I have run the SDR model and finally in the IC raster map some pixel value is “NoData”.
Please help me to interpret this result and what is the solution?
Hi @hilu202 -
In order to help, we could use more information. Please post the log file (.txt) that is produced in your Workspace, and an image where we can see where the pixels are that have NoData values. Thanks.
Also, did you look at the other output files, including those in the “intermediate” folder to see where those NoData pixels originate? That might help explain what you’re seeing.
i have checked different flow accumulation value and as it decreased the Null value in IC map increased and also the sediment export and USLE became more real.
InVEST-Sediment-Delivery-Ratio-log-2022-03-31–12_30_20.txt (64.7 KB)
ic_10.2_0111_02.tif (5.7 MB)
Thanks for the additional information @hilu202. One thing I notice in the log file is that you have the Threshold Flow Accumulation (TFA) value set to 5. This is a very small number, especially since your pixel size is 12.5 meters. My initial guess is that this is causing a lot of your area to be covered in streams, and streams receive a value of NoData in SDR. Compare the areas of NoData with the stream.tif output map, and see if they correspond. If they do, then set the TFA value to a much higher number, start with a value of 1000, and compare the stream.tif output with a real-world stream map. Adjust the TFA value until the modeled streams come close to the real-world streams.
It works and the real stream map mached by the stream out put, there is another problem, the sediment export is not match with the real data, How can i increase it by 3 or 4 times?
Great, it sounds like adjusting the Threshold Flow Accumulation value helped fix that problem.
For calibration, have you read the User Guide section Comparison with Observations? It includes a paper from Hamel (2015) that includes information about calibrating against observed data.
I have checked both Comparison with Observations and Hamel(2015), selecting IC0 and k parameter is still unclear for me. How can I change them till finding the best Sediment EXPORT result? What is the acceptable range of them?
The paper is not the best guide for how to implement calibration, but it does discuss performing a sensitivity analysis over a range of values for many of the model inputs, and using those to inform the calibration. So, you would do multiple model runs where you change, for example, the kb parameter in increments of, say, 10%, within the range of +/- 50%. Then compare the sediment export results to your observed data and see which kb value produces export values that are the closest to observations.
We usually like to do this kind of sensitivity analysis for other input parameters as well, to identify which parameters are the most sensitive in our study area, since it might be the case that a different input parameter needs to be adjusted in order to get close to observations. That said, the paper only uses kb for the final calibration, so that’s a good place to start.
The goal of calibration is to get the modeled results as close to observed as possible, but they probably won’t be exactly the same. I’m not sure if there is an “acceptable range”, that might be up to you to decide. But if, even after calibration, your modeled results are an order of magnitude different, that’s obviously problematic. If they’re more like 10% off, that seems pretty good. If you have observed data for multiple sub-catchments, you can also compare the patterns of high and low between them, to see if the model is capturing their relative export correctly.
I’d also recommend noting some of the issues raised in the Discussion section of the paper, especially regarding the fact that the SDR model only accounts for overland erosion. So if the modeled values are very different even after calibration, it may be that other forms of erosion like in-stream erosion or landslides are a major contributor to the actual sediment observed in streams. In this case, it’s good to understand the sediment budget in the watershed, so you know what proportion of total erosion you’re able to inform with the SDR results.
We’ll work on adding more information to the User Guide about this. Meanwhile, if anyone else has advice, please chime in.
Thanks for your accurate reply. I am looking for the acceptable range of variation of these two parameters.