Mapping Regional Inundation with Spaceborne L-Band SAR

Shortly after the launch of ALOS PALSAR L-band SAR by the Japan Space Exploration Agency (JAXA), a program to develop an Earth Science Data Record (ESDR) for inundated wetlands was funded by NASA. Using established methodologies, extensive multi-temporal L-band ALOS ScanSAR data acquired bi-monthly by the PALSAR instrument onboard ALOS were used to classify the inundation state for South America for delivery as a component of this Inundated Wetlands ESDR (IW-ESDR) and in collaboration with JAXA’s ALOS Kyoto and Carbon Initiative science programme. We describe these methodologies and the final classification of the inundation state, then compared this with results derived from dual-season data acquired by the JERS-1 L-band SAR mission in 1995 and 1996, as well as with estimates of surface water extent measured globally every 10 days by coarser resolution sensors. Good correspondence was found when comparing open water extent classified from multi-temporal ALOS ScanSAR data with surface water fraction identified from coarse resolution sensors, except in those regions where there may be differences in sensitivity to OPEN ACCESS Remote Sens. 2015, 7 5441 widespread and shallow seasonal flooding event, or in areas that could be excluded through use of a continental-scale inundatable mask. It was found that the ALOS ScanSAR classification of inundated vegetation was relatively insensitive to inundated herbaceous vegetation. Inundation dynamics were examined using the multi-temporal ALOS ScanSAR acquisitions over the Pacaya-Samiria and surrounding areas in the Peruvian Amazon.

This two-part data recipe is for users who wish to map regional inundation with spaceborne L-band synthetic aperture radar (SAR). Users will create a) Colorized inundation map b) Inundation animation

A) Background
The all-weather ability of synthetic aperture radar (SAR) to penetrate cloud cover and low-light conditions to acquire imagery of the Earth's surface is well known. Given the recent availability of high-resolution terrain models at 30 meters from the Shuttle Radar Topography Mission (SRTM) and an interest to make SAR data easier to use, the Alaska Satellite Facility (ASF) now offers radiometric terrain corrected (RTC) L-band image products from ALOS PALSAR data. At longer wavelengths such as L-band, SAR can penetrate surface vegetation including the Amazon rain forest, and flooding events can be mapped very accurately at high resolution. With the availability of RTC data, it is now relatively easy to do accurate flood mapping using PALSAR data. In this recipe, we describe methods used to (a) create an inundation image animation from 16 PALSAR RTC products and (b) produce an inundation map that quantifies the number of flooding events from the given set of data. Note: Raster Calculator is part of the Spatial Analyst toolbox, which requires the Spatial Analyst license c) Create output folder HH_db, and specify output filename as Rnnnnn_db.tif where nnnnn is orbit number. d) Compute db = 10. * Log10 ("RTC_HH_granule") ( Figure 1) e) Repeat process for all RTC granules. Process and Classify Images 1. Filter Images to reduce speckle a) In the ArcMap top menu, go to Windows > Image Analysis b) Select the first image. c) Under "Processing," click on "Add Function" (Figure 2). d) In the "Function Template Editor," right-click on "Identity Function," mouse-over "Insert Function," and select "Speckle Function" from the menu. e) In the parameters menu that appears, change the "Filter Type" setting from "Lee" to "Enhanced Lee" and keep the rest of the parameters set to their defaults. f) The files that result from this are temporary. To save them permanently to the hard drive as a GeoTIFF, use the Copy Raster (Data  3. Display images a) Right-click on a feature layer and select Properties. Then click on the <Symbology> tab. Under "Show:" select "Classified." Then, under "Classification," set Classes to "2." b) Click the <Classify> button, and a histogram will display. c) By default, "Classification Method" is "Natural Breaks (Jenks)." Two vertical blue lines appear on the graph, and you should notice a small peak and a large peak indicating a bimodal distribution. d) Set "Classification Method" to "Manual," and specify a starting break point, such as "-14." Click <OK> (Figure 3). e) Under "Color Ramp," double click on the first class Color box and select a blue color. f) Double click on the second class Color box, select black, and click <OK>. The feature layer will be displayed with open water as blue and land as black.

B) Materials for generating inundation animation
Notes: The values for the break points can change significantly between locations and datasets, and will generally require several iterations to determine.
4. Refine the breakpoint value until water is mostly contained within river channels, and very small number of blue dots show on land outside river channels. a) Right-click the image layer, and select "Properties." b) Click on the <Symbology> tab; then click on "Classify…" c) Change the value of the breakpoint, increasing it by one each time (so "-13" is the first increased value). d) Click <OK>. e) Again specify the two colors for the two classes as blue and black; then click <OK> Notes: We found "-11" to be the best value for minimizing water speckle over land.
5. Further refine breakpoint values a) Determine a second breakpoint for areas in the image that are brighter due to double bounce of flooded forests. Here you will determine a breakpoint on the other side of the large peak. b) Start with a higher value such as "0." c) Set the first class color box to black and the second class color box to yellow, for values greater than the breakpoint. Look for yellow in the image and notice how much speckle noise you have away from the bright feature areas. d) Decrease the breakpoint value by one until the amount of speckle is close to what was observed for the blue class.
Notes: We found "-4" to be a good value for the second class e) Select number of classes to be "3," and enter both breakpoint values. Set the first class Color to blue, second class Color to black, and third class Color to yellow. You should see the image classified with water channels as blue and double bounce areas as yellow.
Notes: We found it useful to pick a minimum flood and maximum flood image to verify that the breakpoints work as desired. Some knowledge of the extent of seasonal flooding is useful to confirm that the results are correct.

Save a colorized raster dataset for each classified image for creating an animation. a) Use the Apply Symbology From Layer tool (Data Management Tools > Layers and
Tables Views) to set the symbology of the remaining greyscale images, with the "Symbology Layer" field set to any correctly colored image. b) When each image is colored correctly, right click on the image and select "Data -> Export Data..." This will display the "Export Raster Data" window. c) Set "Extent" to "Data Frame (Current)." d) Set "Output Raster" to "Use Renderer" and "Force RGB." e) Specify an output directory by clicking the folder button; then browse to the desired output directory. f) In the "Name:" field, specify the output filename, such as class_xxxxx_db.tif, and "Format:" "TIFF. g) Click the <Save> button.

Generate and View Animated GIF
1. Generate animated GIF to show the sequence of classified images showing inundation a. Open Adobe Photoshop b. Select "File -> Scripts -> Load Files into Stack…" to load the color classified imagery. c. Use the <Browse> button to navigate to the directory you used for the "Export_Data" step in ArcMap (step 6). Select all input files. The data files will load in the layers control panel. d. Once the data are all loaded, click on the Create Frame Animation button in the <Timeline> tap section located below the image. This will create a single frame instance. e. In the bottom right corner of the image display window, click on the menu button that looks like an arrowhead next to four horizontal lines. A timeline menu is displayed (Figure 4). f. In the timeline menu, select "Make Frames From Layers" to populate the timeline with all image frames. The frames are numbered 1 through 16. There is a delay setting below each frame. g.
Notes: It is preferred to set the delay to increments of 0.5 seconds per cycle of PALSAR data. Each cycle is 671 orbits; if there are gaps in the temporal coverage, then a longer delay can be set as a multiple of 0.5 seconds. In our case, we set delays of 0.5, 0.5, 1.5, 1.0, 0.5, 0.5, 0.5, 1.5, 2.0, 0.5 1.5, 1.0 0.5, 0.5, and 0.5 respectively. The delay of 1.0 was when one cycle was skipped, 1.5 for when two cycles were skipped, etc.

Figure 4: Location of the timeline menu (red arrow) and "Make Frames From Layer" (blue arrow) in Photoshop
h. Click on the "0 sec." dropdown menu under each frame, which offers a choice of delay settings. If the specific delay you want is not listed, select "Other" and enter the delay. i. If you want the animation to loop more than once, click on the <Once> button located below the timeline images and select "3 times," for example . j. To save this animation, select "File -> Save For Web." Notes: A warning may be displayed due to the size of the images; click yes if it does. In the image size section, type 15% to reduce the image size of the output.
k. Click the <Save> button, specify the output filename, and click save again to save the animated GIF image.
2. Using your file browser, navigate to the output directory containing animated GIF. Double click on file to execute. Your animation should be displayed.
UAF is an AA/EO employer and educational institution and prohibits illegal discrimination against any individual: www.alaska.edu/nondiscrimination

D) Output Example of Inundation Animation
Click the following link to see the resulting gif in your browser: https://www-dev.asf.alaska.edu/data-tools/data-recipes/how-to-map-inundation/  Notes: The combined image will display with random colors assigned. It is a good idea to set the color for 0 and 1 to black. Usually values of one are due to remaining speckle in the data after filtering. You can change colors by right-clicking on a color in the layers panel, which brings up a matrix of standard colors. Select colors that are easily distinguished and in some logical order. g) When you are happy with the color settings, right click on the filename and select Data/Export Data. An export raster data panel will display. h) Set Data Frame (current) for Extent. i) Click check box for Use Renderer, and Force RGB. j) Specify an output directory, click the folder button, and browse to the output directory for this classified image. k) Also specify the output filename, such as overlap_count.tif, and format: TIFF. l) Click the save button.
Notes: This creates a full resolution image of inundation. You should also use the snipping tool to grab a low-resolution version of the overlap count image and the color key to be used with either image.