RADAR Practical
Contents
1. and consequently why we need to correct RADAR imagery produced for the impact of terrain Figure displays some output from the process m River Road ___ Deforestation Events N cA ee L i 3 i l Figure 1 Identifying deforestation events in Malawi a 5m resolution optical SPOT imagery regions of higher biomass are visible as red and lower biomass as pale green Forest fire scars are dark green b estimated carbon stocks tc ha 2007 where darker green equates to higher biomass c estimated carbon stocks tc ha 2010 where darker greens equate to higher biomass and d Overlaid on each image are polygons black outlines that identify regions of deforestation Notice the large deforestation event to the North West of the image Explanation In a large number of field plots in Mozambique Miombo woodland biomass was estimated from tree count tree height and width These biomass values were then compared to the Sigma0 values given in an ALOS PALSAR image of the area which show the ratio of the power that is scattered from the ground to the power sent and a relationship was identified between the Sigma0 values and Biomass Here we use this relationship to estimate the biomass values from a Sigma0 image in Malawi characteristic of Miombo woodland in both 2007 and 2010 Once we have estimates of biomass for both years we can identify the changes in biomass between them The Process Figure 2 bel2. i pyre er l Ys wg NC MOZAMBIQUE Lilongwe s J ts Sda A i aes i E i i i i nan ZIMBABWE D 4 0 25 50 100 es kM Figure 3 Study Area The extent of the study area red on a base map of Malawi its districts yellow primary roads black and cities black Forest Reserves Because the University is involved in an on going contract looking at deforestation within Malawi we will have to limit the detailed analysis to a small region of the ALOS PALSAR scene We will focus only on the Michuri Forest Reserve situated to the North East of Blantyre city f l A Milare Forest ee i Rese me National Park l e l Chik vwawa p l Amalika Forest Reserve 4 Figure 4 Region of interest the Michuri Forest Reserve and its surrounding are Black on a base map of Malawi its districts yellow primary roads black and cities black Step 1 Calibration from ALOS PALSAR Digital Number Format to Sigma0 values and Terrain Correction The ALOS PALSAR product is in Digital Numbers These need to be calibrated to Sigma0 values that can then indicate the biomass on the ground These Sigma0 values then need to be corrected for both the geometric and radiometric terrain distortions layover shadow foreshortening that you have been introduced to This can all be done using a piece of software called MapReady produced by the Engineering group at the Alaska Satellite Facility par
3. that span between 0 and 2 Cells that have a value near 1 have not changed much between 2007 and 2010 neither a loss nor gain in Biomass Cells that have a value between 1 and 2 have gained biomass between 2007 and 2010 Cells that have a value between 0 and 1 have lost biomass between 2007 and 2010 Calculate the change map using the raster calculator Within the expression box enter Biomass Raster _2010 Biomass_ Raster 2007
4. 2007 232 QGI5S _ 2007 232 merge alos_chain DEM_test ASTER_sw s0 2007 232 ALPSR POS3696860 H1 1 A LTS _SIGMA PHASE HMtif alos_chain DEM_test ASTER_sw s0 2007 232 ALPSR a DAS Ieoee ww AIT CCMA AMD KF Pier RSF Oi TS Ser ULC a TS Sere pe OK Help ten cf tye GDL AM fies Cancel i S ALPORPOGSOED HLI_ A LTS SOMA FHASE HNO GOBAN HI mer BT kta cal Figure 14 a Accessing the Merge tool from the Raster Menu and b c Merging multiple Sigma0 images from one pass Only continue to this stage once you have merged the datasets from each pass 2 passes from 2010 and 2 passes from 2007 You should have executed the merging process four times The following two steps use a QGIS plugin called RasterCalc that can be found under the Raster Menu Step 4 Average Map Algebra This step involves averaging the two merged datasets from each pass To average the Sigma0 values from the two passes we first need to create a binary raster that acts as a mask that equals 1 where there are valid Sigma0 values in both images and equals 0 for all other areas Valid Sigma0 values will be above 0 0001 Use the relevant buttons on the raster calculator to create the expression shown in Figure 15a This expression essentially says find all locations that have a Sigma0 value above 0 0001 in Pass 1 and that have a Sigma0 value above 0 0001 in Pass 2 Those locations that have a value of 0 0001 or lower
5. in either one of the images are not of interest to us The output raster will be binary raster shown in Figure 15b a tilted rectangle of 1 s white surrounded by 0 s black 3 Raster calculator Raster bands Result layer QGIS_2010_240_merge 1 Output layer ham REDD GIS Data 2010 QGIS_2010_step1 OGIS merge_2010 2661 Current layer extent X min 627615 26957 720938 26957 Y min 86201018 64952 330243 64952 Columns 3973 5169 Output format GeoTIFF Add result to project Operators Raster calculator expression QGIS_2010 240 _merge 1 gt 0 0001 AND OGIS_merge_2010_286 1 gt 0 0001 Expression valid Ok Cancel Figure 15a the expression used to calculate the binary raster in the Raster Calculator dialogue and b an image displaying the binary raster This stage is required because the two consecutive passes within each year do not capture identical areas and we want to focus only on the area that is captured in both passes The next stage is to take the average of the Sigma0 values from the consecutive passes that lie in the regions of the binary layer that have 1s This is done by multiplying each pass by the binary layer and then dividing by 2 Make sure you understand what is going on here and when you do enter the expression into raster calculator that is shown in Figure 16 Raster calculator Raster bands QGIS_2007_232_merge2 1 QGIS_2007_278_merge 1 step_A 1 Result layer Output laye
6. method re pe 7 ie Figure 13 a Two separate Sigma0 images from the one satellite pass and b a single image created merged from the initial separate images To merge multiple datasets browse to the file location that the Sigma0 images are located Notice that there are 4 type of file for each image SIGMA AMP HH SIGMA AMP HV SIGMA PHASE HH SIGMA PHASE HV Select all the SIGMA AMP HV images Figure 14c Set the No data value to 0 Database Web Help Raster calculator Georeferencer A_1TS_ SIGMA AMP H tif Select AXAWAS Sai Output file OGIS 2007 _232_merge Select No data value 0 Interpolation Projections Conversion Extraction Analysis Layer stack Use intersected extent Grab pseudocolor table from the first image Miscellaneous Build Virtual Raster Catalog GdalTools settings Merge Creation Options Nme LAs HL Information a Build overviews Pyramids gu Tile index Remove ot the fies tes Merge io ses Ji Die ehan OEM seer ASTER m eth OO oO CUBE ke hy Computer arene ge tt Be tt serene cobb tl raed E ALRARMGSSO H1LI_ A Leta E ALPSRPIGSCSSO H1 A LTS SGA AN a aH A T A AAH BPP a LT P HL amp AUER Se LA LTI SOMA PHAE Hi E ALPSRPGSE 0 H A LTE SUPA aL A LTS SG Al et ee en ee amp ARREDATI LA LTS AASE RLG Load into canvas when finished gdal_merge bat n 0 of GTiff o alos_chain DEM_test ASTER_sw s0
7. F FE DEM holes with interp AC Refine Geolocation Oniy fe Apply Terrain Cormection Autornatically Mask C Mack from Fr LED ALPSEPOTGSSSET HLI A HH HW ALPSRPORESMEMO HId Ajpg LED AL GRP e H OA HH Hw ALPSRPOA DOOLA DEM File ZA DEMA ASTER aster sa TMF Ft E amoebae Sigma0 pales in locations of es Se Select Sigma from the drop down list This will convert the image from Digital Numbers DNs to units of Sigma 0 Sigma values represent the ratio of the power that is scattered from the Earth s surface to the power sent from the satellite This value provides an indication of the characteristics of the land surface Browse to the Digital Elevation Model used for the terrain correction Sometimes DEMs have holes where there is no data Mapready makes sensible guesses as to what these values are such that it can apply the terrain correction algorithm in these locations Even after geometric terrain correction the brightness of the slopes leaning towards the sensors is still too bright while those facing away are still too dark Radiometric terrain correction adjusts the brightness of the geometrically terrain corrected image according to the slopes layover or shadow Figure 8 Terrain Correction Tab The Sigma0 images are corrected for distortions caused by the terrain using a Digital Elevation Model Select Fill DEM holes with interpo
8. LPSRP083696860 H1 1_A HH HV ALPSRP083696860 H1 1_ A tif LED ALPSRP083696870 H1 1_A HH HV ALPSRP083696870 H1 1_ A tif LED ALPSRP090406860 H1 1_A HH HV ALPSRP090406860 H1 1_ A tif LED ALPSRP090406870 H1 1_A HH HV ALPSRP090406870 H1 1_ A tif term at mnannasnnnenen s wa a a rms ime ar Reman sen en s iwa a a are S Process a Stop Processing Figure 10 Export Tab Export the data as a GeoTIFF format and export all bands as separate images Show Completed Files The Sigma0 images are separated in the Data Folder according to the year that they were taken and the time of year they were taken the pass To begin we need to merge the Sigma0 images from each pass We will first import the images from each pass and then merge them Only import and merge images from one pass at a time or you it will become confusing which images belong to which pass and QGIS will slow down Step 2 Importing the Files into QGIS To import the GeoTiff files into QGIS from a pass select the Add Raster Layer icon Figure 11 and navigate to the location where the calibrated corrected image is saved to Figure 11 Add Raster Layer icon The Raster will appear as a plain grey image on the map canvas To scale the colour of the image according to the Raster values open the Raster properties dialogue via the Raster layer s menu Figure 12a select the custom min max values option Figure 12b set the min to 0 amp max to 0 5 Figure 12c and Stret
9. REDD QGIS workshop Using RADAR imagery to map Forests The purpose of this practical is to introduce the process of using L band RADAR imagery to map changes in carbon stocks in Miombo Woodland It centres on an area from Southern Malawi as a case study Specifically we will be looking at abrupt severe biomass loss events between 2007 and 2010 Within this practical we use the following definitions Forested Land Land with Biomass gt 10tC ha Abrupt severe forest loss events Land that was forested in 2007 but that has lost at least 50 of its biomass between 2007 and 2010 If performing this process in the future for your organisation it is important to a Carefully define forest in terms of biomass for the particular country project b Understand what extent of biomass change 1s of interest e g a loss of 50 or a loss of 10 By default due to the image processing required this practical will also introduce Working with Raster datasets in QGIS importing and setting their properties Merging geographically adjacent Raster datasets using QGIS GDAL plugin Using QGIS Raster Calculator to create new Raster datasets Reclassifying Raster Datasets Exploring Raster values within pre defined areas 1 e polygon features using QGIS Zonal Statistics plugin It assumes a basic knowledge of RADAR theory how RADAR 1s used to identify types of land cover amp land cover change how terrain influences the returned signal
10. ch to MinMax Figure 12d The reason for scaling the colour to values 0 0 5 is because most Sigma0 values lie within this bracket Layer Properties ALPSRP083696860 H1 1_A LTS SIGMA AMP HV w Zoom to Layer Extent Zoom to Best Scale 100 Stretch Using Current Extent Show in Overview e Remove Set Layer CRS Set Project CRS from Layer Properties Rename Copy Style Add New Group TI Expand All T Collapse All Update Drawing Order Use standard deviation Note Minimum Maximum values are estimates user defined or calculated from the current extent Load min max values from band Estimate faster Actual slower Current extent Default No Stretch Restore Default Style Figure 12 a Accessing the Raster properties dialogue and b c d scaling the colour of the image to the Sigma0 values The grey image on the screen will now change to an image with a full range of grey shades as those in Figure 13 to reflect the Sigma0 value for that part of the scene Step 3 Mosaic each pass The next step of the process looking back at the image processing flow chart in Figure 2 is to mosaic each pass to create a continuous image from the two adjacent ALOS PALSAR files see Figure 13 To do this we use the merge tool from GDAL plugin For QGIS version 1 8 0 this plugin should already be integrated into the Raster Menu See Figure 14a If not load install the plugin using the relevant
11. lated values Apply Terrain Correction Perform co registration Apply Radiometric Correction and Interpolate layover shadow regions Refer to the ASF User Manual online for more detailed explanation Set the Coordinate System ag 5 an Pte For Malawi use WGS84 Datum F Poel Sze Cat the nivel size 4 an zl _ UTM 36 Set the pixel size to 25m Set the resampling method to Bilinear LED ALPRRPIJGNASAJO HLI A HH HV ALPSRPOTEMZETO HI_Ajpg LEO AUP See 0 H LA HH AY ALPSRPOGGRSRR HOt Ajpg JI Green Berre Figure 9 Geocode Tab Used to project the data according to either a geographic or projected coordinate system and resample the output image to a user specified resolution E Shoes Completed File Summary Format CEOS L1 Data type Sigma Terrain Correction Yes i J P Geometric amp Radiometric correction Output data in byte format instead of floating point DEM aster_sw_UTM tif Sample mapping method Statistical 2 Sigma Hong aii geocode Export All Bands as Separate Images Intepolate Export RGB Image according to Polarimetric selection Export Multiple Bands in a Single RGB Image User Defined True Color False Color Red Band ae Y Green Band 4 Zi Blue Band Add ceos tert SD Browse inpul File LED ALPSRP076986860 H1 1_A HH HV ALPSRP076986860 H1 1_ A tif LED ALPSRP076986870 H1 1_A HH HV ALPSRP076986870 H1 1_ A tif LED A
12. ow outlines the processing chain that is used to create the biomass maps and change maps Take time to familiarise yourself with each stage of the process and ask questions if you are unsure of any stage Support different DEMs Enables various Radiometric and Geometric Terrain Correction Check contig Files for all necessary Sigma rasters Sort into hierarchical folder structure Imagery within year and pass Mosaic for each pass Larger Sigma image Inputs Calibrate images using semi pseudo invariant objects Calibrated Sigma0 image Convert to Biomass according to relationship between sO values and Miombo woodlands in Mozambique Ryan 2012 Biomass Rasters tC ha Create change maps Change Raster Thresholding Map Creation Allow Various Calculate Statistics Various Degradation Deforestation Statistics Figure 2 Flow chart to summarise the components of the image processing chain used to identify areas of deforestation The configuration file holds information about which settings should be applied in MapReady Worked Example Area of Interest The satellite scene used in this case study lies to the West of Blantyre City covering the Mwanza district entirely and parts of the Blantyre Chikwawa and Ntcheu districts 32 0 0 E 33 0 0 E 34 0 0 E 35 0 0 E 36 0 0 E 37 0 0 E TANZANIA i ji i i f S pi AEA gn _ i N et Swe y t cee
13. r st ASTER_swi s0 2007 QGIS 2007 average Current layer extent min 62040 7 47202 xMax Foe P92 47202 Y min 6201257 12563 Y max 6550457 12563 Columns 3973 Rows 5168 Output format Add result to project Operators Raster calculator expression QGIS_ 2007 232 merge2 1 step A 1 QGIS_2007_278_merge 1 step A 1 2 Figure 16 the expression used to calculate the average Sigma0 value from two passes using the binary raster to constrain the extent of the calculation to where both passes have relevant values The outcome of this will be a raster of mean Sigma0 values of the region For the 2007 passes it is saved as a QGIS_ 2007 average in Figurel6 Repeat this for each year i e average the two passes for each year Step 5 Conversion to Biomass To convert the Sigma0 values now the average of two passes to estimated biomass we use the best relationship identified between biomass in Miombo woodland in Mozambique and Sigma0 This is Biomass Sigma0 1517 18 To create a raster with estimated biomass use the Raster Calculator For each year make the following calculation Averaged Sigma0 Raster 1517 18 Step 6 Creating a change map Now that we have a Raster of estimated biomass in 2007 and another for 2010 we want to understand how much change there has been in the three year period between To do this we divide the 2010 Biomass Map by the 2007 Biomass Map This will create a Raster of values
14. ry Terrain Correction Geocode Export Select Processing Steps Ti Import Data required Temporarily keep intermediate files Keep intermediates Show full path names About MapReady Generate and show thumbnails Save Settings Load Settings Input Files Output Settings Destination Folder Add Output File Prefix or Suffix LED ALPSRP076986860 H1 1_A HH HV ALPSRP0O76986860 H1 1_Ajpg LED ALPSRP076986870 H1 1_A HH HV ALPSRP076986870 H1 1_A jpg LED ALPSRP083696860 H1 1_A HH HV ALPSRP083696860 H1 1_A jpg v 4 b Process All Stop Processing Her Figure 6 General Tab This tab is used to specify which processes we want MapReady to run which SAR files we want to calibrate and how where to save the data Select Terrain Show Completed Files Correct Geocode To Map Projection and Export to a Graphics File Format Browse to the location where the SAR files are saved and select the files These will appear at the bottom of the tab once loaded F Apply ERS2 Gain Coerection to ERS data LEAL Pee A HH H ALPERT O HI Apg LED ALPS eH A HH HV ALPS POE ORO H11_Ajpg GS Process All Gop Procepunmg Figure 7 Calibration Tab Show Completed Fis Uses the Digital Elevation Model to improve refine the geo location of the SAR image fem Poleimeny Terrain Correction Geocode Export
15. t of the Geophysical Institute at the University of Alaska Fairbanks NB This practical will simply summarise the software settings required to process the ALOS PALSAR imagery for biomass estimation in the southern scene of Malawi using an ASTER DEM It will provide a small selection of notes that describe the key processes that we want the software to run but it is HIGHLY RECOMMENDED to spend time reading the ASF Mapready User Manual in detail as the required settings will vary depending on the type of raw SAR imagery used type of project characteristics of the Digital Elevation Model location of SAR imagery and land cover type e g proportion of water bodies within the area of interest The following five images associated captions and labels will now introduce the settings required for Calibration and Terrain Correction of the ALOS PALSAR scene z p ASF MapReady Version 3 0 6 re ft i Summary Format CEOS L1 Data type Amplitude Terrain Correction Yes Geometric correction only DEM lt none gt Run External Program Interpolate Layover Yes Geocoding UTM Polarimet f d pol SLC data g FF paa A Nace p come Zone lt from metadata gt Iv Terrain Correct with Digital Elevation Model Datum WGS84 f Geocode to a Map Projection Resampling Method bilinear i Export JPEG byte V Export to a Graphics File Format fing ee Sigma Fe Keep no intermediate files General Calibration Etern Polarimet
Download Pdf Manuals
Related Search