Interpreting the n_load parameter

Hello INVEST community! :seedling:

I’m looking for clarification about the n_load parameter.

I’m a little confused and unsure if there is contradictory information in the NDR documentation. I interpreted n_load as the N input associated with each land use type. For example, for agricultural land, N input would be the amount of fertilizer, manure, and atmospheric deposition. Locally, we have agricultural N use data, and I have used N inputs as the n_load without applying a correction factor to account for on-pixel nutrient retention. With this approach, and with modeling proportions of subsurface N, I’ve gotten a pretty good fit when calibrating against empirical N export measurements (slope: 0.92, intercept: 8,800 kg). I based this approach/understanding on the below excerpt from page 131 of the INVEST documentation.

"Note 2: Load values may be expressed either as the amount of nutrient applied (e.g. fertilizer, livestock waste, atmospheric deposition); or as “extensive” measures of contaminants, which are empirical values representing the contribution of a parcel to the nutrient budget (e.g. nutrient export running off urban areas, crops, etc.) In the latter case, the load should be corrected for the nutrient retention from downstream pixels of the same LULC. For example, if the measured (or empirically derived) export value for forest is 3 kg.ha-1.yr-1 and the retention efficiency is 0.8, users should enter 15(kg.ha-1.yr-1) in the n_load column of the biophysical table; the model will calculate the nutrient running off the forest pixel (n_export) as 15(1-0.8) = 3 kg.ha-1.yr-1."*

However, I just read this excerpt from the bottom of pg. 126-127, which makes me believe that I should apply a correction factor to the n_load values I have used.

“Loads are the sources of nutrients associated with each pixel of the landscape. Consistent with the export coefficient literature (California Regional Water Quality Control Board Central Coast Region, 2013; Reckhow et al., 1980), load values for each LULC are derived from empirical measures of nutrient export (e.g., nutrient export running off urban areas, crops, etc.). If information is available on the amount of nutrient applied (e.g. fertilizer, livestock waste, atmospheric deposition), it is possible to use it by estimating the on-pixel nutrient use (and apply this correction factor to obtain the load parameters).”

Which is the correct approach? ​Reading similar posts from other users, I’m inclined to say the second excerpt is accurate, but I wanted to double check.

If I followed the wrong approach, is there any validity in what I’ve done, or will I need to reassign N loads and recalibrate the model?

1 Like

Hi @atershy -

Yes, I agree that those paragraphs give contradictory information. I’ve only used the NDR model a few times, and reading these made me doubt my interpretation. So I’m trying to get someone on our science team to verify which is correct, and I’ll update the user guide appropriately and get back to you. Apologies for the delay.

~ Stacie

Hi again @atershy -

Well, your question led to a lot of consternation here at NatCap. The colleagues I’ve talked with all have used applied values of fertilizer (and/or atmospheric deposition etc) as the n_load parameter. We have not adjusted these using the retention coefficient.

But the contradiction in the User Guide text makes us wonder if something different is actually going on in the code that we weren’t aware of. The person who created the model has moved on from NatCap, and we are in the process of looking at the code and figuring things out. Apologies for the confusion, but given that we’re all juggling multiple projects right now, it might take a while before we have a definitive answer.

In the meantime, I suggest using fertilizer application without correction, since that’s been standard practice here. It’s good to hear that this approach gives a pretty good fit with calibration. We’ll update this thread with our findings.

~ Stacie