Validation: what happens when extent shp is larger than raster

[cecileherr]

niche_vlaanderen version: dev 1.1 Python version: 3 Operating System: Win

I tested the validation module with a vegetation shapefile (bwk) with a (slightly) larger extent than the input rasters of the model and I think there might be a problem in this case.

• With the calculation of polygon_count: My guess is that polygon_count is based on the full extent of the shp while it should be based on a smaller extent, that is: the extent overlapping with the input rasters.

• With the calculation of the scores When I use the same bwk in 2 examples, but one with a version clipped to the extent of the rasters and the other one with a larger extent (and polygons extending outside the extent of the rasters), I get unexpectedly very different validation scores. Are zones with NoData (= outside extent niche model) considered as zones with no potentialities instead of NoData?

Could you check these points please?

Example based on Brasschaat + shp in attachment (you will need to change the paths for the bwk though) bwk_fake_brasschaat.zip

simplebr = nv.Niche()
simplebr.run_config_file(path_nv + "tests/data/bwk/niche_brasschaat/simple.yaml")

from niche_vlaanderen.bwk_overlay import NicheOverlay

# with extent = same as rasters
no_simplebr_fkok = NicheOverlay(niche=simplebr, map= path_ch + "testcase_extra/brasschaat/bwk/bwk_fake_extentok.shp")
no_simplebr_fkok.overlay()

no_simplebr_fkok.summary

# with extent = larger than rasters
no_simplebr_fkl = NicheOverlay(niche=simplebr, map= path_ch + "testcase_extra/brasschaat/bwk/bwk_fake_extentlarger.shp")
no_simplebr_fkl.overlay()

no_simplebr_fkl.summary

[johanvdw]

In this case: the second polygon (half covering the niche model): the area of the polygon is 600 ha, but the overlap with the map is only 70% so the area used for area_effective should be reduced. In the last polygon (outside the niche domain) effective area should become 0.

• area_pot and area_nonpot should be unaffected (based on zonal statistics)
• area_pot_perc is unaffected as only based on area_pot and area_nonpot
• area_nonpot_optimistc and area_pot_perc_optimistic will be adjusted because the effective area changes.

[cecileherr]

just as a reference in the future: this a better version of test shapefiles bwk_fake_extentok_mini.zip