Home

A user's guide for identifying rice paddocks using GIS and remote

1. amp CROP Type Field Selection Choose Field used for validation Props petrearerteereetee A new window should reveal where the output excel file is located In this file is the threshold information and the estimation of accuracy OUTPUT FILE i Output file is c vanniel tmp ace1 2 csv The output from running this program should look like this Name NumNRasR NumRasNR TotalCount Threshold OverallAccuracy KappaAccuracy Nov_bnd5 2 1 63 75 5838 0 952381 0 903226 Validation Paddocks file c vanniel tmp rueben_rice_2002_2003 shapes 02_03_validation shp Rice Training Set file C VanNiel tmp Rueben_rice_2002_2003 shapes 02_03_r_trn shp CSIRO Land and Water 17 Non rice Training Set file C VanNiel tmp Rueben_rice_2002_2003 shapes 02_03_nr_trn shp Common Validation Field Crop2003 Time Mon Aug 04 13 04 15 2003 e The NumNRaskR reveals the number of NonRice paddocks that were classified as Rice errors of commission the NumRasNR shows how many Rice paddocks were classified as NR errors of omission the Threshold will be used later to classify the rice and the Overall Accuracy as a proportion is the estimate of the accuracy based on the validation paddocks The Kappa accuracy is another accuracy statistic useful when the sample sizes are very uneven CSIRO Land and Water 18 5 Updating Paddock Boundaries 5 1 Introduction The methods outlined for identification of rice depend on placing the remot
2. BEKRA 7 40504007 pees ERES etal F 02_02padds shp A Water_class shp _ 02_03_padds_to_e f 02_02_validstion s 02_03_1_tn she _ 02_03_nr_trn shp Nov_bnd5 227 255 No Data _ Gpse6 shp N A Nov_bnd5bsa CSIRO Land and Water 22 e However if the match is not good e g due to a change like adding more rice bays in this year when compared to previous years see below then these boundaries need to be UPDATED In this example a previous year s air photos might be used to extend the boundary as a fundamental change to the boundary has probably not occurred For other circumstances the updating of the boundary will require a GPS unit in the field e Every area of classified water will need to be checked against paddock boundaries and similar decision will need to be made 2 ArcView GIS 3 2 DER File Edit View Theme Analysis Surface Graphics Window Help 0 ee AN Bawa So E w eS 644640018 Ki Water_class shp _ 02_03_padds_to_c _ 92_03_validstion s _ 02_03_r_trn shp _ 02_03_nr_trn shp e 2 Nov_bnd5 bsq CSIRO Land and Water 23 6 Identifying Summer Rice Paddocks Optional Step The procedures in this chapter are optional For the 2003 04 season all rice paddocks were classified by looking at the map of standing water generated by the steps in the last chapter and using it to select p
3. NDVI Band4 Band3 Band4 Band3 The NDVI is a normalised index which ranges between 1 and 1 and is directly related to greenness This equation can be calculated in the Analysis Map Calculator see pic below using the following equation and replacing the red variables with the correct names based on your own data The float term maintains the floating point precision in the calculation otherwise everything is converted to either 1 s or 0 s by default The 1000 term is a scalar scales the values of the NDVI between 1000 and 1000 instead of between 1 and 1 and the int term converts this result to integer data The resulting data is integer data with a precision of 3 decimal places because the data is scaled between 1000 and 1000 That is a value of 459 equals a real NDVI value of 0 459 Oct_bnd4 float Oct_bnd3 Oct_bnd4 Oct_bnd3 1000 int Map Calculation 1 Seles Logarithms x Layers 3 d E Exp Log LJ Exp10 Log10 Oct_bnd4 float Oct_bnd3 0ct_bnd4 Oct_bnd3 1000 int a Evaluate Using the Theme Convert to Grid command you can convert this grid to your local hard disk with a relevant name like octndvi CSIRO Land and Water 29 2 ArcView GIS 3 2 File Edit View Theme Analysis Surface Graphics Window Help ike KASENG a e 3 Scale fiean E No Dsts Map Calculation 1 i No Dsta Oct_bnd4 H
4. summarise training sets get mean and std for NON RICE training set and for nr training set Summarise Non Rice Run ZonalStats on Index Grid for non rice training set nr_ts_ FTab nr_ts_theme GetFtab nrFN av GetProject GetWorkDir MakeTmp zstat dbf nr_zoneField nr_ts_FTab FindField vidnStrng If r_zoneField Nil then MsgBox Info Non rice training set field not the same as paddock theme exiting ERROR Return Nil End nr_zt Index_Theme GetGrid ZonalStatsTable nr_ts_Ftab aPrj nr_zoneField FALSE nrFN if nr_zt HasError then return NIL end nrzCountField nr_zt FindField Count nrzSumField nr_zt FindField Sum nrzTotCount 0 nrzTotSum 0 nrzCount 0 nrzSum 0 For each nr in nr_zt nrzCount nr_zt ReturnValue nrzCountField nr nrzSum nr_zt ReturnValue nrzSumField nr nrzTotSum nrzTotSum nrzSum nrzTotCount nrzTotCount nrzCount End nr_ts_ Mean nrzTotSum nrzTotCount Calculate Threshold value based on these means CSIRO Land and Water 47 Threshlid r_ts_ Mean nr_ts_Mean 2 Calculate Accuracy of Rice vs NonRice based on Threshld values Assumes certain field naming convention Initialise Variables TotalCount 0 NumClassR 0 NumClassNR 0 NumNRasR 0 NumRasNR 0 TotalNumR 0 TotalNumNR 0 Get Paddock Shapefile from user Query Ftab based on Threshld np_vidnFld newPaddF Tab FindField vidnStrng MeanFld newPaddFtab FindField Mean If MeanF
5. these two paddocks were not in our validation set so our estimate of accuracy remains the same In other words we don t really know whether we have just improved the accuracy or made it worse However from past studies it is reasonably safe to assume that we have just made it better If we look at the mean November Band5d values for these two paddocks they were just barely under our November Band5d threshold of 75 one was 67 and the other was 74 this also makes us more confident that we have done the right thing since not only did both paddocks show a strong crop signal in October but neither paddock showed a very strong water signal in our November image CSIRO Land and Water 37 8 Rice Administration Once the rice areas have all been completed and you are satisfied that they are correct you can start to generate the information from them required for rice administration 8 1 Adding attributes to rice areas The first step is to add a field listing areas in hectares to the attributes table of the rice area shapefile This can be achieved by running the AreaMeasurement script a copy of which can be found at m esri AV_GIS30 ARCVIEW Samples Scripts This is a very simple script to use This script will add two new fields to the shapefile area in square metres and perimeter in metres To view the area of each polygon in hectares add a new field called Area_ha with one decimal place You will first n
6. of Remote Sensing 20 2443 2460 Van Niel TG McVicar TR 2000 Assessing and improving positional accuracy and its effects on areal estimation at Coleambally Irrigation Area Cooperative Research Centre for Sustainable Rice Production P1 01 00 Yanco NSW Australia Van Niel TG McVicar TR 2001 Assessing positional accuracy and its effects on rice crop area measurement an application at Coleambally Irrigation Area Australian Journal of Experimental Agriculture 41 557 566 CSIRO Land and Water 40 Van Niel TG McVicar TR 2002 Experimental evaluation of positional accuracy estimates from a linear network using point and line based testing methods nternational Journal of Geographical Information Science 16 455 473 Van Niel TG McVicar TR 2003 A simple method to improve field level rice identification Toward operational monitoring with satellite remote sensing Australian Journal of Experimental Agriculture 43 379 387 Van Niel TG McVicar TR 2004a Current and potential uses of optical remote sensing in rice based irrigation systems a review Australian Journal of Agricultural Research 55 155 185 Van Niel TG McVicar TR 2004b Determining temporal windows of crop discrimination with remote sensing a case study in south eastern Australia Computers and Electronics in Agriculture In Press Van Niel TG McVicar TR Fang H Liang S 2003 Calculating environmental moisture for per field discrimination of rice crop
7. 0625 0000 J Polygon 2000_2001 VN 36875 00 I Polygon 1624 2000_2001 i 330625 0000 NR Polygon 1660 2000_2001 i 285000 0000 NR 4 3 Defining the rice and non rice training sets Next two shapefiles containing a small number of paddocks e g 5 whose boundaries are known to be current and whose crop types are also known will be defined These small subsets of paddocks are known as training sets and will be used to derive the threshold that will be used to define which pixels or paddocks are rice and which are non rice These training sets can be defined in various ways For example they can come from landholder surveys or from field observations In this case we used landholder surveys to determine paddocks where the crop types were known as above From these 5 rice and 5 non rice paddocks were arbitrarily selected and saved to separate shapefiles see figure below e g 02_03_r_trn shp and 02_03_nr_trn shp where 02 03 represents the year r and nr stands for rice or non rice and trn stands for training set The non rice training set was selected with a stratified sample in mind That is various crop types were selected to very roughly represent their proportion in the non rice crops 2 maize 1 soybean 1 sorghum 1 millet It is important that both the rice and non rice training sets do not include paddocks that are grossly unrepresentative of the overall class response For example a very small pro
8. 4 Barrs and Prathapar 1996 However highly accurate estimation of total rice area often exceeding 99 e g McCloy et al 1987 Barrs and Prathapar 1996 can be attributed to compensating negative and positive errors at the paddock level Barrs and Prathapar 1996 Van Niel and McVicar 2001 Wide discrepancies between rice area estimations at both the district level McCloy et al 1987 and the individual holding level Barrs and Prathapar 1996 can be exceedingly large and in the past has prevented the adoption of satellite remote sensing for the yearly monitoring of rice areas in both the Murrumbidgee Irrigation Area MIA and the CIA CSIRO Land and Water ziS Figure 1 The idea of temporally variant and Rice Rice invariant features can allow for efficient and accurate mapping of crops at the Year 1 Coleambally Irrigation Area CIA Most farms in the Soybeans CIA are typified by paddock boundaries which change very little from year to year but where the crop type inside the boundary may change frequently In this Rice Rice hypothetical example which looks at the same farm over Yea r 2 three consecutive summer growing seasons only one of the paddock boundaries changes depicted by the dashed line in year 3 However the crops grown in these static boundaries tend to change often If the Soybeans Rice boundaries are updated yearly with a GPS then moderately coarse remote Year 3 s
9. 4 Centre Latitude 34 37 28 S Centre Longitude 146 38 29 E Cloud Cover UL UR 0 0 ap e f you click on one of the quicklook small images you go to a new page where you can zoom in on the image to make sure there are no small clouds or haze over the CIA see below ACRES Landsat 4 5 and 7 Australian Centre for Remote Sensing search result 420021019235026 jpg Display Image Map Date Time Centrel UT 2002 10 19 23 50 26 Sateliits I Landsat 7 CSIRO Land and Water 5 e Make sure the Image radio button is selected see above Now you can zoom in to the CIA by clicking near its centre a couple of times This image looks good We could buy it if we wanted All we need is the date and Path Row these are listed in the table see the above image e Now order the image CICL purchases imagery from Resource Industry Associates RIA in Canberra John Lee is the contact 02 6260 5377 John accepts the ACRES order form see a copy of the order form in Appendix A The imagery can be paid for by credit card using the form that the RIA provide this has to be arranged through the Company Secretary e By going through this process for the 2002 2003 growing season we purchased both an early October image henceforth called the October image and a late November image henceforth called the November image These two images are referred to throughout this text the No
10. 5 A new window should reveal where the output excel file is located In this file is the threshold information and the estimation of accuracy which is meaningless for the winter cereal application CSIRO Land and Water 32 OUTPUT FILE x i Output file is c vanniehtmpsacct4 csy The output excel csv file should look something like this Remember only use the threshold value from this file Name NumNRasR NumRasNR TotalCount Threshold OverallAccuracy KappaAccuracy Octndvi 0 39 381 182912 0 897638 0 Validation Paddocks file c vanniel tmp 02_03_classif_rice shp Rice Training Set file c vanniel tmp rueben_rice_2002_2003 shapes 02_03_wc_trn shp Non rice Training Set file c vanniel tmp rueben_rice_2002_2003 shapes 02_03_nwc_trn shp Common Validation Field Crop2003 Time Mon Aug 25 11 50 56 2003 7 44 Calculate mean NDVI for each paddock using the sum_grid ave program As before the classification is put into a field context by using a paddock boundary shapefile to calculate the mean NDVI value These mean paddock values are then compared to the threshold calculated above However an additional step is required if you are using a shapefile that already contains the statistics fields in it that is if you have already run sum_grid ave on the input paddock theme You will need to make these fields invisible so ArcView does not get confused later as to which mean field to use To do this simply make the p
11. 9E 344516 3703S 325051 989 6152498 930 1065 655 OFFSET 336 ORIENTATION ANGLE 0 00 SUN ELEVATION ANGLE 59 2 SUN AZIMUTH ANGLE 72 3 e Now add the image theme into ArcView make sure the Data Source Types is set to Image Data Source The bil image should be visible File Edit View Theme Analysis Surface Graphics Window Help H E ZAKE eRLLEE Flat Sa Z View1 2 Add Theme Directory c vanniel tmp rueben_rice_2002_2003 images nov2002 0K E band4 bsq r CEE E band bsq ji E band8 bsq E rueben_rice_2002_2003 amp images Directories C Libraries Data Source Types Drives Image Data Source x e zl CSIRO Land and Water 9 e When the image is added to the view you will need to double click on the legend to change it to single band Band 5 e The image should appear as a grey scale with approximate map coordinates AMG66 see below 4 A File Edit View Theme Analysis Surface Graphics Window Help H Ee J GAA EE A e This process will need to be repeated until it matches with the DGPS data described below 3 3 Comparison of imagery with DGPS roads GIS data Now that the initial geometric coordinates are placed in the header file and the image is read into ArcView we want to compare it to a very accurate GIS layer This layer is the roads network digitised using NSW Department of Agriculture s DGPS unit in February 2001 These lines should be
12. A user s guide for identifying rice paddocks using GIS and remote sensing at Coleambally Irrigation Area NSW Thomas G Van Niel and Tim R McVicar CSIRO Land and Water CSIRO Land and Water Client Report A report for Coleambally Irrigation Co operative Limited CICL for research conducted by the CRC for Sustainable Rice Production Project 1105 November 2004 A user s guide for identifying rice paddocks using GIS and remote sensing at Coleambally Irrigation Area NSW Thomas G Van Niel Tim R McVicar a CSIRO Land and Water Private Bag No 5 Wembley WA 6913 Australia b CSIRO Land and Water GPO Box 1666 Canberra ACT 2601 Australia c Cooperative Research Centre for Sustainable Rice Production Yanco NSW 2703 Australia CSIRO Land and Water Client Report November 2004 Copyright and Disclaimer 2004 CSIRO To the extent permitted by law all rights are reserved and no part of this publication covered by copyright may be reproduced or copied in any form or by any means except with the written permission of CSIRO Land and Water Important Disclaimer CSIRO Land and Water advises that the information contained in this publication comprises general statements based on scientific research The reader is advised and needs to be aware that such information may be incomplete or unable to be used in any specific situation No reliance or actions must therefore be made on that information without seeking pri
13. H No Dsts Polygon i Oct_bnd2 Polygon J 0 28 7 3 Define small training set of known winter cereal and non cropped paddocks Next two more training sets winter cereal and non cropped will be used to derive the threshold that will be used to define which pixels or paddocks are winter cereal and which are not Again these training sets can be defined in various ways but you will most likely receive plenty of choices from the landholder surveys or from field observations In this case we used landholder surveys to determine paddocks where the crop types were known as above From these 5 winter cereal and 5 non winter cropped paddocks were arbitrarily selected and saved to separate shapefiles see figure below e g 02_03_wc_trn shp and 02_03_nwc_trn shp where 02 03 represents the year wc and nwc stands for winter cereal or non winter cereal and trn stands for training set The non winter cereal training set needs to be selected from paddocks that did not have any previous winter crop these should have a relatively low NDVI value compared to winter cropped paddocks in October CSIRO Land and Water 30 2 ArcView GIS 3 2 02_02_nwe_tn i A Octndvi HM 098 546 il 545 m 392 El 239 Now the threshold needs to be calculated This can be done in a number of ways For example you could run the sum_grid ave script and then calculate the average of the means between the two training sets ma
14. Mather 1999 Stuckens et al 2000 Furthermore spatially accurate paddock boundaries delineated from either fine resolution remote sensing or Global Positioning System GPS data can then be merged with this high accuracy crop identification information In this example the temporally invariant features the paddock boundaries need to be mapped very accurately only once or infrequently Conversely the temporally variant features uniform crop types found within the invariant boundaries can be mapped at a more coarse spatial resolution without loss of classification accuracy In this way temporally variant and invariant features can be mapped more efficiently while resultant crop classifications maintain both high attribute and spatial accuracy see figure 1 This is the common situation for most farms at the Coleambally Irrigation Area CIA where paddock boundaries change very little from year to year but the crop attribute is frequently changing The accurate measurement of rice area to the individual holding level is important for management of groundwater recharge in the irrigation areas of southern New South Wales NSW Barrs and Prathapar 1994 Humphreys et al 1994 and has resulted in yearly rice administration responsibilities of the local irrigation managers Both the accurate identification of rice and the estimation of its total area have been achieved in NSW with satellite remote sensing McCloy et a 1987 Barrs and Prathapar 199
15. Navigate to the table you exported The table will appear when you add it in In the table properties ensure only the farm numbers property IDs and outlet numbers are displayed Make the ID heading active Open the attributes table for the rice areas and make the ID CSIRO Land and Water 38 heading active It is important that you make the ID heading for the rice areas table active after that for the dbase table Click Table Join to append the attributes from the dbase table to the attributes table of the rice areas Close the tables when you are satisfied the table join has occurred successfully e Convert the rice areas shapefile to a new shapefile This will preserve the attributes that were generated by the table join in the new shapefile All the required attributes have now been appended to the rice areas shapefile 8 2 Location of rice paddocks on suitable ground The location of rice paddocks on suitable ground as delineated by the rice soil suitability classification can be determined by using the GeoProcessing Wizard Use the clip option to identify if any rice was grown on unsuitable ground The same option can be used to determine which rice areas were grown on marginal 1 year in 4 ground For these areas further overlays are required to see if they were grown with rice in any of the previous three years To perform each clip the rice suitability shapefile should be displayed by type using the theme prop
16. South Wales NSW is an important aspect of yearly management This process relies upon identification of rice paddocks and summarisation of each landholder s rice area In the past rice administration has been strictly based on high resolution aerial photography because although satellite remote sensing can accurately identify rice 1 its spatial accuracy has not been suitable for paddock level management and 2 image processing has required a great deal of training Recently greatly simplified techniques have been developed to accurately classify rice removing one of the impediments to using satellite remote sensing for rice administration If these results can be merged with spatially accurate paddock boundaries then the other impediment is also removed This report describes the methods required to both accurately classify rice from satellite remote sensing and merge these results with highly accurate vector paddock boundaries thus providing high attribute and spatial accuracy in a way that is both more automated and less expensive The format of this paper and the description of the procedures are informal and non technical wherever possible The examples provided are from data acquired in the 2002 03 growing season and are conducted solely in ArcView the generic GIS software already used by Coleambally Irrigation Co operative Limited CICL However these general techniques could easily be adapted to other GIS or image processing softwar
17. addock measurements from previous years for this year s rice The per paddock classification process is still described in detail for completeness and is thus available for use in this or other similar applications if the user so chooses There is no need to perform this task however if the per pixel visual inspection approach is taken Remember if you are not going to perform the following per paddock classification then it is a good idea to make use of the Waterways database that CICL records for every farm In this database the landholder s estimated rice area and the volume of water applied to rice are recorded and can be used as a guide in determining the area of rice for each farm When we know the area determination of rice paddocks by visual assessment is easier 6 1 Introduction As discussed before the methods outlined for identification of rice depend on placing the remotely sensed signal within a per paddock context which requires the updating of some boundaries every year see above After the publication of research using a moisture index based on the depth of ETM band 5 reflectance compared to that of ETM bands 4 and 7 it was found that simply using ETM band 5 alone provides just as accurate or better results Also the added benefit is that using a single band is even easier to process The methods here depend on the previous sections already being accomplished That is the ground validation should be compiled the rudi
18. addock theme active in this case 01_03 _classif_rice shp and then open it s theme table Go to the Table menu and click on Properties When this new menu pops up uncheck the checkboxes next to all of the statistics fields that were generated from the previous run of sum_grid ave most importantly the Mean field see below CSIRO Land and Water 33 amp Table Properties x Title Attributes of 02_03_classif_rice shp Creator Cancel Creation Date Monday 25 August 2003 12 24 34 PM Comments Visible Field Alias Now that ArcView does not see these fields we can run sum_grid ave on this paddock theme and the statistics that come out will be based on the new index theme in this case the NDVI for October Choose the set of paddocks shapefile in this case those classified as rice should all be selected Paddock Selection Choose the Paddock Theme Choose the GRID file representing October NDVI image CSIRO Land and Water 34 amp Index Selection Choose the Index Grid Themes to Summarize f aA Cancel Map Calculation 1 Oct_bnd4 Oct_bnd3 Nov_bnd5 After completion of running this script a new shapefile is added to the view This shapefile is a duplicate of the paddock theme input into it except that a set of fields are added to the dbf These new fields are the statistics we will use to refine the classification of rice specifically the mean valu
19. addocks as either rice or non rice Therefore we want at least all of the boundaries suspected of being rice defined previously In this example we use all of the paddock boundaries Remember the paddock theme should have all of its records selected Also make sure that the GRID of ETM band 5 is in the view e First the script will need to be loaded compiled and ran see ESRI and previous section for details CSIRO Land and Water 24 e Once the script is run follow the prompts Choose the full set of paddocks shapefile should all be selected amp Paddock Selection Choose the Paddock Theme 02 O3padds shp After completion of running this script a new shapefile is added to the view This shapefile is a duplicate of the paddock theme input into it except that a set of fields are added to the database file dbf These new fields are the statistics we will use to classify rice specifically the mean value for each paddock see below 43160 0000 CSIRO Land and Water 25 6 3 Classifying each paddock to rice or non rice Previously ground validation section 4 the threshold was calculated as 75 5838 This threshold will be used to select any paddocks that are less than or equal to this value Itis these paddocks that we will classify as rice There are various ways to approach this last step so this is more of a guide than a definitive list of instructions In this case we use th
20. alled By View GUI Calls None Assumptions Assumes that the view was the last active document before running the script Must have a selection on the polygon theme Initialize Variables theView av GetActiveDoc thethemes theView GetThemes GrList FList aPrj theView GetProjection Put themes in separate lists for feature and Grid themes For each Thm in theThemes If Thm ls GTheme then GrList add Thm Elseif Thm Is fTheme then FList add Thm End End ERROR CHECK LISTS If Flist Count 0 then System Beep MsgBox info No feature Themes to Select From must add Polygon Theme to the View ERROR Return NIL CSIRO Land and Water 51 End If GrList Count 0 then System Beep MsgBox info No Grids to Select From must add a Grid to the View ERROR Return NIL End Get info from User and Error Handle Get Polygon Theme Plygn_theme MsgBox Choice FList Choose the Paddock Theme Paddock Selection Check for Cancel Exit if true If Plygn_theme NIL then System Beep Return NIL End Make sure there is a feature selection on PlygN Theme PlygnFtab Plygn_Theme GetFtab OrigPlygnBit PlygnFtab GetSelection Clone If OrigPlygnBit Count 0 then System Beep MsgBox info No Polygon Theme features selected please try again ERROR Return NIL End Put Fields from Plygn theme in a list and allow user to select the field to summarize Make sure they are onl
21. ating Paddock Boundaries 19 5 1 Introduction 19 5 2 Defining flooded areas on the imagery 20 6 Identifying Summer Rice Paddocks Optional Step 24 6 1 Introduction 24 6 2 Calculating the mean ETM band 5 value for every paddock 24 6 3 Classifying each paddock to rice or non rice 26 7 Identifying Winter Cereal Paddocks for Exclusion Optional Step 28 7 1 Introduction 28 7 2 Calculating the NDVI on an October ETM image 28 7 3 Define small training set of known winter cereal and non cropped paddocks 30 7 4 Calculate mean NDVI for each paddock using the sum_grid ave program 33 7 5 Identify and eliminate rice paddocks that had a winter crop 35 8 Rice administration 38 8 1 Adding attributes to rice areas 38 8 2 Location of rice paddocks on suitable ground 39 8 3 Total rice area and rice water usage 39 References 40 Appendix A 42 Appendix B 43 Appendix C 51 CSIRO Land and Water iv CSIRO Land and Water 1 Introduction The strengths of moderate to coarse resolution satellite remote sensing in both identifying crop types and estimating crop area has resulted in the widespread use of this technology for the monitoring of agriculture e g Barbosa et al 1996 Fang 1998 Although the spectral information and cost of these remote sensing data are attractive their spatial resolutions are often perceived as being inadequate for agricultural management at either the individual holding or the paddock level Quarmby et al 1992 Conversely fine reso
22. cView GIS 3 2 Fie Edit View Theme Analysis Surface Graphics Window Help E2 ZAJRKEKE Scale 1 253 892 eis etree F 142 170 Gl 171 198 TMM ico 226 227 255 HM No Dats e This GRID can then be saved to a temporary shapefile using Theme Convert to Shapefile command Once this is complete use the shapefile Query Builder to select only the polygons with Gridcode 1 as seen below File Edit View Analysis Surface Graphics Window Help Bld ele AHK Baek Ese E be sa aoa AN A Grideode 1 New Set Add To Set Select From Set 57 85 R 113 E 114 141 EE lt 2 170 171 198 199 226 EE 227 255 HM No Dste Y Gps shp N x vi Nov_bnd5 bsq CSIRO Land and Water 21 e Run Theme Convert to Shapefile again with this selection and this will result in a shapefile of the areas predominantly covered by water on this image date see below Theme Analysis Surf KE Scale 1 520 154 seta A Water_ciassshp 2_03_validation s EE No Dats _ Gps 6 shp Af Nov_bnd5 bsq 1 ag e This shapefile can then be used to compare the extent of water for this current date to previously digitised paddocks If the match is good then the old paddock boundary is OK to use see below for an example of a good match between three previously defined boundaries in red to the water classification of this year Surface Graphics Window Help
23. e for each paddock see below Attributes of Octndv2 shp 260 7338 40153 0000 7 5 Identify and eliminate rice paddocks that had a winter crop Now that we have the mean paddock values and the threshold it is just a matter of identifying which paddocks from those already classified as rice are now identified as having a winter cereal This can be accomplished by querying the output shapefile above with the query builder the hammer icon in the view GUI As before select those paddocks that meet the criteria defined by the threshold In this case however we are interested in those paddocks that are greater than or equal to the threshold those paddock previously classified as rice that had very high NDVI in October CSIRO Land and Water 35 First use the query builder to select any paddocks in the new theme generated above step 1 of this section that are greater than or equal to the previously defined threshold of 182 912 of coarse this threshold will change every year or if different training sets are used 2 Octndv2 shp 421 6911 ixl V Update Values Mean gt 182 912 re Add To Set Select From Set In this case two paddocks are selected see image below 2 ArcView GIS 3 2 File Edit Table id Window Help E Ge SAL a ASE E EL Sle E a _ Theme21 shp _ Octndv4 shp Octndvi HMB ooe 02_03_clsssif_rice Nov_bn15 shp 02_03padd
24. e packages in the future Also these techniques are specifically designed to work best under the management practices existing at the CIA at the time of publication That is under the current management practices rice is easily identified from all other crops in late November when the water of the permanently flooded rice paddocks is drastically different from the other paddock s surfaces If management practices significantly change in the future e g if rice is predominantly grown on beds then the specifics of this research will need to be adjusted For more details see the associated scientific publications of this work regarding 1 spatial accuracy of aerial photography and GIS data Van Niel and McVicar 2000 2001 2002 and 2 rice identification using satellite remote sensing Van Niel and McVicar 2003 Van Niel et al 2003 CSIRO Land and Water iii Table of Contents Copyright and Disclaimer Acknowledgements Executive Summary 1 Introduction 1 2 Selection and Purchase of TM Imagery from ACRES Website 3 2 1 Introduction 3 2 2 ACRES website 3 3 Rudimentary Geometric Offset 7 3 1 Introduction 7 3 2 Importing Images into ArcView 7 3 3 Comparison of imagery with DGPS roads GIS data 10 4 Ground Validation Classification Accuracy and Training Sets 13 4 1 Introduction 13 4 2 Defining the ground validation dataset 13 4 3 Defining the rice and non rice training sets 14 4 4 Estimating classification accuracy 15 5 Upd
25. e query builder to select only those paddocks that are less than or equal to the threshold and then save them out to a separate shapefile First use the query builder to select any paddocks in the new theme generated above step 1 of this section that are less than or equal to the previously defined threshold of 75 5838 of coarse this threshold will change every year or if different training sets are used 2 Attributes of Nov_bn15 shp DEAR Fields Values IV Update Values Add To Set Select From Set e Click on New Set This results in a subset of the original number of paddocks being selected see below a OSE BOE m AA a oo Be A 381 of 2845 selected kee 02_02padds shp Water_class shp 02_03_nr_trn shp Nov_bnd5 J 0 28 E waw ane Se Stam e iie iie an T a 192 7704 37783 0000 129 3649 20 9714 38292 0000 121 1934 6 2849 47629 0000 CSIRO Land and Water 26 e Now that the rice paddocks are selected if we click on Theme Convert to Shapefile the rice paddocks can be saved out to a separate shapefile called mine 02_03 classif_rice shp see below oy File Edit Vi Graphics Window Help ZAKR P 408 974 97 Scale 12815 614998824 Theme Analysis Surf _ Nov_bn15 shp 02_02padds shp _ Water_class shp _ 02_03_padds_to_c 02_03_validstion s These are our final rice paddocks Optio
26. e than about 0 5 of a pixel or in this case 12 5 metres 2 ArcView GIS 3 2 File Edit View Theme Analysis Surface Graphics Window Help Ok i eene eT Seale 1 20000 SEE e AE View1 E LL did this and was happy with the following coordinates see pic below ULXMAP 351762 500 ULYMAP 6168937 500 This represents a shift of 50 metres in the x direction and 100 metres in the y direction resulting in a good match CSIRO Land and Water 11 e Note that an addition to the X coordinate will result in the image moving right and a subtraction will result in it moving left An addition to the Y coordinate will result in it moving up and a subtraction will result in it moving down 2 ArcView GIS 3 2 File Edit View Theme Analysis Surface Graphics Window Help E ZAKR Pe ekl eene TL Seale 1 20000 615112528 m o AE A View1 Gps66 shp N Nov_bnd5 bsq LL LL CSIRO Land and Water 12 4 Ground Validation Classification Accuracy and Training Sets 4 1 Introduction One of the most important things in the classification of rice is defining an appropriate ground validation or reference dataset This reference dataset will be used for two purposes i estimation of rice classification accuracy and ii defining two small training sets which will be used to determine the threshold used to classify paddocks as either rice or non rice The following topics then need more detailed discuss
27. e the Inc field e Load the Xtools extension in the ArcView project A window will appear in which you can change the default units to metres and hectares Close the window e An Xtools option will appear as one of the headings at the top of the ArcView window When you click on the heading there is a long list of options Click on Convert shapes to centroids Follow the instructions and select the rice areas theme as the theme to convert to centroids Store the new file Cntrs in X Temp e Open the attributes table of Cia_external Click on Table Properties and un tick the data that you do not need As stated earlier the attributes you will need are farm numbers property ID and outlet numbers Click OK e You can now use the Geoprocessing Wizard to perform a spatial join You may need to load the Geoprocessing extension Then go to View GeoProcessing Wizard Select the option Assign data by location Spatial Join and click Next Assign data to Cntrs and assign data from Cia_external Click Finish e Open the attributes table of Cntrs You will see that the three fields we require have been appended with data for every record in the table If you wish to redo the spatial join for some reason remove the join by clicking on Table Remove All Joins e You now need to export the attributes table for Cntrs as a dbase file to X Temp e Inthe project window click on Tables and then Add
28. eed to start editing the table and make the heading Area_ha active Under the Field Calculate option you can run a calculation in which you divide the values in the Area field by 10 000 All the values will be automatically filled in Next you will need to append the attributes from the shapefile cia_external to the rice areas In particular the attributes that we require are farm numbers property ID and outlet numbers It might be expected that this could be achieved by straightaway doing a simple spatial join using the Geoprocessing Wizard However this option relies on the rice polygons being entirely contained within the farm boundaries Due to slight differences in the georeferencing of our GIS layers including the aerial photographs this is not possible You will therefore need to convert the rice areas to centroids points at the centre of each polygon to enable the spatial join to occur The spatial join is performed as follows e Ensure that there is a field in the attributes table of the rice areas called ID and that every record in the attributes table has a unique ID number You can add these numbers to the table if they are not already there by running the increment script This can be found in m esri AV_GIS30 ARCVIEW Samples Scripts This will generate a field called Inc By doing a field calculation as described above for hectares you can give the field called ID the same numbers as in Inc Delet
29. ely sensed signal within some sort of context Basically this means that we don t just use the information from all the pixels in the image separately per pixel classification but rather we also make use of the information that is surrounding these points at the same time One of the most common ways to put remote sensing data into context for agricultural applications is by summarising the entire paddock per paddock classification This requires a GIS layer of paddock boundaries As usually the hardest part of a mapping project is defining boundaries this per paddock method has a distinct advantage over per pixel classifications That is since we know the potential boundaries of the phenomena that we wish to map before we map it the hardest part of the mapping procedure is done before we start This is also one of the reasons why we are able to get such high accuracies using such simple techniques there are also other reasons like homogeneous management practices of the landholders and the fact that we are really mapping standing water versus non water which is not too hard compared to lots of other things Unfortunately there is also a distinct disadvantage to this per field method if the GIS boundaries are not spatially accurate or up to date temporally accurate then the results can be worse than a per pixel classification Therefore assuring that the paddocks are both spatially accurate as well as appropriate for the curren
30. ensing data can be used to map the crop type every Fallow year for less expense than high resolution aerial photographs Maize Because high spatial accuracy is important for management both the MIA and CIA have primarily used the interpretation of high resolution aerial photography to monitor rice crop areas every year However recent research at the CIA regarding 1 spatial accuracy of aerial photography and GIS data Van Niel and McVicar 2000 2001 2002 and 2 rice identification using satellite remote sensing Van Niel and McVicar 2003 Van Niel et al 2003 has moved satellite based remote sensing one step closer to an operational system in the region This research has been further refined and simplified in collaboration with staff from the Coleambally Irrigation Co operative Limited CICL over the 2002 03 and 2003 04 growing seasons during which satellite remote sensing was used to complete rice administration at the CIA The reader is also referred to a review of remote sensing of rice based systems where these and other issues are described Van Niel and McVicar 2004a This report then outlines the steps required to merge the high spatial accuracy paddock boundaries defined from either high resolution digital aerial photography or a GPS in the field with high attribute accuracy rice classifications from moderately coarse and relatively inexpensive satellite imagery in order to accomplish yearly rice ad
31. ensing to estimate summer crop areas in the Coleambally Irrigation Area NSW CSIRO Division of Water Resources Consultancy Report 96 17 Fang HL 1998 Rice crop area estimation of an administrative division in China using remote sensing data International Journal of Remote Sensing 19 3411 3419 Humphreys L Van Der Lely A Muirhead W Hoey D 1994 The development of on farm restrictions to minimise recharge from rice in New South Wales Australian Journal of Soil and Water Conservation 7 11 20 McCloy KR Smith FR Robinson MR 1987 Monitoring rice areas using Landsat MSS data International Journal of Remote Sensing 8 741 749 Moody A 1997 Using landscape spatial relationships to improve estimates of land cover area from coarse resolution remote sensing Remote Sensing of Environment 64 202 220 Pedley MI Curran PJ 1991 Per field classification an example using SPOT HRV imagery International Journal of Remote Sensing 12 2181 2192 Quarmby NA Townshend JRG Settle JJ White KH Milnes M Hindle TL Silleos N 1992 Linear mixture modelling applied to AVHRR data for crop area estimation International Journal of Remote Sensing 13 415 425 Stuckens J Coppin PR Bauer ME 2000 Integrating contextual information with per pixel classification for improved land cover classification Remote Sensing of Environment 71 282 296 Tso B Mather PM 1999 Crop discrimination using multi temporal SAR imagery International Journal
32. er of the view in the corresponding direction The and icons at the far right make the map image larger or smaller or the Configure twiddle can be opened for numeric size control and to select different map layers to display Depending on your preference settings either the boundaries of the current view will define the search area or an Input menu may be present used for marking a point or region of interest Select Search at the bottom of this form to begin a search of the specified area Use the Areas of the Interest list below the map to select a place or region of interest Selecting an item from the list automatically re displays the map with new search area indicated Dd Location Map coordinates Y Date eee es back to top SPATIAL INFORMATION FOR THE NATION security amp privacy disclaimer site map Commonwealth of Australia 2002 CSIRO Land and Water 3 e Now Click on the Databases tab along the top of the page e On the databases page select ACRES Landsat 4 5 and 7 Australian Centre for Remote Sensing but DO NOT click on Search yet e Next click on the Refine tab along the top of the page e On the refine page adjust the search so only Landsat 5 and TM are highlighted for the 2002 03 growing season Landsat 7 was still operating and was chosen see picture below For 2003 04 season and after Landsat 5 should be chose
33. erties to isolate unsuitable areas from marginal areas and vice versa 8 3 Total rice area and rice water usage The rice areas can now be exported for use in rice administration Open the attributes table of the new shapefile and export to an appropriate location on the network The file can be modified and saved in Excel The information exported can be used to sum the rice areas for each farm enabling comparison with the allowed rice areas for the current season A report can be obtained from the Waterways database with rice water use by block outlet number The outlet numbers relating to the rice areas can be matched with the outlet numbers relating to rice water use and the rice water usage in ML ha can then be calculated for each block CSIRO Land and Water 39 References Aplin P Atkinson PM Curran PJ 1999 Fine spatial resolution simulated satellite sensor imagery for land cover mapping in the United Kingdom Remote Sensing of Environment 68 206 216 Barbosa PM Casterad MA Herrero J 1996 Performance of several Landsat 5 Thematic Mapper TM image classification methods for crop extent estimates in an irrigation district International Journal of Remote Sensing 17 3665 3674 Barrs HD Prathapar SA 1994 An inexpensive and effective basis for monitoring rice areas using GIS and remote sensing Australian Journal of Experimental Agriculture 34 1079 1083 Barrs HD Prathapar SA 1996 Use of satellite remote s
34. g KappaAccuracy AsString AccuracyLst Add AccuracySirng End Write accuracy info to a text outFN av GetProject GetWorkDir MakeTmp acc csv outFile LineFile Make outFN FILE_PERM_WRITE outFile WriteElt Name NumNRasR NumRasNR TotalCount Threshold OverallAccuracy KappaAccuracy For Each Strng in AccuracyLst outFile WriteElt Strng End Write metadata info at the end CSIRO Land and Water 49 outFile WriteElt outFile WriteElt Validation Paddocks file padd_theme GetSrcName GetFileName asString outFile WriteElt outFile WriteElt outFile WriteElt Common Validation Field vidnFld asString outFile WriteElt Time Date Now AsString Rice Training Set file r_ts_theme GetSrcName GetFileName asString Non rice Training Set file nr_ts_theme GetSrcName GetFileName asString Close output file outFile Close Tell user where output file is MsgBox Info Output file is outFN asstring OUTPUT FILE av ClearStatus av ShowMsg Output file is outFN asstring CSIRO Land and Water 50 Appendix C DiskFile Sumgrid sumgrd ave Programmer Tom Van Niel Created 2002 Revisions 11 June 2004 TVN added more error traps and made code i more general Function Summarises a grid dataset by each polygon in a polygon dataset Performs a zonalstats resulting in min mean max majority etc values of the grid for ea polygon C
35. ich paddocks may need updating and requires a per pixel classification Since water absorbs light in the wavelengths around 1650 nm where band 5 is centred the rice areas during this time of year have very low values Also non rice paddocks are predominantly some much drier surface e g soil stubble or green crops that are not near canopy closure revealing very high values in ETM band 5 Therefore it is generally very easy to classify water versus non water which is really what we are doing we assume that all the flooded paddocks are rice We will put these values in a per paddock context later Be careful to note whether a heavy rainfall occurred just prior to image acquisition as this might influence the results of this analysis e First the GRID of ETM band 5 generated in the Ground Validation section above is classified using the threshold value determined in that same section As determined above the threshold between rice non rice from our training sets was 75 5838 e This entails using the Map Calculator in the Analysis menu Set up the equation to define the classes based on the above threshold see figure below Map Calculation 1 Exp Log Exp2 Log2 Exp10 Log10 Nov_bnd5 lt 75 5838 and Nov_bnd5 gt 0 Evaluate e This will result in the calculation of a GRID that has a two classes zero s for non water areas and one s for water see below CSIRO Land and Water 20 i Ar
36. ield MUST contain unique numbers within any one shapefile or the zonal stats command used in this program chokes The validation field is a field describing which paddocks withing the validation shapefile are rice or non rice The naming convention of R for rice and NR for non rice must be used for the program to work properly Sumgrid summarise grid statistics for input shapefile output new shapefile with statistics Initialize Variables theView av GetActiveDoc thethemes theView GetThemes GrList FList CSIRO Land and Water 43 AccuracyLst aPrj theView GetProjection Put themes in separate lists for feature and Grid themes For each Thm in the Themes If Thm ls GTheme then GrList add Thm Elseif Thm Is fTheme then FList add Thm End End ERROR CHECK LISTS If Flist Count 0 then System Beep MsgBox info No feature Themes to Select From must add Paddock Theme to the View ERROR Return NIL End If GrList Count 0 then System Beep MsgBox info No Grids to Select From must add a Grid to the View ERROR Return NIL End Get info from User and Error Handle Get Paddock Theme Padd_theme MsgBox Choice FList Choose the Paddock Theme Paddock Selection Check for Cancel Exit if true If Padd_theme NIL then System Beep Return NIL End Get Rice Training Set theme r_ts theme MsgBox Choice FList Choose the Rice Training Set Theme RICE t
37. ile to view newPlygnFTheme FTheme Make newPlygnFTab theview addtheme newPlygnFTheme End CSIRO Land and Water 53
38. ion e Defining the ground validation dataset e Defining the rice and non rice training sets and e Estimating classification accuracy 4 2 Defining the ground validation dataset The first step is to define the base ground validation set of paddocks This can be done by sending out simple surveys to landholders Since this has been done for several years ina row there is a set of Word documents to use including the surveys and farm maps These can be faxed to the landholders who usually respond quickly if the survey is relatively simple Maps can be found in NRE on Sol Rice Survey forms cropping Before faxing and mailing the forms out to the landholders it is necessary to update the year at the top of the form and the covering note For the 2002 2003 growing season we had a good response where the crop types of 73 paddocks were known Since 40 of these were rice a fairly even sample of rice and non rice was also achieved which is one of the goals Arbitrarily it would be good to have at least 30 rice and 30 non rice paddocks to estimate the classification accuracy from Since 5 rice paddocks and 5 non rice paddocks are used for training the classification see below and cannot be used for both training the classification and estimating it s accuracy we will ideally need at least 35 rice and 35 non rice paddocks from this survey If we do have 30 of each class for accuracy estimation it means that although a bit course the preci
39. ld NIL then MsgBox ERROR Could not find Mean field MEAN FIELD ERROR Return NIL End For each record in newPaddFtab TotalCount TotalCount 1 vidnValue newPaddFTab ReturnValue np_vidnFld record MeanValue newPaddFTab ReturnValue MeanFld record If vidnValue R then TotalNumR TotalNumR 1 Else TotalINumNR TotalNumNR 1 End Determine Direction of comparison i e if rice mean is lower than nr mean than use lt symbol If r_ts Mean lt nr_ts_Mean then If MeanValue lt Threshld then NumClassR NumClassR 1 If vidnValue lt gt R then NumNRasR NumNRasR 1 End Else CSIRO Land and Water 48 NumClassNR NumClassNR 1 If vidnValue R then NumRasNR NumRasNR 1 End End Else If MeanValue gt Threshld then NumClassR NumClassR 1 If vidnValue lt gt R then NumNRasR NumNRasR 1 End Else NumClassNR NumClassNR 1 If vidnValue R then NumRasNR NumRasNR 1 End End End End NumRasR TotalNumR NumRasNR NumNRasNR TotalNumNR NumNRasR Prop_observed NumRasR NumNRasNR TotalCount Prop_expected TotalNumR NumClassR TotalNumNR NumClassNR TotalCount 2 OverallAccuracy NumNRasNR NumRasR TotalCount KappaAccuracy Prop_observed Prop_expected 1 Prop_ expected AccuracyStrng Index_Theme AsString NumNRasR AsString NumRasNR AsString TotalCount AsStri ng Threshld AsString OverallAccuracy AsStrin
40. lution remote sensing e g aerial photography very often contain spatial detail that will allow management decisions to be made at the paddock level Van Niel and McVicar 2001 but these data can be expensive to acquire and subsequent manual digitisation of crop areas is labour intensive when done every year Because agricultural systems are generally very structured when compared to natural systems land cover and management practices can be assigned within discrete paddock boundaries Also the more regulated agricultural systems e g irrigation areas have paddock boundaries that are practically temporally invariant The combination of these characteristics can result in both very high attribute and spatial accuracies when fine resolution imagery is used However mapping relatively large areas every year to sucha fine detail can be particularly costly Also it is an inefficient spatial exercise to map invariant features at a high spatial resolution more than once Given the ordered nature of agricultural systems the relative strengths of fine and coarse resolution spatial data can often be integrated successfully for optimising accuracy cost and effort of crop area assessment although this is rarely done High crop identification accuracy can be achieved using relatively coarse resolution satellite remote sensing data especially when it is merged with contextual information Pedley and Curran 1991 Moody 1997 Aplin et al 1999 Tso and
41. mentary geometric offset should be applied the rice and non rice training sets should be defined the threshold should be calculated and the paddock boundaries should be updated The basic steps that will be covered in this section include e Calculate the mean ETM band 5 value for every paddock and e Classify each paddock into a rice or non rice class by comparing the threshold defined from the training sets to the mean paddock values Using a subset of the r_nr ave Avenue script called sum_grid ave see Appendix C the mean value of each paddock will be summarised and will be saved to a separate shapefile saved to your ArcView working directory to change this look in File Set Working Directory The inputs are a paddock theme this time it is the full set of paddocks over the entire CIA instead of the validation set of paddocks and the GRID of ETM band 5 6 2 Calculating the mean ETM band 5 value for every paddock Previously in the ground validation section we used a similar script to calculate a per pixel classification for the purposes of updating paddock boundaries Now we will rerun the same basic steps using the information from that original processing i e the threshold determined from the training sets and the ETM band 5 GRID Whatever paddock boundaries are input into this sum_grid ave program will be summarised statistically including the mean of each paddock which is what we will subsequently use to classify the p
42. ministration It also demonstrates that by taking advantage of certain strengths of coarse and fine scale spatial data the relationship between cost and accuracy can be improved by the use of satellite remote sensing in conjunction with highly accurate paddock boundaries CSIRO Land and Water 2 2 Selection and Purchase of TM Imagery from ACRES Website 2 1 Introduction The timing of image acquisition can greatly influence the results of satellite image classifications Van Niel and McVicar 2004b The first problem in any remote sensing application is defining when discrimination of the particular target of interest e g rice is best We have seen at the CIA that best rice discrimination occurs in late November due to the water of the ponded rice paddocks being easily identified when compared to all other areas The second problem is acquiring an image during the desired temporal window This is not usually a problem at the CIA however because it is completely contained within the overlap area of two Landsat scenes This means that the main CIA is imaged about every 8 days instead of every 16 days like in non overlap areas This provides twice as many opportunities to acquire cloud free imagery in the desired timeframe and thus greatly enhances the operational management of satellite remote sensing at the CIA To define what images have been acquired and to order an image the Australian Centre for REmote Sesning ACRES website is ex
43. n e Then type in the valid paths and row Path 92 to 93 and Row 84 e Choose the Month and Year you would like to search under the Season section e g Oct and Nov 2002 e Then Click on the Search button see the picture below VY ACRES Landsat 4 5 and 7 Australian Centre for Remote Sensing Landsat 4 Satellite Sensor Y world Reference System Path Row Path 92 to 93 Row 84 lto 7 Cloud Cover average Oguadrant Average Cloud 0 100 0 to 50 YW Season Sep a Month Dec 7 Sort Results Primary Field v Ascending Descending Secondary Field v Ascending Descending Tertiary Field O Ascending Descending e This brings up a list of images within the specified criterion with relevant info in the table on the left and a quicklook jpeg image on the right see below CSIRO Land and Water 4 Results of your last search 1 database searched 7 matching records were found ACRES Landsat 4 5 and 7 Australian Centre for Remote Sensing search results Y 7 matching records were found Items 1 7 are shown Date Time Centre UT 2002 10 03 23 50 22 Satellite Landsat 7 Sensor ETM Path Row 92 84 Centre Latitude 34 37 22 S Centre Longitude 146 40 27 E Cloud Cover UL UR 20 50 a Date Time Centre UT 2002 10 19 23 50 26 Satellite Landsat 7 Sensor ETM Path Row 92 8
44. nally these can be adjusted by preventing paddocks which contained previous winter cereals from being classified as rice This optional section is discussed below CSIRO Land and Water 27 7 Identifying Winter Cereal Paddocks for Exclusion Optional Step 7 1 Introduction An optional processing tactic is to adjust the previously classified rice paddocks defined from the step above by not allowing any paddocks to be classified as rice if they contained a crop during the preceding winter growing season The premise is that the winter crops are not usually harvested until late November since the recommended time for planting rice is early October almost no farmers will ever plant a rice crop immediately after a winter crop Since it is very easy to define cropped areas from non cropped areas using the Normalised Difference Vegetation Index NDVI we are able to detect any paddocks that were winter cereals and prevent them from being classified as summer rice An October image should be ideal for this since the winter crops should still be green while the summer paddocks should be in some state of preparation for planting if they have not been sown yet but not yet green Also during this time most winter pasture should be dried out and being prepared if there is going to be a summer crop If you have purchased an October image then all of the previously discussed steps will need to be applied to this image That is importing in
45. newPaddFtab Padd_Theme ExportToFTab av GetProject GetWorkDir MakeTmp p_basename shp zoneField newPaddFTab FindField Id aFN av GetProject GetWorkDir MakeTmp zstat dbf zt Index_Theme GetGrid ZonalStats Table newPaddFtab aPrj zoneField FALSE aFN if zt HasError then return NIL end ztJoinFld zt FindField Id Join Shape file s FTab to zt newPaddFtab Join zoneField zt ztJoinFld Add new padd shapefile to view newPaddFTheme FTheme Make newPaddF Tab theview addtheme newPaddFTheme summarise training sets get mean and std for rice training set and for nr training set Summarise Rice first Run ZonalStats on Index Grid for rice training set r_ts FTlab r_ts theme GetFtab rFN av GetProject GetWorkDir MakeTmp zstat dbf r_zoneField r_ts_FTab FindField vidnStrng If r_zoneField Nil then MsgBox Info Rice training set field not the same as paddock theme exiting ERROR Return Nil End r_zt Index_Theme GetGrid ZonalStatsTable r_ts_Ftab aPrj r_zoneField FALSE rFN if r_zt HasError then return NIL end rzCountField r_zt FindField Count rzSumField r_zt FindField Sum rzTotCount 0 CSIRO Land and Water 46 rzTotSum 0 rzCount 0 rzSum 0 For each rinr_zt rzCount r_zt ReturnValue rzCountField r rzSum r_zt ReturnValue rzSumField r rzTotSum rzTotSum rzSum rzTotCount rzTotCount rzCount End r_ts Mean rzTotSum rzTotCount
46. nually Otherwise you can run the r_nr ave script and disregard the accuracy information That is have the script calculate the threshold for you If using the script to calculate the threshold ignore the accuracy statistics they are meaningless See the menus below for the input for this option If using the r_nr ave script the rice and non rice prompts will need to be replaced in your mind by winter cereal and non winter cereal Choose the shapefile containing the paddocks that have already been classified as rice should all be selected 2 Paddock Selection Choose the Paddock Theme fei cedt nade Cancel Then the winter cereal training set CSIRO Land and Water 31 RICE training set Selection Choose the Rice Training Set Theme 02 03 we tm shp x Cancel Then the non winter cereal training set Non Rice Training Set Selection Choose the Non Rice Training Set Theme 3 03 nwo Ta A Then choose the field that contains the R and NR information This actually doesn t matter for the winter cereal processing but has to be filled in for the script to finish 2 CROP Type Field Selection Choose Field used for validation Prop tits AAE SEEE EENE EEEE EEEREN And finally choose the GRID file representing the NDVI of the October image Index Selection Choose the Index Grid Theme s to Summarize Map Calculation 1 Oct_bnd4 Oct_bnd3 Nov_bnd
47. or expert professional scientific and technical advice To the extent permitted by law CSIRO Land and Water including its employees and consultants excludes all liability to any person for any consequences including but not limited to all losses damages costs expenses and any other compensation arising directly or indirectly from using this publication in part or in whole and any information or material contained in it Cover Photograph Description Enhanced Thematic Mapper ETM image over the Coleambally Irrigation Area acquired on 13 February 2002 Wavelengths displayed are 0 7199 um red colour gun 0 6614 um green colour gun and 0 5610 um blue colour gun 2004 CSIRO CSIRO Land and Water i Acknowledgements This research was funded by the CRC for Sustainable Rice Production Rice CRC and CSIRO Land and Water Please note mention of commercial products in this work does not imply an endorsement of these products by either CSIRO or the Rice CRC Many thanks to Reuben Robinson and David Klienert of Coleambally Irrigation Co operative Limited CICL who provided helpful comments on the manuscript and learned rice identification from remote sensing over the telephone The new section 8 about rice administration was written by Rueben Robinson and David Klienert during their review of the manuscript CSIRO Land and Water ii Executive Summary Rice administration at the Coleambally Irrigation Area CIA New
48. or each paddock in the three shapefiles It then calculates the threshold value as the average of the mean rice and mean non rice responses from the two training sets Then it compares the validation paddocks known crop types to what this newly defined threshold would classify the paddock as in order to estimate accuracy The script called r_nr ave should be loaded into a script editor window and then compiled Make sure that all three of the above shapefiles are loaded into the ArcView View and that all the records in the validation shapefile are selected f there is no selection the script will complain The script also assumes that the view is the last thing to be active prior to running the script this is due to the first line in the script where the view variable is set to the active document using the av GetActiveDoc request Therefore if an error occurs right CSIRO Land and Water 15 off the bat and say s something like A n Table does not recognise request GetThemes then make the View active then make the script active and then run it again it should work this time Another solution would be to attach this script to a button directly on the View s GUI see ArcView help for details Also make sure that a GRID of ETM band 5 is in the view To convert the image to a GRID simply click on Theme Convert to Grid from the View s GUI and follow the prompts the ArcView Spatial Analyst extension must also be alread
49. ou need to put the initial ULXMAP and ULYMAP coordinates based on the ACRES geometric correction into the hdr file if not already there e This requires copying and pasting the coordinates from the file called header hrf which is found in the same directory as the imagery for 27 November 2002 The header hrf file is a text file and can be opened in any text editor or even Word The contents of this file for the November image are shown below with the UL coordinates coloured red These are the values that need to be pasted into nov_bnd5 hdr As can be seen the template was made with these coordinates so we don t have to worry about actually changing it this first time but you will need to do it next year or for this year s October image CSIRO Land and Water 8 0 000000000000000 0 000000000000000 0 000000000000000 0 000000000000000 0 000000000000000 0 000000000000000 0 000000000000000 0 000000000000000 0 000000000000000 0 000000000000000 0 000000000000000 0 000000000000000 GEOMETRIC DATA MAP PROJECTION UTM ELLIPSOID AustralianNational DATUM AGD66 USGS PROJECTION PARAMETERS 6378160 000000000000000 6356774 719195305400000 0 000000000000000 USGS MAP ZONE 55 UL 1452257 5446E 343641 3657S 351712 500 6168837 500 UR 1460728 1294E 343708 6552S 419737 500 6168837 500 LR 1460710 2050E 350515 7448S 419737 500 6116862 500 LL 1452224 4454E 350447 9781S 351712 500 6116862 500 CENTER 1450519 027
50. portion of the rice paddocks are not flooded by late November So if by chance we include one of these paddocks in our training set the statistics developed from it will not really represent the normal case and may result in poor classification accuracy Therefore particular care must be taken to select a representative training set in order to achieve the required accuracies CSIRO Land and Water 14 ArcView GIS 3 2 SE File Edit Table Field Window Help BOI w MINED E EI GI amp D of 5 selected 1 aK new y 02_03_r_tnshp 4 nF og 3 7 m s Views ii 02_03_nr_trn shp 02_03_known_pad amp Nov_bnd5 T Attributes of 02_03_nr_trn shp e Once these shapefiles are organised we will estimate the classification accuracy of rice based on the remaining known paddocks The final validation shapefile will include all of the original paddocks from step 1 above minus those paddocks used for training in step 2 That is the paddocks used for training the classification should not be used to estimate it s accuracy since they are no longer independent 4 4 Estimating classification accuracy Finally these three shapefiles will be used in conjunction with an Avenue script and the GRID representation of ETM band 5 to define both the threshold used to classify rice versus non rice and the accuracy defined from the final validation set of paddocks The script summarises the mean ETM band 5 value f
51. raining set Selection CSIRO Land and Water 44 Check for Cancel Exit if true If r_ts_theme NIL then System Beep Return NIL End Get Non Rice Training Set theme nr_ts theme MsgBox Choice FList Choose the Non Rice Training Set Theme Non Rice Training Set Selection Check for Cancel Exit if true If nr_ts_theme NIL then System Beep Return NIL End Get the field to use for validation FildLst Padd_theme GetFtab GetFields vidnFld MsgBox Choice FidLst Choose Field used for validation CROP Type Field Selection vidnStrng vidnFld AsString If vidnFid NIL then System Beep Return NIL End Make sure there is a feature selection on Paddock Theme PaddFtab Padd_Theme GetFtab OrigPadaBit PaddFtab GetSelection Clone If OrigPaddBit Count 0 then System Beep MsgBox info No Paddock Theme features selected please try again ERROR Return NIL End Get Index GRID theme to summarize IndexLst MsgBox MultiList GrList Choose the Index Grid Theme s to Summarize Index Selection CSIRO Land and Water 45 Check for Cancel or no selection Exit if true If IndexLst Nil or IndexLst Count 0 then System Beep Return Nil End Run ZonalStats on Index Grid for paddocks theme Export paddock theme to a new shapefile so it can be joined with zstats later For each Index_ Theme in IndexLst p_basename Index_Theme GetGrid GetSrcName GetFileName GetBaseName
52. s International Journal of Remote Sensing 24 885 890 CSIRO Land and Water 41 Appendix A ORDER FORM LANDSAT 7 ETM DATA ZN Australian Centre for Remote Sensing ACRES moe o ooo o o o o S a PLEASE ee babes BUSINESS AND INDUSTRY SECTOR AND APPLICATIONS FROM THE OPTIONS BELOW Business Sector 7 Brea emne eray Body hesto ony Overniment Private Caen g0 5 Appicsions dustry Sector AgnoutureS Fisnenes Conservation amp Envronmen Deine Emergency SEVE aston ony Li Education Research OU Pisving 5 Land Deveooment U Geosciences mee Li Land nitrmstion Soecisics ORGS PIG C S UU Trensporstion L Otter please speci ji aS ake E E aeeone i Facne J tys Emengency Disaster may C Expiraton ieee ca amp Logistics Managemer __ ei Graphic Presentation as recess ON iap Production ONS A ri JORET Assessment L Ores Wore PLEASE ENTER YOUR PREFERRED dtl ic orate THE PARAMETERS BELOW Saetiteseneor __ _ eens lt tr we _ o ENOTE aoe dae s TE SECTION A RAW DATA OR PATH IMAGE Select emer RAW DATA OF RAW IMAGE OF PATHIMAGE eee __ pasm paor i ween LIe LO FRAMING L Ener atte on path cerai Te pa A SECTION B MAP ORIENTED OR ORTHOCORRECTED IMAGE seecterer MAP PI CENTE OF wim WINDOW SCENE SIZE f fe aus a PHOTOGRAPHIC PRODUCT PRINT BE Transparency Orech uay Eje STANDARD 25m data requested ackno
53. s shp Water_class shp _ 02_03_padds_to_c 02_03_validation s p CSIRO Land and Water 36 These two paddocks had mean values well above the threshold both greater than 375 and are most probably not rice that is they were heavily vegetated in October Because of this the final shapefile representing the classification of rice should be saved out with these adjustments To do this from the Table GUI click on the Edit Switch Selection option and then from the View GUI click on Theme Save to Shapefile This selects all the other paddocks except these two which have been classified as winter cereals and then saves them to a separate shapefile Remember to use a name for the output that makes sense to you have used the name 02_03_classif_rice_wc_mask shp And as you can see below these two paddocks are not included in my new shapefile 2 ArcView GIS 3 2 DEK Fie Edit View Theme Analysis Surface Fe Window Help E 5 Se ALS 4 zz Oe CE w w F 7 5 Scale 112616 ga BOOT _ Octndv5 shp _ Theme21 shp _ Octndv4 shp _ 02_03_clsssif_rice shp _ Nov_bn15 shp a _ 02_02psdds shp Ol _ Water_class shp _ 02_03_padds_to_classify shp _ 02_03_validstion shp If these paddocks the ones that have been removed coincide with the original paddocks that were used for validation i e accuracy assessment then our estimation of accuracy will now need to be reassessed In this case
54. sion of our accuracy measurement is good enough to detect changes of 1 67 1 60 That is the smallest increment of change in our accuracy estimate is 1 67 so if one paddock out of the 60 is misclassified our estimate will be 100 1 67 or 98 33 if 2 are misclassified it will be 100 3 33 or 96 67 e The known paddocks should be put into a shapefile These known paddocks should also have current boundaries Therefore the boundaries might need updating for details see section 5 regarding updating paddock boundaries e Also it is good to make sure that the attributes of this shapefile are in order For the subsequent classification accuracy assessment below a field of crop types using the code of R for rice and NR for non rice is required In my shapefile this field is called Crop2003 see the figure below It is not a bad idea also to keep track of which areal photos the paddock was digitised from who digitised it and other details as you see fit e g variety if known etc CSIRO Land and Water 13 ArcView F O X File Edit Table Field Window Help 40 of 73 selected Be Z View1 N il 02_02_known_padi Nov_bnd5 i f yi T Attributes of 02_03_known_padds shp hye a Sue te 2w aw eee Te Polygon 2001_2002 lyg Z Polygon 793 2001_2002 f CICL Staff 241250 0000 s 4 i i g Polygon 2001_ 2002 i CICL Staff 565625 0000 Polygon 2000_2001 TVN 51
55. t time period growing season is essential Normally drastic differences to paddock boundaries do not occur from year to year at the CIA which means that keeping the GIS boundaries up to date is not a huge job However the more common changes where parts of paddocks e g rice bays are alternately included or omitted can be enough to alter the mean statistics of the paddock enough to misclassify the paddock Therefore these changes need to be updated in the GIS paddock boundary every year or the resulting accuracies will not be as high as expected There are a number of ways that this updating of paddocks could take place so this section is not really a definitive recipe book for updating paddocks The methods introduced here can be altered as need be The methods we used concentrated on updating ONLY the boundaries of those paddocks that were suspected of being rice Since this included any paddocks that had any standing water it was a pretty safe starting point 1 First flooded areas on the November ETM image are defined using per pixel classification of band 5 1650nm and subsequently converted to an ArcView shapefile 2 Once these areas suspected of being ponded water i e rice are defined they are then compared to the previously digitised paddock boundary GIS dataset 3 Next a decision is made about whether or not each of these paddocks needs to be updated using the following rather subjective rule e f one of the pre
56. this case ETM bands 3 red and 4 NIR into ArcView calculating and applying the geometric offset it should be the same for all bands within the October image and defining two more training sets one winter cereal and one non cropped training set Please see the previous notes on all of these steps The steps for doing this follow the methods for identifying rice so they will only be covered very briefly The steps include Calculation of NDVI on an October ETM image e Define small training set of known winter cereal and non cropped paddocks and determine threshold for classifying winter cereals from the October NDVI image Calculate mean NDVI for each paddock using the sum_grid ave program and e Identify and eliminate paddocks previously classified as rice which are also classified as having a winter cereal crop these are most likely not rice 7 2 Calculating the NDVI on an October ETM image Using the techniques discussed previously ETM bands 3 red waveband and 4 NIR waveband of the October image should be imported into ArcView and a different geometric offset should be applied Note the steps for calculating the geometric offset will be the same but will need to be redone for the different image before we were using a November image Then convert both of these bands to GRIDs as before and add them to the view The formula for the NDVI is NDVI NIR red NIR red or for ETM bands CSIRO Land and Water 28
57. tremely useful because they provide estimates of cloud cover and a quicklook image of the entire scene In this section we describe the steps necessary to access the ACRES website and how to define imagery for purchasing 2 2 ACRES website Go to the ACRES website http www ga gov au acres Click on the Digital Catalogue link on the left hand side of the homepage Click on the Please Log In link in the green tab along the bottom of this page Click on the Guest User radio button and Click Connect Now that you are into the main search page the first thing to do is to close the Location tab so your screen looks like the picture below This is so we can refine the search ACRES Digital Catalogue Where amp When K Meleon Bile Edit View Go Bookmarks Favorites Hotlist Layers Help 7 Aaas Layers ACRES Di URL http acs ausig gov au cg bin user server41 informs ontact Us Media News Vay GEOSGCIENGE regis SAE IE SS AUSTRALIA oe onn rats Man Dmahases Roine Resuns cart LEI Use the Zoom menu and select a point on the map to magnify and or center a specific feature or landmark The two icons to the left of this menu increase decrease the view magnification by a factor of two keeping the same center point The Earth icon below this sets the view to its minimum magnification usually the whole world The ring of arrow icons on the left shift the cent
58. ved by line bil band interleaved by pixel bip and band sequential bsq image data These images however require a standard header text file specifying image parameters To demonstrate we will be using Band 5 central wavelength 1650 nm for the late November 2002 image you purchased as this will be the main input into the rice non rice classification later This image was in bsq format but your image will likely be in bil format Whenever you see a bsq image in the pictures in this document you will need to be aware that this also refers to your bil image This requires a few steps e Copy the image data from the ACRES CD to a local directory The imagery comes as band interleaved by line BIL This format allows one file to contain the data for all bands of reflectance e Next open the template header file using a text editor e g Word Notepad The contents of the template header file should look something like this ArcView Image Information NCOLS 2722 NROWS 2080 NBANDS 1 NBITS 8 LAYOUT BSQ BYTEORDER SKIPBYTES 0 MAPUNITS METERS ULXMAP 351712 500 ULYMAP 6168837 500 XDIM 25 00000 YDIM 25 00000 CSIRO Land and Water 7 Where NCOLS NROWS NBANDS and NBITS represents the number of columns rows bands and bits represented in the image LAYOUT is the file format ULKMAP ULYMAP are the upper left x and y coordinates in map space and XDIM and YDIM are the pixel size in map units i e metres e Now y
59. vember image is used to identify ponded areas i e rice while the October image can then be used to refine this classification of rice by eliminating confusion between the previous winter cereal crops and the current summer rice crop In the 2003 2004 growing season only one image was purchased near the late November temporal window as it was decided by CICL staff that the optional October image would not be used CSIRO Land and Water 6 3 Rudimentary Geometric Offset 3 1 Introduction The ACRES map oriented data is provided with a geometric model already applied to it and results in reasonably accurate positioning e g within 50 to 100 metres within the projection defined in the selection and purchase of imagery e g AMG66 see above However when comparing any subsequent paddock boundaries to the imagery especially if more than one image is considered it is advisable to have a better correction that what is provided The geometric model applied by ACRES results in good internal geometry which means that a simple x y offset can be applied to the imagery in order to nudge the image into place This nudge is easy it only requires a positionally accurate GIS dataset as a reference ArcView and a text editor In order to apply the shift we will cover e Importing images into ArcView and e Comparison of imagery to DGPS roads GIS layer 3 2 Importing Images into ArcView ArcView recognises uncompressed band interlea
60. vious year s rice boundaries matches well with the area of ponded water defined from step 1 above then this boundary can be used in the classification e However if the previously defined boundary and the classification from step 1 above provides a poor match then the boundary will need to be updated Ultimately these boundaries need to be of high accuracy which means digitising from previous aerial photos or by using a GPS unit in the field However if the analyst chooses the initial digitisation of these changed paddocks can occur from the satellite imagery which allows for an approximate boundary for classification CSIRO Land and Water 19 These approximate boundaries should be identified separately so that the ones that are classified as rice can be updated in the future from either a previous aerial photograph or by using a GPS in the field When the difference is small for example due to a missing rice bay in a paddock this defines a change that can probably be fixed from past aerial photos When the change is too big to be identified properly from previous photos someone will need to update the boundary using a GPS unit in the field Remember the date of update who recorded the GPS data and who updated the GIS boundary could be recorded in the GIS paddock boundary file These three steps are outlined below 5 2 Defining flooded areas on the imagery The purpose of this initial pass is only to assess wh
61. within about 1 to 2 metres of the centreline of almost every road within the main CIA boundary so they provide lots of reference points to compare the placement of our imagery The roads shapefile is also in AMG66 For details of this dataset see Van Niel and McVicar 2002 This step requires adding the GPS roads theme to the same view that the imagery is in changing the colour to something easily seen and then altering the ULXMAP and ULYMAP values in the image header file until you are satisfied with how the roads lines up with the image Please note the match should be good throughout the CSIRO Land and Water 10 entire main CIA boundary Also be careful not to mix up canals and side roads with the main roads which are represented in the shapefile e Add the GPS roads shapefile to the view e Zoom in to an appropriate scale make sure the view map units are set to meters like about 1 20000 and usually start somewhere around Anderson and Frazer Roads but this is arbitrary e Ascan be seen below the ACRES correction is not too bad but the roads don t quite line up It looks like the Y direction is worse than the X e Change the coordinates until you are happy with the match over the whole CIA Note you will have to delete the theme and add it again to the view each time you change the header file coordinates or ArcView will not make note of the change to the coordinates Also it probably doesn t make sense to be more precis
62. wledge that have been provided with a copy of the end user licence terms applicable to these data products and am aware that these terms may be inspected at www auslig gov 2u scres referenc licences htm The terms are accepted by the end user End usersis Signature ACRES is a business unit of AUSLIG Department of Industry Science and Resources Scrivener Building Dunlop Court Fern Hil Park Bruce ACT 2617 POBox 28 Belconnen ACT 2616 Tel 02 6201 4107 Fax 02 62014199 Email gcres austig gov ay CSIRO Land and Water 42 Appendix B DiskFile r_nr ave Rice_non rice Programmer Tom Van Niel Orgainization CSIRO Land and Water Created 24 June 2003 Revisions Function Summarise accuracy of rice non rice classification based on a validation paddock shapefile and two training paddock shapefiles one rice and one non rice Outputs f to an excel comma separated text file Reports overall amp Kappa accuracy amp the threshold defined from the training i sets Called By View GUI Notes Assumes 4 input files 1 validation paddock shapefile 1 rice paddock training shapefile 1 non rice paddock training shapefile and at least 1 grid The mean grid value is calculated for each paddock in the shapefiles The average of the two training sets is used as a threshold for classifying rice vs non rice Note there also must be a common field in all of the shapefiles i called Id This f
63. y loaded The script is included in Appendix B as a backup of the original file and can be cut and pasted into the script window directly The script depends on a common field defined by the user in each of the three shapefiles This common field must contain an R in it for any paddocks that are rice and an NR in it for any paddocks that are non rice Currently if the field has something different then this script will need to change Also make sure that these shapefiles all have unique ids or the summary may not be correct For example when a new paddock is added to a current shapefile the attrioutes of the new record will be set to zero if 5 new paddocks are added and their ids are all left zero the last four will be classified the same as the first rather than each separately resulting in a false estimate of the accuracy reported by this script see Appendix B e Run the script as shown below Choose the validation paddock shapefile This works off of the selection so if you want all the paddocks to be considered within this shapefile then they should all be selected Choose the Paddock Theme RICE training set Selection Choose the Rice Training Set Theme 02 03 r tmshp Then the non rice training set amp Non Rice Training Set Selection Choose the Non Rice Training Set Theme 02 03 nr tm shp Cancel CSIRO Land and Water 16 Then choose the field that contains the R and NR information
64. y numeric fields tmpFldList PlygnFTab GetFields theFldList For each Fld in tmpFidList If Fld IsTypeNumber then theFldList Add Fld End End theFld MsgBox List theFldList Choose Field to summarize Summary Field Get Index GRID theme to summarize CSIRO Land and Water 52 IndexLst MsgBox MultiList GrList Choose the Index Grid Theme s to Summarize Index Selection Check for Cancel or no selection Exit if true If IndexLst Nil or IndexLst Count 0 then System Beep Return Nil End Run ZonalStats on Index Grid For Each Index_Theme in IndexLst aFN av GetProject GetWorkDir MakeTmp zstat dbf zoneField PlygnFtab FindField theFld AsString If zoneField NIL then Return NIL End zt Index_Theme GetGrid ZonalStats Table PlygnFtab aPrj zoneField FALSE aFN if zt HasError then System Beep MsgBox info Problem calculating zonal stats please try again or contact author ERROR Return NIL End ztJoinFld zt FindField theFld AsString If ztJoinFld NIL then System Beep MsgBox info ztJoinFld not found please try again or contact author ERROR Return NIL End Temporarily join Shape file s FTab to zt PlygnFtab Join zoneField zt ztJoinFld Export joined FTAB to a new shapefile basename Index_Theme GetGrid GetSrcName GetFileName GetBaseName newPlygnFtab Plygn_Theme ExportToFTab av GetProject GetWorkDir MakeTmp basename shp Add new Plygn shapef

A user's guide for identifying rice paddocks using GIS and remote

Contents

Download Pdf Manuals

Related Search

Related Contents