My goal is to evaluate water purification on a global-scale with the InVEST NDR model. As known, this model requires a biophysical table with information pertaining specific LULC classes and its respective LULC raster.
I am currently using the HILDA+ LULC raster that basically distinguishes the globe into 8 classes, one of which is cropland. As of now, I ran the NDR model assuming that all the cropland in the world has the same nitrogen load. I took the nitrogen load value from the database given by InVEST.
The result of this exercise is that the total nitrogen load globally seems to differ by a factor of 10000, compared to a recent study by Chaplin Kramer et al 2019. Therefore, I tried to breakdown the LULC cropland into country specific cropland, for example: cropland Switzerland, cropland USA, etc. Each of these specific croplands have their own nitrogen load value. This value was calculated by taking an average of different nitrogen loads from a raster showing nitrogen input to agriculture.
This results an “expansion” of my initial LULC classes from 8 to almost 250 (based on the total number of countries globally).
From the nitrogen load raster, each pixel has different load values. So actually, it would be more accurate to use this raster as an input for nitrogen load for the NDR model. However, to my understanding, the NDR model only calculates by looking at a group of pixels, the group being the LULC class, instead of a pixel-by-pixel calculation. Hence, aggregating the nitrogen load raster into a specific region and then defining these regions and LULC classes is the only way to operate the NDR model.
Does my existing approach make sense? Is it possible to input nitrogen load information from a raster directly?
Thank you all for your time!