Documentation for semi-automated wetland gain, loss, and type change detection. Poley 10/20/2021

Pre-processing:

1. Normalize worldview images
   • normalize_WV.py
   • Author: Mike Billmire
   • Purpose: Normalize worldview images that cover the same extent to each other
   • Inputs:
     • Reference WV image
     • WV images to be normalized
     • PIF.shp (shapefile in same coordinates as images. White and black features. Shapefile needs a ‘pif’ attribute.)
   • Output:
     • New directory with normalized images including reference image
2. Radiometric change
   • wv_CVA.py
   • Author: Andrew Poley
   • Purpose: calculate change magnitude, angle, and tasseled cap greenness (TCG)
   • Inputs:
     • WV images from year 1 & year 2
     • Output name and directory
   • Output:
     • Single image with 3 bands (magnitude, angle, and TCG)
3. Image segmentation
   • Segmentation.py
   • Author: Andrew Poley
   • Purpose: segment radiometric change images using scikit-image SLIC segmentation
   • Inputs:
     • Output image from radiometric change code
     • Output directory
   • Output:
     • Segmented image with mean values for each segment. 4-band image: mean-magnitude, mean-angle, mean-TCG, and segment number
     • NOTE: code takes a very long time to run on larger areas. Run ‘tile_raster.py’ on radiometric change image before running segmentation to speed up processing.
     • Tile raster code needs input image and output directory for image tiles and will create image tiles. Input tile directory into segmentation code. Run ‘mosaic_tiles.py’ on segmented tile outputs to re-mosaic the image for the next step.

Wetland Gain/Loss/Type change

1. Wetland_gain_loss.py

   • Author: Andrew Poley
   • Purpose: Calculate wetland gain, loss, and type change from segmented radiometric change
   • Inputs:
     • Mosaiced segmented radiometric change image
     • Change thresholds:
       • CVA magnitude used to determine change/no-change
       • TCG_upper is used to determine wetland gain
       • TCG_lower is used to determine wetland loss
     • Land cover classification of year 1
   • Output:
     • Classified map of wetland gain, loss, and type change
       • -1 = wetland loss
       • 1 = wetland gain
       • 100 = wetland change type

   NOTES:

   • Brief code description
     1. Use provided thresholds to determine change/no change in CVA image
     2. Reclassify land cover classification to only include wetland classes
     3. Intersect change/no-change and wetland classification to estimate which wetlands are changing
     4. Intersect wetland change/no-change with thresholded TCG layer to get gain, loss, and type change
     5. Reclassify change into gain & loss
   • IMPORTANT:
     • Input radiometric change and classification images must be in the same coordinate system and must have the same number of rows and columns.
     • Assumes input classification has been reclassified into the following classes/numbers: 1 = urban 2 = suburban 3 = barren land 4 = agriculture 5 = grasslands 6 = deciduous 7 = evergreen 8 = shrubs 9 = woody wetland 10 = emergent wetland 11 = floating aquatic 12 = water 13 = detritus 17 = typha 18 = phragmites