Calibrating Borselli vector in SDR

I’m running the SDR model on a basin composed of two watersheds. whithin the basin there are two dams and for each dam a watershed is delineated so that all the sediments that reachs the flow ends in the dam reservoir. when trying to calibrate based on the evolution of the two dams retention volumes, I can’t find the right value for the k parameter. the more I approch of the value of exported sediments in one dam, the more the value in the other watershed become different than what it should be according to the volume of sediments in the dam.
Is it possible to access the script of the SDR model and for each watershed give a specific k parameter value so that the sediment export become equal to the volume of sediments in the dams?

Hi @amin -

InVEST is open source, so you’re free to work with the source code. Go here for links to the API and Bitbucket repository.

However, I do wonder if you can simply run SDR twice, once for each watershed, so you can set the k parameter separately for each run. Recall that the model doesn’t do any in-stream processing, nor take into account retention by upstream dams. It just routes sediment until it reaches a stream, then aggregates the per-pixel sed_export.tif within each watershed provided as input. So I think you’d be safe running them separately.

This brings up another point regarding dam retention. It is likely that the upstream dam is retaining sediment, so in reality the downstream dam may be receiving less sediment than the model calculates. If you have information about the retention efficiency of the upstream dam, you can use that to adjust the total sed_export arriving from that upper watershed, to get a more accurate idea of what is being passed on to the lower reservoir.

@RafaSchmitt, @adrianvogl, any additions or corrections?

~ Stacie

1 Like

Hi Amin, in principal you can run the model for the contributing area of each dam separately, with a different set of k values for each contributing area. However, usually we would expect that the same k values work well for the entire area. Otherwise there might be a structural problem with the model. I am not sure how the reservoirs are handled in your model - but Stacy’s comment about sediment trapping might be relevant. If you still want to try going ahead with two separate models, you could also assign a specific land use classes in each watershed (e.g., forest_watershed_1, forest_watershed_2) and assign them different k values and then run a single model.

Thank you all for your responses. I think I will go with the two seperate models and give each watershed a specific k value.