Extreme LS values and disrupted flow accumulation in SDR erosion model 3.3.1

Extreme LS values and disrupted flow accumulation in SDR erosion model

I am encountering some very high LS values (>500) in my SDR model outputs. These lead to very high USLE values (>500) and often to very high sed_export values. The same pixels also seem to be associated with disjunctions in the flow accumulation layer, where instead of the flow accumulating it splits across multiple pixels and then the flow comes back together. This then leads to the stream layer being broken, since the flow does not accumulate continuously. I do not get this kind of result when I use the same DEM file and the ArcMap flow accumulation tool.

I am assuming all these oddities are connected, but I don’t know precisely how. Has anyone encountered similar issues? I have filled my DEM, using ArcMap. I even did it multiple times just to be sure, but same results. I ran into some similar issues in earlier version of InVEST, but I had to take a break from it for a while and am just now getting back into it. Thanks.



  • RichRich Administrator, NatCap Staff
    Hi Clinton, I'll let Perrine comment/follow up on the LS values, but the effect you're seeing with flow accumulation is because our model uses a d-infinity flow were you might be using D8 with ArcMap.  Perrine might comment on this further too, but we prefer d-inf since it's better at modeling flow across the landscape despite the occasional divergent streams.  D8 on the other hand tends to bias toward long diagonal flows since diagonal pixels are likely to be further from the center and hence have a larger elevation gradient than the 4 adjacent.
  • OK, that makes sense about the flow accumulation. It's not such a big deal since I have a hydrology layer that I use for the streams rather than the derived streams using flow accumulation. It does seem odd that the strangely high LS values tend to co-occur with these differing areas in the flow accumulation.
  • RichRich Administrator, NatCap Staff
    I agree, it seems odd.  Perrine is travelling at the moment (actually many NatCap scientists are) but I suspect she'll be able to help further when she gets back.

    More soon!
  • PerrinePerrine Moderator, NatCap Staff
    Hi Clinton, 

    Sorry for the slow reply. 
    We've observed these high LS values in landscapes with very high slopes, where the USLE doesn't work great. Unfortunately, we are not aware of a single empirical relationship that we could use to cap the value of the LS factors for high slopes: if you have one to suggest, please feel free to share. 

    In the mean time, you can also correct your high slopes values in post-processing: because soil loss per pixel is the product of 5 factors, you can just modify one of these factors (LS) as needed and use the raster calculator in your GIS program to recompute the new soil loss.



Sign In or Register to comment.