SDR model v.s. Sediment retention model

Dear all,
   In the Sediment retention model found in InVEST v3.0.1 and earlier, the total retained sediment has two parts, which equals to the sum of the sediment removed by the pixel itself and the sediment removed through routing filtration.
   But in SDR model found in InVEST v3.1, it does not account for the retention from upstream sediment flowing through the given pixel?
   So if I want to map the ecosystem service of soil retention, it would better to use the Sediment retention model found in InVEST v3.0.1 and earlier, in order to accout for the two parts, which just like the eqution in the attached image?
   Thanks a lot.

899 x 359 - 89K


  • PerrinePerrine Moderator, NatCap Staff

    The SDR sediment model (v3.1) does account for retention from upstream sediment, but you are right that it does not provide a map of the sediment retained. The closest output is the sed_retention index, but as noted in other posts, it is only an index, and does not represent the amount of sediment deposited on a pixel.

    I would not recommend using the v3.0 model for this variable either: the reason we upgraded the model was that we had low confidence in the old algorithm (it does not represent the physics of sediment transport in a watershed).

  • Hi Perrine,
       Thanks a lot for you kind help. I am not quite sure about the differece between the output of sed_retention.tif and sed_retention_index.tif.  And which one would you recommend to use for mapping ecosystem service of soil retention ?
    Thanks again.

  • RichRich Administrator, NatCap Staff
    Hi, one other note (not related to retention vs. retention index), we fixed a bug in SDR last Friday that overestimates sediment export on pixels with slopes > 9%.  If you're using the SDR model now, I'd recommend upgrading to this development version until we can release it in an offical InVEST patch:
  • PerrinePerrine Moderator, NatCap Staff
    Hi Cao, 

    The raster sed_retention_index.tif is calculated as follows: R*K*LS*(1-CP)*SDR. See details here:

    The raster sed_retention.tif is, unless I'm mistaken, an intermediate output representing the difference between sed_export for the current LULC and the bare soil watershed. The model aggregates (sum) this raster at the subwatershed scale to compute the "sed_retention" field in the attribute table of the output shapefile.


  • Hi Perrine,
        Thanks for you answer. 
        According to the equation, I thought sed_retention_index could be seen as  the sediment retained on each pixel. However, the documentation of InVEST just pointed out that it is NOT.  So if I use sed_retention_index to map the ecosystem service of soil retention, it would be inapproprate? I am a little confused....

  • Hi Rich,
       Thanks for your note. I will try it and look forward the new version of InVEST.

  • PerrinePerrine Moderator, NatCap Staff
    Hi Cao, 

    This is a good point, and it depends on the definition of "retention".
    If you are talking about the avoided soil loss from the pixel itself, then it's fine to use the index (because as you note, the index represents the difference between soil loss from a bare soil pixel and from the current vegetation, weighted by the SDR factor).

    However, it may be more accurate to define "retention" as the avoided soil loss on the pixel (the amount describe above) PLUS the amount of sediment from upstream pixels that is deposited on the pixel. Unfortunately, this amount is not directly calculated by the model, this is why we use the subwatershed scale proxy (i.e. sed_retention in the shapefile). 
    For your purposes, you can use the sed_retention value at the subwatershed scale or the sed_retention_index, as long as you are clear about the definition.
    I hope this helps
  • Hi Perrine,
        Thanks a lot, and I got it~
        I remembered that the Sediment retention model found in InVEST v3.0.1 and earlier could calculate the retention which include the two parts. However, as the documentation said, the model is not unstable. I am really looking forward to the release of a stable version~

  • RichRich Administrator, NatCap Staff
    Hi Cao, feel free to use the version I posted above.  The only difference between that and the 3.3.3 release is that the version number will print as "3.3.3" when you run it.

    Otherwise, keep an eye out on the forums for when we post it.  It might even be tomorrow!
  • Hi Rich, Got it~  bravo!
  • Hello sir,
    i'm using SDR to run the Model, having problem with the basins which are getting low sed_export as a result. They are doesn't match with the observed data. i have tried for fill_dem and Route dem but it won't works.kindly give the suggestion to resolve problem..
    thank you

  • swolnyswolny Member, NatCap Staff
    Hi Suresh -

    Have you calibrated the SDR results with observed data? It is normal for the model results to be different from observed data, but if you perform a calibration on the model using your observed data, the results should be close to observed. Here is a paper describing a calibration process done with the SDR model, which might be useful.

    ~ Stacie
  • Dear Mam,

    Thank you for your kind response.

    I have already calibrated the models for 33 basins by increasing the value of Borselli Parameter( kb). However, out of 33 basins. 6 of the basins are not coinciding with the observed data. I also have tried filling and Route the  DEM but the results  still remains the same. 
     I would like to know up to how much the kb value should be increase? Because i have run the model by increasing the kb value upto 240.

    Thank you,

  • PerrinePerrine Moderator, NatCap Staff
    Hi Suresh, 

    One thing to remember is that the InVEST model only looks at the sheetflow erosion. It's possible that you are trying to calibrate the model without accounting for the different sediment sources (river banks, landslides).
    The paper that Stacie sent earlier comprises a discussion of this issue.

    The kb has no strong boundaries but you should look at the watershed-scale sediment delivery ratio (ie. sed_export/usle), and this value should typically be between 0.01 and 0.1. (See also the paper for a discussion).

    Hope this helps!

Sign In or Register to comment.