This forum is shutting down! Please post new discussions at

Nutrient Delivery Ratio - Missing information in User Guide on Nutrient Runoff Proxy

Hi all,

I'm getting prepared for using the Nutrient Delivery Ratio model to assess nitrogen retention. The version of InVEST I have is the latest one to date (I believe) V3.3.0 (*86), downloaded 12 March 2016.

I have a question related to a required input parameter: Nutrient Runoff Proxy (raster). I went through the User Guide and I can't find out any detail on this input data...
The NDR model help displays "Weighting factor to nutrient loads. Internally this value is normalized by its average values so a variety of data can be used including precipitation or quickflow" (attached).

Could I use a rainfall raster (total mm rain/year)? Or maybe something else is more relevant?

Thanks for your help!


  • RichRich Administrator, NatCap Staff
    Hi Emilie, yes you can use a precipitation raster here.  And embarrassing we don't have a description of that parameter in the user's guide.  Hopefully Perrine can comment further on the science behind that parameter.
  • PerrinePerrine Moderator, NatCap Staff
    Hi Emilie and Rich, 

    Apologies for the inconsistency in the terms used in the User's guide and the user interface. We did have a short description of the runoff proxy but it was simply termed "runoff proxy".

    I have fixed this term and added a paragraph in the data needs section:

    3. Nutrient runoff proxy (required). A GIS raster dataset representing the spatial variability in runoff potential, i.e. the capacity to transport nutrient downstream. This raster can be defined as a quickflow index (e.g. from the InVEST seasonal water yield model) or simply as annual precipitation. The raster is normalized (by dividing by its average value) to compute the runoff potential index (RPI, see Eq. 1).

    Please let me know if you have questions.
  • Hi Rich and Perrine,

    Thanks a lot for your quick answers! It's now much clearer...

    The way this forum works is great, thanks also for that!
  • Perrine and Rich,

    I'm using the NDR for my research, and the model is working fine with all my data. However, something came to my attention: the user guide says "The raster is normalized (by dividing by its average value) to compute the runoff potential index (RPI, see Eq. 1)". 

    Should I use a raster calculator tool to divide all pixels of the raster by the average of the raster values? In this case, instead of use a raster with the annual precipitation per pixel, I'd use a raster with an index, with numbers probably ranging from 0.1 or so to 2 or 3. Which approach would be correct?

    Thank you!
  • swolnyswolny Member, NatCap Staff
    Hi @fabiosecanho -

    You don't need to normalize the runoff proxy, the model will do that. I just give it a regular annual precipitation raster, with values in millimeters.

    ~ Stacie
  • Stacie,

    Thank you. 
    I have one more question on those lines. I'm using ArcMap to interpolate annual precipitation of weather stations located across my study area. I'm using the Geostatistical Analyst tool to do that and the layer I created looks good (Figure on the left). However, when I export this data to create a raster (GA layer to grid), my raster looks messed up with weird lines crossing the area (Figure on the right). I looked in some forums online and people are saying that's normal. Did you ever try to do that? Do you think I could use this raster as my precipitation grid? 


    1013 x 875 - 20K
    1013 x 871 - 29K
  • swolnyswolny Member, NatCap Staff
    Hi Fabio -

    No, I've not used that particular process. And I'm not sure which interpolation tool you're using, but can you output directly to a raster instead of a GA layer? 

    ~ Stacie
  • Hi Stacie,

    The reason I'm interpolating to a layer is that in this way I can see the statistics of the process, so I can decide which method yields better results. If I interpolate direct to raster I don't have this information, so I have to randomly pick a method. In my case, I tried five different methods and chose cokriging as the best. 
    The raster itself is fine, I can use as an input data for the NDR model. What bothers me is the weird looking of it.

    Thank you,
  • swolnyswolny Member, NatCap Staff
    Once you've chosen cokriging, can you go back and just run that method only direct to raster?

    Or, if the look of it bothers you, have you tried using QGIS?

    Sorry I can't be of more help here. I do understand wanting the inputs to be as correct as they can be - that's the goal!

    ~ Stacie
  • Stacie,

    Although the look of it bothers me, I'm confident that the raster is correct. I tried it in my analysis and the results look consistent. 

    Thank you,
  • CharChar Member
    edited April 2018

    Hi Perrine and Rich,

    I have a question about the subsurface critical length (SCL) and subsurface maximum retention efficiency(SMRE) in the NDR model. When setting the proportion of the nutrient load that travel via subsurface flow , the value is set to zero by default on the assumption that all nurtients travel via surface or shallow subsurface flow. So based on this, could the SCL and SMRE be set to zero if there are no data about these two parameters? The NDR model can't run without these two parameters . Thank you for your time and looking forward to your reply.

    Post edited by Char on
Sign In or Register to comment.