InVEST SDR sediment export overestimation

Hi,

I am currently using InVEST SDR to estimate the sediment export of some watersheds in France. I am using a 25 meters mesh and a 100 meters mesh. The results are pretty close between these two meshes but I overestimate my validation data by 10 each time.

I have compared InVEST SDR with a manual way to compute sediment export and it’s doing good compare to my validation data. The only difference between the two methods is that the LS factor is really high compare to the one I am manually computing with QGIS (Desmet and Govers 1996 equation).

I would to know if someone has met the same issue and if the LS factor could be the only reason of the overestimation of my sediment export estimation.

Thank you and cheers from France.

1 Like

Hi @TheoMzr,

Welcome to the forum and thanks for writing in about your experience with SDR results.

Is the model overestimating your validation data by 10 times?

Recently, one of our hydrologists (@RafaSchmitt) began investigating a potential issue with the LS-factor cap in this model. It is currently hardcoded at a value of “333”. It has been observed that the model may significantly overestimate export in very large watersheds. We are now testing a developmental build where we can manually vary the LS-factor cap and have been considering “122” instead of “333”. So this definitely could be the sole cause of your overestimation.

I apologize that you ran into this problem. Perhaps Rafa can offer more details.

-Jesse

1 Like

Hi
I also have this problem, The SDR value is close to real data observed in hydrometric station but the erosion which resulted from USLE method is overestimated (almost 5 times higher than real). The main reason is high value of calculated LS by the model and there is no way to introduce the LS raster instead of model computation.
Please let me know to face this problem

hi
I reduce the L max value to 1 and also threshold _ flow_ accumulation to 50, it looks suitable for estimation both SDR and USLE in my watershed!!! ,. Is it right and acceptable??

HI @hilu202 -

I don’t know much about changing the Lmax value, but I suspect that a value of 1 is unrealistically low, since the User Guide says “Its default value is 122 but reasonable values in literature place it anywhere between 122-333 see Desmet and Govers, 1996 and Renard et al., 1997.”

And a definition of the LS factor is
“LS is the slope length-gradient factor. The LS factor represents a ratio of soil loss under given conditions to that at a site with the “standard” slope steepness of 9% and slope length of 22.13 m (72.6 ft). The steeper and longer the slope, the higher the risk for erosion.”

So with a value of 1, you’re saying that the maximum slope length in your study area is 1m, which is probably not really the case.

For Threshold Flow Accumulation (TFA), it is best to set it to a value that creates a stream.tif layer that comes close to matching your real-world stream network. If 50 is the value that creates streams that are close to reality, then it is a good value. If it doesn’t create a fairly accurate stream network, then I’d be very wary of using it. I would do a sensitivity analysis of the other inputs to the USLE, see which one(s) have the largest impact on results, and adjust those. TFA is not used for USLE.

~ Stacie

2 Likes

Hi @swolny
The mentioned watershed has a lot of steep slopes. You are right it is better to calibrate TFA according to real streams. However in this case other parameters of USLE is correct according to literatures but the LS parameters is high and it reduces by changing the L max. As the model result come to real by changing the L max below the suggested 122 for steep slopes watersheds we can use it.

A post was split to a new topic: SDR sediment export overestimatino