assemble training data from hydraulic model rasters

[skylerlewis]

For each training site (Yuba, Deer, etc.):

• arcgis: Based on RAS geometry, create high resolution polygons delineating each COMID reach across the model domain
• Extract depth, velocity, cover as multiband raster for the model domain
• Calculate pixelwise suitability based on multiple approaches:
  • simple: 1 if within required depth range and velocity range, else 0
  • weighted: calculate suitability scores (between 0 and 1) for depth, velocity, and cover (see Yuba report Table 8), then combined using formula [geometric mean of depth and velocity scores] * [cover score]
• Sum the suitability raster, for each COMID reach polygon, as percent of total area.
• Assemble table of % Suitability by COMID; this is our response variable for the model

Yuba

• [ ] Awaiting data from CBEC

Deer

• [x] Obtain model files
• [x] Export land cover
• [ ] Export depth and velocity (in progress)

Stanislaus:

• [x] Fix the couple COMIDs that do not have IDs; re-run and export
• [ ] Figure out class definitions for land cover layer (we have IDs and Manning's n values) -- asked Paul too b/c he developed model

San Joaquin:

• [ ] Prepare spatial data (in progress)
• [ ] Locate cover data (note that cover layer is very simple on this model, only 2 classes n=0.035 and n=0.045)
• [ ] Add to Rmd (basically exact same processing steps as Stanislaus b/c also a SRH2D model)