I am currently working with the InVEST Nutrient Delivery Ratio (NDR) model in a predominantly flat watershed, and I have a conceptual question regarding the interpretation of pixel-level nitrogen retention. When estimating retention as the difference between modified_load_n and n_total_export, I observe that land uses with high nitrogen loads, such as croplands, appear as the areas with the highest nitrogen retention, while natural land covers show lower values. This raises questions about whether this pattern is ecologically meaningful or simply reflects that the metric represents absolute retention (kg) rather than retention efficiency. This issue is particularly relevant because I am attempting to use pixel-level retention to identify priority protection zones, and I am unsure whether this approach is appropriate in such a flat landscape or if relative metrics or land-use–aggregated indicators would be more suitable. I would appreciate any insights or similar experiences.
Hello @Fernbg77 -
In the NDR model it is generally the case (even in sloping land) that where more nutrient is produced, there is correspondingly more nutrient to retain. This model is very simple, and relies on mainly loading and efficiency values, without a lot of other nuance. It also currently doesn’t provide an explicit retention calculation (but we have been working on it), so the retention estimation of taking the difference between modified_load_n and n_total_export is intended to illustrate the patterns of high and low retention more than the absolute values.
So i would agree that it is most appropriate to use relative metrics for identifying priority zones. Also a reminder that in general the absolute values of models are only really to be trusted if you’ve done a calibration exercise to validate the model results against real-world data.
~ Stacie
Thank you very much for your response. It has been very encouraging for my research.
Would an appropriate relative metric be (modified_load_n − n_total_export)/modified_load_n?
Additionally, for the other ecosystem service modules used to identify priority areas, such as SDR, which already provide a retention-related output, would it be appropriate to use that output directly, or would you recommend applying a relative metric as well?