Hello, InVEST community! 
I’m using a calibrated NDR model to compare N retention in riparian buffers and N export from the watershed in two scenarios: current (C) and increased buffers (S2). After looking at the results, I was perplexed as to why S2, a scenario with 80% more buffer area, had a cumulative lower N retention (kg).
I followed a recommendation from a previous post to calculate N retention. As recommended, I used the modified load raster minus the N export raster to calculate the N retention for both scenarios. Then I isolated the riparian buffer pixels and summed the values of all pixels to get kilograms of N retained in riparian buffers.
After getting perplexing results, I reread the hyperlinked post and now think that I may have misinterpreted it. I assumed that the calculation modified load raster minus the N export raster would give me the total amount of N retained in riparian buffers pixels, that is, per pixel N loads and the N leaching to riparian pixels from nearby agricultural fields.
Is it the case that doing the above raster calculation will only get me the per-pixel N retention? If so, is there another way to calculate the total N retention by riparian pixels?
Disclaimer: I know that the NDR model is not necessarily designed to do an in-depth analysis, but my PI is interested in testing its application 
Thank you!