frequency domain IIR Correction Z R A 200 correction yes filter b 1 6 VPR gt No Atmospheric Z R a 200 b 1 6 attenuation correction Z R a 200 b 1 6 RLAN filtering in development Is any quality NO in development NO NO map available Description of National composite SRI National composite CAPPI Projection instanteneous Projection azimutal CAPPI 2 km Azimuthal Equidistant radar product equidistant standard Projection Mercator Resolution 250 m used in HSAF elipsoid Resolution 1 Resolution
75. 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 135 183 HSAF Product Validation Report PVR 01 8 Annex 1 Status of working group H SAF Product Validation Programme Working Group 1 Rain gauge data Coordinator Federico Porc University of Ferrara supported by Silvia Puca DPC Italy Proposal completed first report available Participants Emmanuel Roulin and Angelo Rinollo Belgium Gergana Kozinarova Bulgaria Claudia Rachimow and Peter Krahe Germany Emanuela Campione Italy Rafal Iwanski and Bozena Lapeta Poland Ibrahim Sonmez and Ahmet Oztopal Turkey Working Group 2 Radar data Coordinators Gianfranco Vulpiani DPC Italy and Eszter Labo HMS Hungary Proposal completed first report available Participants Rafal Iwanski Poland Emmanuel Roulin and Angelo Rinollo Belgium Marian Jurasek Luboslav Okon Jan Kanak Ladislav Meri Slovakia Firat Bestepe and Ahmet Oztopal Turkey Working Group 3 INCA products Coordinator Jan Kanak SHMU Slovakia Proposal completed first report available Participants Claudia Rachimow and Peter Krahe Germany Rafal Iwanski and Bozena Lapeta Poland Silvia Puca Italy Working Group 4 COSMO grid Coordinators Angelo Rinollo RMI Belgium supported by Federico Porc University of Ferrara and Lucio Torrisi CNMCA Italy Proposal completed First report avail
76. 02 0 99 0 74 1 10 1 02 1 02 sD MAE 0 62 0 61 0 83 1 28 0 72 0 67 0 61 0 70 0 67 0 67 MAE MB 0 63 0 56 1 30 2 88 0 94 0 66 0 75 0 81 0 71 0 71 MB cc 0 08 0 06 0 13 0 21 0 09 0 07 0 09 0 08 0 07 0 07 cc RMSE 0 95 0 90 1 39 2 25 1 14 1 02 0 78 117 1 05 1 05 RMSE RMSE lt imm h 200 192 12 313 80 487 12 246 218 64 156 35 254 79 229 RMSE NS 1 10mmh 2656 16684 10585 3737 35216 15170 1067 7273 29505 29505 NS NR 1 10mm h 1563 15893 5219 929 23604 18721 1593 4688 27839 27839 NR ME 1 10mm h 1 38 1 46 0 15 257 1 01 1 55 1 33 0 91 1 36 1 36 ME sD 1 10mm h 1 46 1 67 2 76 4 30 2 00 2 12 1 87 2 22 2 09 2 09 sD Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 128 183 Product Validation Report PVR 01 F MAE 2 01 1 81 2 01 3 21 1 92 2 08 1 74 1 94 1 98 1 98 MAE MB 0 40 0 30 0 91 2 49 0 53 0 34 0 45 0 49 0 39 0 39 MB cc 0 15 0 09 0 19 0 26 0 12 0 09 0 24 0 16 0 11 0 11 cc RMSE 2 59 2 23 2 78 5 01 2 49 2 63 2 29 2 45 2 52 2 52 RMSE RMSE h 105 39 95 83 156 85 318 04 BIS 102 12 88 94 123 92 106 106 RMSE NS JE 168 275 0 627 470 2 0 795 795 NS NR 26 96 12 0 134 297 37 0 341 341 NR ME 11 79 12 67 12 819 12 51 14 23 1436 14 25 14 25 ME sD 3 68 4 69 6 401 4 64 5 59 8 54
77. 03 31 92 94 119 51 98 51 OBYEBIS RMSE Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 129 183 Product Validation Report PVR 01 F NS 210mmh 15 16 13 0 48 292 0 144 10 489 NS NR 210mm h 18 34 5 0 57 960 13 13 6 992 NR ME 210mm h 7 4369 10 44 11 815 9 61 14 76 24 17 2 90 12 43 10 48 ME sD 210mm h 3 4783 4 88 1 952 4 18 10 03 37 22 6 91 1 76 7 57 sD MAE 210mm h 7 4687 10 44 11 815 9 62 15 07 2417 5 67 12 43 10 71 MAE MB 210mm h 0 4573 0 25 0 048 0 30 0 12 0 04 0 74 0 10 0 08 MB cc 210mm h 0 00 0 10 0 10 0 07 0 03 0 10 0 62 0 47 0 01 cc RMSE 210mm h 8 8241 11 55 11 975 10 73 117 85 4496 7 24 12 56 12 94 RMSE RMSE _ 210mm h_ 58 39 76 86 95 30 725 90 55 93 17 6300 90 80 622594 RMSE Table 41 The main stati ical scores evaluated by PPVG for H01 during the autumn period Rain rates lower than 0 25 mm h have been considered as no rain A general precipitation underestimation by H01 is reported in Table 41 using both rain gauge and radar data for all the precipitation classes The statistical scores obtained during this season with both radar data RMSE Cl1 186 Cl2 91 CI3 73 and rain gauge RMSE Cl1 211 Cl2 107 Cl3 62 are the best ones of all the
78. 05 Flechtdorf 51 18 43 08 48 12 Hannover 52 27 47 09 41 54 Neuheilenbach 50 06 38 06 32 59 Emden 53 20 22 07 01 30 T rkheim 48 35 10 09 47 02 Neuhaus 50 30 03 11 08 10 Dresden 51 07 341 13 46 11 Table 15 Location of the 16 meteorological radar sites of the DWD Radarverbund des DWD mit 150 km Radien A omise Radasoniote C Rosara 180 en Figure 25 Left radar compound in Germany March 2011 Right location of ombrometers for online calibration in RADOLAN squares hourly data provision about 500 circles event based hourly data provision about 800 stations 4 http www dwd de bvbw appmanager bvbw dwdwwwDesktop _nfpb true amp windowLabel dwdwww main book amp T1460994925114492118088 igsbDocumentPath Navigation 2FWasserwirtschaft 2FUnsere _Leistungen 2FRadarniederschlagsprodukte 2FRADOLAN 2Fradolan_node ht ml 3F_nnn 3Dtrue amp switchLang en amp pageLabel _dwdwww_spezielle_nutzer_forschung_fkradar Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 54 183 HSAF Product Validation Report PVR 01 The flowchart of online calibration method applied in RADOLAN is depicted in the figures below Preprocessing I 5 min intervalls refined Z R relation
79. 1 Page 12 183 Product Validation Report PVR 01 AMSU AMSU A AMSU B ATDD AU BIG CAF coop CESBIO CM SAF CNMCA CNR CNRS DMSP DPC EARS ECMWF EDC EUM EUMETCast EUMETSAT FMI FTP GEO GRAS SAF HDF HRV H SAF DL IFOV IMWM IPF IPWG IR IRM ISAC mu LATMOS LEO LSA SAF M t o France METU MHS MSG MvIRI Mw NESDIS NMA NOAA NWC SAF NWP NWP SAF O3M SAF omsz Acronyms Advanced Microwave Sounding Unit on NOAA and MetOp Advanced Microwave Sounding Unit A on NOAA and MetOp Advanced Microwave Sounding Unit B on NOAA up to 17 Algorithms Theoretical Definition Document Anadolu University in Turkey Bundesanstalt f r Gewasserkunde in Germany Central Application Facility of EUMETSAT Continuous Development Operations Phase Centre d Etudes Spatiales de la BlOsphere of CNRS in France SAF on Climate Monitoring Centro Nazionale di Meteorologia e Climatologia Aeronautica in Italy Consiglio Nazionale delle Ricerche of Italy Centre Nationale de la Recherche Scientifique of France Defense Meteorological Satellite Program Dipartimento Protezione Civile of Italy EUMETSAT Advanced Retransmission Service European Centre for Medium range Weather Forecasts EUMETSAT Data Centre previously known as U MARF Short for EUMETSAT EUMETSAT s Broadcast System for Environmental Data European Organisation for the Exploitation of Meteorological Satellites Finnish Meteorologica
80. 1 km Threshold Rain Gauge Validation km Threshold 5 dBZ No Threshold 31 5 dBZ Correction with limited rain gauge correction No rain gauge number of gauges correction Description of Acc Periods 1 6 24h Acc periods 3 6 12 Acc periods 1 3 6 12 24h accumulated National composite PAC 24h Projection Azimuthal radar product Projection azimuthal National composite Equidistant used in HSAF equidistant standard CAPPI 2 km Resolution 250 m Validation elipsoid Resolution 1 Projection Mercator Threshold km Treshold 0 1 mm No Resolution 1 km Rain gauge correction rain gauge correction Threshold 31 5 dBZ No rain gauge correction applied for 1h Rain Acc Table 12 Inventory of the main radar data and products characteristics in Poland Slovakia and Turkey Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 43 183 Product Validation Report PVR 01 4 4 Rain gauge and radar data integrated products in PPVG In order to investigate the possible improvement of the ground precipitation field estimation a WG INCA WG has been introduced in the validation activities of PPVG In this section the first results with some considerations of the INCA WG Annex 5 are reported Within the WG participating countries Slovakia Poland and Germany there are two types of systems providing precipitatio
81. 10 mm h 60 40 20 Accuracy RMS lt 1 mm h 200 100 50 Table 1 Simplified compliance analysis for product PR OBS 1 Requirements Result of PR OBS2 threshold target optimal validation Accuracy RMS gt 10 mm h 50 30 15 Accuracy RMS 1 10 mm h 60 40 20 Accuracy RMS lt 1 mm h 120 80 40 Table 2 Simplified compliance analysis for product PR OBS 2 Requirements Result of PROBS threshold target optimal validation Accuracy RMS gt 10 mm h 80 40 20 Accuracy RMS 1 10 mm h 160 80 40 Accuracy RMS lt 1 mm h 320 120 80 Table 57 Simplified compliance analysis for product PR OBS 1 2 3 Obviously it is RMSD gt RMSE since RMSD is inclusive of Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 179 183 HSAF Product Validation Report PVR 01 the error of satellite measurements RMSEsat that is what we would like to know from validation the error of ground measurements RMSEground that should be known by the owners of the stations the error of the comparison methodology RMSEcomparison that should be estimated by metrologists Then we should consider It should be RMSD RMSEsat RMSEground RMSEcomparison 2 In the final part of the H SAF Development Phase attempts have been made to evaluate RMSEgr
82. 183 HSAF Product Validation Report PVR 01 Algorithms Step 2 Step 3 Step 4 Barnes 32 411 52434 68 43 Kriging 35412 58 36 77 62 NN 56 20 77445 96 50 IDS 63 37 71441 81 4 43 Table 59 RMSE and standard deviation of interpolation algorithms for 3 different regular grids In the cases studied Barnes appears to be the algorithm with the lower mean value of RMSE and their standard deviation than the other interpolation algorithms and the error of interpolation can be evaluated in the 30 for the step 2 that means an ideal condition were the rain gauge are disposed long a regular grid with a distance that the half of phenomenon length The structure of precipitation depends from precipitation typology time and spatial resolution therefore phenomenon length cannot be considered absolute An irregular distribution of perfect measurements has been considered also For each step the number of perfect measurement has been redistribuited randomly to simulate the raingauge network In the figure below the white circles mean the position of perfect measurement points In the figure below the white circle mean the position of perfect measurement points in the best case step 2 The results shown again the Barnes tecnique the best choice to reproduce the field Sinsed egal dats a 7 2 D 08 9 0 D w w No Figure 123 Randomly distribution of perfect measurem
83. 2 66 URD 1 10mm h 94 49 9601 155 28 311 57 MMASI 104 05 102 35 132 89 116 77 NS 210mm h 449 12316 1284 273 23487 1332 94 371 574 5090 NR 210mm h 138 826 247 13 1224 2300 203 70 111 2684 ME 9 84 10 53 9 23 15 77 393 13 89 16 61 8 80 13 13 sD 4 31 7 11 10 66 15 91 7 61 7 70 13 31 4 84 10 32 7 15 MAE 11 37 11 46 12 79 18 47 11 79 14 10 16 95 9 53 13 88 12 58 MB 0 30 0 26 0 41 2 27 0 32 0 11 0 20 0 29 0 21 0 11 cc 0 25 0 06 0 03 0 09 0 08 0 01 0 18 0 32 0 05 0 14 RMSE 12 42 12 90 14 76 22 40 13 32 15 98 21 54 10 32 17 09 14 41 URD 82 50 84 21 89 77 193 75 86 31 90 91 82 55 83 11 90 46 November 2010 Rain rates lower than 0 25 mm h have been considered as no rain The yearly averages obtained by all the countries using both radar and rain gauge data are quite similar The worst RMSE has been evaluated for light precipitation comparing H01 precipitation estimations with radar data In this case there is a precipitation overestimation by the satellite product but in general a clear precipitation underestimation is reported Table 42 Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 130 183 Product Validation Report PVR 01 Je usar 6 3 The multi categorical statistic Two sets of validation have been performed e one set for Countries Teams that has compared satellite dat
84. 8 Small scale surface soil moisture by radar scatterometer SM OBS 3 H 16 Large scale surface soil moisture by radar scatterometer SM DAS 2 H 14 _ Liquid root zone soil water index by scatterometer assimilation in NWP model SN OBS 1 H 10 Snow detection snow mask by VIS IR radiometry SN OBS 2 H 11 Snow status dry wet by MW radiometry SN OBS 3 H 12 Effective snow cover by VIS IR radiometry SN OBS 4 H 13 Snow water equivalent by MW radiometry Table 1 H SAF Products List 2 Introduction to product PR OBS 1 2 1 Sensing principle Product PR OBS 1 is fundamentally based on the instruments SSM I and SSMIS flown on the DMSP satellites These conical scanners provide images with constant zenith angle that implies constant optical path in the atmosphere and homogeneous impact of the polarisation effects see Figure 3 Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 16 183 Product Validation Report PVR 01 Single PAKU Naa Figure 3 Geometry of conical scanning for SSMIS Also conical scanning provides constant resolution across the image though changing with frequency It is noted that the IFOV is elliptical with major axis elongated along the viewing direction and the minor axis along scan approximately 3 5 of the major Its size is dictated by the antenna diameter actually the antenna is slightly elliptical to partially
85. But in some places in Italy and for the Turkish radars the maximum range distance used is 120 km or even less e g 80 km Range resolution is 250 m in Belgium 250 340 225 and sometimes 500 m for the Italian radars 500 m for one of the Hungarian radars and 250m for the other two Polish radars can work with 125 m and 250 m resolution and in Turkey it is 250 m for all the radars All in all the scan strategies within the PPVG countries are well balanced and similar to each other though they vary from one radar to the other even within countries All radars are regularly maintained and calibrated which is a good indicator of the continuous supervision of quality of radar data and the important element to sustain radar data quality Overview of radar products used for validation in the HSAF project The Table at the end of the report is provided to summarize the available products generated from radar measurements and the processing chain used to generate them Finally the list of the radar products used for the validation work is included in the last row Radar rainfall products are obtained after processing the measured radar reflectivities at different elevations of the radar scan strategy After each elevation the PPI Plan Position Indicator products and the CAPPI Constant Altitude PPI products are calculated PPI is the measurement of the radar antenna rotating 360 degrees around the radar site at a fixed elevation angle CAPPI
86. Chapter 4 the average scores reported in the following tables have been obtained on measurements collected in heterogeneous geographical orographical and climatological conditions 6 2 The continuous statistic There are three sets of columns e one set for Countries Teams that has compared satellite data with meteorological radar in inner land areas Belgium BE Germany DE Hungary HU and Slovakia SL and their average weighed by the number of comparisons e one set for three Countries Teams that has compared satellite data with rain gauges in inner land areas Italy IT Germany DE Poland PO and Turkey TU and their average weighed by the number of comparisons e one column for Turkey TU that has compared satellite data with rain gauges in coastal areas In order to highlight the seasonal performances of H01 the statistical scores have been presented not only for yearly average but also for seasons averages The seasons are reported in the following table Winter Spri Summer Autumn Dec 2009 Jan and Feb March April and June July and August Sept Oct and Nov 2010 May 2010 2010 2010 Table 37 split in four sections one for each season reports the Country Team results side to side With NR has been indicated the number of ground samples radar or rain gauge and with NS the number of satellite samples 6 2 1 The winter period
87. DOLAN RW left filled raster 2010 08 07 05 50 UTC and station data right dots 2010 08 07 06 00 UTC Table 24 Results of the categorical statistic of the validation for whole month August 201 Table 25 Continuous statistic Table 26 Results of the categorical validation statistic of case study 3rd June 2010 Table 27 Results of the categorical statistic of the validation for whole month June 2010 Table 28 Continuous statistic Table 29 Results of the categorical validation statistic of case study5 6th December 2010 Table 30 Results of the categorical statistic of the validation for whole month December 2010 Table 31 Continuous statistic Table 32 Results of the categorical statistics obtained for PR OBS 1 Table 33 Results of the categorical statistics obtained for PR OBS 1 Table 34 Scores for continuous statistics Table 35 Scores for dichotomous statistics Table 36 Statistic scores for H01 Table 37 split in four sections one for each season reports the Country Team results side to side 125 Table 38 The main statistical scores evaluated by PPVG for HO1 during the winter period Rain rates lower than 0 25 mm h have been considered no rain Table 39 The main statistical scores evaluated by PPVG for HO1 during the spring period Rain rates lower than 0 25 mm h have been considered as no rain Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 7 183
88. Duchon C E and C J Biddle 2010 Undercatch of tipping bucket gauges in high rain rate events Adv Geosci 25 11 15 Leitinger G N Obojes and U Tappeiner 2010 Accuracy of winter precipitation measurements in alpine areas snow pillow versus heated tipping bucket rain gauge versus accumulative rain gauge EGU General Assembly 2010 held 2 7 May 2010 in Vienna Austria p 5076 Sevruk M Ondr s B Chv la 2009 The WMO precipitation measurement intercomparisons Atmos Res 92 376 380 Sugiura K D Yang T Ohata 2003 Systematic error aspects of gauge measured solid precipitation in the Arctic Barrow Alaska Geophysical Research Letters 30 1 5 Schutgens N A J and R A Roebeling 2009 Validating the validation the influence of liquid water distribution in clouds on the intercomparison of satellite and surface observations J Atmos Ocean Tech 26 1457 1474 Tokay A D B Wolff K R Wolff and P Bashor 2003 Rain Gauge and Disdrometer Measurements during the Keys Area Microphysics Project KAMP J Atmos Oceanic Technol 20 1460 1477 Wagner A 2009 Literature Study on the Correction of Precipitation Measurements FutMon C1 Met 29 BY 32 p available at www futmon org 10 Annex 3 Working Group 2 Radar data PROPOSAL Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 143 183 HSAF Product Validation Report PVR 01
89. F PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 162 183 Product Validation Report PVR 01 Group of information litem GERMANY POLAND SLOVAKIA domaint SLOVAKIA domain2 Dokumentation received availabilty of documentation for INCA or ee similar German system YesiNo Documentation available Documentation available Documentation should be from ZAMG rom ZAMG issued in future meeting Definition of geographical area covered by INR or sels Corman ayaton PY Grd size in pixels 900x900 Tanesi so1xa01 19951 min longitude as93E 1382 15 99231 E 89953784943 E max longitude 15712495E 25334E 23 09630 E 259996967316 E min iatitude asas719 N 48 728 N 4713585 N 4510027313232 N max atitude 54 73662 N 55 029 N 50 14841 N 53 000579834 N Space resolution 1km 1km 1km akm EEN Composite of 16 national Composite of 8 nationa Composite of 2 national Composite of Hauta international radars lumber of precipitation stations oland oni 507 RERO OPA ZANG mo Number ot precipitation static 130 475 Poland only pt Blacklist for precipitation stations fea ies 7a IvesiNo 2 i a gt EEEE IMap of density of precipitation stations r a z Density ot raingauge stati eee instantaneous precipitation based oniy output data lon raingauge network time resolution Smin No Yes 15 min timelines instantaneous precipitation based oniy lon
90. F PVR 01 1 1 Issue Revision Index 1 1 Date 30 09 2011 Page 62 183 Radar data processing chain adar Clutter Propagation Melting Layer conditions ht Data removal VPR Rec Figure 33 Schematic representation of radar data processing chain Attenuation correction and hydrometeor classification Polarimetric radar systems enable the use of reliable algorithms for correcting rain path attenuation Based on the paradigm that specific attenuation ah dp and specific differential phase Kdp Kdp 0 5 dFdp dr are linearly related in rain ah dp g h dp Kdp cumulative attenuation effects can be corrected through the use of Fdp Carey et al 2000 Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 63 183 Product Validation Report PVR 01 a Observed Z 482 60 5 40 2 H 20 3 w o 180 Lio 180 10 so s 100 180 Distance East km b Corrected Zn BZ 150 T 80 50 100 40 50 30 z o 8 3 20 z 50 10 1007 0 150 z 10 150 100 50 0 50 100 150 Distance East km Figure 34 Measured upper panel and attenuation corrected lower panel PPI 1 0 deg of reflectivity observed on 09 14 08 at 0500 U T C by the polarimetric radar operated by Piemonte and Liguria regions HSAF Product Validation Report PVR 01 Product H01 PR
91. Figure 69 Synopsis for Central Europe for 05th December 2010 FU Berlin http wkserv met fu berlin de Over a period of 4 days precipitation sum reached 100 mm next figure Init Sot 04DEC2010 osz 3 Valid Wed 08DEC2010 oz Cesaminiederschiag bis zum Termin mm d ba GFS Modell des US Wetterdenstas Figure 70 96h totals of precipitation Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 99 183 Product Validation Report PVR 01 Figure 71 Hourly preci tion sum mm for H01 satellite data crosses time stamp 2010 12 05 07 027 UTC and for Figure 72 Hourly precipitation sum mm for HO1 satellite data crosses time stamp 2010 12 06 06 49 UTC and for RADOLAN RW left filled raster 2010 12 06 06 50 UTC and station data right dots 2010 12 06 07 00 UTC Data used PR OBS1 data for Bavaria in the given period were available for st December 4 59 5 51 and 7 02 UTC and for 6 December 04 43 05 38 06 49 and 16 13 UTC Only these data are analysed in this case study Comparison Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 100 183 HSAF Product Validation Report PVR 01 A first look to the results Fig 2 13 2 14 shows that rain rates detected by satellite product are in
92. H clouds Surface station observations T precipitation Radar measurements reflectivity currently 2 d 3 d in development Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 44 183 HSAF Product Validation Report PVR 01 Satellite data CLM Cloud type in development for use in precipitation analysis Elevation data high resolution DTM indication of flat and mountainous terrain slopes ridges peaks The INCA system provides High resolution analyses interest of validation WG 3 Nowcasts Improved forecasts of the following variables Temperature 3 d field Humidity 3 d Wind 3 d Precipitation 2 d interest of validation WG 3 Cloudiness 2 d Global radiation 2 d The INCA precipitation analysis is a combination of station data interpolation including elevation effects and radar data It is designed to combine the strengths of both observation types the accuracy of the point measurements and the spatial structure of the radar field The radar can detect precipitating cells that do not hit a station Station interpolation can provide a precipitation analysis in areas not accessible to the radar beam The precipitation analysis consists of the following steps i Interpolation of station data into regular INCA grid 1x1 km based on distance weighting only nearest 8 stations are taken into accoun
93. HMU directly for the H SAF precipitation validation but they are utilized as the input to the INCA precipitation analysis system which is supposed to become a new Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 70 183 HSAF Product Validation Report PVR 01 validation tool Prior the INCA analysis the rain gauge data are interpolated onto the regular 1x1 km grid using the inverse distance squared IDS interpolation method Only the 8 nearest rain gauge stations are taken into account in the interpolation in order to reduce occurrence of precipitation bull eyes artifact SHM performs the offline automatic and manual quality check of the rain gauge data In frame of the INCA system a quality control technique called blacklisting has been developed which avoids the data from systematically erroneous rain gauges to enter the analysis Currently the blacklisting is used in manual mode only 4 11 2 Radar data The network The Slovak meteorological radar network consists of 2 radars see next figure One is situated at the top of Maly Javornik hill near city Bratislava and second one is on the top of Kojsovska hola hill close to the city Kosice Both are Doppler C band radars the newer one at Kojsovska hola is able to measure also the dual polarization variables non operational Figure 40 Map of SHM radar network the rings represent maximum operati
94. HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 120 183 Product Validation Report PVR 01 Figure 95 Comparison of radar precipitation field from 15 August 2010 at 07 05 UTC in original 1 km resolution left and upscaled into the satellite grid of the 07 04 UTC passage of DMSP16 Conclusions In this intense convective precipitation event the H01 product strongly overestimated the precipitation as compared to radars especially in case of higher precipitation rates This conclusion was made by visual comparison of the precipitation fields and confirmed by high values of continuous statistical scores as Mean Error and Multiplicative bias The strong overestimation of heavy precipitation by H01 could have resulted from scanning horizontally small but vertically developed radar echo tops about 14 km convective cells by instruments with different scanning geometry Thus the IFOVs of the microwave instrument observing at relatively high incident angle more than 45 degrees could have captured much larger area of the cells volume than area of the cells projected into the radar CAPPI product Almost all precipitation detected by radars was captured by H01 POD very close to 1 but on the other hand the satellite product falsely detected a lot of light precipitation FAR gt 0 5 Despite the H01 overestimation and false detection the overall spatial consistency of the HO1 and radar field
95. HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 4 183 HSAF Product Validation Report PVR 01 4 10 Ground data in Poland IMWM 67 4 10 1 Rain gauge 4 11 Ground data in Slovaki 4 11 1 Rain gauge 4 11 2 Radar data 4 12 Ground data in Turkey 4 12 1 Rain gauge 4 13 Conclusions 5 Validation results case study analysis 5 1 Introduction 5 2 Case study analysis in Belgium IRM 5 2 1 Case study 14 17 of August 2010 5 2 2 Case study 22 24 of August 2010 5 2 3 Case study 12 15 of November 2010 h01 5 3 Case study analysis in Germany BfG 5 3 1 Case study 7 of August 2010 River Nei e Oder Spree and Elbe catchments 5 3 2 Case study 3 of June 2010 River Danube catchment 5 3 3 Case study 5 6 of December 2010 River Rhine catchment 5 4 Case study analysis in Hungary OMSZ 5 4 1 Case study 5 of May 2010 5 4 2 Case study 18 of July 2010 5 4 3 Case study 10 of September 2010 5 5 Case study analysis in Italy Uni Fe 5 5 1 Case study 6 of July 2010 5 6 Case study analysis in Poland IMWM 5 6 1 Case study 15 of August 2010 5 6 2 Case study 17 of May 2010 5 7 Case study analysis in Slovakia 5 7 1 Case study 15 of August 2010 5 8 Case study analysis in Turkey ITU 5 8 1 Case study 20 of October 2010 5 9 Conclusions 6 Validation results long statistic analysis 6 1 Introduction 6 2 The continuous statisti
96. MAX Available rain rate 120 Km velocity PPI Products 120 Km CAPPI 2 5 km MAX 240 Km VIL VVP2 Windprofiles ETops Hail Probability Base Hail Probability 24h HailProbability Overview 1 3 24 Hr Rainrate accumulation Is any quality NO YES NO map available Processing Clutter removal time Clutter suppression by RLAN wifi filter Clutter chain domain Doppler filtering Fuzzy Logic scheme using removal atttenuation and static clutter map Clutter map Velocity correction beam Z R a 200 b 1 6 Texture blocking correction gt Z R a 200 b 1 next Year 2012 VPR correction under VPR gt No testing Z R a 200 b 1 6 Description of PCAPPI 1500m Cartesian Nationale composite National composite instanteneous grid CAPPI 2 km CAPPI 3 km CMAX radar product 600m resolution CAPPI 5 km VMI SRI Projection stereographic used in HSAF Projection Mercator S60 Validation Resolution 1 km Resolution 2 km Threshold No Threshold 7dBZ No rain gauge correction Description of 24 h accumulation with Acc periods 1 3 6 12 Acc periods 3 6 12 24h accumulated range dependent gauge 24h National composite radar product adjustment Projection Mercator CMAX used in HSAF Cartesian grid Resolution 1 km Projection stereographic Brod Validation R PVR 01 Doc No SAF HSAF PVR 01 1 1 E H SAF roduct Validation Report n Issue Revision Index 1 1 Product H
97. N 0 90 and in comparison with rain gauges 0 85 The different results are due to the fact that RADOLAN data produce more valid pairs of satellite ground points A valid pair is given if for a satellite observation point fixed date time at least one ground observation point can be found within a surrounding area formed by a search ellipse of 2 5 km x 2 5 km Also the False Alarm Rate FAR is slightly different For RADOLAN a FAR of 0 66 and for rain gauge of 0 62 is estimated These values are higher than those for whole month August 2010 Only for the RADOLAN data there was one valid pair in the class RR gt 10 mmh so that for this class we have no statement on validation with rain gauge data 2837 576 1 756 127 0 0 90 0 59 0 02 0 85 0 39 0 00 0 66 0 29 0 00 0 62 0 50 0 33 0 48 0 02 0 35 0 28 0 00 Table 23 Hourly precipitation sum mm for H01 satellite data crosses time stamp 2010 08 07 05 43 UTC and for RADOLAN RW left filled raster 2010 08 07 05 50 UTC and station data right dots 2010 08 07 06 00 UTC Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 i Product H01 PR OBS 1 Date 30 09 2011 Page 92 183 HSAF Product Validation Report PVR 01 78959 15085 105 18925 3800 26 0 72 0 49 0 06 0 63 0 43 0 05 0 70 0 46 0 84 0 64 0 59 0 92 0 27 0 35 0 05 0 30 0 26 0 03 Table 24 Results of the categorical statistic of
98. OBS 1 Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Date 30 09 2011 Page 64 183 Although several approaches with different degree of sophistication have been proposed in the last years the procedure named APDP proposed in Vulpiani et al 2007 has been chosen to be implemented for its physical adaptability and operationally oriented architecture APDP Adaptive PhiDP method is an iterative correction of attenuation based on the use of Fdp that taking advantage from the classification of hydrometeors Marzano et al 2006 2007 adapt the coefficients g h dp to the observed physical conditions As an example Figure 34 shows the 1 0 degree PPI of measured upper panel and attenuation corrected lower panel reflectivity observed on 09 14 08 at 0500 U T C by the polarimetric radar located in mount Settepani operated by Piemonte and Liguria regions Figure 6 shows the hydrometeor classes detected by the classification algorithm corresponding to the event illustrated in Figure 33 Note LD Large Drops LR Light Rain MR Moderate Rain HR Heavy Rain R H Rain Hail mixture HA Hail G H Graupel or small Hail DS Dry Snow WS Wet Snow IC Ice Crystals Hydrometeors 150 100 50 Distance North km 50F 100 Ic ws Ds GH HA HR MR LR Lo 150 150 100 50 o 50 Distance East km 100 150 Figure 35 Hydrometeor classes as detected by
99. Product Validation Report for product H01 PR OBS 1 Version 1 1 30 September 2011 EUMETSAT Satellite Application Facility on Support to Operational Hydrology and Water Management The EUMETSAT Network of Satellite Application Facilities Support to Operational Hydrology and Water Management Product Validation Report PVR 01 for product H01 PR OBS 1 Precipitation rate at ground by MW conical scanners Reference Number SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Last Change 30 September 2011 About this document This Document has been prepared by the Product Validation Cluster Leader with the support of the Project Management Team and of the Validation and Development Teams of the Precipitation Cluster Je usar Product Validation Report PVR 01 Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 2 183 DOCUMENT CHANGE RECORD Issue Revision Date Description Baseline version prepared for ORR1 Part 2 1 0 16 05 2011 i f P Obtained by PVR 01 delivered during the Development Phase lS 30 09 2011 Updates acknowledging ORR1 Part 2 review board recommendation Minor adjustments 1 2 16 01 2012 e Document reference number as PVR 01 instead of PVR Doc No SAF HSAF PVR 01 1 1 HSAF Product Validation Report PVR 01 Issue Revision Index 1 1 Bi Product H01
100. SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Date 30 09 2011 Page 79 183 5 2 Case study analysis in Belgium IRM The following three case study analysis are here presented August 14 17 of 2010 August 22 24 of 2010 November 12 15 of 2010 5 2 1 Case study 14 17 of August 2010 Description of the event This event has been select because convective precipitation occurred during 14 17 August and covering large parts of the study area during 15 and 16 August A low was moving from Germany to The Netherlands see next figure Warm air from Central Europe was lifted over oceanic cold air over the study area Figure 45 Synoptic situation on 15 August 2010 at 6 UTC zoom in the surface map Satellite and ground data used Products H01 from 6 00 UTC of August 14 to 18 00 UTC of August 17 have been considered The total is 19 satellite passages distributed as follows 2 in the morning of August 14 3 in the morning of August 15 3 in the afternoon of August 15 3 in the morning of August 16 3 in the afternoon of August 16 3 in the morning of August 17 2 inthe afternoon of August a7 The ground data used for validation are the Wideumont radar instantaneous measurements without rain gauge adjustment Radar data are available within 5 minutes around the satellite passage Comparison Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1
101. SMO model grid in a rotated coordinate system and to develop software tools for a common validation methodology of the product Activities First step defining the best validation strategy for PR ASS 1 depending on the resolution of the ground data used Implementation of prototype softwarefor grid cutting and ground data up scaling in the COSMO grid with the help of Lucio Torrisi CNMCA Start Time End time November 2010 December 2010 First Report 20 of December 2010 Second step up scaling software tools dissemination and checks by the different validation teams Eventual improving and refining if needed Start Time End time January 2011 February 2011 Final Report 28 of February 2011 Codes delivery and related documentation 28 of February 2011 Composition of the working group Coordinator Angelo Rinollo RMI Belgium supported by Federico Porc University of Ferrara Italy and Lucio Torrisi CNMCA Italy Participants Belgium Bulgaria Germany Italy Hungary Slovakia Turkey Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 171 183 HSAF Product Validation Report PVR 01 H SAF project WP 6100 Working Group 4 Development of a common procedure for validation of PR ASS 1 in the native COSMO model grid A Rinollo RMI Belgium F Porcu Universit di Ferrara Italy L Torrisi CNMCA Italy 1 Val
102. Table 6 Table 6 Accuracy requirements for product PR OBS 1 RMSE Precipitation range threshold target optimal gt 10 mm h 90 80 25 1 10 mm h 120 105 50 lt 1 mmhh 240 145 90 This implies that the main score to be evaluated has been the RMSE However in order to give a more complete idea of the product error structure several statistical scores have been evaluated as reported Mean Error Standard Deviation SD and Correlation Coefficient CC Probability Of Detection POD False Alarm Rate FAR and Critical Success Index CSI These scores have been defined in Section 3 7 The long statistic results obtained in Belgium Hungary Germany Italy Poland Slovakia and Turkey will be showed in the next sections The country validation results are here reported in order to respond not only to the question whether the product meets the requirements or not but also where meets or approaches or fails the requirements The average performance of HO1 for all sites is presented in a compact synoptic way in this chapter The contents of the monthly statistical scores have been provided by the individual Countries Teams and verified by the Validation Cluster Leader step by step as described in the Chapter 3 As stressed in Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 125 183 Product Validation Report PVR 01 F
103. The Radar WG Annex 3 is now working to define quality index static or dynamic in order to select the more reliable radar fields and to associate an error structure to the radar data Quality information should take into account the radar site geographical areas event type radar products The study performed by the Slovakian team Annex 4 and the scheme published by J Szturcn et all 2008 on the quality index evaluation are under consideration by the Radar WG In the future the satellite product testing will be carried out using only the data having a sufficient quality but the validation results showed in this document have been obtained using radar data which passed only data owner institute controls Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 24 183 Product Validation Report PVR 01 500 L oos MB PPV Countries Weather radar units Horizontal beam extent of 100 km O Horizontal beam extent of 200 km OE Figure 7 The networks of 54 C band radars available in ther H SAF PPVG The studies that have been carried out in the PPVG on comparison of radar data with rain gauge data have shown that RMSE error associated with radar fields depends considerably on radar minimum visible height above the rain gauge in mountainous terrains like Slovakia but less importantly in flat terrains like Hungary In Slovakia the RMSE error see Section 3 7 of r
104. The accuracy and reliability of the raingauge stations significantly affect final precipitation analysis of the INCA or INCA like systems and therefore need to be checked In order to solve this problem an automated blacklisting technique is going to be developed at SHM currently blacklisting is used in manual mode The case studies presented in the report comparing the INCA analyses with corresponding input precipitation fields from radars and raingauges pointed out the benefits of the INCA system It has been shown that the system has potential to compensate errors due to effects like radar beam orographical blocking but also to correct instantaneous factors affecting radar measurement quality like radar beam attenuation in heavy precipitation what can not be achieved by standard methods of climatological radar data adjustment First attempts to utilize the INCA analyses as a ground reference data for the H SAF products validation have been done by statistical analysis of the PR OBS 2 product during selected precipitation events The software for upscaling the INCA precipitation field into the H SAF products grid will have to be developed Since the grids of INCA and RADOLAN have similar horizontal resolution to the common radar grid the radar upscaling techniques can be applied also on the INCA or RADOLAN data In frame of the unification of the validation methodologies the same common upscaling software could be shared between both radar
105. UTC is presented The map was constructed on the base of data from Polish Lighting Detection System PERUN Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 110 183 Product Validation Report PVR 01 Tine one 1570872010 14 i 13 08 2010 14 i 13 03 010 14 i 13 08 2010 14 i 13 03 010 14 w 13 08 2010 L4 i 15 08 2010 15 i 13 08 2010 15 i 15 08 2010 5 Eai 15706 2010 13 em 13 08 2010 rs i 15 08 2010 15 X f 5 Cet oon X J i 4 z Lat STN 3790 Lon 01E 0038 4 4 Os Figure 84 Total lighting map of Poland showing electrical activity between 1445 and 1515 UTC on 15th of August 2010 Data and products used Reference data data from Polish automatic rain gauges network IMWM H SAF product PR OBS 1 Ancillary data used for case analysis e Polish Lighting Detection System PERUN IMWM e Weather charts courtesy of Wetterzentrale Comparison This event is dominated by convective systems of limited spatial scales moving across Poland The average rain rate measured by rain gauges during this event is about 2 6 mm h while the PR OBS 1 averaged rain rate is of 7 9 mm h The highest peak measured by rain gauges is of about 51 6 mm h negligibly reduced by IFOV averaging while PR OBS 1 shows a peak value of 56 2 mmh On the next figure the PR OBS 1 product is visualized for afternoon overpass For comparison the
106. VR 01 HSAF Product H01 PR OBS 1 only covering the western part of Anatolia 3 Data processing The partners of the Validation Group have been using a variety of different strategies to treat gauge data and to compare them with satellite estimates Some are using interpolation algorithms to get spatially continuous rainfall maps while others process directly the measurements of individual gauges All the data in the network except for cold months in Poland are quality controlled there is no information about the techniques used but usually quality control rejects data larger than a given threshold and in case of too high rainrate difference exceeding given thresholds among neighbouring gauges and between subsequent measures of the same instrument Table 50 summarizes the data pre processing performed in different Countries while Table 51 and Table 52 reports the different matching approaches for HO1 HO2 and H03 H05 respectively As for the temporal matching the used approaches are rather homogeneous within the Groups instantaneous measurements are matched with next ground cumulated values over the different available intervals ranging from 1 minute Turkey to 1 hour Italy Germany Cumulated estimates obviously are compared to ground measured rain amounts over the same cumulation intervals As for spatial matching different approaches are considered also taking into account the different spatial st
107. WG2 and INCA WG3 working groups in the future References T Haiden A Kann G Pistotnik K Stadlbacher C Wittmann Interated Nowcasting through Comprehensive Analysis INCA System description ZAMG Vienna Austria 4 January 2010 Andr Simon Alexander Kann Michal Ne tiak Ingo Meirold Mautner kos Horv th K lm n Csirmaz Olga Ulbert Christine Gruber Nowcasting and very short range forecasting of wind gusts generated by deep convection European Geosciences Union General Assembly 2011 Vienna Austria 03 08 April 2011 Ingo Meirold Mautner Benedikt Bica Yong Wang INCA CE A Central European initiative in nowcasting applications Central Institute for Meteorology and Geodynamics Hohe Warte 38 1190 Vienna Austria Ingo Meirold Mautner Yong Wang Alexander Kann Benedikt Bica Christine Gruber Georg Pistotnik Sabine Radanovics Integrated nowcasting system f or the Central European area INCA CE Central Institute for Meteorology and Geodynamics ZAMG Hohe Warte 38 1190 Vienna Austria Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 170 183 HSAF Product Validation Report PVR 01 13 Annex 6 Working Group 4 PR ASS 1 COSMO grid validation PROPOSAL The aim of the group is to find in cooperation with the developing team of PR ASS 1 the most reliable way to validate the PR ASS 1 product which is provided on the CO
108. Zentralanstalt f r Meteorologie und Geodynamik of Austria Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 14 183 Product Validation Report PVR 01 Je usar 1 The EUMETSAT Satellite Application Facilities and H SAF The EUMETSAT Satellite Application Facility on Support to Operational Hydrology and Water Management H SAF is part of the distributed application ground segment of the European Organisation for the Exploitation of Meteorological Satellites EUMETSAT The application ground segment consists of a Central Application Facility CAF and a network of eight Satellite Application Facilities SAFs dedicated to development and operational activities to provide satellite derived data to support specific user communities See fig 1 Systems of the EUM NOAA other data Cooperation sources Data Acquisition EUM Geostationary and Control Systems Data Processing EUMETSAT HQ Meteorological Products Extraction EUMETSAT HQ USERS Figure 1 Conceptual scheme of the EUMETSAT application ground segment Next figure reminds the current composition of the EUMETSAT SAF network in order of establishment eee Z Ee NWC SAF OSI SAF 03M SAF CM SAF NWP SAF GRAS SAF LSA SAF H SAF TST ETE oceananasearce SORENESS T Gina montorng MATSNGRIWT Gras tnwariogy Lana Surfac
109. a horizontal cross section of data at constant altitude The CAPPI is composed of data from several different angles that have measured reflectivity at the requested height of CAPPI product The PPVG group uses mostly CAPPI products for calculation of rainfall intensities except for Hungary which uses the CMAX data maximum radar reflectivity in each pixel column among all of the radar elevations for deriving rainfall intensities However the rest of the countries have also chosen different elevation angles for the CAPPI product which provides the basis for rain rate estimations Additionally we have to say that the countries apply different techniques of composition of radar data that were not specified in this questionnaire The composition technique is important in areas which are covered by more than one radar measurements Also the projection applied is varying from one country to the other To sum up the radar products used are not harmonized different techniques are applied However each of them is capable to grasp rainfall and to estimate rainfall intensity As for the accumulated products we see that Belgium uses 24 hourly accumulations with rain gauge correction Italy uses 3 6 12 24h accumulations without gauge correction in Hungary 3 6 12 24h data is used but only the 12h and 24 hourly accumulations are corrected by rain gauges in Poland and Slovakia no rain gauge correction is applied Poland has only 6 and 24 hourly data
110. a ion of H SAF precipitation products Table 3 Number and density of raingauges within H SAF validation Group Table 4 Summary of the raingauge characteristics Table 5 Data pre processing strategies Table 6 Classes for evaluating Precipitation Rate products Table 7 Number and density of raingauges within H SAF validation Group Table 8 Summary of the raingauge characteristics Table 9 Data pre processing strategies Table 10 Matching strategies for comparison with H01 and H02 Table 11 Inventory of the main radar data and products characteristic n Belg m Italy and Hungary 42 Table 12 Inventory of the main radar data and products characteristics in Poland Slovakia and Turkey 42 Table 13 INCA Questionnaire Table 14 Precipitation data used at BfG for validation of H SAF products Table 15 Location of the 16 meteorological radar sites of the DWD Table 16 Main characteristics of the Hungarian radar network Table 17 Characteristics of the three radar instruments in Hungary Table 18 Characteristics of the SHMU radars Table 19 QA flags descriptions modified from Shafer et al 1999 Table 20 Scores obtained with the comparison with radar data in mm h 1 Table 21 Scores obtained with the comparison with radar data in mm h 1 Table 22 Scores obtained with the comparison with radar data in mm h 1 Table 23 Hourly precipitation sum mm for H01 satellite data crosses time stamp 2010 08 07 05 43 UTC and for RA
111. a with meteorological radar in inner land areas Belgium BE Germany DE Hungary HU and Slovakia SL e one set for Countries Teams that has compared satellite data with rain gauges in inner land areas Italy IT Germany DE Poland PO and Turkey TU Each Country Team contributes to this Chapter by providing the monthly contingency table and the statistical scores The Validation Cluster Leader has collected all the validation files has verified the consistency of the results and evaluated the monthly and yearly contingency tables and the statistical scores 6 3 1 Radar validation Dec 09 Jan 10 Feb 10 Mar 10 Apr 10 May 10 Jun 10 Jul 10 Aug 10 Sep 10 Oct 10 Nov 10 _tot POD with RR20 25 mm h 0 18 0 30 0 20 0 09 0 17 0 46 0 63 0 74 0 72 0 60 0 23 0 20 0 FAR with RR gt 0 25mm h 0 84 0 93 0 96 0 86 0 82 0 71 0 74 0 74 0 73 0 78 0 90 0 69 OH CSI with RR20 25mm h 0 09 0 06 0 03 0 06 0 09 0 22 0 22 0 24 0 25 0 19 0 07 0 14 POD with RRz1mm h_ 0 23 0 28 0 18 0 14 0 15 0 38 0 52 0 49 0 52 0 56 0 27 0 28 0 41 FAR with RREimm h 0 95 0 99 0 99 0 94 0 88 0 67 0 60 0 59 0 52 0 58 0 91 0 83 0 87 CSI with RR gt 1mm h_ 0 04 0 01 0 01 0 04 0 07 0 21 0 30 0 29 0 33 0 31 0 07 0 12 0 11 Table 43 The averages POD FAR and CSI deduced comparing H01 with radar data
112. able in ftp ftp meteoam it hsaf WP6000 WP6100 precipitation WG_groups software developed WG CLOSED Participants Emmanuel Roulin Eszter Labo and Judit Kerenyi Testing over Belgium successful procedure already generalized in a way that can be tested and used by all groups and delivered Testing by other members in progress Working Group 5 Geographical error map Coordinator Bozena Lapeta IMGW Poland Proposal completed Participants Silvia Puca Italy Ibrahim Sonmez and Ahmet Oztopal Turkey 9 Annex 2 Working Group 1 Rain gauge data PROPOSAL Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 136 183 Product Validation Report PVR 01 Je usar The ground reference does not exits The common validation methodology inside the H saf project has been based on the hydrologist reference end users constituted mainly by rain gauge and then by radar data During the Precipitation Product and Hydrological Validation workshop held in Bratislava the 20 22 of October 2010 the Precipitation Product Validation Group PPVG has decided to set up a working group for the definition of the correct verification of satellite precipitation product performances using the rain gauges data available inside the PPVG The main aims of this working group are to identify the more suitable techniques to compare rain gauge data with
113. adar accumulated fields is between 70 90 whereas in Hungary it is slightly lower between 60 80 Dataset for May September 2010 have been used to derive these parameters In PPVG it is under investigation INCA WG annex 5 the possibility to use ground data integrated software to produce precipitation field The results obtained by INCA WG are reported in the chapter 4 3 5 Spatial interpolation for rain gauges The partners of the Validation Group have been using a variety of different strategies to treat gauge data Some are using interpolation algorithms to get spatially continuous rainfall maps while others process directly the measurements of individual gauges Table 5 The first approach seems to be more convenient especially when the large IFOV of H01 are concerned Country Type of interpolation Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 i Product H01 PR OBS 1 Date 30 09 2011 Page 25 183 HSAF Product Validation Report PVR 01 Belgium Barnes over 5x5 km grid Bulgaria Co kriging Germany Inverse square distance Italy Barnes over 5x5 km grid Poland No Turkey No Table 5 Data pre processing strategies One of the next step of the Rain Gauge WG will be to harmonize the different spatial interpolation techniques among partners developing a common software for the validation collaborating with the GeoMap WG Annex 7 3 6 Te
114. adars b interpolated raingauge data c INCA analysis and d PR OBS 3 product 5 57 UTC supplemented with map of minimum visible height above surface level of the SHMU radar network e 15 August 2010 08 00 UTC The case from 08 00 UTC next figure gives an example of partial correction of radar beam orographical blocking by the INCA analysis The radar precipitation field in the north western part of Slovakia next figure a is affected by orographical blocking as indicated by relatively high minimum elevations of radar beam above this location in next figure e Also in this case information from raingauge network next figure b supplemented the radar field in the resulting INCA analysis next figure c Product Validation Report PVR 01 Product H01 PR OBS 1 Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Date 30 09 2011 Page 165 183 Figure 115 As in previous figure except for 8 00 UTC Statistical analysis of the PR OBS 2 product on selected precipitation events As a first step towards utilizing the INCA precipitation analyses for the H SAF validation it has been decided to perform at SHMU a statistical analysis of the H SAF products using the precipitation fields from INCA radars and raingauges as a ground reference data for selected precipitation events Since this task required modification of the SHMU software currently used for upscaling radar data until now results for t
115. al cumulation interval min of ATS network rain gauges is set for 10 minutes and can be adjusted according to given needs There is possibility to have very short cumulation intervals for case studies theoretically 1 minute but not on every given preci post It depends on local DCS settings The data processing As stated above the data quality control can be achieved by comparison on two rainfall datasets collected by two independent rain gauges at the same ATS post It is done operationally during summertime There is no such possibility during the winter because of lack of non heated rain gauge dataset In case that one pair of rain gauges at the same ATS post provide two different rainfall readings the higher one is taken into account No specialization technique is used for standard validation process However for some case studies the Natural Neighbor technique is applied for satellite and ground precipitation data To match the precipitation information with satellite data spatial and temporal matching are applied e Spatial matching for each given satellite pixel the posts situated within that pixel were found The pixel size was taken into account however its shape was assumed to be Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 69 183 HSAF Product Validation Report PVR 01 rectangular If more than one rain gauge were found within one
116. ann and M Peura Quality information for radars and radar data Technical rapport 2005 EUMETNET OPERA OPERA_2005_19 77p Our main purpose for the first step was to collect characteristics polarization beam width maximum range range resolution scan frequency geographical coordinates scan strategy elevations of the radar networks which composes the PPVG adopted processing chain and the generated products including the quality map if any This report is intended to present the results of the overview of different radar capacities and instruments in each of the participating countries Radar sites and radars In the PPVG group we have all together 54 radars used or in the plan to be used Their distribution in the countries is gt Belgium 1 radar Germany 16 radars not BfG products Hungary 3 radars Italy 18 radars Slovakia 2 radars Poland 8 radars Turkey 6 radars These radars cover wide range of geographical area from the longitude 5 50562 in Wideumont Belgium to the most Eastern area with longitude 32 58 15 in Ankara Turkey and from the Northern latitude of 54 23 03 17 in Gda sk Poland to the latitude of 36 53 24 in Mugla Turkey and lat 37 462 in Catania Italy Radars are built at different elevations above the sea level In mountainous countries they are placed at elevations more than 1000m above sea level whereas in flat countries like Hungary or Belgium their height position is not ex
117. are the Wideumont radar instantaneous measurements without rain gauge adjustment Radar data are available within 5 minutes around the satellite passage Comparison Here is an example of H01 files compared with radar data upscaled to the same grid of the afternoon of August 23 E i A Segale hcs Figure 51 H01 image of August 23th 2010 at 16 18 left compared with upscaled radar at 16 20 right The scale corresponds to thresholds of 0 1 1 and 10 mm h 1 The product matches the rainfall pattern quite good but underestimates higher rain amounts As in the other summer case there are also files with a coarser resolution than the normal 128 pixels per row One of them referring to the morning of the same day August 23 and showing a rather good matching is reported here below Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 84 183 Product Validation Report PVR 01 Figure 52 H01 image of August 23th 2010 at 5 53 left compared with upscaled radar at 5 55 right The scale corresponds to thresholds of 0 1 1 and 10 mm h 1 Also in this case characterized by a coarser resolution the matching is good but also underestimation is more evident We can see that in both cases the product can make a rather correct reconstruction of rainfall patterns but underestimates the am
118. arm months where small scale convection dominates to 50 km in cold months when stratified and long lasting precipitation mostly occur In next figure the value of the linear correlation coefficient is computed between each raingauge pair in the Italian hourly 2009 dataset as function of the distance between the two gauges 200812 200901 200902 200903 200904 200905 200906 200907 200908 200909 200910 200911 Figure 101 Correlation coefficient between raingauge pairs as function of the distances between the gauges Colours refer to the months of the year 2009 Assuming these values significant for the other Countries involved in this study we can conclude the distribution of gauges is capable to resolve the spatial structure of rain patterns only for stratified systems but it is inadequate for small scale convective events Country Total number Average minimum of gauges distance km Belgium 89 11 2 Bulgaria 378t T Germany 1300 17 Italy 1800 9 5 Poland 330 475 13 3 Turkey 193 27 Table 49 Number and density of raingauges within H SAF validation Group the number of raingauges could vary from day to day due to operational efficiency within a maximum range of 10 15 only in the Wallonia Region only in 3 river basins Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Date 30 09 2011 Page 140 183 Product Validation Report P
119. ase of too high rainrate difference exceeding given thresholds among neighbouring gauges and between subsequent measures of the same instrument Table 9 summarizes the data pre processing performed in different Countries while Table 10 reports the different matching approaches for HO1 HO2 respectively As for the temporal matching the used approaches are rather homogeneous within the Groups instantaneous measurements are matched with next ground cumulated values over the different available intervals ranging from 1 minute Turkey to 1 hour Italy Germany Cumulated estimates obviously are compared to ground measured rain amounts over the same cumulation intervals As for spatial matching different approaches are considered also taking into account the different spatial structure of the satellite IFOVs Two basic ideas are pursued pixel by pixel matching or ground measure averaging inside satellite IFOV The second approach seems to be more convenient especially when the large IFOV of HO1 and H02 are concerned Probably it is mandatory for H02 also Je usar Product Validation Report PVR 01 Product H01 PR OBS 1 Date 30 09 2011 Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Page 36 183 take into account that the size of the IFOV changes across the track and could become very large The first approach e g nearest neighbour can be more effective for H03 and HOS products
120. ate 30 09 2011 Page 28 183 Product Validation Report PVR 01 Je usar The large statistic analysis in PPVG is based on the evaluation of monthly and seasonal Continuous verification and Multi Categorical statistical scores on one year of data 2010 for three precipitation classes see Table 6 It was decided to evaluate both continuous and multi categorical statistic to give a complete view of the error structure associated to HO1 Since the accuracy of precipitation measurements depends on the type of precipitation or to simplify matters the intensity the verification is carried out on three classes indicated by hydrologists during the development phase see Table 6 i 7 2 3 1 10 mm h gt 10 mm h rec medium precipitation intense precipitation Table 6 Classes for evaluating Precipitation Rate products Precipitation Classes The rain rate lower than 0 25 mm h is considered no precipitation The main steps to evaluate the statistical scores are e all the institutes up scale the national radar and rain gauge data on the satellite native grid using the up scaling techniques before described e all the institutes compare H01 with the radar precipitation intensity and the rain gauge cumulated precipitation e all the institutes evaluate the monthly and seasonal continuous scores below reported and contingency tables for the precipitation classes prod
121. atellite motion i e from scan line to scan line is invariably 12 5 km dictated by the satellite velocity on the ground and the scan rate EAN Sgio mam NSA n Aiiwdo OPi N 45 Deg Sean Angle seen 2 60 Beam Diameter so o7 an Figure 8 Geometry of conical scanning left and IFOV right of SSMI e usar Product Validation Report PVR 01 Product H01 PR OBS 1 Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Date 30 09 2011 Page 26 183 The main steps followed by the PPVG to face the up scaling of radar data versus the instantaneous PR OBS 1 rainfall rates are average of high resolution ground validation data and smoothing of radar precipitation a Average of hi res ground validation Radar instruments provide many measurements within a single SSMI or SSMI S pixel Those measurements should be averaged following the SSMI antenna pattern that means SSMI S 15 5 Km along Track x 13 2 Km establish the size in km of the axis for each FOV FXn Fyn SSMI 15 Km along Track x 13 Km define a 2 dimensional Gaussian surface matrix G NXxN having resolution R pixel size Rsradar resolution which full width at half maximum FWHM is an ellipse with axes EXn Eyn of size equal to ones of a single FOV i e Fxn EXn e Fyn Eyn see Figure 9 f E P ipg Xipx Figure 9 Left Gaussian filter Right section of gaus
122. auge Brangauge g mradar mradar 0 75 4 INCA INCA 074 0 65 RR 0 25 mmh RR2 1 mmh RR gt 0 25 mmh RR2 1 mm h Threshold Threshold Critical success index 1 09 0 8 o7 _ 06 Braingauge Bos mradar 0 4 INCA 03 024 014 o RR 0 25 mmh RR21mmh Threshold Figure 116 Comparison of selected statistical scores for the PR OBS 2 product obtained by different ground reference data valid for event 1 convective Je usar Product Validation Report PVR 01 z Product H01 PR OBS 1 Page 168 183 Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Date 30 09 2011 Mean Error Multiplicative Bias 08 25 06 04 02 Braingauge 15 Braingauge E 0 radar g mradar H o2 INCA 1 ia nca 04 0s 06 08 o O 25sPR lt 1 1 lt PR lt 10 PR210 PR20 25 0 25sPR lt 1 1sPR lt 10 PR210 PR20 25 Class mm h Class mm h Relative RMSE Correlation Coefficient 400 1 350 08 3009 ae T 04 ae 250 mak De z in gauge raingauge a 200 H m radar g ofm H lmradar 150 onca 02 nca 0 4 100 E
123. auge values compared to 5 estimated at the minute averaged 3kmX3km grid rain for Temporal structure within matching satellite IFOV by using semi variogram Table 10 Matching strategies for comparison with H01 and H02 Belgium and Bulgaria use raingauges only for cumulated precipitation validation 4 2 4 Some conclusions After this inventory some conclusion can be drawn First it seems the raingauge networks used in this validation activities are surely appropriated for the validation of cumulated products 1 hour and higher while for instantaneous estimates the use of hourly cumulated ground measurements surely introduces intrinsic errors in the matching scores that can be estimated as very large The validation of instantaneous estimates should be carried on only when gauges cumulation interval is 10 to 15 minutes as in Poland Values cumulated over shorter Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 37 183 Product Validation Report PVR 01 Je usar intervals 5 or even one minute as it is done in Turkey are affected by large relative errors in cases of low moderate rainrates Different approaches for the estimates matching are considered and probably could be a good idea to harmonize them among partners The ground data up scaling procedure indicated in Section 3 5 has been already developed by E Roulin Van de Vyver H
124. ble are better than the long term results The POD reaches values very close to 1 for both intensity thresholds i e almost all precipitation detected by radars was also observed by H01 On the other hand FAR values are also relatively high supporting the finding from visual comparison that HO1 detected some light precipitation that was not observed by radars Comparison of original and upscaled radar data next figure demonstrates significant spill of convective cells in upscaled image because of big ratio of radar and satellite resolution 1 15 This effect is strengthened also due to a typical size of convective cells which does not exceed the satellite IFOV Capturing of convective cores by satellite IFOV or in upscaled radar image is then strongly dependent on the mutual position of convective core and IFOV centers The purpose of this case study is to evaluate the H01 product by means of radar measurements as ground reference But it should be noted that the evaluation results can be strongly affected by validation methodology itself Upscaling method for radar data is applied on 2 dimensional CAPPI 2km radar product and if we consider the horizontal radar beam width of 1 degree atmospheric volume from which radar signal is coming differ from the volume represented by microwave satellite measurements Different atmospheric volumes can differ also in water content detected and transformed into precipitation intensities Doc No SAF
125. ble 19 QA flags descriptions modified from Shafer et al 1999 The precipitation data QA tests are summarized as follows Range Test This test is used to see if any individual precipitation observation falls within the climatological lower and upper limits The test procedures applied in the study are as follows IF LiMiower lt Obserjt lt LiMupper THEN Obser flag is Good IF Obser gt Limypper OR Obser lt LiM ower THEN Obser flag is Failure Lim tower and Limupper thresholds are separately determined for each station on a monthly basis At any specific site all the observed monthly data is considered for determination of the upper and lower limits By applying this test each observation is flagged either by Good or Failure label depending on the comparison tests mentioned above Step Test It is used to see if increment decrement between sequential observations in time domain is in acceptable range or not The applied test procedure is IF Obserj Obserj 1 lt Step THEN Obser flag is Good IF Obser Obser 1 gt Step THEN Obser flag is Suspect Step threshold is determined again for each site on a monthly basis For each site the dataset containing the absolute difference of the sequential observations is determined by considering the observations for the matching month The 99 9 cumulative histogram value of the dataset is set as the Step threshold for the related site and mon
126. c 6 2 1 The winter period 6 2 2 The spring period 6 2 3 The summer period 6 2 4 The autumn period 6 2 5 The annual average 6 3 The multi categorical statistic 6 3 1 Radar validation 6 3 2 Rain gauge validation 6 4 User requirement compliance 7 Conclusions 7 1 Summary conclusions on the status of product validation 7 2 Next steps Doc No SAF HSAF PVR 01 1 1 HSAF Product Validation Report PVR 01 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 5 183 8 Annex 1 Status of working group 135 9 Annex 2 Working Group 1 Rain gauge data 135 10 Annex 3 Working Group 2 Radar data 142 11 Annex 4 Study on evaluation of radar measurements quality indicator with regards to terrain visibility 150 12 Annex 5 Working Group 3 INCA Precipitation for PPV 156 13 Annex 6 Working Group 4 PR ASS 1 COSMO grid validation 170 14 Annex 7 Working Group 5 Geographical maps distribution of error 173 15 Annex 8 Comments on the Validation Results for Products PR OBS 1 PR OBS 2 And PR OBS 3 178 Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 6 183 HSAF Product Validation Report PVR 01 List of tables Table 1 H SAF Products LiSt sesesssssesisessss js Table 2 List of the people involved in the valid
127. c No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 122 183 Product Validation Report PVR 01 H01 RAIN RATE mm h 20 10 2010 04 57 GMT aio L 1 A L L 1 i 41 04 40 0 Rain rate mm h E 0 25 to 1 00 J 1 01 to 4 00 z E 4 01 to 7 00 7 01 to 10 00 L JJ 10 01 to 14 00 38 04 LATITUDE 35 04 26 5 265 305 325 345 365 385 405 425 44 5 LONGITUDE RG RAIN RATE mm h 20 10 2010 04 57 GMT L L f f f a z i i I L Rain rate mm w Ie Cora i E 1 01 to 4 00 r zf 4 01 to 7 00 ni 7 01 to 10 00 J 10 01 to 14 00 LATITUDE 28 5 305 325 345 365 3865 405 425 445 LONGITUDE Figure 98 Comparison of H01 product and rain gauge RG According to next figure there is an overestimation for this case study H01 rain rate mm h 8 10 RG rain rate mm h Figure 99 Scatter diagram of rain gauge and H01 product Red line is 45 degree line Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 123 183 HSAF Product Validation Report PVR 01 Statistical scores Statistics scores can be seen from next table Correlation coefficient is 0 70 for HO1 product POD FAR and CSI are respectively 0 81 0 12 and 0 73 for this case study All these indicators are acceptable NS NR ME sD MEA MB cc RMSE
128. cation vertical profile of reflectivity correction and rainfall estimation will be treated in the following sections Radar data quality As known any fruitful usage of radar data either for quantitative precipitation estimation or just for operational monitoring must deal with a careful check of data quality Figure 4 schematically shows the operational processing chain that is applied within the system DATAMET software system for radar remote control product generation visualization system maintenance and data archive developed by DATAMAT S P A Ground clutter anomalous propagation beam blockage effects are routinely mitigated through the application of the decision tree method proposed by Lee et al 1995 for single polarized systems Dual polarized systems provide additional observables such as differential reflectivity correlation coefficient and their texture that can be used to further reinforce the traditional techniques Furthermore as soon as the polarimetric systems directly managed by DPC will be operational end of summer 2008 the property of the rain medium at vertical incidence are planned to be used for differential reflectivity calibration according to the procedure proposed by Gorgucci et al 1999 Redundancy of polarimetric variables will also be used for absolute calibration Gourley et al 2005 Product Validation Report PVR 01 HSAF Product H01 PR OBS 1 Doc No SAF HSA
129. ceeding 400m This information collected will be useful in the future steps of the Working Group to assess the partial or total beam shielding by mountains in the propagation way of the radar signals vvvvvv All radars are C band radars working at frequency in C band at 5 6 GHz This is important to know that our radar system is comparable All radars are equipped by Doppler capacity which means that ground clutters can be removed from the radar data measurements effectively however not all of them have dual polarization which would be important to correct rain path attenuation Scan strategies We have explored the scan strategy for each of the radars used In this matter all countries have shared their information on the number of elevations minimum and maximum elevations scan frequency maximum nominal range distance and range resolution Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 146 183 Product Validation Report PVR 01 Height N Azimuth lt gt xX Slant Range Elevation Volume Scan Procedure Ss We can conclude that the scan frequency ranges from 5 minutes in Belgium Germany and Slovakia to 10 minutes in Turkey and Poland and 15 minutes in Hungary and varying frequency for Italian radars The number of elevation stays between 4 and 15 in average around 10 The range distance used is 240 km in general
130. chniques to make observation comparable up scaling technique for radar data From the first Validation Workshop in 2006 it has been decided that the comparison between satellite product and ground data has to be on satellite native grid Generally one or two rain gauges are ina SSMI pixel but radar instruments provide many measurements within a single SSMI pixel For this reason an up scaling technique is necessary to compare radar data with the HO1 precipitation estimations on the satellite native grid The precipitation data in the retrieval product H01 follows the scanning geometry and IFOV resolution of SSM I and SSMI S flown on the DMSP Satellites These conical scanners provide images with constant zenith angle that implies constant optical path in the atmosphere and homogeneous impact of the polarisation effects see next figure Also conical scanning provides constant resolution across the image though changing with frequency It is noted that the IFOV has a constant elliptical dimension with major axis elongated along the viewing direction and the minor axis along scan approximately 3 5 of the major Its size is dictated by the antenna diameter actually the antenna is slightly elliptical to partially compensate for the panoramic distortion but also by the portion of antenna effectively illuminated As for the pixel i e the area subtended as a consequence of the bi dimensional sampling rate the sampling distance along the s
131. cipitation products for test catchments made Precipitation Validation Group to set up a Working Group for creating geographical maps of error distribution The main goals of this working group are e to investigate the opportunity to create geographical maps of error distribution for H SAF validation e to define if necessary the methodology for spatial representation of precipitation products errors e to produce a well referenced documentation on the methodology defined e to produce two short reports on the results obtained first by 31 of March 2011 and second by 30 of November 2011 e to develop if necessary the code to be used in the PPVG for a correct generation of the defined geographical maps of error distribution Activities First step to define the methodology e selection of the appropriate methodology for spatial distribution of precipitation products errors taking into consideration spatial and temporal characteristics of each product e first study performed for selected Polish test catchments as well as Polish territory Start Time End time December 2010 March 2011 First Report 31 of March 2011 Second step To define the precipitation products errors maps for country members of PPVG e collection of collocated ground data and satellite products for selected period possibly through 6300 e creation of the error maps for territory of PPVG country members for selected period e analysis of the achieved re
132. comparison it appears that for this winter situation the HO1 product could not apart from few cases reproduce the rainfall patterns and amounts sometimes even missing them at all 5 3 Case study analysis in Germany BfG 5 3 1 Case study 7 of August 2010 River Nei e Oder Spree and Elbe catchments Description at7 August 2010 there was a baroclinic zone reaching from the Baltic sea across Poland and Czechia until Austria where sub tropical air was advected from south to north at the eastern flank of the associated low pressure During the 7 3 August 2010 the precipitation reached about 35 mmh 150 mm in 48 hours in parts of Germany especially in Saxony causing floods in the upper parts of the rivers Nei e Spree and Elbe with catastrophic damages 5 Zur Rolle des Starkniederschlages am 7 9 August 2010 im Dreil ndereck Polen Tschechien Deutschland bei der Entstehung der Hochwasser von Nei e Spree und Elbe Bissolli at all Rapp Friedrich Ziese Weigl Nitsche Gabriele Malitz Andreas Becker Floods in Eastern Central Europe in May 2010 FU Berlin 2010 Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Product Validation Report PVR 01 i Waning Page 90 183 T Vy at US i j Figure 59 Synopsis for Central Europe for 07th August 2010 FU Berlin http wkserv met fu berlin de
133. compensate for the panoramic distortion but also by the portion of antenna effectively illuminated this enables to obtain the same IFOV for a group of different frequencies if co registration is a strong requirement As for the pixel i e the area subtended as a consequence of the bi dimensional sampling rate the sampling distance along the satellite motion i e from scan line to scan line is invariably 12 5 km dictated by the satellite velocity on the ground and the scan rate Along scan the sampling rate is selected differently for different frequencies or set of frequencies as necessary to fulfil the radiometric accuracy requirement and to minimise aliasing For more information please refer to the Products User Manual specifically volume PUM 01 2 2 Algorithm principle The baseline algorithm for PR OBS 1 processing is described in ATDD 01 Only essential elements are highlighted here Fig 04 illustrates the flow chart of the SSM I SSMIS processing chain There is an off line activity to prepare the Cloud Radiation Database CRD and a real time activity to exploit the satellite data for the product retrieval The off line activities consist of e collecting well documented meteorological events analysis or re analysis e applying a Cloud Resolving Model CRM to simulate the cloud microphysics missing in the analysis e applying a Radiative Transfer Model RTM to convert the cloud pattern in a pattern of monochromatic
134. cores for both satellite passages have been computed Totally 684 radar satellite pairs have been included in the computation Results of the scores for continuous and dichotomous statistics are presented in next two tables respectively 025 4 Number of satellite values Number of radar values Mean error mm h Standard deviation mm h Mean absolute error mm h Multiplicative bias Correlation coefficient Root mead square error mm h URD RMSE Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 119 183 HSAF Product Validation Report PVR 01 Table 34 Scores for continuous statistics In agreement with visual comparison of the precipitation fields the scores of continuous statistics first table exhibit strong overall overestimation of the HO1 product in these events All of the error based scores are higher by one order than results of long term statistics for August 2010 or whole year 2010 However this is not true for the results of correlation coefficient which are even better than the long term values except for 0 25 lt PR lt 1 The relatively high correlation coefficient values reflect good spatial consistency of the compared fields Precipitation threshold mm h 20 25 POD FAR CsI Table 35 Scores for dichotomous statistics All obtained scores of dichotomous statistics previous ta
135. ctual effective local station density and also on the variance of the differences of the bias between radar and rain gauge measurements On the whole we can say that our correction method is efficient within a radius of 100 km from the radar In this region it gives a final underestimation of about 10 while at bigger distance the underestimation of precipitation fields slightly increases Besides we also produce 12 hour total composite images first the three radar data are corrected separately and then the composite is made from them The compositing technique consists of weighting the intensity of each radar at a given point according to the distance of the given point from the radars This is also true for the 24 hourly accumulations Resolution projection threshold of detection The resolution of the radar data used for validation is 2km by 2km This is true for the accumulated and the instantaneous products as well As We have already mentioned the threshold of detection in Hungary is 7dB Hungarian radar data is available operationally in stereographic S60 projection References P ter N meth Complex method for quantitative precipitation estimation using polarimetric relationships for C band radars Proceed of 5th European Radar Conference ERAD Helsinki Finland http erad2008 fmi fi proceedings extended erad2008 0270 extended pdf Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS
136. currently blacklisting is used in manual mode Twelve case study analysis of HO1 have been here reported in Chapter 5 Stratiform and convective precipitations during summer and winter periods have been analysed in different countries Rain gauges with 10 minutes refresh time radar data and nowcasting tools have been used to highlight different characteristics of the satellite product The case studies proposed have pointed out that different statistical score values are obtained during summer and winter period problems on coast Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 133 183 HSAF Product Validation Report PVR 01 line and parallax shift It has been also showed a case study Poland where the ground data have been unable to catch the precipitation system while the satellite product reproduced more correctly the precipitation area In Chapter 6 the validation results of the H01 long statistic analysis obtained for the period 1 12 2009 31 11 2010 have been presented To assess the degree of compliance of the product with user requirements Each Country Team has provided the monthly contingency tables and the statistical scores The results have been showed for radar and rain gauge land and coast area in the three precipitation classes defined in fig 11 of Chapter 3 The rain rates lower than 0 25 mm h have been considered as no rain The
137. d F S Marzano 2008 The Italian radar network within the national early warning system for multi risks management Proceed of 5th European Radar Conference ERAD Helsinki Finland http erad2008 fmi fi proceedings extended erad2008 0184 extended pdf Vulpiani G M Montopoli L Delli Passeri A Gioia P Giordano and F S Marzano 2010 On the use of dual polarized C band radar for operational rainfall retrieval in mountainous areas submitted to J Appl Meteor and Clim http www erad2010 org pdf oral tuesday radpol2 5_ERAD2010 0050 pdf Hungary P ter N meth Complex method for quantitative precipitation estimation using polarimetric relationships for C band radars Proceed of 5th European Radar Conference ERAD Helsinki Finland http erad2008 fmi fi proceedings extended erad2008 0270 extended pdf Slovakia D Kotlarikova J Kali k and Strmiska Radar horizon modelling as a requirement of SHMI radar network enhancement Physics and Chemistry of the Earth Volume 25 Issues 10 12 2000 Pages 1153 1156 First European Conference on Radar Meteorology doi 10 1016 S1464 1909 00 00170 2 Poland Szturc J O r dka K and Jurczyk A 2008 Parameterization of QI scheme for radar based precipitation data Proceedings of ERAD 2008 http erad2008 fmi fi proceedings extended erad2008 0091 extended pdf Szturc J O r dka K and Jurczyk A 2009 Quality index scheme for 3D radar data volumes 34 Co
138. d with radar data upscaled to the same grid Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 87 183 Product Validation Report PVR 01 aida inl 1 1 1 1 bree 1 ieee Lu 1 Figure 55 H01 image of November 13th 2010 at 5 28 left compared with upscaled radar at 5 30 right The scale corresponds to thresholds of 0 1 1 and 10 mm h 1 on state o i AON NN 9 Flat cout E ies he E ROAR TS ary nana ys Akane MAN San CU 3 DOTEA E NN ORR MEN ntl i bes Figure 56 H01 image of November 13th 2010 at 15 17 left compared with upscaled radar at 15 15 right The scale corresponds to thresholds of 0 1 1 and 10 mm h 1 We can see that in both cases the satellite product misses or dramatically underestimates the rainfall As in the summer case there are files with a coarser resolution than the normal 128 pixels per row Some of them show a better matching compared to the examples seen just above Here are two of them Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 88 183 Product Validation Report PVR 01 Figure 57 H01 image of November 13th 2010 at 6 03 left compared with upscaled radar at 6 05 right The scale corresponds to thresholds of
139. de Figure 60 two day totals ending at 9th August O UTC interpolated on a 1 x1 evaluation grid as derived from SYNOP messages Global Precipitation Climatology Centre GPCC operated by DWD 90 Figure 61 Hourly precipitation sum mm for H01 satellite data crosses time stamp 2010 08 07 05 43 UTC and for RADOLAN RW left filled raster 2010 08 07 05 50 UTC and station data right dots 2010 08 07 06 00 UTC Figure 62 Contingency table statistic of rain rate mmh 1 for PR OBS1 vs radar data Left for 7th August 2010 Right for whole August 2010 Figure 63 Contingency table statistic of rain Rate mmh 1 for PR OBS1 vs rain gauge data Left for 7th August 2010 Right for whole August 2010 Figure 64 Synopsis for Central Europe for 03rd June 2010 FU Berlin http wkserv met fu berli Figure 65 12h totals of precipitation ending at 3rd June 2010 7 UTC Figure 66 Hourly precipitation sum mm for H01 satellite data crosses time stamp 2010 06 03 07 17 UTC and for RADOLAN RW left filled raster 2010 06 03 07 50 UTC and station data right dots 2010 06 03 08 00 UTC oe Figure 67 Contingency table statistic of Rain Rate mmh 1 for PR OBS1 vs radar data Figure 68 Contingency table statistic of rain rate mmh 1 for PR OBS1 vs rain gauge data Figure 69 Synopsis for Central Europe for O5th December 2010 FU Berlin http wkserv met fu berlin de Doc No SAF HSAF PVR 01 1 1 Issue Revision Inde
140. dex 1 1 i Product H01 PR OBS 1 Date 30 09 2011 Page 148 183 HSAF Product Validation Report PVR 01 Belgium Goudenhoofdt E and Delobbe L Evaluation of radar gauge merging methods for quantitative precipitation estimates Hydrol Earth Syst Sci 13 195 203 doi 10 5194 hess 13 195 2009 2009 http radar meteo be en 3302595 Publications html Berne A M ten Heggeler R Uijlenhoet L Delobbe Ph Dierickx and M De Wit 2005 A preliminary investigation of radar rainfall estimation in the Ardennes region Natural Hazards and Earth System Sciences 5 267 274 http radar meteo be en 3302595 Publications html Italy Fornasiero A P P Alberoni G Vulpiani and F S Marzano Reconstruction of reflectivity vertical profiles and data quality control for C band radar rainfall estimation Adv in Geosci vol 2 p 209 215 2005 http www adv geosci net 2 index html R Bechini L Baldini R Cremonini E Gorgucci Differential Reflectivity Calibration for Operational Radars Journal of Atmospheric and Oceanic Technology Volume 25 pp 1542 1555 2008 http journals ametsoc org doi pdf 10 1175 2008JTECHA1037 1 Silvestro F N Rebora and L Ferraris 2009 An algorithm for real time rainfall rate estimation using polarimetric radar Rime J Hydrom 10 227 240 Vulpiani G P Pagliara M Negri L Rossi A Gioia P Giordano P P Alberoni Roberto Cremonini L Ferraris an
141. dle 2010 Drifting wind can also greatly reduce the size of the effective catching area if rain does not fall vertically resulting in a rainrate underestimation quantitatively assessed in about 15 for an average event Duchon and Essenberg 2001 Further errors occur in case of solid precipitation snow or hail when frozen particles are collected by the funnel but not measured by the buckets resulting in a temporal shift of the measurements since the melting and the measure can take place several hours or days depending on the environmental Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 35 183 HSAF Product Validation Report PVR 01 conditions after the precipitation event Leitinger et al 2010 Sugiura et al 2003 This error can be mitigated by an heating system that melts the particles as soon as are collected by the funnel All these errors can be mitigated and reduced but in general not eliminated by a careful maintenance of the instrument A number of a posteriori correction strategies have been developed in order to correct precipitation data measured by raingauges but mainly apply at longer accumulation intervals daily to monthly Wagner 2009 Country Minimum detectable Maximum detectable Heating system cumulation rainrate rainrate mm h Y N interval min Belgium 0 1 mm N A N 60 Bulgaria
142. duct with the different reference fields for selected precipitation events are also included Summary of the INCA system survey As a first step of survey experts contact persons were identified inside the INCA group community as listed in the next table Country Contact person expert ail address Slovakia Jozef Vivoda jozef vivoda shmu sk Michal Ne amp tiak michal nestiak shmu sk Poland Rafal lwanski rafal iwanski imgw pl Germany Claudia Rachimow rachimow bafg de Peter Krahe krahe bafg de Table 53 List of contact persons Within the participating countries there are two types of systems providing precipitation analyses usable for H SAF validation INCA developed by ZAMG Austria and RADOLAN DWD Germany Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 159 183 Product Validation Report PVR 01 The INCA system is currently under development as INCA CE Central Europe and is used in pre operational mode in Slovakia and Poland The RADOLAN system is used in Germany operationally and is already utilized for the H SAF products validation TENG Figure 111 Coverage of Europe by the INCA and RADOLAN systems Here below a brief description of the INCA and RADOLAN systems follows More information on both systems can be found in the documentation which is available on the H SAF ftp server hsaf WP6000 precipitation WG_groups WG3 inca docum
143. ducts it is necessary to know errors distribution of used ground reference data in case of SHMU it is precipitation intensity and accumulated precipitation measured by Slovak radar network For this purpose a study called SHMU study on evaluation of radar measurements quality indicator with regards to terrain visibility has been elaborated To find distribution of errors in radar range next steps had to be done e simulations of terrain visibility by radar network using 90m digital terrain model e statistical comparison of radar data against independent rain gauge data measurements e derivation of dependence regression equation describing the errors distribution in radar range with regard to terrain visibility based on rain gauge and radar data statistical evaluation computation of error distribution maps using regression equation and terrain visibility Main results of this study are shown in next figure It is evident that the best visibility of SHMU radars corresponds to the lowest URD RMSE of 60 displayed by light violet colors URD RMSE is of quite homogeneous distribution with average of 69 in prevalent lowlands of Slovakia displayed by bluish colors But in central and north west mountainous areas this error exceeds 100 Figure 41 Map of relative RMSE left and Mean Error right over the SHMU radar composite Similar studies that have been carried out in the PPVG on comparison of radar data with rain gauge data have shown in gen
144. e aya SSS TOS Short Range Forecasting Chemistry Monitoring Prediction Figure 2 Current composition of the EUMETSAT SAF network in order of establishment Conceptual scheme of the EUMETSAT application ground segment The H SAF was established by the EUMETSAT Council on 3 July 2005 its Development Phase started on 1 September 2005 and ended on 31 August 2010 The SAF is now in its first Continuous Development and Operations Phase CDOP which started on 28 September 2010 and will end on 28 February 2012 The list of H SAF products is shown in Table 1 Doc No SAF HSAF PVR 01 1 1 HSAF Product Validation Report PVR 01 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 15 183 Acronym Identifier Name PR OBS 1 H 01 Precipitation rate at ground by MW conical scanners with indication of phase PR OBS 2 H 02 Precipitation rate at ground by MW cross track scanners with indication ofl phase PR OBS 3 H 03 Precipitation rate at ground by GEO IR supported by LEO MW PR OBS 4 H 04 Precipitation rate at ground by LEO MW supported by GEO IR with flag forl phase PR OBS 5 H 05 Accumulated precipitation at ground by blended MW and IR PR OBS 6 H 15 Blended SEVIRI Convection area LEO MW Convective Precipitation PR ASS 1 H 06 Instantaneous and accumulated precipitation at ground computed by a NWP model SM OBS 2 H 0
145. e rain rate for moderate precipitation rarely exceeding the 4 mm h value Rain rate H 01 mm h Rain rate RG mm h Figure 91 Scatter plot for measured RG and satellite derived H 01 rain rate obtained for all PR OBS 1 data on the 17th of May 2010 Finally the analysis of rain classes was performed The categories were selected in accordance with the common validation method Next figure shows the percentage distribution of satellite derived precipitation categories within each precipitation class defined using ground measurements One can easily notice that most of the precipitation cases was missed and only 17 and 30 of pixels in respectively light and moderate precipitation classes were properly allocated On the other hand more than 30 of no rain pixels were recognized as precipitation Doc No SAF HSAF PVR 01 1 1 H SAF Issue Revision Index 1 1 peia Product H01 PR OBS 1 Date 30 09 2011 Page 116 183 Product Validation Report PVR 01 100 90 80 70 60 50 40 4 30 4 20 10 mm h m 110 m 0 25 1 0 m 00 25 H 01 Percentage contribution 00 25 0 251 0 110 Rain rate RG mm h Figure 92 Percentage distribution of PR OBS 1 precipitation classes in the rain classes defined using rain gauges RG data on the 17th of May 2010 Some conclusions The analysis performed for situation with stratiform
146. e the temperatures are between 22 29C while in North West Europe only 6 14 C are measured ID J R SI HELYZET 2010 05 05 00 UC hee EZA weet FOKON HOSSZTARTO Data used Figure 75 Synoptic chart at 00 UTC on 5 May 2010 Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 103 183 HSAF Product Validation Report PVR 01 ae ae athe Figure 76 Precipitation rate from the Hungarian radar network at its original resolution upper right panel HO1 product upper left panel operational png lower left panel SAFNWC Cloud Type CT product lower right Comparison H01 well detected the thunderstorm systems over the country Rainfall a little bit is overestimated if we compare the values Conclusions Note that the same blue colours in the radar and the HO1does not correspond to the same rain rate H01 dark blue 1 2 mm light blue 4 5 mm radar dark blue 0 1 mm light green 5mm The H01 well detects the precipitation area but it overestimates the precipitation values mainly the light rain values 5 4 2 Case study 18 of July 2010 Description At Iceland a cyclone multiple centre derives the weather of Europe Along the front lot of clouds with rain develope thunderstorms are also observed Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011
147. eating system cumulation rainrate mm h rainrate mm h Y N interval min Belgium 0 1 N A N 60 Bulgaria 0 1 2000 ki 120 1440 Germany 0 05 3000 Y 60 Italy 0 2 300 Y N 60 Poland 0 1 300 Y 10 Turkey 0 2 288 Y 1 Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 i Product H01 PR OBS 1 Date 30 09 2011 Page 23 183 HSAF Product Validation Report PVR 01 Table 4 Summary of the raingauge characteristics only 300 out of 1800 gauges are heated An inventory on radar data networks and products used in PPVG Chapter 4 has pointed out that all the institutes involved in the PPVG declared the system are kept in a relatively good status and all of them apply some correction factors in their processing chain of radar data In Figure 7 there is the map of the 54 C band radars available in their H SAF PPVG Only the radar data which pass the quality control of the owner Institute are used by the PPVG for validation activities However these correction factors are diverse in the countries depending on their capacities and main sources of error in the radar measurements This also means that the corresponding rainfall estimates are different products in nature and the estimation of their errors cannot be homogenized for all the countries of the PPVG However each county can provide useful information of the error structure of its rainfall products based on its own resources
148. ected precipitation events overall convective 1864 observations stratiform 2251 mixed 3409 the obtained results may not be representative enough Therefore it is questionable if any conclusion about dependence of the investigated ground reference data on the long term validation results can be made It is proposed that statistical analysis using longer validation period will have to be performed HSAF Product H01 PR OBS 1 Product Validation Report PVR 01 Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Date 30 09 2011 Page 167 183 Mean Error Multiplicative Bias 6 4 4 3 5 2 0 3 pa margae 25 ae radar Be us janca 15 10 1 12 i Le 16 of O25sPRe1 1 lt PRe10 PR210 PR20 25 0 25sPR lt 1 1sPR lt 10 PR210 PR20 25 Class mm h Class mm h Relative RMSE Correlation Coefficient 1200 1 1000 800 A e Braingauge j E raingauge 600 mradar g E mradar H INCA INCA 400 ai tn 0 om A 0 25sPR lt 1 1sPR lt 10 PR210 PR20 25 O 25sPR lt 1 1sPR lt 10 PR210 PR20 25 Class mm h Class mm h Probability of detection False alarm rate 09 0 85 4 a 28 Braing
149. eld is about 1 km but values upscaled into the satellite grid using the IFOV Gaussian filter are presented For statistical scores computation only the data lying inside the 120 km rain effective range of both radars are considered Comparison The HO1 and upscaled radar precipitation fields for both satellite passages are presented in Fig 2 In the situation corresponding to the 07 04 UTC passage Fig 2 top row a general overestimation of the precipitation by H01 compared to radars is clearly seen This is obvious especially in case of higher precipitation intensities The maximum value observed by radars is 17 mm h while by HO1 it is 58 mm h for the identical area or within the same cell Also some light precipitation detected by H01 is not present in the radar field On the other hand a good match of the local maxima can be seen in the precipitation cell located in the eastern Austria It should be noted that the comparison in the north west of Slovakia where a local maximum of about 45 mm h was measured by H01 is complicated by poor radar visibility in this region Contrary to the intensities a good spatial consistency between the two fields can be seen Not only the precipitation rate maxima but also the patterns of light precipitation e g in the northern and eastern parts of Austria were localized quite well Only small dislocation of centers of intense precipitation cells over Slovakia can be observed probably caused by the parallax shift
150. elevation is reaching 3000m in northern central part of composite picture no rain gauge station was available in this region Only rain gauge stations with minimum visible heights in the interval Om 1100m were available in this study Im Minimum visible height of radar beam above the rain gauge m Temes NaN POHRONSKA POLIORA poesins Lavova Figure 105 Distribution of rain gauges according to the minimum visible height of radar beam To understand dependence of radar precipitation estimations and rain gauge values on gauge altitude above the sea and on radar beam altitude the scatterplots of log R G versus station altitude shown on Next figure and log R G versus radar beam altitude shown on Fig 6 were generated Quite wide scattering can be observed but quadratic polynomial trend lines indicate that in general radar underestimates precipitation and this underestimation is proportional to station elevation and radar beam elevation weaned log Radar Gauge versus station elevation above the sea level V 3E 07x 6E 05x 0 2656 Figure 106 Scatterplot of log R G versus station altitude shows general underestimation of precipitation by radar Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 154 183 Product Validation Report PVR 01 stn log Radar Gauge versus elevation of rada
151. ent to remap the field on a regular grid Assuming the best condition step 2 for the regular grid an evaluation of spread of RMSE respect the structure of precipitation field has been done In the figure 125 below the Barnes and Kriging tecniques show a low dependence from the standard deviation of field ie the level of inomogeneity of field The performance of up down scaling tecniques are reported in table below Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 183 183 Product Validation Report PVR 01 INTERPOLATION Step 2 150 g i Bames Y Kriging Nearest Neighbor 2 Inverse Distance ita ty tal Pn Re la oa x v y ay e z Pee lt ae wef 60 TEMA Syl i ee E3 seer 5 oe Brig y vy yor VW ky y y o i i i i i W 2 amp 4 E E amp f Standard Deviation original field mm Figure 124 STD vs RMSE for interpolations by step 2 Taking into account the results discussed before is possible to define a range of uncertainty that is necessary to consider when comparing the results of validation with operational requirements More effort has to be done to understand if exist a link between the error of remap procedure and precipitation intensity but the preliminary study shows that in the best case an error of 30 has to be considered for the up down scaling remapping procedure Using the previo
152. entation Brief description of the INCA system The INCA Integrated Nowcasting through Comprehensive Analysis analysis and nowcasting system is being developed primarily as a means of providing improved numerical forecast products in the nowcasting and very short range forecasting It should integrate as far as possible all available data sources and use them to construct physically consistent analyses of atmospheric fields Among the input data sources belong NWP model outputs in general P T H clouds Surface station observations T precipitation Radar measurements reflectivity currently 2 d 3 d in development Satellite data CLM Cloud type in development for use in precipitation analysis Elevation data high resolution DTM indication of flat and mountainous terrain slopes ridges peaks The INCA system provides High resolution analyses interest of validation WG 3 Nowcasts Improved forecasts of the following variables Temperature 3 d field Humidity 3 d Wind 3 d Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 160 183 Product Validation Report PVR 01 nsar Precipitation 2 d interest of validation WG 3 Cloudiness 2 d Global radiation 2 d The INCA precipitation analysis is a combination of station data interpolation including elevation effects and radar data It is desi
153. eral that RMSE error associated with radar fields depends considerably on radar minimum visible height above the rain gauge especially in mountainous countries In lowlands this dependence is not so significant but no negligible The reason can be the location of radar sites at the top of hills and impossibility of the lowest elevation to reach the lowland s surface In case of Slovakia The URD RMSE error of radar accumulated fields is between 60 90 with an average URD RMSE value of 69 3 Mean Error specified for 24 hours cumulated precipitation is 4 42mm or converted into instantaneous precipitation 0 184 mm h RMSE specified for 24 hours cumulated precipitation is 9 48mm or converted into instantaneous precipitation 0 395 mm h Complete SHMU study is available on the H SAF ftp server hsaf WP6000 WP6100 precipitation WG_groups WG2 radar WG 2 3_radar quality indication_v1 doc Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 i Product H01 PR OBS 1 Date 30 09 2011 Page 73 183 HSAF Product Validation Report PVR 01 4 12 Ground data in Turkey 4 12 1 Rain gauge The network 193 Automated Weather Observation Station AWOS located in the western part of Turkey are used for the validation of the satellite precipitation products in the HSAF project The average distance between the AWOS sites is 27 km The locations of the AWOS sites are shown in next figure 42 5 N 40 0 N 37 5 N
154. ers measured 1999 Budapest Dual polarimetric Z ZDR Doppler radar 2003 Napkor Dual polarimetric Z Zpr Kpp Opp Doppler radar 2004 Poganyvar Dual polarimetric Z ZpR Kpp Opp Doppler radar Table 16 Main characteristics of the Hungarian radar network The instruments The Hungarian radar network is composed by three Doppler radars which are measuring in the C band mainly at same frequencies The scan strategy is the same for all the radars the Budapest radar has a resolution lower than the two other radars which are newer types The parameters of the instruments and the measurement campaigns are listed in next Table Budapest Napkor Poganyvar Frequency band C Band 5625MHz C Band 5610MHz C Band 5610MHz Polarization single single single Single Double e e 8 Doppler capability Yi Yi Yi Yes No es es es Scan strategy scan freq 15 min scan freq 15 min scan freq 15 min elevations A inal Elevaions deg 00 5 Elevaions deg 00 5 Elevaions deg 00 5 maximum nia 1 1 1 8 2 7 3 8 5 1 1 1 1 8 2 7 3 8 5 1 1 1 1 8 2 7 3 8 5 1 range istance 66 8 5 6 68 5 6 68 5 range resolution Range 240 Km Range 240 Km Range 240 Km Resolution 500m Resolution 250m Resolution 250m Table 17 Characteristics of the three radar instruments in Hungary The data processing Radar measurements are influenced by many error sources that should be minimized as much as possible As such in case of the Hun
155. esults for POD see above 0 19 0 31 0 10 0 00 0 09 0 06 0 28 0 17 1 21 1 33 1 07 1 23 1 34 1 46 1 41 1 46 0 97 1 06 0 75 0 84 1 12 1 19 0 97 1 04 1 34 1 55 0 81 1 00 0 90 1 08 0 64 0 79 0 06 0 05 0 12 0 09 0 02 0 00 0 11 0 15 1 23 1 36 1 07 1 23 1 34 1 46 1 43 1 47 0 74 0 74 0 84 0 72 17 85 17 90 1 40 1 58 1 73 1 84 7 7 9 32 6 23 1 46 1 59 1 65 1 70 7 x 17 85 17 90 0 48 0 54 0 49 0 57 7 0 00 0 00 0 06 0 05 0 06 0 13 1 58 1 74 1 92 1 98 amp 20 14 18 95 Table 31 Continuous statistic Conclusions This case study has worst results in comparison with the two examined summer cases The false alarm rates were higher than the probability of detection There were no detected rain rates between usar Product Validation Report PVR 01 Product H01 PR OBS 1 Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Date 30 09 2011 Page 102 183 0 2mmh and Immh while about 15 of radar rain gauge data fell in this class Averaging in this winter period we get an underestimation by satellite data 5 4 Case study analysis in Hungary OMSZ 5 4 1 Case study 5 of May 2010 Description A front from North East to South West Europe defines the weather Close to the front at Alps Carpathian basin North Italy South France several thunderstorms are developed In Carpathian basin South East Europ
156. evolution of fraction area with rain measured by radar gt 0 25 mm h and Equitable Threat Score ETS during the present case study Figure 51 HO1 image of August 23th 2010 at 16 18 left compared with upscaled radar at 16 20 right The scale corresponds to thresholds of 0 1 1 and 10 mm h 1 The product matches the rainfall pattern quite good but underestimates higher rain amounts 83 Figure 52 H01 image of August 23th 2010 at 5 53 left compared with upscaled radar at 5 55 right The scale corresponds to thresholds of 0 1 1 and 10 mm h 1 84 Figure 53 Time evolution of fraction area with rain measured by radar gt 0 25 mm h and Equitable Threat Score ETS during the present case study Figure 54 Surface map on 13 November 2010 at 06 UTC MSLP and synoptic observations Figure 55 HO1 image of November 13th 2010 at 5 28 left compared with upscaled radar at 5 30 right The scale corresponds to thresholds of 0 1 1 and 10 mm h 1 Figure 56 H01 image of November 13th 2010 at 15 17 left compared with upscaled radar at 15 15 right The scale corresponds to thresholds of 0 1 1 and 10 mm h 1 Figure 57 HO1 image of November 13th 2010 at 6 03 left compared with upscaled radar at 6 05 Figure 58 Time evolution of fraction area with rain measured by radar gt 0 25 mm h and Equitable Threat Score ETS during the present case study Figure 59 Synopsis for Central Europe for 07th August 2010 berlin
157. f the INCA analysis Original radars INCA Raingauges 201008151500 9 J 201008151500 INCA Radars raingauges 101 201008151500 ay 1201008151500 A Cana Figure 113 Precipitation intensity field from 15 August 2010 15 00 UTC obtained by a radars b interpolated raingauge data c INCA analysis and d PR OBS 1 product Visual comparison of the precipitation fields In this section two case studies from 15 August 2010 focused on performance of the INCA analyses are presented 15 August 2010 06 00 UTC This case illustrates potential of the INCA system to correct errors in radar precipitation measurements due to radar beam attenuation in heavy precipitation As can be seen in next figure part a the radar measured precipitation near centre of the circled area was relatively weak However as next figure part c suggests the precipitation was probably underestimated by radars because an intense Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 164 183 HSAF Product Validation Report PVR 01 convective cell occurred directly in path of the radar beam dashed line in next figure part c The raingauge network next figure part b captured the intense precipitation underestimated by radars and improved the INCA analysis next figure part c b wos Figure 114 Precipitation intensity d from 15 August 2010 6 00 UTC obtained by a r
158. garian radar data many correction methods are applied or planned to be applied int he near future to filter out false radar reflectivity measurements Clutter removal and WLAN filter is already implemented int he processing chain of all three radar data and a filter to disregard signals below 7dBz is also applied because in general these data is not coming from real rain drops but false targets According to experiences beam blockage can result in serious underestimation of precipitation amounts e g behind the B rzs ny mountains at the north of Budapest So the bleam blockage correction is planned to be implemented during year 2012 Also the attenuation correction the Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 56 183 HSAF Product Validation Report PVR 01 attenuation of electromagnetic waves in water environment water drops is planned for 2012 Hungary does not apply VPR Vertical Profile Reflectivity correction Precipitation intensity is derived from radar reflectivity with the help of an empirical formula the Marshall Palmer equation R a Z b where a 200 b 1 6 From the three radar images a composite image over the territory of Hungary is derived every 15 minutes applying the maximum reflectivity in one column method in order to make adjustments in overlapping regions Description of instantaneous and accumulated radar product used i
159. gher while for instantaneous estimates the use of hourly cumulated ground measurements surely introduces intrinsic errors in the matching scores that can be estimated as very large The validation of instantaneous estimates should be carried on only when gauges cumulation interval is 10 to 15 minutes as in Poland Values cumulated over shorter intervals 5 or even one minute as it is done in Turkey are affected by large relative errors in cases of low moderate rainrates Different approaches for the estimates matching are considered and probably could be a good idea to harmonize them among partners As an example for H02 a document was delivered by the developers where the best estimate ground reference matching strategy was indicated and also Angelo Rinollo delivered few years ago the code for the Gaussian weight of the antenna pattern in the AMSU MHS IFOV Anyway different approaches over different Countries are leading to very similar values in the considered skill scores indicating probably two things 1 none of the considered approaches can be considered as inadequate and more important 2 the differences between ground fields and satellite estimates are so large that different views in the data processing do not results in different numbers 5 References Duchon C E and G R Essenberg G R 2001 Comparative rainfall observations from pit and aboveground gauges with and without wind shields Water Resour Res 37 3253 3263
160. ghest relative error for radar measurements at rain rate values lower than 1 mm h around 150 following annex8 Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 132 183 HSAF Product Validation Report PVR 01 Precipitation Requirement RMSE radar gauge gauge class thresh target optimal land land coast gt 10 mmh 90 80 25 1 10 mm h 120 105 50 lt 1 mmh 240 145 90 Table 47 User requirement and compliance analysis for product H01 As reported in 15 the results obtained by the current validation procedure represent the convolution of at least three factors the satellite product accuracy the accuracy of the ground data used and the limitations of the comparison methodology e g errors of space and time co location representativeness changing with scale etc Therefore the results currently found are by far pessimistic in respect of what is the real product performance 7 Conclusions 7 1 Summary conclusions on the status of product validation The HO1 product has been validated by the PPVG on one year of data 1st of December 2009 30th of November 2010 Each Country Team have provided case study and long statistic analysis using radar and rain gauge following the validation methodology reported in Chapter 3 The results of the Precipitation Validation Programme are reported in this Product Validatio
161. gned to combine the strengths of both observation types the accuracy of the point measurements and the spatial structure of the radar field The radar can detect precipitating cells that do not hit a station Station interpolation can provide a precipitation analysis in areas not accessible to the radar beam The precipitation analysis consists of the following steps e Interpolation of station data into regular INCA grid 1x1 km based on distance weighting only nearest 8 stations are taken into account to reduce bull eyes effect e Climatological scaling of radar data by means of monthly precipitation totals of raingauge to radar ratio partial elimination of the range dependance and orographical shielding e Re scaling of radar data using the latest rain gauge observations e Final combination of re scaled radar and interpolated rain gauge data e Elevation dependence and orographic seeding precipitation In the final precipitation field the raingauge observations are reproduced at the raingauge station locations within the limits of resolution Between the stations the weight of the radar information becomes larger the better the radar captures the precipitation climatologically Important factor affecting the final precipitation analysis is accuracy and reliability of the raingauge stations In order to eliminate the influence of raingauge stations providing evidently erroneous data the SHMU is developing the blacklisting technique which te
162. gress 6 2 4 The autumn period r J PR OBS 1 BE u fse rer r Po i l autumn autumn autumn autumn autumn autumn autumn autumn autumn autumn autumn 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 Version 1 4 radar radar radar radar radar gauge gauge gauge gauge gauge gauge NS 7005 58393 37706 18556 121660 18609 466 25600 14548 59223 NS NR 5999 37962 11950 4320 60231 28185 3017 19920 12023 63145 NR ME lt 0 11 0 17 0 18 0 57 0 04 0 21 0 05 0 11 0 14 0 16 ME sD lt imm h 0 20 0 77 0 94 1 01 0 76 0 95 0 91 1 08 0 84 0 97 sD MAE n 0 63 0 64 0 66 0 87 0 66 0 65 0 69 0 69 0 63 0 66 MAE MB 0 78 0 68 1 33 2 25 0 93 0 60 0 90 0 80 0 72 0 70 MB cc 0 09 0 09 0 18 0 27 0 12 0 09 0 14 0 12 0 10 0 11 cc RMSE 0 89 0 79 0 97 1 16 0 87 0 97 0 92 1 09 0 86 0 99 RMSE RMSE 195 98 164 84 209 53 286 40 216 10 172 75 226 82 184 78 RMSE NS h 5510 25846 14176 5887 51419 20169 1370 12878 7907 42324 NS NR 2278 22034 5397 399 30108 27042 1633 11989 5435 46099 NR ME 0 91 1 07 0 05 1 20 0 83 1 63 0 66 0 84 0 91 1 30 ME sD 1 14 1 47 1 62 1 31 1 47 2 38 1 70 2 27 1 54 2 23 sD MAE 1 37 1 51 1 23 1 45 1 45 2 22 1 46 1 90 1 49 2 02 MAE MB 1 10mm h 0 52 0 45 1 02 1 78 0 58 0 40 0 68 0 59 0 49 0 47 MB cc 1 10mm h 0 36 0 24 0 44 0 43 0 29 0 23 0 27 0 30 0 21 0 25 cc RMSE 1 10mm h 1 78 1 82 1 63 1 78 1 79 2 89 1 84 2 44 1 80 2 61 RMSE RMSE _ 1 10mm h 92 40 89 05 96 76 127 39 QING 1
163. gt No Z R a 200 b 1 6 Description of PCAPPI 1500m Nationale composite National composite instanteneous Cartesian grid CAPPI 2 km CAPPI 3 km CMAX radar product 600m resolution CAPPI 5 km VMI SRI Projection used in HSAF Projection Mercator stereographic S60 Validation Resolution 1 km Resolution 2 km Threshold No Threshold 7dBZ No rain gauge correction Description of 24 h accumulation with Acc periods 1 3 6 12 Acc periods accumulated range dependent gauge 24h 3 6 12 24h radar product adjustment Projection Mercator National composite used in HSAF Cartesian grid Resolution 1 km CMAX Validation 600m resolution Threshold No Projection No rain gauge correction stereographic S60 Resolution 2 km Threshold 7dBZ Rain gauge correction applied for 12 24 hourly data POLAND SLOVAKIA TURKEY List of PPI PCAPPI RHI MAX CAPPI 2 km MAX igus Doc No SAF HSAF PVR 01 1 1 HSAF Product Validation Report PVR 01 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 150 183 Available EHT SRI PAC VIL VVP Etops PPI Products HWIND VSHEAR PPI 0 2 CAPPI HSHEAR LTB SWI Base VIL MESO WRN Cmax ETOPS List of non operational Hmax EBASE products LMR CMAX VIL RAIN Acumulation UWT VAD SHEAR SWI Precip Intensity 1h 1 3 6 12 24h MESO
164. h Institute in addition to the large statistic verification produces a case study analysis based on the knowledge and experience of the Institute itself Each institute following a standard format here reported decides whether to use ancillary data such as lightning data SEVIRI images the output of numerical weather prediction and nowcasting products The main sections of the standard format are description of the meteorological event comparison of ground data and satellite products visualization of ancillary data discussion of the satellite product performances indications to Developers indication on the ground data if requested availability into the H SAF project More details on case study analysis will be reported in the Chapter 5 4 Ground data used for validation activities 4 1 Introduction In the following sections the precipitation ground data networks used in the PPVG are described radar and rain gauge data of eight countries Belgium Bulgaria Germany Hungary Italy Poland Slovakia and Turkey H01 has been submitted to validation in all these countries except Bulgaria Until now the Bulgarian data are used only for HOS validation activity according to the Project Plan Their uses in the next months is under consideration It is well know that radar and rain gauge rainfall estimation is influenced by several error sources that should be carefully handled and characterized before using these data as reference for g
165. he PR OBS 2 product are only available In order to eliminate interpolation artifacts in the areas outside the raingauge network occuring in the INCA analyses only the PR OBS 2 data falling inside the Slovakia territory were taken into account in the statistical analysis Overall five precipitation events with different prevailing type of precipitation have been selected for the statistical analysis as listed in next table Period UTC Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 166 183 Product Validation Report PVR 01 15 August 2010 00 00 21 00 convective 16 August 2010 06 00 23 45 convective 15 September 2010 15 00 mixed 18 September 2010 09 00 21 November 2010 20 00 stratiform 22 November 2010 23 45 28 November 2010 15 00 stratiform 29 November 2010 10 00 Table 55 List of precipitation events selected for statistical analysis For each precipitation event and each ground reference data a set of continuous and dichotomous statistical scores was computed The scores and thresholds of the precipitation classes were adopted from the H SAF common validation methodology As an example the results of selected statistical scores obtained with different reference data for the event 1 and 4 are shown in next two figures respectively Due to the small number of compared PR OBS 2 observations during the sel
166. he development of more advanced products e to characterise the product error structure in order to enable the Hydrological validation programme to appropriately use the data e to provide information on product error to accompany the product distribution in an open environment after the initial phase of distribution limited to the so called beta users Validation is obviously a hard work in the case of precipitation both because the sensing principle from space is very much indirect and because of the natural space time variability of the precipitation field sharing certain aspects with fractal fields that places severe sampling problems It is known that an absolute ground reference does not exist In the H saf project the validation is based on comparisons of satellite products with ground data radar rain gauge and radar integrated with rain gauge During the Development phase some main problems have been pointed out First of all the importance to characterize the error associated to the ground data used by PPVG Secondly to develop software for all steps of the Validation Procedure a software available to all the members of the PPVG Three months ago the radar and rain gauge Working Group WG have been composed in order to solve these problems The first results obtained by the working groups are reported in the following sections and a complete documentation is available as annex 1 7 of this document In addition to the rada
167. he error sources through the construction of appropriate quality maps As requested by the hydro meteorological community many operational institutions already provide such information others are currently working on this task The main aims of this WG are to describe the characteristics and generated products of PPVG radar networks to produce a referenced documentation on minimal requirements for certifying the radar products quality radar rainfall products testing and the procedure for satellite products validation to develop the code to be used in the PPVG for satellite products validation Activities First step collect e characteristics polarization beam width maximum range range resolution scan frequency geographical coordinates scan strategy elevations number of integrated samples etc of the radar networks which composes the PPVG e adopted processing chain e generated products including the quality map if any Start Time End time December 2010 February 2011 First Report 10 of Febrary 2011 Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 144 183 HSAF Product Validation Report PVR 01 Second step define on the base of published papers and studies of the characteristics of the radar data available inside the PPVG 5 minimal requirements for certifying the radar products quality 6 radar rainfal
168. he next two figures the PR OBS 1 product is visualized for two morning overpasses For comparison the distributions of 10 minute precipitation obtained from RG data measured at closest to the given time slots are presented The RG derived precipitation maps were prepared using Near Neighbor method a Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 114 183 Product Validation Report PVR 01 HO1 rain rate 17 05 2010 0453 UTC ee eee Ce ee a se 6 5 55 s 1 452 J 51 im 25 ae A 1 t as A EREEREER a 6 6 7 B 8 4 2 A m Figure 89 PR OBS 1 at 0453 UTC on the 17th of May 2010 right panel and 10 minute precipitation interpolated from RG data from 0500 UTC left panel RG rain rate 17 05 2010 0500 UTC b RG rain rate 17 05 2010 0550 UTC HOL rain rate 17 05 2010 0546 UTC 5 or mb s a 2 2 a all 5 amp 7 8 H HF 2 BM Figure 90 PR OBS 1 at 0546 UTC on the 17th of May 2010 right panel and 10 minute precipitation interpolated from RG data from 0550 UTC left panel Although precipitation maps obtained for two time slots with the use of ground data are very similar the distributions achieved on the base of PR OBS 1 are totally different Most of the rainfall observed on the 17 of May 2010 at 0453 UTC was missed by satellite product leaving only few precipitating spots left panel
169. he physics in the retrieval model 4 The case for continuing the validation activity essentially as it is now or improving it if considered cost effective is also very strong since it is necessary to continuously watch that the product generation chain works correctly
170. his problem an automated blacklisting technique is going to be developed at SHMU currently blacklisting is used in manual mode Product Validation Report PVR 01 Product H01 PR OBS 1 Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Date 30 09 2011 Je usar Page 48 183 The software for upscaling the INCA precipitation field into the H SAF products grid will have to be developed Since the grids of INCA and RADOLAN have similar horizontal resolution to the common radar grid the radar upscaling techniques can be applied also on the INCA or RADOLAN data In frame of the unification of the validation methodologies the same common upscaling software could be shared between both radar and INCA working groups in the future 4 5 Ground data in Belgium IRM 4 5 1 Radar data The network Belgium is well covered with three radars see next figure Further radar is currently under construction in the coastal region KTI THE NETHERLANDS J N J GERMANY 4 Zaventem Bosgorontn bat BE LW IA A E kd ont Ru FRANCE a WRG 50100 km rassaa larri errsilisrirrisi lari ahad Figure 19 Meteorological radar in Belgium The instruments These are Doppler C band single polarization radars with beam width of 1 and a radial resolution of 250 m Data are available at 0 6 0 66 and 1 km horizontal resolution for the Wideumont Zaventem and Avesnois radars respectively
171. his set of processes is described by hydrological discharge models and by river discharges measured by hydrological equipments Moreover validation of precipitation products can not be overcasted by only Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 157 183 HSAF Product Validation Report PVR 01 an evaluation of methods describing transformation of precipitation into river discharge For this reason a common validation methodology to compare satellite precipitation estimations with ground data radar and rain gauge inside the H saf project has been defined The validation of precipitation field is a difficult task and a continuous study of possible validation methodology improvement is necessary The Precipitation Product Validation Group decided during the last internal workshop held in Bratislava from 20 22 of October 2010 to set up various working groups for the investigation of possible improvement of the validation methodology One of this working groups is INCA precipitation for PPV group Definition of INCA Precipitation Products INCA system consists of computational modules which enable us to integrate various sets of precipitation data sources raingauge network radar network NWP models outputs and climatological standards into common precipitation product which can describe well the areal instantaneous and cumulated precipitation fields
172. icture of minimum visible height above the surface over the whole radar network Compositing algorithm selects the minimum value from both radar sites Figure 105 Distribution of rain gauges according to the minimum visible height of radar beam 153 Figure 106 Scatterplot of log R G versus station altitude shows general underestimation of precipitation by radar Figure 107 Scatterplot of log R G versus radar beam altitude shows increased underestimation of radar for high and close to zero radar beam elevations Figure 108 Relative RMSE left and Mean Error right computed independently for each rain gauge station in radar range and corresponding trend lines extrapolated for beam elevation up to 1500m 155 Figure 109 Final relative root mean square error map of radar measurements with regard to terrain visibility by current radar network of SHM Figure 110 Final mean error map of radar measurements with regard to terrain visibility by current radar network of SHMU General underestimation of precipitation by radars is observed Figure 111 Coverage of Europe by the INCA and RADOLAN systems Figure 112 Procedure of the RADOLAN online adjustment hourly precipitation amount on 7 August 2004 13 50 UTC Figure 113 Precipitation intensity field from 15 August 2010 15 00 UTC obt interpolated raingauge data c INCA analysis and d PR OBS 1 product Figure 114 Precipitation intensity field from 15 August 2010 6 00 UTC obtained by a radars b interpola
173. idation technique depends on data resolution The task of the present group is to develop a common validation procedure for the PR ASS 1 product characterized by the COSMO model native grid which is built up in a rotated coordinate system Depending on the resolution of the ground data we decided to suggest two different approaches in case of ground data with a spatial sampling similar to the one of COSMO that is the typical case of raingauge networks the nearest neighbor approach is suggested In this case no upscaling is needed in case of a resolution of the ground data much finer than the one of COSMO that is the case of many radar products then the upscaling to the native COSMO grid is recommended For this case we are currently working on a common upscaling procedure Methodology The main issue in this task is the fact that PR ASS 1 is based on the rotated coordinate system of the source model COSMO while the ground observations are normally based on geographical coordinates For this reason in case upscaling is needed a regular portion i e a fixed number of rows and columns is extracted from the COSMO grid Then all the coordinates of the ground data are converted in the rotated system and associated to the grid cell in which they fall in At this stage upscaling technique is straightforward the upscaled value associated to every grid cell is simply the arithmetical average of all the ground observations falling int
174. in IFOV amount IFOV amount Je usar Product Validation Report PVR 01 Product H01 PR OBS 1 Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Date 30 09 2011 Page 141 183 Poland mean gauges value each overpass is mean gauges value each overpass is over the IFOV area compared to the over the IFOV area compared to the rectangular next 10 minutes rain rectangular next 10 minutes rain amount amount Turkey weighted mean each overpass is weighted mean each overpass is semi variogram compared to the semi variogram compared to the gauges value corresponding 1 gauges value over corresponding 1 centred on satellite minute rain rate centred on satellite minute rain rate IFOV IFOV Belgium and Bulgaria use raingauges only for cumulated precipitation validation Table 51 Matching strategies for comparison with H01 and H02 H03 H05 Country Spatial matching Temporal matching Spatial matching Temporal matching Belgium N A N A Nearest neighbour rain amounts in the same number of hours are compared 24 hours Bulgaria N A N A Nearest neighbour rain amounts in the same number of hours are compared 3 and 24 hours Germany matching gauges are each overpass is matching gauges are rain amounts in the searched on a radius compared to the searched on a radius same number of of 2
175. in PUM 01 Here are the main highlights The horizontal resolution Ax descends from the instrument Instantaneous Field of View FOV the sampling distance pixel the Modulation Transfer Function MTF and number of pixels to co process for filtering out disturbing factors e g clouds or improving accuracy Conclusion for PR OBS 1 e resolution Ax 30km sampling distance 16 km The observing cycle At depends on the instrument swath and the number of satellites carrying the addressed instrument For PR OBS 1 there are 4 DMSP satellites but because of the limited instrument swath they provide a total service equivalent to that one of two satellites around 7 00 and 18 00 LST In average the observing cycle over Europe is At 6 h with actual interval ranging from 2 to Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 18 183 Product Validation Report PVR 01 Je usar 10 hours Gaps are filled by product PR OBS 2 that also has observing cycle At 6 h but LST around 9 30 and 14 00 with actual intervals ranging from 4 5 to 7 5 hours Conclusion e for PR OBS 1 as stand alone i e from DMSP satellites cycle At 6 h sampling 2 10 h e for the composite PR OBS 1 PR OBS 2 system cycle At 3 h sampling 2 4 5 h The timeliness 6 is defined as the time between observation taking and product available at the user site assuming a defined dissemi
176. ing rain gauges RG data on the 17th of May 2010 116 Figure 93 Synoptic situation on 15 August 2010 at 0 00 UTC 116 Figure 94 Instantaneous precipitation field on 15 August 2010 at 07 05 UTC top row and 15 00 UTC second row derived by SHMU radar network left column and HO1 product right column 118 Figure 95 Comparison of radar precipitation field from 15 August 2010 at 07 05 UTC in original 1 km resolution left and upscaled into the satellite grid of the 07 04 UTC passage of DMSP16 44 120 Figure 96 Atmospheric condition 20 10 2010 06 00 GMT Figure 97 Atmospheric condition 20 10 2010 12 00 GMT Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 11 183 HSAF Product Validation Report PVR 01 Figure 98 Comparison of H01 product and rain gauge RG 122 Figure 99 Scatter diagram of rain gauge and H01 product Red line is 45 degree line 122 Figure 100 Rain gauge networks in PPVG 137 Figure 101 Correlation coefficient between raingauge pairs as function of the distances between the gauges Colours refer to the months of the year 2009 139 Figure 102 Distribution of rain gauges according their altitude above the sea level 151 Figure 103 Radar horizon model output for Maly Javornik left and KojSovska hola right radar sites Figure 104 Composite p
177. ing the wintertime precipitation information might be burdened by a slightly bigger measuring error The number of rain gauges available for H SAF validation activities varies from day to day due to operational efficiency of ATS network in Poland and depends on large number of independent factors Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 68 183 HSAF Product Validation Report PVR 01 It can be stated that the number varies between 330 and 475 rain gauges for each day of operational work Mean minimum distance between precipitation measuring ATS posts between each pair of rain gauges in Polish national network is 13 3 km N A Legend ATS gauges SYNOP gauges o 400 200km Main rivers Figure 38 ATS national network in Poland All rain gauges working within Polish ATS national network are MetOne tipping bucket type instruments Minimum detected quantity that can be measured by those rain gauges is 0 1 mm h which means that each tilt of rain gauge bucket adds 0 1mm to the total sum of the measured precipitation During very heavy precipitation events MetOne rain gauges tend to underestimate real precipitation by factor of 10 Maximum measured rainrate mmh by MetOne instruments in Poland was recorded in 5 06 2007 at ATSO Koscielisko Kiry at the foot of Tatra Mountains The recorded values reached 65 mm h Operation
178. instruments The measurement instruments are precipitation sensors OTT PLUVIO of Company ow They continually and precisely measure quantity and intensity of precipitation in any weather based on balance principle with temperature compensation heated funnel and by an electronic weighing cell The absolute measuring error is less than 0 04 mm for a 10 mm precipitation amount and the long term 12months stability is better than 0 06 mm The operating temperature ranges from 30 C to 45 C The minimum detected quantity sensitivity is 0 05 mmh The maximum possible measured rain rate is 3000 mmh The operational accumulation interval theoretically is one minute http www ott com web ott_de nsf id pa_ottpluvio2_vorteile html OpenDocument amp Click Precipitation amount and intensity measurements with the Ott Pluvio Wiel Wauben Instrumental Department INSA IO KNMI August 26 2004 e HSAF Product Validation Report PVR 01 z Product H01 PR OBS 1 Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Date 30 09 2011 Page 52 183 The data processin Continuous automatic measurement of liquid and solid precipitation data are collected accumulated intervals from 1hour until 1day and provided as SYNOP tables by DWD These data are error corrected and quality controlled in four steps with checks of completeness temporal spatial consistency and marginal checks climatologic
179. ion the weighting functions are always the same Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 77 183 HSAF Product Validation Report PVR 01 from site to site and time to time However in reality it is expected that the weights should reflect to a certain extent the regional and temporal dependence behavior of the phenomenon concerned For the validation activities the point cumulative semi variogram technique proposed by en and Habib 1998 is used to determine the spatially varying weighting functions In this approach the weightings not only vary from site to site but also from time to time since the observed data is used In this way the spatial and temporal variability of the parameter is introduced more realistically to the validation activity Matching approach The temporal and spatial matching approaches are applied separately in the validation of the satellite products As for the temporal matching the product time is taken into account and 5 minute window t 2 to t 3 is considered for estimation of the average rainrate for each site For the spatial matching the mesh grid size of 3kmX3km is constructed for each IFOV area For each grid point the rainrate is estimated by taking the 5 minute averaged rainrate amounts observed at the nearby AWOS sites within the radius distance of 45 km for convective type or 125 km for stratiform
180. ion estimation with the rain gauge precipitation cumulated on different intervals the Polish and Turkish data will be used for this purpose The rain gauge inventory has also pointed out that different approaches for the estimates matching are considered in the PPVG The second steps in the next future will be to define the rain gauge spatial interpolation technique and to develop the related software The radar data in the PPVG is composed by 54 C band radars across the 7 countries Belgium Germany Hungary Italy Slovakia Poland Turkey The rain gauge network responsible declared that the systems are kept in a relatively good status The rain gauge inventory pointed out that different correction factors are applied This means that the corresponding rainfall estimates are diverse and the estimation of their errors cannot be homogenized The first step in PPVG Radar WG will be to define a quality index on the base of the study performed by the Slovakian team Annex 4 and the Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 78 183 HSAF Product Validation Report PVR 01 scheme published by J Szturc et all 2008 The main difficulty consists on the definition of a quality index computable for every radar networks of PPVG The evaluation of this quality index will allow to evaluate the rain gauge error in the same way and to select the more reliable radar
181. ision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 158 183 HSAF Product Validation Report PVR 01 e select extreme weather events and make case studies on comparison the INCA and H SAF relevant precipitation products e in case of positive case studies to perform batch validation of H SAF products and provide standard validation statistical outputs continuous and multicategorica statistics Start Time End time April 2011 November 2011 Second Report 31 of July 2011 Final Report 30 of November 2011 Composition of the working group Coordinator Jan Kanak SHMU Participants members of H SAF consortium which are in parallel involved in development of INCA products Belgium Germany Italy Hungary Slovakia Turkey FIRST REPORT Coordinator J n Ka k Slovakia Participants Claudia Rachimow and Peter Krahe Germany uboslav Okon Jozef Vivoda and Michal Ne tiak Slovakia Rafal lwanski and Bozena Lapeta Poland Silvia Puca Italy Introduction This report presents outcomes of the initial activities performed within the INCA products working group In the first part information on the INCA or INCA like systems available in the participating countries are summarized The second part of the report presents several case studies comparing precipitation fields estimated by radars raingauges and the INCA system Results of the statistical comparison of the PR OBS 2 pro
182. k and the approach to match gauge data with the satellite estimates The results are summarized in the next pages 1 000 km _ a V Countries Rain gauges 7 We wE wE wE Figure 100 Rain gauge networks in PPVG 1 The Instruments Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Date 30 09 2011 Page 138 183 Product Validation Report PVR 01 HSAF Product H01 PR OBS 1 Most of the gauges used in the National networks by the Precipitation Product Validation Group PPVG Partners are of the tipping bucket type which is the most common device used worldwide to have continuous point like rainrate measurement Nevertheless several source of uncertainty in the measurements are well known but difficult to mitigate First very light rainrates 1 mm h and less can be incorrectly estimated due to the long time it takes the rain to fill the bucket Tokay et al 2003 On the other side high rainrates above 50 mm h are usually underestimated due to the loss of water during the tips of the buckets Duchon and Biddle 2010 Drifting wind can also greatly reduce the size of the effective catching area if rain does not fall vertically resulting in a rainrate underestimation quantitatively assessed in about 15 for an average event Duchon and Essenberg 2001 Further errors occur in case of solid precipitation snow or hail when frozen particles are collected by the funnel but not mea
183. l Institute File Transfer Protocol Geostationary Earth Orbit SAF on GRAS Meteorology Hierarchical Data Format High Resolution Visible one SEVIRI channel SAF on Support to Operational Hydrology and Water Management Interactive Data Language Instantaneous Field Of View Institute of Meteorology and Water Management in Poland Institut f r Photogrammetrie und Fernerkundung of TU Wien in Austria International Precipitation Working Group Infra Red Institut Royal M t orologique of Belgium alternative of RMI Istituto di Scienze dell Atmosfera e del Clima of CNR Italy istanbul Technical University in Turkey Laboratoire Atmosph res Milieux Observations Spatiales of CNRS in France Low Earth Orbit SAF on Land Surface Analysis National Meteorological Service of France Middle East Technical University in Turkey Microwave Humidity Sounder on NOAA 18 and 19 and on MetOp Meteosat Second Generation Meteosat 8 9 10 11 Meteosat Visible and Infra Red Imager on Meteosat up to 7 Micro Wave National Environmental Satellite Data and Information Services National Meteorological Administration of Romania National Oceanic and Atmospheric Administration Agency and satellite SAF in support to Nowcasting amp Very Short Range Forecasting Numerical Weather Prediction SAF on Numerical Weather Prediction SAF on Ozone and Atmospheric Chemistry Monitoring Hungarian Meteorological Service HSAF
184. l products testing 7 identification of the test bed scenario for satellite products validation Start Time End time January 2011 July 2011 Second Report 31 of March 2011 Final Report 31 of July 2011 Third step code possible Matlab realization for e comparison between radar data and satellite products on SSMI AMSU B and SEVIRI satellite grid Start Time End time June 2011 November 2011 Codes delivery and related documentation 30 of November 2011 Composition of the working group Coordinators Estezr Labo HMS Hungary and Gianfranco Vulpiani DPC Italy Participants Belgium Germany Hungary Italy Slovakia Turkey FIRST REPORT AND SECOND REPORT Reported by Eszter Labo Hungarian Meteorological Service Contributors Gianfranco Vulpiani DPC Italy Angelo Rinollo Belgium Jan Kanak and Luboslav Okon Slovakia Firat Bestepe Turkey Rafal lwanski Poland Claudia Rachimow Germany Description of tasks In the HSAF project satellite based precipitation estimations are compared regularly with the radar derived precipitation fields However radar rainfall products are influenced by several error sources that should be carefully analyzed and possibly characterized before using it as a reference for validation purposes However we have to emphasize that the radar data used for validation purposes is not developed by the validation groups themselves They are developed within specialized radar working tea
185. lidation workshop are joined with the Hydrological validation group The results of the Product Validation Programme are reported in this Product Validation Report PVR and are published in the validation section of the H SAF web page A new structure and visualization of the validation section of H SAF web page is in progress to take into account the user needs This validation web section is continuously updated with the last validation results and studies coming from the Precipitation Product Validation Group PPVG In the last Validation Workshop hosted by Slovensk Hydrometeorologick stav in Bratislava 20 22 October 2010 it has been decided to introduce several Working Groups to solve specific items of Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 20 183 HSAF Product Validation Report PVR 01 validation procedure and to develop software used by all members of the validation cluster The coordinators and the participants of the working groups are members of the PPVG or external experts of the institutes involved in the validation activities The first results obtained by the Working Groups are here reported 3 2 Validation objects and problems The products validation activity has to serve multiple purposes e to provide input to the product developers for improving calibration for better quality of baseline products and for guidance in t
186. lt RR lt 025 025 lt RR lt 1 1 lt RR lt 10 AR gt 10 Oe RR lt 025 025 RR lt 1 t lt RR lt 10 RA gt 10 RR Radar RR Radar Figure 73 Contingency table statistic of Rain Rate mmh 1 for PR OBS1 vs radar data Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 101 183 HSAF Product Validation Report PVR 01 wo w w w w w v wo o Zu Eo 0 v x0 0 a 2 1 w o o clas 1 clase 2 clase 3 clase 4 class 1 class 2 class 3 class a O lt RR lt 0 25 025 lt RR lt 1 1 lt RR lt 10 ARS 10 O lt AR lt 025 025 lt RR lt 1 t lt RR lt 10 RA gt 10 RR Rain Gauge RR Rain Gauge Figure 74 Contingency table statistic of rain rate mmh 1 for PR OBS1 vs rain gauge data Results of the continuous statistic Table 9 show positive Mean Error ME in the period 5 6 December with both kind of ground data in the first class in opposite to the whole month which means that H SAF product overestimated small precipitation amounts For the other classes there is generally an overestimation Standard deviation SD with 1 46 mmh for the class RR gt 0 25 mmh is the highest for validation with radar data for 5 6 December the correlation coefficient CC with mostly less than 0 1 is more worse than for results in summer analogue to the r
187. lt 1 1 lt RR lt 10 RR gt 10 RR Radar RR Radar Figure 62 Contingency table statistic of rain rate mmh 1 for PR OBS1 vs radar data Left for 7th August 2010 Right for whole August 2010 HSAF Product Validation Report PVR 01 Product H01 PR OBS 1 Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Date 30 09 2011 Page 93 183 m RRsat in class 1 MRRsatin class2 MWRRsatinclass3 W RRsat in class 4 class 4 Oe RR 0 25 class 2 clase 3 025 lt RR lt 1 1 lt RR lt 10 RR Rain Gauge class a RR gt 10 class 1 0 lt RR lt 0 25 class 2 0 25c RRet RR Rain Gauge clase 3 t lt eARCIO class 4 RR gt 10 Figure 63 Contingency table statistic of rain Rate mmh 1 for PR OBS1 vs rain gauge data Left for 7th August 2010 Right for whole August 2010 Results of the continuous statistic see next Table show negative Mean Error ME for detection of precipitation RR gt 0 25 mmh which means that H SAF product underestimates the fact of precipitation generally Standard deviation SD with 2 97 mmh for this class is the highest for validation with radar for 7 August nevertheless the correlation coefficient CC with 0 46 is the best analogue to the results for POD see above Conclusions The detection of precipitation RR
188. lts were some worse than for the case of 7 August in summary the results are similar The detection of precipitation RR gt 0 25 mmh in comparison with both radar and rain gauge data was quite good for higher rain rates the probability of detection is lower although lower false alarms with one exception for comparison with ground data The quantitative precipitation amounts were overestimated for small amounts and underestimated generally for rain rates greater Immh 5 3 3 Case study 5 6 of December 2010 River Rhine catchment Description Intense rains on 5 6 December 2010 lasting over 72 hours fell along an air mass boundary lying across France and Germany It was a result of subtropical air from south west and polar cold air over Central Europe moving forward to south First precipitation as snow and rain were observed on 5 in relation to the cyclone Liane in northern parts of Germany On the evening the precipitation deflected to the south of Germany In higher regions of the river Rhine they fell as snow In the night to 6 December in south of river Danube the snow changed to rain 7 Der Wetterservice f r NRW und Deutschland R ckblick Starkniederschlage Hochwasser West Mitteleuropa 05 12 09 12 2010 Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 98 183 Product Validation Report PVR 01 sna A RAEI ERN
189. me of the countries are testing new procedures for dealing with VPR Italy and Partial Beam Blockage PBB effects VPR Vertical Profile of Refelctivity used in Turkey v vyv This means that the corresponding rainfall estimates are diverse and the estimation of their errors cannot be homogenized However each county can provide useful information of the error structure of its rainfall products based on its own resources e g if they have already defined Quality Indicators or estimations of errors based on studies of comparison of radar and rain gauge data in the country itself In the future possible separation of reliable and quasi reliable radar fields would be possible Separation would be based on radar site geographical areas event type radar products Selected cases will be suitable enough to be used as a reference for the H SAF products validation Satellite product testing will be carried out in areas with higher reliability Statistical results will be evaluated and compared to previous data As such the accuracy of statistical results of PPVG with radar data as ground reference will be able to be established References References have been collected from each country describing radar data radar data quality and radar data quality estimation techniques This list will be the baseline for further work of the Radar WG The following list of references has been set up Doc No SAF HSAF PVR 01 1 1 Issue Revision In
190. mporarily excludes such stations from the analysis Currently the stations can be put into the blacklist only manually but development of the automated blacklisting is expected in near future Brief description of the RADOLAN system RADOLAN is a routine method for the online adjustment of radar precipitation data by means of automatic surface precipitation stations ombrometers which has started on a project base at DWD in 1997 Since June 2005 areal spatial and temporal high resolution quantitative precipitation data are derived from online adjusted radar measurements in real time production for Germany The data base for the radar online adjustment is the operational weather radar network of DWD with 16 C band sites on the one hand and the joined precipitation network of DWD and the federal states with automatically downloadable ombrometer data on the other hand In the course of this the precipitation scan with five minute radar precipitation data and a maximum range of 125 km radius around the respective site is used for the quantitative precipitation analyses Currently from more than 1000 ombrometer station approx 450 synoptic stations AMDA I lIl and AMDA III S of DWD approx 400 automatic precipitation stations AMDA III N of DWD approx 150 stations of the Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 161 183 Product Validation Report PVR 01 Je usa
191. ms in many of the countries Therefore it should not be the aim of the work of the Radar WG to improve the radar data used however it is specifically expected from the current activities to characterize radar data and error sources of the ground data coming from the radar networks of the Precipitation Validation Group PPVG Main error sources of radar rainfall estimations are listed in the Radar Working Group description document e system calibration e contamination by non meteorological echoes i e ground clutter sea clutter clear air echoes birds insects W LAN interferences e partial or total beam shielding e rain path attenuation e wet radome attenuation Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 145 183 HSAF Product Validation Report PVR 01 e range dependent errors beam broadening interception of melting snow e contamination by dry or melting hail hot spots e variability of the Raindrop Size Distribution RSD and its impact on the adopted inversion techniques Moreover several studies have been on radar quality assessments like S alek M Cheze J L Handwerker J Delobbe L Uijlenhoet R 2004 Radar techniques for identifying precipitation type and estimating quantity of precipitation COST Action 717 Working Group 1 A review Luxembourg Germany or Holleman I D Michelson G Galli U Germ
192. n HSAF Validation Activities Rain gauge correction The non corrected precipitation field can be corrected by rain gauge measurements In Hungary we do not make corrections to instantaneous 15 minutes radar data In our institute we only use a correction for the total precipitation for 12 and 24 hour periods For the 3h and 6h accumulated products we use a special method to accumulate rainfalls we interpolate the 15 minutes measurements for 1 minute grid by the help of displacement vectors also measured by the radar and then sum up the images which we got after the interpolation It is more precise especially when we have storm cells on the radar picture because a storm cell moves a lot during 15 minutes and thus we do not get continuous precipitation fields when we sum up only with 15 minutes periods This provides satisfying results However there is still a need for rain gauge adjustment because there are obviously places behind mountains that the radar does not see The radars are corrected with rain gauge data every 12 hours The correction method using rain gauge data for 12 hour total precipitation consists of two kinds of corrections the spatial correction which becomes dominant in the case of precipitation extended over a large area whereas the other factor the distance correction factor prevails in the case of sparse precipitation These two factors are weighted according to the actual situation The weighting factor depends on the a
193. n PPVG The first step was to collect characteristics polarization beam width maximum range range resolution scan frequency geographical coordinates scan strategy elevations of the radar networks which composes the PPVG adopted processing chain and the generated products including the quality map if any The results of the overview of different radar capacities and instruments in each of the participating countries are here reported 4 3 2 The instruments In the PPVG group there are 54 C band radars used or in the plan to be used Their distribution in the countries is Belgium 1 radar Germany 16 radars not BfG products Hungary 3 radars Italy 18 radars Slovakia 2 radars Poland 8 radars Turkey 6 radars These radars cover wide range of geographical area from the longitude 5 50562 in Wideumont Belgium to the most Eastern area with longitude 32 58 15 in Ankara Turkey and from the Northern VVVVVVV Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 39 183 Product Validation Report PVR 01 Je usar latitude of 54 23 03 17 in Gda sk Poland to the latitude of 36 53 24 in Mugla Turkey and lat 37 462 in Catania Italy The Radars are built at different elevations above the sea level In mountainous countries they are placed at elevations more than 1000m above sea level whereas in flat countries like Hu
194. n Report PVR A precipitation product validation section of the H SAF web page is under development This validation web section will be continuously updated with the last validation results and studies coming from the Precipitation Product Validation Group SPVG It is well know that radar and rain gauge rainfall estimation is influenced by several error sources that should be carefully handled and characterized before using these data as reference for ground validation of any satellite based precipitation products A complete inventory of the precipitation ground networks instruments and data available inside the PPVG has been provided in Chapter 4 in order to highlight the main error sources and to present possible methodology for selecting the ground data more reliable Annex 1 7 In the last months the first example of precipitation fields integration has been also provided Section 4 4 INCA and RADOLAN products The INCA system a tool for the precipitation products validation is available in Slovakia and Poland in both countries being run in pre operational mode In Germany similar precipitation analysis system called RADOLAN is being run operationally The study performed in the PPVG Annex 5 showed that the accuracy and reliability of the raingauge stations significantly affect final precipitation analysis of the INCA or INCA like systems In order to solve this problem an automated blacklisting technique is going to be developed at SHMU
195. n analyses usable for H SAF validation INCA developed by ZAMG Austria and RADOLAN DWD Germany The INCA system is currently under development as INCA CE Central Europe and it is used in pre operational mode in Slovakia and Poland The RADOLAN system is used in Germany operationally and it is already utilized for the H SAF products validation Both systems consist of computational modules which enable to integrate various sets of precipitation data sources raingauge network radar network NWP models outputs and climatological standards into common precipitation product which can describe well the areal instantaneous and cumulated precipitation fields Here below a brief description of the INCA and RADOLAN systems follows More information on both systems can be found in the documentation which is available on the H SAF ftp server hsaf WP6000 precipitation WG_groups WG3 inca documentation Google Figure 16 Coverage of Europe by the INCA and RADOLAN systems 4 4 1 INCA system The INCA Integrated Nowcasting through Comprehensive Analysis analysis and nowcasting system is being developed primarily as a means of providing improved numerical forecast products in the nowcasting and very short range forecasting It should integrate as far as possible all available data sources and use them to construct physically consistent analyses of atmospheric fields Among the input data sources belong NWP model outputs in general P T
196. n near the High Tatras mountain in the northern part of Slovakia where only low precipitation rates were observed by radars Fig 1 a The resulting INCA analysis is shown in Fig 1 c The corresponding PR OBS 1 field Fig 1 d shows overestimation even when compared with the rain gauge adjusted field of the INCA analysis Original radars INCA Raingauges 2010081515005 j f201008151500 INCA Radarstraingauges HOI 201008151500 i 4 201008151500 Rain intensity Cwn hi Figure 18 Precipitatio ensity field from 15 August 2010 15 00 UTC obtained by a radars b interpolated raingauge data c INCA analysis and d PR OBS 1 product 4 4 3 Some conclusions The INCA system as a potential tool for the precipitation products validation is available in Slovakia and Poland in both countries being run in pre operational mode It is still relatively new system undergoing continuous development More sophisticated algorithms of the precipitation analysis e g assimilation of the 3 D radar data can be expected from its development in frame of the ongoing INCA CE project In Germany similar precipitation analysis system called RADOLAN is being run operationally This tool is already used for validation of the H SAF precipitation products in this country The accuracy and reliability of the raingauge stations significantly affect final precipitation analysis of the INCA or INCA like systems and therefore need to be checked In order to solve t
197. nation mean The timeliness depends on the satellite transmission facilities the availability of acquisition stations the processing time required to generate the product and the reference dissemination means In the case of PR OBS 1 it is strongly conditioned by the availability of DMSP data at CNMCA through NOAA and UKMO The outcome is e timeliness 6 2 5 h The accuracy RMS is the convolution of several measurement features random error bias sensitivity precision To simplify matters it is generally agreed to quote the root mean square difference observed reference values The accuracy of a satellite derived product descends from the strength of the physical principle linking the satellite observation to the natural process determining the parameter It is difficult to be estimated a priori it is generally evaluated a posteriori by means of the validation activity 3 Validation strategy methods and tools 3 1 Validation team and work plan To evaluate the satellite precipitation product accuracy a Validation Group has been established by the beginning of the Validation Phase in the H SAF project The Precipitation Product Validation team is composed of experts from the National Meteorological and Hydrological Institutes of Belgium Bulgaria Germany Hungary Italy Poland Slovakia and Turkey Figure 5 Hydrologists meteorologists and precipitation ground data experts coming from these countries are involved in the pr
198. nce the beginning of 2008 whereas six C band radars including two dual polarized systems will be operational by the end of 2008 see Figure 1 As an example the national mosaic CAPPI at 2000 m is shown in Figure 2 relatively to the event of 04 18 08 at 0015 U T C 18 04 2008 ore 00 15 Figure 31 Graphical mosaic of reflectivity CAPPI at 2000 m for the event of 04 18 08 at 0015 U T C As depicted in Figure 3 each Doppler Radar System either dual or single polarized PDRS or DRS are connected by satellite links to the two National Radar Primary Centres RPC located in Roma DPC and Savona CIMA Research Foundation in order to mainly ensure the remote control through the RRC server and products generation through the RPG server The RPC located in Savona works as backup centre in order to continuously ensure the system functioning The subsystem RAC Radar Archive Centre is devoted to archive and manage radar data and products by means of a relational database The generated products are then disseminated to all institutions composing the national network Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 61 183 Product Validation Report PVR 01 Figure 32 Architecture of the Italian radar network Data processing Data processing and product generation are here briefly described In particular attenuation correction hydrometeor classifi
199. nd Nearest Neighbor NN Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 181 183 Product Validation Report PVR 01 ta hb w e mo o o m o 0 0 D 8 7 al D 1 N STEP 2 Sampled and Interpolated data Bh senpics cote Interpolated data Figure 122 Example of sampled data for a regular grid In right on the upper part a detail of field studied below the original grid of field for step 2 Note from figure above From the field the white circles mean the data removed from the map The black squares mean the position of perfect measurement The techniques of up down scaling reproduce the field only from the perfect measurements The algorithms used in the validation group are similar to the Barnes algorithm This like Barnes algorithm creates a grid of regular step where each node contains the data calculated from all data weighted by distance from the node itself The calculation is done several times through successive iterations in order to minimize errors in the precipitation field In the following table 65 are reported the values of RMSE were sat is the sampled and captioned data and true is the value of perfect field RMSE 100 N is the total number of pairs data in which the reference value is different by 0 Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 182
200. nd beam attenuation is reduced by limiting the validation area to rain effective range of 120 km for both radars in the composite The instantaneous precipitation products are provided in Mercator projection with approximately 1 km resolution Threshold for precipitation detection is 0 02 mm h Time resolution of the current instantaneous products is 5 minutes for the products prior to April 2010 it was 10 minutes and prior to August 2009 15 minutes Precipitation accumulation in case of 3 hourly interval is based on integration of 5 10 or 15 minutes instantaneous measurements in time period of 3 hours Accumulated precipitation for intervals of 6 12 and 24 hours is calculated as a sum of the 3 hourly accumulated precipitation At least 92 of instantaneous measurements must exist in relevant time period for the 3 hourly accumulated product to be produced No rain gauge correction of the accumulated precipitation is applied but the same limitation of validation area is used as for the instantaneous product Threshold for precipitation detection of the 3 Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 72 183 Product Validation Report PVR 01 Je usar hourly accumulated product is 0 5 mm Geographical projection and space resolution of the accumulated products are the same as those of instantaneous product see above For validation of H SAF precipitation pro
201. nf on Radar Meteorology Proceedings AMS 5 9 10 2009 Williamsburg VA USA Katarzyna Osrodka Jan Szturc Anna Jurczyk Daniel Michelson Gunther Haase and Markus Peura Data quality in the BALTRAD processing chain Proceed of 6th European Radar Conference ERAD 2010 Sibiu Romania http www erad2010 org pdf oral wednesday dataex 06_ERAD2010_ 0240 pdf igus Doc No SAF HSAF PVR 01 1 1 HSAF Product Validation Report PVR 01 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 149 183 Szturc J Osrddka K and Jurczyk A Quality index scheme for quantitative uncertainty characterization of radar based precipitation Meteorological Applications 2010 doi 10 1002 met 230 BELGIUM ITALY HUNGARY List of Rain rate 240 Km CMAX Available rain rate 120 Km PPI Products velocity 120 Km CAPPI 2 5 km MAX 240 Km VIL VVP2 Windprofiles ETops Hail Probability Base Hail Probability 24h HailProbability Overview 1 3 24 Hr Rainrate accumulation Is any quality NO YES NO map available Processing Clutter removal time Clutter suppression by RLAN wifi filter chain domain Doppler filtering Fuzzy Logic scheme using Clutter removal and static clutter map Clutter map Velocity atttenuation Z R a 200 b 1 6 Texture correction beam Z R a 200 b 1 blocking correction gt VPR correction under next Year 2012 testing VPR
202. ng the distribution beyond the area defined by stations In order to evaluate the quality of the error distribution the cross validation was performed and the results are presented on the next figure Product Validation Report PVR 01 Product H01 PR OBS 1 Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Date 30 09 2011 Page 176 183 kriging Natural Neighbour d Estimated mm Estimated mm Real mm Real mm Estimated mm Real mm Figure 121 Cross validation results obtained for three different methods for spatial interpolation For all methods the results are similarly scattered around the perfect estimation however for IDW 2 some underestimation was found for negative ME values The values of Mean Residual and Mean Absolute Residual defined as mean and mean absolute difference between Estimated and Real values of ME are presented in the next table Mean Residual Mean Absolute Residual Kriging Natural Neighbour IDW 2 Table 56 Mean Residual and Mean Absolute Residual values obtained for three algorithms for spatial interpolation using cross validation approach The lowest value of Mean Absolute Residual was found for Natural Neighbour method what indicates that application of this algorithm may allow for minimizing the systematical error introduced
203. ngary or Belgium their height position is not exceeding 400m This information collected will be useful in the future steps of the Working Group to assess the partial or total beam shielding by mountains in the propagation way of the radar signals All radars are C band radars working at frequency in C band at 5 6 GHz All radars are equipped by Doppler capacity which means that ground clutters can be removed from the radar data measurements effectively however not all of them have dual polarization which would be important to correct rain path attenuation The scan strategy for each of the radars used has been investigated In this matter all countries have shared their information on the number of elevations minimum and maximum elevations scan frequency maximum nominal range distance and range resolution Height N Azimuth lt lt gt x Slant Range w eee E Volume Scan Procedure S Figure 15 Radar scan procedure In the PPVG the scan frequency ranges from 5 minutes in Belgium Germany and Slovakia to 10 minutes in Turkey and Poland and 15 minutes in Hungary and varying frequency for Italian radars The number of elevation stays between 4 and 15 in average around 10 The range distance used is 240 km in general But in some places in Italy and for the Turkish radars the maximum range distance used is 120 km or even less e g 80 km Range resolution is 250 m in Belgium 250 340 225 and sometimes 50
204. notice very good ability of HO1 to recognize both no rain and heavy precipitation situations more than 80 of ground cases was properly allocated by satellite product The light precipitation is strongly overestimated more than 70 of cases is allocated in the moderate and Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 112 183 HSAF Product Validation Report PVR 01 heavy precipitation classes The moderate precipitation is properly recognized in more than 50 of cases mm h E m gt 10 110 nn m 0 25 1 0 m 00 25 H 01 Percentage contribution o o o a S a amp e 00 25 0 251 0 110 gt 10 Rain rate RG mm h Figure 87 Percentage distribution of PR OBS 1 precipitation classes in the rain classes defined using rain gauges RG data on the 15th of August 2010 Some Conclusions To sum it up the analysis performed for situation with convective precipitation showed very good ability of PR OBS 1 product in recognition of precipitation of no rain and heavy ones rain rate gt 10mm h while the light precipitation is strongly overestimated The displacement of the maximum precipitation was also found 5 6 2 Case study 17 of May 2010 Description Between 15 and 20 of May 2010 Poland was within a centre of low pressure moving from Northern Italy to Hungarian Lowlands and Ukraine Due
205. o that cell Software development extraction of a regular subset in the PR ASS 1 files The first program we developed useful to all groups with both the validation approaches allows to select a fixed number of rows and columns in the PR ASS 1 files given the geographical extremes of the chosen validation area In this way it s possible to process uniquely the data falling in and around the region of interest Software development upscaling of fine resolution data to the COSMO grid A prototype version of the upscaling procedure has been developed and successfully tested over Belgium It consists of two programs the first creates a lookup table a file which states a correspondence between every point of the observational grid radar in this case and the corresponding cell of the chosen subsection of the COSMO grid in which it falls The second upscales every observational file to the COSMO grid given the lookup table and it is part of the Belgian validation procedure previously developed by E Roulin RMI Belgium Preliminary testing results Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 172 183 HSAF Product Validation Report PVR 01 Here are the preliminary results obtained by testing the upscaling procedure over Belgium On the first image the original images from Wideumont radar RMI Belgium and on the second one the corresponding u
206. o the overall RMSEcomparison WG and VS started to evaluate the sources of errors awaiting the final results is it possible to reconsider the requirements like to understand the thresholds of requirements in table 63 Then it is need to anticipate the likely size of these errors The very low POD values and very high FAR values as well as the invariably poor values of the correlation coefficient indicate that RMSEcomparison could be dominant in the error partitioning with RMSEsat and RMSEground An estimate of the errors due to the various effects impacting the RMSEcomparison is not difficult It is not necessary to build a large statistics but just perform experiments using a few campaigns carried out over one dense rain gauge network and one well calibrated radar In fact the purpose is simply to evaluate the size of RMSEcomparison not to reduce it that would require a large effort probably unproductive For the sake of providing an example it is noted that if the three contributions RMSEsat RMSEground and RMSEcomparison were of comparable size equipartitioning of the error would improve the RMSD by a factor 3 1 7 and the figures resulting from the current validation would match at least the threshold requirements In order to obtain an estimate of RMSEcomparison and then a more accurate estimation of RMSEsat CNR ISAC performed an experiment based on its polarimetric C band radar Polar 55C located close to Rome surrounded by a net
207. oduct validation activities Table 2 H01 has been submitted to validation in all these countries except Bulgaria Until now the Bulgarian data are used only for HO5 validation activity according to the Project Plan Their use in the next months is under consideration Figure 5 Structure of the Precipitation products validation team Validation team for precipitation products Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 19 183 Product Validation Report PVR 01 Slovensky Hydrometeorologicky Ustav Jan Ka k SHM Slovakia _ jan kanak shmu sk Slovensk Hydrometeorologick stav uboslav Okon SHM Slovakia luboslav okon shmu sk Slovensk Hydrometeorologick stav Mari n Jurasek SHM Slovakia marian jurasek shmu sk Table 2 List of the people involved in the validation of H SAF precipitation products The Precipitation products validation programme started with a first workshop in Rome 20 21 June 2006 soon after the H SAF Requirements Review 26 27 April 2006 The first activity was to lay down the Validation plan that was finalised as first draft early as 30 September 2006 After the first Workshop other ones followed at least one per year to exchange experiences problem solutions and to discuss possible improvement of the validation methodologies Often the Precipitation Product Va
208. on evaluation regards to terrain visi Jan Ka k uboslav Okon SHM of radar measurements quality indicator with For validation of H SAF precipitation products it is necessary to know errors distribution of used ground reference In this case precipitation intensity or accumulated precipitation measured by SHM radar network is considered as a ground reference To find distribution of errors in radar range next steps must be done Je usar Product Validation Report PVR 01 Product H01 PR OBS 1 Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Date 30 09 2011 Page 151 183 e simulations of terrain visibility by radar network using 90m digital terrain model e statistical comparison of radar data against independent rain gauge data measurements e derivation of dependence regression equation describing the errors distribution in radar range with regard to terrain visibility based on rain gauge and radar data statistical evaluation e computation of error distribution maps using regression equation and terrain visibility 24 hour cumulated precipitation measurements from 68 automatic precipitation stations from the period 1 May 2010 30 September 2010 were coupled with radar based data Distribution of gauges according their elevation above the sea level is shown in next figure Iml Rain gauge altitude m HUTY Figure 102 Distribution of
209. on map Statistical scores The ability of HO1 product to recognize the precipitation was analysed using dichotomous statistics parameters performed for all overpasses available for the 15 of August 2010 The 0 25mm h threshold was used to discriminate rain and no rain cases In the Table 1 the values of Probability of Detection POD False Alarm Rate FAR and Critical Success Ratio CSI are presented Parameter Scores POD 0 93 FAR 0 62 csi 0 37 Table 32 Results of the categorical statistics obtained for PR OBS 1 Higher value of POD than the value of FAR indicate that the product ability to recognize the convective precipitation is quite good The quality of H01 in estimating the convective precipitation is presented on the next figure One can easily notice that HO1 overestimates the rain rate for light and moderate precipitation 60 50 ee E 30 3 6 Eal 2 ee sd 10 a 3 o 10 20 30 40 50 60 RG mm h Figure 86 Scatter plot for measured RG and satellite derived H01 rain rate obtained for all H01 data on the 15th of August 2010 Finally the analysis of rain classes was performed The categories were selected in accordance with the common validation method Next figure shows the percentage distribution of satellite derived precipitation categories within each precipitation class defined using ground measurements One can easily
210. onal range 240 km for radar at Maly Javornik left 200 km for radar at Kojsovska hola right The instruments The radars are operated and technically maintained by SHMU Receivers of radars are calibrated regularly by means of internal test signal generator TSG In case of radar at Maly Javornik calibration is performed every 3 months and in case of radar at Kojsovska hola every 1 month The basic parameters of both SHMU radars are summarized in next table Maly Javornik Kojsovska hola Frequency band C Band 5600 MHz C Band 5617 MHz Polarization Single Double but so far only single pol Single Double 8 products generated frequency elevations maximum nominal range distance range resolution Elevations deg 0 2 0 7 1 42 5 3 85 4 7 3 9 5 13 0 17 0 25 0 Range 240 Km Resolution 1000m d lidati Doc No SAF HSAF PVR 01 1 1 HSAF Product Validation Report PVR 01 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 71 183 Doppler capability Yes Yes Yes No Scan strategy scan Scan frequency 5 min Scan frequency 5 min Elevations deg 0 5 0 0 0 5 1 0 1 5 2 5 4 0 6 0 10 0 20 0 Range 200 Km Resolution 125m Table 18 Characteristics of the SHMU radars The data processing For ground clutter removal the Doppler filtering is used In case of radar at Maly Javornik the frequency domain IIR filter is used at Kojsov
211. ore ETS during the present case study Conclusions From qualitative and statistics comparison it appears that for this case study summer storm characterized by convective rainfall the HO1 product could reproduce the rainfall patterns and amounts with quite good confidence The qualitative location of the precipitating cells is correct in 1 particular in the range between 1 and 10 mm h 5 2 2 Case study 22 24 of August 2010 Description of the event Figure 50 Surface map on 22 August 2010 at 06 UTC MSLP and synoptic observations This event has been chosen because thunderstorms with intense precipitation resulted in local flooding in Belgium The country was at the edge of a large anti cyclone that was moving away towards South East next figure Warm but humid and unstable air was brought from South West whereas a cold front was moving from West Data used HSAF Product Validation Report PVR 01 f Product H01 PR OBS 1 Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Date 30 09 2011 Page 83 183 Products H01 from 6 00 UTC of August 22 to 12 00 UTC of August 24 have been considered The total is 10 satellite passages distributed as follows 1 in the morning of August 22 2 inthe afternoon of August 22 3 in the morning of August 23 1 in the afternoon of August 23 3 in the morning of August 24 The ground data used for validation
212. otprint Next figure presents satellite foot print FOV centers of the HO1 and H02 products an elliptical footprint for the corresponding center area within the yellow dots and Awos ground observation sites The comparison statistic can be performed by considering just the sites in the footprint area Although this approach is reasonable on the average but it is less useful in spatial precipitation variability representation The comparison is not possible when no site is available within the footprint area Awos sites FOV center Figure 43 H01 and H02 products footprint centers with a sample footprint area as well as the Awos ground observation sites Alternatively the point to area approach is more appealing for the realistic comparison of the precipitation product and the ground observation This approach is simply based on the determination of the reference precipitation field underneath the product footprint area To do so the footprint area is meshed and precipitation amounts are estimated at each grid point by using the precipitation observations at the neighboring Awos sites as shown in next figure A 3x3 km grid spacing is considered for the products with elliptical geometry while 2x2 km spacing is considered for the Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 76 183 Product Validation Report PVR 01 Je usar produc
213. ound All validation groups not only for precipitation but also for soil moisture and snow have been requested to quote figures to characterize the errors of the ground reference that they used The various teams did this after consultation with the operational units in charge of the observing networks in their institutes For precipitation the following figures were quoted Rain gauge UniFerrara Radar BIG Rain gauge SHMU Radar OMSZ Radar IRM Rain gauge interpolated Gauge adjusted radar IMWM Gauge TSMS Gauge Table 58 Errors of the ground reference provided by all validation groups The values of table above apart from details indicates that the errors due to the ground reference are of the same order than the threshold requirements It is interesting to note that the validation activity has indicated that the results from rain gauge and radar are comparable whereas the error of radar should be definitively higher This means that radar is favored in the third error type RMSEcomparison RMSEcomparison is in reality a composition of several errors It refers to the limitations of the comparison method that in spite of all efforts envisaged and implemented by the validation teams has left residual errors difficult to be further reduced but needing evaluation by in depth investigation A short list is upscaling downscaling processes to make compatible the instrument resolution and the
214. ounts especially the higher ones Scores evaluation The scores obtained for the present case study Table 2 are better than the long period scores but poorer if compared with the other summer case study In particular here the product is remarkably underestimating Probability of detection is still high but also false alarm ratio is unlike the other case This might be connected with the fact that in this case the fraction of area interested by the rainfall is smaller Sample 10 Mean error 0 77 Standard deviation 1 98 Mean absolute error 1 23 Multiplicative bias 0 56 Correlation coefficient 0 27 Root mean square error 2 07 URD RMSE 1 45 POD 0 83 FAR 0 53 CSI 0 43 Table 21 Scores obtained with the comparison with radar data in mm h 1 The time evolution of the fraction area with rain measured by radar gt 0 25 mm h and the Equitable Threat Score ETS is reported in next figure Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 85 183 Product Validation Report PVR 01 v gt pa E J E 8 E 2 al 8 q 5 3 3 4 5 k E g A 23 24 25 p E E E A P ee 23 24 25 Figure 53 Time evolution of fraction area with rain measured by radar gt 0 25 mm h and Equitable Threat Score ETS during the present case study Conclusions From qualitative and statistics compari
215. ours refer to the months of the year 2009 Assuming these values significant for the other Countries involved in this study we can conclude that the gauge network in PPVG is capable to resolve the spatial structure of rain patterns only for stratified systems but it is inadequate for small scale convective events Country Total number of gauges Average minimum distance km Belgium 89 11 2 Bulgaria cT ii 7 Germany 1300 17 Italy 1800 9 5 Poland 330 475 13 3 Turkey 193 27 Table 7 Number and density of raingauges within H SAF validation Group the number of raingauges could vary from day to day due to operational efficiency within a maximum range of 10 15 only in the Wallonia Region only in 3 river basins only covering the western part of Anatolia 4 2 2 The instruments Most of the gauges used in the National networks by the PPVG Partners are of the tipping bucket type which is the most common device used worldwide to have continuous point like rainrate measurement Nevertheless several source of uncertainty in the measurements are well known but difficult to mitigate First very light rainrates 1 mm h and less can be incorrectly estimated due to the long time it takes the rain to fill the bucket Tokay et al 2003 On the other side high rainrates above 50 mm h are usually underestimated due to the loss of water during the tips of the buckets Duchon and Bid
216. panel HO1 product left panel at 6 45 UTC Figure 81 HO1 precipitation map at 15 42 UTC Figure 82 Raingauges hourly precipitation cumulateda at t16 00 UTC right of 06 July 2010 Please note different colour scales 108 Figure 83 Synoptic chart at 1200 UTC on 15th of August 2010 109 Figure 84 Total lighting map of Poland showing electrical activity between 1445 and 1515 UTC on 15th of August 2010 Figure 85 PR OBS 1 at 1459 UTC on the 15th of August 2010 right panel and 10 minute precipitation interpolated from RG data from 1500 UTC left panel 110 Figure 86 Scatter plot for measured RG and satellite derived HO1 rain rate obtained for all HO1 data on the 15th of August 2010 111 Figure 87 Percentage distribution of PR OBS 1 precipitation classes in the rain classes defined using rain gauges RG data on the 15th of August 2010 112 Figure 88 Synoptic chart at 0000 UTC on 17th of May 2010 Source MWM 113 Figure 89 PR OBS 1 at 0453 UTC on the 17th of May 2010 right panel and 10 minute precipitation interpolated from RG data from 0500 UTC left panel Figure 90 PR OBS 1 at 0546 UTC on the 17th of May 2010 right panel and 10 minute precipitation interpolated from RG data from 0550 UTC left panel Figure 91 Scatter plot for measured RG and satellite derived Ha 01 rain rate obtained for all PR OBS 1 data on the 17th of May 2010 Figure 92 Percentage distribution of PR OBS 1 precipitation classes in the rain classes defined us
217. papers and the characteristics of the ground data available inside the PPVG the main next steps are foreseen in order to improve the validation methodology e quantitative estimation of the errors introduced in the validation procedure comparing the instantaneous satellite precipitation estimation with the rain gauge precipitation cumulated on different intervals e definition of a rain gauge and radar data quality check e application of the data quality check to all radar and rain gauge data used in the PPVG e definition of the optimal and minimal spatial density of rain gauge stations to be representative of the ground precipitation in the view of satellite product comparison e development of the three software for raingauges radar and INCA products up scaling vs SSMI and SSMI S grids e definition and code implementation of the technique for the temporal matching of satellite rain rate with rain gauge and radar data e selection of the appropriate methodology for spatial distribution of precipitation products errors taking into consideration spatial and temporal characteristics of each product for selected areas as test catchments HSAF Product Validation Report PVR 01 Product H01 PR OBS 1 Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Date 30 09 2011 Page 134 183 All these activities will be developed and coordinated inside the Working Groups Annex 1 7 Doc No SAF HSAF PVR
218. parameters the first step in the creation of maps of H SAF precipitation products error was the selection of the interpolation algorithm Commonly used Ordinary Kriging Inverse Distance Weighted and Natural Neighbour methods were tested firstly The analysis was performed for monthly average mean error of H 05 3 h cumulated precipitation for selected months In the analysis data from Polish rain gauges were used In the next figure the example mean error maps for July 2010 obtained using three mentioned above algorithms are presented Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 175 183 HSAF Product Validation Report PVR 01 H 05 3h July 2010 ME Ordinary Kriging HOS 3h July 2010 ME Natural Neighbour po gg Rs o1 a j 01 04 aad 01 03 p 03 05 a 54 01 01 8 03 R 05 LJ 07 5 7 09 aa t 13 T T 15 687 8 8 Dn RD mM Figure 120 Distribution of the monthly average H 05 3 h cumulated precipitation Mean Error calculated for July 2010 using three methods a Ordinary Kriging b Natural Neighbour and c IDW 2 One can see that the obtained maps do not differ significantly however for the map created with the use of Natural Neighbour method the maximum and minimum values are less pronounced that on the other two maps Moreover application of Natural Neighbour method does not allow for extrapolati
219. pitation developers of CNR ISAC and it has been adopted by the PPVG One of the Radar WG and Rain Gauge WG next steps is to develop a common code for the up scaling of radar data versus SSMI and SSMI S grids following this technique The code will be an evolution and optimization of the code already available by Belgium Van de Vyver H and E Roulin 2008 All participants of validation task will use not only the same technique but the same software 3 7 Temporal comparison of precipitation intensity Taking into account the revisiting time of the PR OBS 1 3 4 hours it was decided during the first validation workshop in 2006 to perform a direct comparison between the satellite and radar precipitation intensity maps The revisiting time of the product does not allow to have a sensible accumulated precipitation map on 1 24 hours In the PPVG the satellite product is compared with the closest up scaled radar and rain gauge data in time The satellite time is considered the time in the BUFR4 file provided by CNMCA when validation area is first reached 3 8 Large statistic Continuous and multi categorical The large statistic analysis allows to point out the existence of pathological behaviour in the satellite product performance It requires the application of the same validation technique step by step in all the institutes take part of the PPVG Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 D
220. products are derived from this by taking into account the radar displays which give a horizontal cross section of data at constant altitude The CAPPI is composed of data from several different angles that have measured reflectivity at the requested height of CAPPI product The PPVG group uses mostly CAPPI products for calculation of rainfall intensities except for Hungary which uses the CMAX data maximum radar reflectivity in each pixel column among all of the radar elevations for deriving rainfall intensities However the rest of the countries have also chosen different elevation angles for the CAPPI product which provides the basis for rain rate estimations Additionally we have to say that the countries apply different techniques of composition of radar data that were not specified in this questionnaire The composition technique is important in areas which are covered by more than one radar measurements Also the projection applied is varying from one country to the other To sum up the radar products used are not harmonized different techniques are applied However each of them is capable to grasp rainfall and to estimate rainfall intensity Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 147 183 HSAF Product Validation Report PVR 01 As for the accumulated products we see that Belgium uses 24 hourly accumulations with rain gauge correction
221. pscaled images The images appear correctly upscaled Figure 118 the Wideumont radar image of 1 2 2010 cumulated rainfall in the previous 24 hours raingauge adjusted M viii shins L Figure 119 The Wideumont radar image of 1 2 2010 cumulated rainfall in the previous 24 hours raingauge adjusted and upscaled to the COSMO grid Adaption of the software to all the groups and delivery for testing present status After successful testing over Belgium the software has been adapted for common use by all the other groups and then delivered for testing Some feedback from Hungary and Slovakia Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 173 183 Product Validation Report PVR 01 Je usar has already been received and used for improvements The testing by all the groups is still in progress References About the COSMO model rotated grid see http www cosmo model org content model documentation core cosmoDyncsNumcs pdf pages 21 27 14 Annex 7 Working Group 5 Geographical maps distribution of error PROPOSAL Validation activities performed during Development Phase for land and coast areas showed the difference in H SAF precipitation products quality depending on geographical localisation Those first achievements as well as the request from Hydrological Validation Group to provide the error characteristic of pre
222. r densification measurement network of the federal states the hourly measured precipitation amount is used for the adjustment procedure In advance of the actual adjustment different preprocessing steps of the quantitative radar precipitation data are performed These steps partly already integrated in the offline adjustment procedure contain the orographic shading correction the refined Z R relation the quantitative composite generation for Germany the statistical suppression of clutter the gradient smoothing and the pre adjustment Further improvements of these procedures are being developed Precipitation distribution of Precipitation distribution of RADOLAN precipitation the rain gauge point the areal original radar product measurements measurements TON C P sa w j i Figure 112 Procedure of the RADOLAN online adjustment hourly precipitation amount on 7 August 2004 13 50 UTC In order to collect more detailed information about both types of systems a questionnaire was elaborated and completed by Slovakia Poland and Germany The questionnaire provided details such as geographical coverage input data inventory or availability of different instantaneous and cumulated precipitation products The final version of the questionnaire is shown in the next table and is also available on the H SAF ftp server hsaf WP6000 precipitation WG_groups WG3 inca questionnaire Doc No SAF HSA
223. r and rain gauge WG other WG have been composed on integrate various sets of precipitation data sources raingauge network radar network NWP models outputs and climatological standards into common precipitation product which can describe the areal instantaneous and cumulated precipitation fields INCA WG and to investigate the opportunity to create geographical maps of error distribution for providing information on test catchments to the Hydrological Validation Group GEO MAP WG 3 3 Validation methodology From the beginning of the project it was clear the importance to define a common validation procedure in order to make the results obtained by several institutes comparable and to better understand their meanings The main steps of this methodology have been identified during the development phase inside the validation group in collaboration with the product developers and with the support of ground data experts The common validation methodology is based on ground data radar and rain gauge comparisons to produce large statistic multi categorical and continuous and case study analysis Both components large statistic and case study analysis are considered complementary in assessing the accuracy of the implemented algorithms Large statistics helps in identifying existence of pathological behaviour selected case studies are useful in identifying the roots of such behaviour when present The main steps of the validation procedure
224. r beam above the rain gauge y 1E 06 0 001 0 5262 eationofosanabovethe riage in Figure 107 Scatterplot of log R G versus radar beam altitude shows increased underestimation of radar for high and close to zero radar beam elevations Polynomial trend line on the Figure 106 differs from trend line of Figure 107 While in case of rain gauge altitudes the lowest underestimation by radar can be observed for the lowest rain gauge altitudes in case of radar beam altitudes the lowest underestimation by radar is observed for radar beam elevation about 500m Stronger underestimation for rain gauges with close to zero radar beam elevation can be explained by partial signal blocking by terrain obstacles These are the cases when rain gauge station is close to the top of terrain obstacle Finally set of statistical parameters for each single rain gauge station was computed mean error standard deviation mean absolute error multiplicative bias correlation coefficient RMSE and relative RMSE Relative RMSE and Mean Error were selected to be specified for radar precipitation measurement over the whole radar range For this purpose quadratic or linear polynomial trend lines were created as is shown on next figure Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 155 183 Product Validation Report PVR 01 Je usar Slovak radar network URD_RMSE based on
225. r errors cannot be homogenized Je usar Product Validation Report PVR 01 Product H01 PR OBS 1 Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Date 30 09 2011 Page 41 183 However each county can provide useful information of the error structure of its rainfall products based on its own resources e g if they have already defined Quality Indicators or estimations of errors based on studies of comparison of radar and rain gauge data in the country itself The study performed by the Slovakian team Annex 4 and the scheme published by J Szturc on the quality index evaluation are under consideration by the Radar WG In the future possible separation of reliable and quasi reliable radar fields would be possible Separation would be based on radar site geographical areas event type radar products Selected cases will be suitable enough to be used as a reference for the H SAF products validation A study on evaluation of radar measurements quality indicator with regards to terrain visibility has been conducted by the Slovakian team see Annex 4 Satellite product testing will be carried out in areas with higher reliability Statistical results will be evaluated and compared to previous data As such the accuracy of statistical results of PPVG with radar data as ground reference will be able to be established BELGIUM ITALY HUNGARY List of Rain rate 240 Km C
226. radar n Slovak radar network Mean Error based on radar uro RMSE minimum visible height above the rain gauge Imh minimum visible height above the rain gauge Radar minimum visible height above the rain gauge m Radar minimum visible height above the rain gauge m Figure 108 Relative RMSE left and Mean Error right computed independently for each rain gauge station in radar range and corresponding trend lines extrapolated for beam elevation up to 1500m Relative RMSE and Mean Error can be specified for each pixel of radar network composite map using regression equations which describe dependence on minimum radar beam elevation above the surface This can be considered as quality indicator maps of radar measurements with regard to terrain visibility by current radar network of SHM as is shown in next two figures Figure 109 Final relative root mean square error map of radar measurements with regard to terrain visibility by current radar network of SHM Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 156 183 Product Validation Report PVR 01 Figure 110 Final mean error map of radar measurements regard to terrain vi y by current radar network of SHMU General underestimation of precipitation by radars is observed Conclusions Considering the fact that reference precision of rain gauges used in this study is
227. radar network time resolution Smin No Yes 5 minute Yes 5 minute timelines instantaneous precipitation based on combined raingauge and Smin Yes 10 minutes Yes 5 minutes Yes 5 minutes k time resolution timelines je precipitation based only on Yes min 5 min available Yes min 5 min available aingauge network tme tervals S min 136121828 hours No pype prr ionan Cumutative precipitation based only on aS Yes min 5 min available Yes min 5 min available radar network time intervals timelines 5 1316121824 ho ba 413 612 24 hours 1 36 12 24 hours Cumulative precipitation based on Yes min 10 minutos Yes min 5 min available Yes min 5 min avalable combined raingauge and radar Smin 13612 18 26 hours Yes min 10 minutes sae sae beware lable in futur 413 6 12 24 he 1 36 12 28 he Dates for selected case studies case 1 willbe set No 29 32009 case 2 No 1 362010 Icases No 206 2010 case 4 No 18 16 8 2010 lcases No Availabilty of own software for upscaling ra me i w INCA data into native satelite grid liad n a o2 yes No No No Hos yes No No No Hos no No No No Inos yes No No No Inos yes No No No Table 54 Questionnaire Case studies Several case studies comparing the INCA analyses with their source precipitation fields from radars and raingauges and with selected H SAF products have been elaborated at SHMU The precipitation fields from individual observations have been compared visually
228. recipitating systems dominate and reduces to roughly 10 km when small scale convection is more likely to occur warm months This points out that the distribution of gauges could be able to describe the spatial structures of precipitation fields in case of wintertime rainfall while may be inadequate for spring summer convective events 200812 200901 200902 200903 200904 200905 200906 200907 200908 200909 200910 200911 0 6 0 4 0 2 0 0 20 40 60 80 km Figure 28 Correlation between rainrates detected by two close by stations as function of the distance between the two stations Colors refer to the month along 2009 In the following figure the distribution of working stations over Italy is shown for a given day Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 58 183 Product Validation Report PVR 01 Figure 29 Distribution of the raingauge stations of the Italian network collected by DPC The instruments The following information should be provided in this section Allthe available raingauge are of tipping bucket type Most of the raingauge have a minimum detected quantity of 0 2 mm others have 0 1 mm The maximum rainrate that can be measured by the gauges ranges between 300 and 500 mm over one minute depending on the manufac
229. recipitation area Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 124 183 Product Validation Report PVR 01 Je usar It is also interesting to stress the German case study where ground data of different sources rain gauge and radar have been used for validation exercise over the same region The statistical scores obtained by rain gauge and radar data validation have very similar values 6 Validation results long statistic analysis 6 1 Introduction In this Chapter the validation results of the HO1 long statistic analysis are reported for the period 1 12 2009 31 11 2010 The validation has been performed on the product release currently in force at the time of writing Each Country Team contributes to this Chapter by providing the monthly contingency tables and the statistical scores The results are showed for radar and rain gauge land and coast area in the three precipitation classes defined in fig 11 of Chapter 3 The rain rates lower than 0 25 mm h have been considered as no rain The precipitation ground networks instruments and data used for the validation of H01 have been described in Chapter 4 To assess the degree of compliance of the product with user requirements all the PPVG members provided the long statistic results following the validation methodology reported in Chapter 3 For product H01 the User requirements are recorded in
230. richt Juni 2010 Bayrisches Landesamt f r Umwelt Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 95 183 Product Validation Report PVR 01 e ae Figure 66 Hourly precipitation sum mm for H01 satellite data crosses time stamp 2010 06 03 07 17 UTC and for RADOLAN RW left filled raster 2010 06 03 07 50 UTC and station data right dots 2010 06 03 08 00 UTC Data used PR OBS1 data for Bavaria in the given period were available for 5 43 UTC 6 14 UTC and 7 17 UTC Only these data are analysed in this case study Statistical score evaluted A first look to the results shows that rain rates detected by satellite product are in the same area of Germany as those indicated by the ground data In next two tables the result of the categorical statistic of the validation with both radar and rain gauge data are listed The results for validation with radar data for 3 June are better than for the whole month June Probability Of Detection of precipitation RR gt 0 25 mmh was 0 74 with less False Alarm Rate of 0 39 and Critical Success Index is 0 5 compared with other periods quietly good Since there was not detected hourly precipitation data in both radar and PR OBS1 this class has no amounts and for rain gauge we have got false alarm rate of 100 0 38 0 00 0 61 1 00 0 24 0 00 dy 3rd June
231. rks in PPVG Figure 15 Radar scan procedure Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 8 183 HSAF Product Validation Report PVR 01 Figure 16 Coverage of Europe by the INCA and RADOLAN systems 43 Figure 17 Procedure of the RADOLAN online adjustment hourly precipitation amount on 7 August 2004 13 50 UTC 45 Figure 18 Precipitation intensity field from 15 August 2010 15 00 UTC obtained by a radars b interpolated raingauge data c INCA analysis and d PR OBS 1 product Figure 19 Meteorological radar in Belgium Figure 20 Distribution of the raingauge stations of Iskar River Basin Figure 21 Distribution of the raingauge stations of Chepelarska River Basin Figure 22 Distribution of the raingauge stations of Varbica River Basin Figure 23 Network of rain gauges in Germany Figure 24 Pluvio with Remote Monitoring Module Figure 25 Left radar compound in Germany March 2011 Right location of ombrometers for online calibration in RADOLAN squares hourly data provision about 500 circles event based hourly data provision about 800 stations Figure 26 Flowchart of online calibration RADOLAN DWD 2004 Figure 27 location and coverage of the three Hungarian radars Figure 28 Correlation between rainrates detected by two close by stations as function of the distance between the two stations Colors refer to the month along 2009 Figure 29 Di
232. round validation of any satellite based precipitation products In the last months working groups Annex 1 2 3 4 5 and 7 have been composed in order to provide a complete information on the ground data characteristics and to evaluate the associated errors In this chapter a complete analysis of the ground data available in the PPVG is reported by the rain gauge and radar data in PPVG summaries Section 4 2 and 4 3 the Rain gauge and radar data integrated products in PPVG first report Section 4 4 and a country by country ground data description Section 4 5 4 13 The chapter has the object to provide ground data information and to highlight their error sources 4 2 Rain Gauge in PPVG In this section the complete inventory of the raingauges used in the PPVG with some considerations are reported as first results of the Rain gauge WG Annex 2 4 2 1 The networks The validation work carried on with raingauges uses about 3500 instruments across the 6 Countries Belgium Bulgaria Germany Italy Poland and Turkey as usual irregularly distributed over ground A key characteristic of such networks is the distance between each raingauge and the closest one averaged over all the instruments considered in the network and it is a measure of the raingauge density Instruments number and density are summarized in Table 7 Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 33 183
233. ructure of the satellite IFOVs Two basic ideas are pursued pixel by pixel matching or ground measure averaging inside satellite IFOV The second approach seems to be more convenient especially when the large IFOV of H01 and H02 are concerned Probably it is mandatory for H02 also take into account that the size of the IFOV changes across the track and could become very large The first approach e g nearest neighbour can be more effective for H03 and HOS products Country Type of interpolation Quality control Y N Belgium Barnes over 5x5 km grid Y Bulgaria Co kriging Y Germany Inverse square distance Y Italy Barnes over 5x5 km grid Y Poland No Y except cold months Turkey No Y Table 50 Data pre processing strategies H01 H02 Country Spatial matching Temporal matching Spatial matching Temporal matching Belgium N A N A N A N A Bulgaria N A N A N A N A Germany matching gauges are each overpass is matching gauges are each overpass is searched on a radius compared to the searched on a radius compared to the of 2 5 km from the next hourly rain of 2 5 km from the next hourly rain IFOV centre amount IFOV centre amount Italy mean gauges value each overpass is Gaussian weighted each overpass is over 15x15 km area compared to the mean gauges value compared to the centred on satellite next hourly rain centred on satellite next hourly ra
234. rue _ LS true Gat sat 2 ue True 53 Range 1 to 1 Perfect score 1 CC Root Mean Square Error RMSE N RMSE aD Gat true 3 Range 0 to Perfect score 0 kat Root Mean saae Error percent RMSE used for precipitation since error grows with rate RMSE a antru 100 Range 0 to Perfect score 0 ue k The statistical scores evaluated in PPVG for multi categorical statistic are derived by the following contingency table Contingency Table ground yes no total hits false alarms forecast yes satellite no misses correct negatives forecast no observed yes observed no total where hit event observed from the satellite and also observed from the ground miss event not observed from the satellite but observed from the ground false alarm event observed from the satellite but not observed from the ground correct negative event not observed from the satellite and also not observed from the ground The scores evaluated from the contingency table are Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 31 183 HSAF Product Validation Report PVR 01 Probability Of Detection POD hits hits POD gt hits misses observed yes Range 0 to 1 Perfect score 1 False Alarm Rate FAR ___falsealarms _ falsealarms hits falsealarm
235. ry conclusions can be deducted that HO1 product in February 2010 could provide false detections of snow cover especially in flat non forested areas over the clear atmosphere as a precipitation field with rain intensity of 10mm h Very similar phenomena can be observed over the Gulf of Bothnia in the second half of February The Slovakian team confirmed certain outages of radar measurements in the period from February 19 to 28 so bad statistical scores from first half of February could not be fully suppressed in the validation results for this period 6 2 2 The spring period PR OBS 1 DE st T i spring spring 2010 spring 2010 spring 2010 spring spring spring spring spring spring spring 2010 2010 2010 2010 2010 2010 2010 3010 Version 1 4 radar radar radar radar radar gauge gauge gauge gauge gauge gauge NS lt immin 4950 36210 8667 5224 55051 13988 506 17327 41407 41407 NS NR 3793 29711 12905 3890 50299 24207 2165 12434 49208 49208 NR F Product Validation Report PVR 01 Product H01 PR OBS 1 Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Date 30 09 2011 Page 127 183 ME 0 19 0 23 0 16 0 91 0 04 018 013 0 10 0 15 0 15 ME sD 0 20 0 84 1 37 2 06 1 02 099 074 1 10 1 02 102
236. s forecast yes Range 0 to 1 Perfect score 0 Critical Success Index CSI hits CSI ___________ hits misses falsealarm Range 0 to 1 Perfect score 1 Equitable Threat Score ETS a hits hitS andom l with hits hits misses falsealarm hitS andom ETS ranges from 1 3 to 1 O indicates no skill Perfect score 1 observed yes forecast yes total ETS random Frequency Blas FBI FBI hits falsealarms forecast yes Range 0 to Perfect score 1 hits misses observed yes Probability Of False Detection POFD POFD falsealarms _ falsealarms k Range 0 to 1 Perfect score 0 correctnegatives falsealarms observed no Fraction correct Accuracy ACC hits correct negatives total ACC Range 0 to 1 Perfect score 1 Heidke skill score HSS oe hits correct negatives expected Correct andom N expected correct ndom with expected correct andom i Jobserved yes forecast yes forecast no observed no _ Range 2e to 1 0 indicates no skill Perfect score 1 Dry to Wet Ratio DWR false alarm correct negative _ observed no hits misses observed yes DWR Range 0 to oe Perfect score n a Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 32 183 HSAF Product Validation Report PVR 01 3 9 Case study analysis Eac
237. s is very good as confirmed by results of the correlation coefficient and dichotomous scores 5 8 Case study analysis in Turkey ITU It is here reported the case study analysis of 20 of October 2010 on Turkish territory performed by ITU 5 8 1 Case study 20 of October 2010 Description As it can be seen from next two figures Turkey is in low pressure area and there are respectively warm and stationary fronts rain bands and precipitation in western part of Turkey on October 20 at 06 00 GMT and 12 00 GMT http www wetters deffax Figure 96 Atmospheric condition 20 10 2010 06 00 GMT Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 121 183 Product Validation Report PVR 01 h tp howw wetter3 delfax Figure 97 Atmospheric condition 20 10 2010 12 00 GMT Data used In this case study 193 rain gauges which have specifications as explained in section 4 12 in western part of Turkey has been used HO1 product on October 20 at 04 57 GMT has been compared with gauge observations Moreover synoptic cards from UK MetOffice have been taken for understanding the meteorological situation Comparison Next two figures show comparison of H01 product and rain gauge Values of HO1 product are between 0 25 to 14 00 mm h but they vary from 0 25 to 7 00 mm h for gauge Main patterns of product and gauge are similar in next figure Do
238. satellite precipitation products to analyse the application of these techniques to the rain gauge available inside the PPV to produce a well referenced documentation on the methodology defined to develop the code to be used in the PPVG for a correct verification of satellite precipitation product performances Activiti First step collect e characteristics telemetric spatial distribution temporal resolution quality check applied instrument sensitivity and saturation value and accuracy of the rain gauge networks which composes the PPVG Belgium Bulgaria Germany Poland Italy Turkey Start Time End time December 2010 January 2011 First Report 31 of January 2011 Second step define on the base of published papers and the characteristics of the rain gauge data available inside the PPVG e ground data quality check to be applied to all rain gauge data e optimal spatial density of rain gauge stations to be representative of the ground precipitation in the view of satellite product comparison e optimal time resolution of rain gauge network 15 min 30 min 1 h for a correct comparison with rain rate and cumulated precipitation satellite products e raingauges up scaling techniques vs AMSU B SSMI SEVIRI grids e technique for the temporal matching of rain rate and cumulated precipitation satellite products with rain gauge data Start Time End time January 2011 July 2011 Second Report 31
239. sian filter If the Radar resolution is 1km 1px 1km If the matrix NxN is too large it is reduced to a MxK matrix until the pixels 1 C C 1 N C C N are less than C C 100 Figure 10 ha 0 10 CIL JaN 24 0 25 tw T1 Co CN 023 025 25 025 023 ws 025 CR a nol a NN 0 24 Figure 10 Left Original Gaussian matrix Right Reduced matrix to dimensions M xK Doc No SAF HSAF PVR 01 1 1 HSAF Product Validation Report PVR 01 Issue Revision Index 1 1 i Product H01 PR OBS 1 Date 30 09 2011 Page 27 183 normalize gaussian matrix G by now MxK obtaining G matrix which element sum would be G 1 G mip _ Sime K X G m k m 1k 1 b Smoothing of radar precipitation for each FOV and for each SCANLINE in the file HO1 superimpose gaussian filter G on radar data in such a way that the central pixel C C corresponds to HO1jat HO1jon and the y axis has the same direction of the scanline multiply each element of G by the nearest radar precipitation estimation RRyign lat lon and sum the products M K RRos SG mk RR m 1k 1 Following this procedure it is obtained for each FOV and SCANLINE a value RRiow RRiow FOV SCANLINE which represents the matrix of validation used versus SSMI estimates This scheme has been suggested by the preci
240. ska hola the Doppler filtering is supplemented with moving target identification MTI technique Isolated radar reflectivity and Doppler velocity bins are removed by the Speckle removal filter The data with signal to noise ratio below the specified threshold are also eliminated The measured radar reflectivity is corrected for atmospheric clear air attenuation of the radar beam Neither beam blocking correction nor vertical profile of reflectivity VPR is applied at SHMU However implementation of the beam blocking correction is being considered for the H SAF validation due to complicated orographical conditions in Slovakia Precipitation intensity is derived from radar reflectivity according to the Marshall Palmer equation Z a R4b with constant coefficients valid for stratiform rain a 200 b 1 6 Polarimetric techniques for quantitative precipitation estimation in case of dual polarization radar at Kojsovska hola are not used because the measured polarimetric data are not operational calibration would be required Software filter for the RLAN interference detected by radars is currently in development at SHMU Radar composite based on CAPPI 2 km products from both radars is used for the H SAF validation The composition algorithm used selects the higher value measured by the two radars in the overlapping area No raingauge correction of the derived instantaneous precipitation is applied Effect of elevating radar beam with increasing range a
241. smoothing and the pre adjustment Further improvements of these procedures are being developed Precipitation distribution of the Precipitation distribution of the RADOLAN precipitation rain gauge point areal original radar product measurements measurements Figure 17 Procedure of the RADOLAN online adjustment hourly precipitation amount on 7 August 2004 13 50 UTC In order to collect more detailed information about both types of systems a questionnaire was elaborated and completed by Slovakia Poland and Germany The questionnaire provided details such as geographical coverage input data inventory or availability of different instantaneous and cumulated precipitation products The final version of the questionnaire is shown in next table and is also available as annex 5 Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 z Product H01 PR OBS 1 Date 30 09 2011 Page 46 183 H SAF Product Validation Report PVR 01 Group of information Item GERMANY POLAND SLOVAKIA domain1 SLOVAKIA domain2 avatabity or documentation tor NCA or ipossbile please anach inkor Dokumentation received Documentation avaabie Documentation avatabie Documentation shouldbe similar German system Yes No documentation hg meeting from ZAMG from ZAMG Issued in future Beinion or geographical area coveredby Petition of geographies aes covered by o
242. son it appears that for this case study the HO1 product could reproduce the shape of rainfall patterns but failed in the quantitative reproduction of the actual rainfall amount 5 2 3 Case study 12 15 of November 2010 h01 Description of the event A wide area with low pressure extended from Scandinavia to Great Britain and made a very active precipitating perturbation stay over the country during several days Fig 10 and result in high flows and even flooding Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 86 183 Product Validation Report PVR 01 x 178 ne Figure 54 Surface map on 13 November 2010 at 06 UTC MSLP and synoptic observations Satellite and ground data used Products H01 from November 12 at 0 00 UTC to November 15 at 18 00 UTC have been considered The total is 18 satellite passages distributed as follows 3 in the morning of November 12 3 in the afternoon of November 12 3 in the morning of November 13 Lin the afternoon of November 13 2 inthe early morning of November 14 5 inthe morning of November 15 Lin the afternoon of November 15 The ground data used for validation are the Wideumont radar instantaneous measurements without rain gauge adjustment Radar data are available within 5 minutes around the satellite passage Comparison Here are two examples of H01 files compare
243. statistical scores evaluated by the PPVG reach the thresholds stated in the User Requirements in all cases using rain gauge data as ground reference and in all cases except for precipitation lower than 1 mm h using radar data as ground reference table 6 9 This result might be explained by considering the highest relative error for radar measurements at rain rate values lower than 1 mm h around 150 following annex8 As reported in Annex 8 the results obtained by the current validation procedure represent the convolution of at least three factors the satellite product accuracy the accuracy of the ground data used and the limitations of the comparison methodology e g errors of space and time co location representativeness changing with scale etc Therefore the results currently found are by far pessimistic in respect of what is the real product performance 7 2 Next steps On the base of the development phase it is possible to say that the ground data error characterization is necessary and that a validation of a common protocol is not enough Only the use of the same software can guarantee that the results obtained by several institutes are obtained in the same way To improve the validation methodology and to develop software used by all members of the validation cluster several working groups have been composed during the last Validation Workshop held in Bratislava 20 22 October 2010 see annex 1 2 3 4 5 6 7 On the base of published
244. stribution of the raingauge stations of the Italian network collected by DPC Figure 30 Italian radar network coverage Figure 31 Graphical mosaic of reflectivity CAPPI at 2000 m for the event of 04 18 08 at 0015 U T C 60 Figure 32 Architecture of the Italian radar network Figure 33 Schematic representation of radar data processing cha Figure 34 Measured upper panel and attenuation corrected lower panel PPI 1 0 deg of reflectivity observed on 09 14 08 at 0500 U T C by the polarimetric radar operated by Piemonte and Liguria regions Figure 35 Hydrometeor classes as detected by the classification algorithm starting from the radar variables observed on 09 14 08 at 0500 U T C by the polarimetric radar operated by Piemonte and Liguria regions 64 Figure 36 Measured upper panel and VPR corrected lower panel PPI of reflectivity observed on 03 25 07 at 0930 U T C by the polarimetric radar located in Gattatico Emilia Romagna Italy 66 Figure 37 Cumulated radar rainfall estimates versus gage measurements for the event observed on 06 01 2006 by the dualpolarized radar located in Settepani Liguria Italy Figure 38 ATS national network in Poland Figure 39 Map of SHMU rain gauge stations green automatic 98 blue climatological 586 red hydrological stations in H SAF selected test basins 37 Figure 40 Map of SHMU radar network the rings represent maximum operational range 240 km for radar at Maly Javornik left 200 km for
245. sults emphasizing the errors distribution obtained for test catchments e analysis of the possible solutions for operational creation of the error maps and selection of the best one Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 174 183 Product Validation Report PVR 01 Je usar e creation the software if necessary Start Time End time March 2011 November 2011 Second Report 30 of November 2011 Coordinator Bozena Lapeta IMGW Poland Members Ibrahim Somnez ITU Turkey H SAF project Validation Programme WP 6100 Working Group 5 Geographical maps distribution of error Bozena Lapeta IMGW Poland First report March 31 2011 Introduction The Working Group 5 aims at creating geographical maps of H SAF products error and analyzing its usefulness for H SAF validation The idea of this work stemmed from hydrological validation community that is interested in distribution of the error over the catchments In this report the results obtained during the first step of WG5 activities aiming at selection of the best method for mean error specialization are presented Selection of spatialisation algorithm first results The most important issue in creating geographical distribution of any parameter is the algorithm for spatial interpolation As there is no universal spatial interpolation method that can be applied for any
246. sured by the buckets resulting in a temporal shift of the measurements since the melting and the measure can take place several hours or days depending on the environmental conditions after the precipitation event Leitinger et al 2010 Sugiura et al 2003 This error can be mitigated by an heating system that melts the particles as soon as are collected by the funnel All these errors can be mitigated and reduced but in general not eliminated by a careful maintenance of the instrument A number of a posteriori correction strategies have been developed in order to correct preci data measured by raingauges but mainly apply at longer accumulation intervals daily to monthly Wagner 2009 Country Minimum detectable Maximum detectable Heating system cumulation rainrate rainrate mm h Y N interval min Belgium 0 1mm N A N 60 Bulgaria 0 1mm 2000 k 120 1440 Germany 0 05 mm h 3000 Y 60 Italy 0 2mm N A Y N 60 Poland 0 1mm N A Y 10 Turkey 0 2 mm 720 Y af Table 48 Summary of the raingauge characteristics only 300 out of 1800 gauges are heated information not available at the moment a value about 300 mmh can be assumed for tipping bucket raingauges Most of these shortcomings could be avoided by using instruments based on different principle or mechanisms The German network and a part of the Bulgarian network as an example are equipped by precipitation weigh
247. t to reduce bull eyes effect ii Climatological scaling of radar data by means of monthly precipitation totals of raingauge to radar ratio partial elimination of the range dependance and orographical shielding iii Re scaling of radar data using the latest rain gauge observations iv Final combination of re scaled radar and interpolated rain gauge data v Elevation dependence and orographic seeding precipitation In the final precipitation field the raingauge observations are reproduced at the raingauge station locations within the limits of resolution Between the stations the weight of the radar information becomes larger the better the radar captures the precipitation climatologically Important factor affecting the final precipitation analysis is accuracy and reliability of the raingauge stations In order to eliminate the influence of raingauge stations providing evidently erroneous data the SHM is developing the blacklisting technique which temporarily excludes such stations from the analysis Currently the stations can be put into the blacklist only manually but development of the automated blacklisting is expected in near future 4 4 2 RADOLAN system RADOLAN is a routine method for the online adjustment of radar precipitation data by means of automatic surface precipitation stations ombrometers which has started on a project base at DWD in Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS
248. ted raingauge data c INCA analysis and d PR OBS 3 product 5 57 UTC supplemented with map of minimum visible height above surface level of the SHMU radar network e 164 Figure 115 As in previous figure except for 8 00 UTC Figure 116 Comparison of selected statistical scores for the PR OBS 2 product obtained by different ground reference data valid for event 1 convective Figure 117 As in previous figure except for event 4 stratiform Figure 119 The Wideumont radar image of 1 2 2010 cumulated rainfall n the previous 24 hours raingauge adjusted and upscaled to the COSMO grid Figure 118 the Wideumont radar image of 1 2 2010 cumulated rainfall in the previous 24 hours raingauge adjusted Figure 120 Distribution of the monthly average H 05 3 h cumulated precipitation Mean Error calculated for July 2010 using three methods a Ordinary Kriging b Natural Neighbour and c IDW 2 Figure 121 Cross validation results obtained for three different methods for spatial interpolation 176 Figure 122 Example of sampled data for a regular grid In right on the upper part a detail of field studied below the original grid of field for step 2 Figure 123 Randomly distribution of perfect measurement to remap the field on a regular grid Figure 124 STD vs RMSE for interpolations by step 2 Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 201
249. tellite IFOV centres Other cases of medium large size convective cells have showed a general correct qualitative location and estimation of the precipitation by H01 in particular for the rain rate in the range between 1 and 10 mm h The highest precipitation class rain rate greater than 10 mm h has often been slightly underestimated by the satellite product The dichotomous statistical scores evaluated for the summer cases have the following mean values POD 0 90 FAR 0 50 and CSI 0 40 So the case study analysis has pointed a high capacity of the satellite product to detect precipitation POD 0 90 but also a tendency of the product to detected falsely a lot of light precipitation FAR gt 0 5 During spring period different dichotomous statistical scores have been obtained with lower values in particular for POD During winter period when more stratiform events occur the HO1 product did not apart from few cases reproduce correctly the rainfall patterns and amounts The satellite product misses or strongly underestimates the rainfall In general for these events the FAR has an higher value than POD and the CSI is average 0 20 Some general satellite product characteristics have been highlighted by the case studies here proposed as problems on coast line and parallax shift It has been showed a case study Poland where the ground data have been unable to catch the precipitation system while the satellite product reproduced more correctly the p
250. th Persistence Test Persistence test is used to determine if any group of observations are due to instrument failures The test procedure applied is defined as IF T lt A THEN Flag for all Obser in T Good IF T gt A THEN Flag for all Obser in T Suspect where T is the total number of the sequentially repeating observations forward in time and A is the possible maximum number of sequentially repeating observations As in the other two tests A threshold is determined for each site on a monthly basis For any site the data belonging to the same month is taken into account to determine the repeating number of the sequential observations Then 99 9 cumulative histogram value of the repeating number dataset is assigned as the A amount for the corresponding site and month Since there is a high possibility of no precipitation data zero the sequential zero observations are excluded in this test during the determination of the A threshold amount and application of the test QA Test procedure By applying the control procedures of the QA test mentioned above each individual precipitation observation receives three flags referring to the corresponding test For the corresponding observation if all the test flag is not Good then the observation is excluded from the validation process Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 75 183 Product Validation
251. the validation for whole month August 2010 In comparison with categorical statistic of the whole August 2010 we got better results for POD in all classes for both kinds of ground data than for the 7 August Mostly for validation with rain gauge data we have less POD and less FAR than for validation with RADOLAN data caused on less valid data pairs The critical success index CSI is more stable and differs only by 1 2 between the different validation methods A CSI of 0 33 0 35 means that 33 resp 35 of the predictions H01 of precipitation gt 0 25 mmh of all predicted observed rain events are correct Figure 16 and 17 show the contingency table of four precipitation classes By validation with radar data in the lower three classes over 50 of the H01 data are in the same class in both periods August 7 and for the whole month August For the validation with rain gauge data it was only for the lowest two periods on 7 August Over the whole month of August we ve got worse results for second class but better results for class 1 mmh lt RR lt 10 mmh The validation for whole month August shows analogue results with both kinds of ground data MRRsatin class1 MIRRsatincdass2 MH RRsatinclass3 W RRsat in class 4 00 00 0 s w w n 70 0 s Za Zs o o x0 20 20 0 0 o o class 1 class 2 clase 3 class 4 elass 1 class 2 class 3 class 4 Oe AR lt 0 25 O 25 lt RAct t lt RR lt 10 AR gt 10 Oe RR lt 025 025 lt RA
252. ting gauges that allow continuous precipitation both solid and liquid measurements with higher accuracy Other option could be the use of disdrometers that give more information about the precipitation structure and a more accurate rainrate measure In table 1 relevant characteristics of the raingauges used in the different countries are reported 2 The networks The validation work carried on with raingauges uses about 3000 instruments across the 6 Countries as usual irregularly distributed over the ground A key characteristics of such networks is the distance between each raingauge and the closest one averaged over all the instruments considered in the Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 139 183 HSAF Product Validation Report PVR 01 network and it is a measure of the raingauge density Instruments number and density are summarized in table 2 The gauges density ranges between 7 for Bulgaria where only 3 river basins are considered to 27 km for Turkey These numbers should be compared with the decorrelation distance for precipitation patterns at mid latitude Usually the decorrelation distance is defined as the minimum distance between two measures to get the correlation coefficient Pearson Coefficient reduced to e A recent study on the H SAF hourly data for Italy shows this decorrelation distance varies from about 10 km in w
253. tion is present along the Apennines chain Some of these small systems are expected to grow in the following hours Prod Validation R PVR 01 Doc No SAF HSAF PVR 01 1 1 E H SAF roduct Validation Report n Issue Revision Index 1 1 x Product H01 PR OBS 1 Date 30 09 2011 Page 107 183 Data used Reference data Italian hourly raingauges network provided by DPC Ancillary data used for case analysis SEVIRI images courtesy of University of Dundee NEODAAS Weather charts courtesy of Wetterzentrale Comparison This event is dominate by convective systems over a wide range of spatial scales from the nearly Mesoscale Convectiv System over Central Italy to few SEVIRI pixels sized cells scattered over southern Italy HO1 is expected to perform at its best over the large deep systems and to suffer of rainrate rainarea underestimation in case of sub pixel structures The average rainrate during this event is about 2 3 mmh while the highest peak measured by raingauges is of about 17 mmh negligibly reduced by IFOV averaging while h01 shows a peak value around 20 mmh Statistical scores Statistical indicators show acceptable values for POD 0 71 and FAR 0 46 with a corresponding ETS of 0 32 indicating good capabilities of h01 to detect this kind of summertime convective and intense precipitation The multi category HSS is 0 36 confirming that h01 is able to at least partially catch the s
254. to that low humid and cold air masses of North Atlantic origin covered western part of Poland At the same time eastern part of the country was influenced by very hot and humid air from Mediterranean region Total cloud cover connected with intensive and prolonged precipitation on lowlands brought snow in mountains on South of Poland In the Eastern part of the country storms were reported The highest diurnal sums of precipitation reached 186mm in the South of Poland Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 113 183 HSAF Product Validation Report PVR 01 We i PANE atures Mapa smoiv rma Figure 88 Synoptic chart at 0000 UTC on 17th of May 2010 Source IMWM Data and products used Reference data data from Polish automatic rain gauges network IMWM H SAF product PR OBS 1 Ancillary data used for case analysis Polish Lighting Detection System PERUN IMWM Weather charts IMWM Comparison This event is dominate by stratiform precipitation connected with low pressure system of very large spatial scales resulted in floods in South Poland The average rain rate measured by rain gauges during this event is about 1 9 mm h while the PR OBS 1 averaged rain rate is of 0 6 mm h The highest peak measured by rain gauges is of about 8 4 mm h negligibly reduced by IFOV averaging while PR OBS 1 shows a peak value of 4 9 mm h On t
255. tructure of such precipitating systems In the figures are the results of the 15 42 UTC overpass right and the raingauges hourly cumulated precipitation at 16 00 UTC left with different colour scales Most of the small scale precipitation structures are detected by h01 however the satellite algorithm estimates very light precipitation around 1 mm h in few areas where raingauges don t measure rainfall especially on the surrounding of the main cells Underestimation seems greatly reduced and apparently no coast effects are present A quite large spot is overestimated close the Gulf of Taranto where rainrates of about 8 mm h are estimated but no rain was detected by gauges Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 108 183 Product Validation Report PVR 01 DMSP16 06 07 2010 14 09 rrsursol Figure 82 Raingauges hourly precipitation cumulated at 16 00 UTC right of 06 July 2010 Please note different colour scales Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 i Product H01 PR OBS 1 Date 30 09 2011 Page 109 183 HSAF Product Validation Report PVR 01 Some conclusions This case study with severe convective developments is generally well described by the H01 algorithm in terms of both areal matching and quantitative estimate even if some spurious scattered raining IFOVs are present in so
256. ts with rectangular geometry For any grid point Awos sites within the 45 km for the time period of April September convective type and 125km for the rest stratiform type are taken into consideration At each grid point the precipitation amount is estimated by n WO mn Zi i l n XWin i 1 where Zp is the estimated value and W r m is the spatially varying weighting function between the i th site and the grid point m Zm 4 13 1 Awos sites FOV center Figure 44 Meshed structure of the sample H01 and H02 products footprint Determination of the W rim weighting function in Equation 1 is crucial In open literature various approaches are proposed for determining this function For instance Thiebaux and Pedder 1987 suggested weightings in general as a fi lt R Weg m or Nm 4 13 2 for tjm 2R where R is the radius of influence rim is the distance from point i to point m to the point and q is a power parameter that reflects the curvature of the weighting function Another form of geometrical weighting function was proposed by Barnes 1964 as a Win e 4 13 3 Unfortunately none of these functions are observation dependent but suggested on the basis of the logical and geometrical conceptualizations only They are based only on the configuration i e geometry of the measurement stations and do not take into consideration the natural variability of the meteorological phenomenon concerned In addit
257. turer The rainrate is measured over different cumulation intervals by the different local administrations managing the network but the data disseminated are all integrated over 60 minutes At the moment the National network made available by DPC provides only hourly data Shorter cumulation times could be available for case studies after specific agreements with local management authorities Only a small subset about 300 stations of gauges have heated funnel especially in alpine regions such as Valle d Aosta and Piedmont and this is a clear source of errors in both summer due to hailfall and in autumn winter due to snowfall The data processing No quality control is performed on the data right now In this Project the point like gauges data are interpolated by using the Barnes method Barnes 1964 Koch et al 1983 widely used to interpolate station data It works by defining a regular output grid 5x5 km in our case and a radius of influence of each station in our case it was 10 km The point information from a raingauge is spread in the neighbour by an exponential function limited by the influence radius and the rainfall value for each grid point is computed as the contribution of all the closest measurements Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 59 183 HSAF Product Validation Report PVR 01 The resulting
258. type considering the weighting of each site with respect to the grid point Equation 1 The weighting amounts are derived from the spatially varying weighting functions obtained by using the semi variogram approach Sen and Habib 1998 Finally the Gaussian filter is applied to the estimations at the mesh grid of the IFOV area to get the average rainrate Then this amount is compared with the satellite precipitation product amount for the validation purposes 4 13 Conclusions After these inventories some conclusions can be drawn The rain gauge in PPVG is composed by 3500 instruments across the 6 Countries Belgium Bulgaria Germany Italy Poland and Turkey These data are as usual irregularly distributed over ground and are generally deduced by tipping bucket type instruments Moreover most of the measurements are hourly cumulated So probably the raingauge networks used in this validation activities are surely appropriated for the validation of cumulated products 1 hour and higher while for the validation of instantaneous estimates the use of hourly cumulated ground measurements could introduce a large error Moreover the revisiting time 3 4 hours of HO1 makes impossible or not reasonable to validate the product for 1 24 hours cumulated interval The first object of PPVG Rain Gauge WG in the next future it will be to quantitatively estimate the errors introduced in the validation procedure comparing the instantaneous satellite precipitat
259. ucing numerical files called CS and MC files e all the institutes evaluate PDF producing numerical files called DIST files and plots e the precipitation product validation leader collects all the validation files MC CS and DIST files verifies the consistency of the results and evaluates the monthly and seasonal common statistical results Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 29 183 Product Validation Report PVR 01 HSAF GERMAN Y v Vv v v Vv Vv v v v h M v v Figure 11 Main steps of the validation procedure in the PPVG Statistical scores The statistical scores evaluated in PPVG for continuous statistics are Mean Error ME N ME 1d sat true Range to ee Perfect score 0 kat Mean Absolute Error MAE N MAE gt Isat true Range 0 to gt Perfect score 0 kat Standard Deviation SD N SD 4 Eat true ME Range 0 to Perfect score 0 Multiplicative Bias MBias Doc No SAF HSAF PVR 01 1 1 Issue Revision Index 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 30 183 HSAF Product Validation Report PVR 01 12 sat MB N ae Range ee to ee Perfect score 1 Aye Correlation Coefficient CC 3 dora a gt Gt sat Yue trae gt ag N N with sat y sat and j
260. us equation we can derive RMSEsat RMSD RMSEgroung RMSEcomparison where RMSD is provided by the validation activity RMSEground is provided by tab 64 using the University of Ferrara numbers At the moment for the RMSEcomparison is assumed equal to three values resulting from the ISAC study for validation w r t rain gauges and equal zero for validation w r t radar plus a 30 for the remapping procedure Conclusions 1 It is believed that the results of the validation activity cannot be substantially improved they are most probably consistent with the size of the error sources satellite ground stations and comparison method This needs to be confirmed by evaluating the size of the error associated with the limits of the comparison technique 2 It must be considered that the total RMSD is affected by other than satellite terms one of which RMSEground very difficult to be reduced and the other one RMSEcomparison possibly dominant and also very difficult to be reduced This tells us that the validation figures have a large component which is independent from the structure of the algorithm 3 However the case for continuing algorithm improvement is very strong Data are produced for being used and the better the quality the higher the impact The fact that the current validation methodology cannot completely evaluate the intrinsic error of satellite data is regrettable but should not prevent a better representation of t
261. uthern Italy For this convective case the performances are quite acceptable very likely the indicators could reach even higher values if parallax correction would be applied before matching Probably a final control scheme for isolated raining IFOVs could be implemented 5 6 Case study analysis in Poland IMWM 5 6 1 Case study 15 of August 2010 Description of the case study On the 15 of August 2010 significant cloud layer reaching over Lower Silesia region with its upper constituent belongs to developing low pressure centre That structure is a part of bigger low pressure centre over France and tends to move over Germany to Poland Stripe of clouds extending from Tunis central Italy Adriatic Sea to Austria is a cold front of Atlantic air which is going to reach Poland on Monday 16 of August when bay of low pressure over Germany moves over Poland Mentioned above bay of low pressure extends further over Balkans with significant wind convergence stimulating convection updrafts with large scale moves Moreover the forecast dated on 0000 UTC shows very turbulent night because of development of low pressure centre over Poland aa LS AAA WN Wed 2 JA http www wetterd de Figure 83 Synoptic chart at 1200 UTC on 15th of August 2010 Convective storms where observed over the country on that day The precipitation was accompanied by lightning activity In next figure the lightning activity map for half an hour time spam 1445 UTC 1515
262. work of 14 rain gauges in an area of 14 km x 14 km approximately the pixel of SSM I at 85 5 GHz and of AMSU B MHS generally used for the radar calibration Assuming Polar 55C as reference the spread of rain gauge measurements resulted as follows RR gt 10 mm h 50 1 lt RR lt 10 mm h 80 RR lt 1 mm h 150 A similar experiment with 2 rain gauges in reduced area of 5 km x 5 km approximately the pixel of SEVIRI at middle latitude shows similar results That s means that In order to obtain an estimate of upscaling downscaling and interpolation process theoretical experiment of some methodologies has been implemented Hypothetic perfect fields have been defined and a grid of perfect measurements has been defined The experiment assumed different typologies of precipitation field respect the variance of precipitation intensity in the field To obtain the field of perfect measurements some grid points from the precipitation field are been removed The experiment removed the perfect rain gauge long a regular grid to simulate an unreal distribution of non realistic rain gauges The sampling has been done at different grid spacing 2 3 and 4 time the perfect field to obtain new data at different spatial density Then the algorithm performances of up down scaling procedure to reproduce the original field are been evaluated The work has been implemented for 4 different algorithms Barnes Inverse Distance Squared IDS kriging a
263. x 1 1 Product H01 PR OBS 1 Date 30 09 2011 Page 10 183 HSAF Product Validation Report PVR 01 Figure 70 96h totals of precipitation Figure 71 Hourly precipitation sum mm for H01 satellite data crosses time stamp 2010 12 05 07 027 UTC and for RADOLAN RW left filled raster 2010 12 05 07 50 UTC and station data right dots 2010 12 05 08 00 UTC Figure 72 Hourly precipitation sum mm for H01 satellite data crosses time stamp 2010 12 06 06 49 UTC and for RADOLAN RW left filled raster 2010 12 06 06 50 UTC and station data right dots 2010 12 06 07 00 UTC Figure 73 Contingency table statistic of Rain Rate mmh 1 for PR OBS1 vs radar data Figure 74 Contingency table statistic of rain rate mmh 1 for PR OBS1 vs rain gauge data Figure 75 Synoptic chart at 00 UTC on 5 May 2010 Figure 76 Precipitation rate from the Hungarian radar network at its original resolution upper right panel HO1 product upper left panel operational png lower left panel SAFNWC Cloud Type CT product lower right 103 Figure 77 Synoptic chart at 00 UTC on 18th of July 2010 104 Figure 78 H01 product left panel Cloud type from NWC SAF right panel Precipitation rate from the H ungarian radar network at its original resolution in middle 104 Figure 79 Synoptic chart at 00 UTC on 10th of September 2010 105 Figure 80 Precipitation rate from the Hungarian radar network at its original resolution at 6 45 UTC right

Product User Manual - H-SAF

Contents

Download Pdf Manuals

Related Search

Related Contents