Hello! I am using ArcGIS Pro 2.6.2 and need to run a supervised land cover classification on multiple clips from 30cm x 30cm resolution rasters. I've created buffers around objects of interest and clipped rasters to be within those buffers.
I would like to create a training dataset that pulls values from those several rasters. Is there a way to do this that doesn't involve adding all of these rasters into a mosaic dataset and then copying the raster into a large TIF for analysis? This is the only way I've been able to run this thus far, and copying the raster often results in a file that is too large for my computer to process. When I try to run land cover classifications on the rasters separately, the attempt fails. I feel like there must be a more efficient way to do this!
I do this using Spatial Analyst, so it takes that license. You don't have to clip or buffer the rasters unless there is some other reason to do so. I like to use model builder. You could build one like this and then copy and paste it and replace the value raster with the successive rasters and change the field in the calculation to polygon field.
I'm sure there is a way to do this with an iterator (iterate through the polygons or something like that) and or in python, but the general idea would be the same: do zonal statistics to a table, the join that table with the polygon layer and use the calculate tool to transfer the statistics you want from the table to the polygon layer and then remove the join.
I believe that if there is more than one polygon in the polygon layer this will work unless they overlap. I think that If they overlap you will need to iterate through the features.