Hi folks
I am using InVEST model since 2019 for sediemnt export quantification. While running this model in curent version 3.14.0 for similar dataset which was employed in 3.8.0, the significant differnce in the sediement export was observed for similar paramaeters.
The earlier sediment export results (0.75 t/ha/yr) were validated using the observed data (got very good fit though), but currently the values are underestimated (0.16 t/ha/yr) for similar input data and parameters in V 3.14.0.
I have also gone through the āRecent Modificationā section of SDR-InVEST userguide, and understad that the changes are expected.
The underestimation is huge that, the results gerenared from this version will certainly lead to some bias.
Kindly share your thoughts to overcome this issue.
Thanks for bringing this up, and Iām sorry for the frustration! The most likely causes for this discrepancy are the changes we made to the LS factor calculations. Specifically:
The LS Factorās upstream contributing area is now calculated as an approximation of the specific catchment area ā¦ instead of the absolute area upstream. This change is in keeping with the 1-dimensional situations for which the LS factor equations (and USLE more generally) were developed and parameterized, where the āupstream areaā in USLE is not a real area but a length. Thus, this change to InVEST SDR creates more realistic results and is better aligned with the literature and the LS approach taken in other software packages. For a full discussion of this change, please see the InVEST development record about this. (InVEST UG reference)
Unfortunately in your case, you will likely want to re-calibrate the model in order to get more realistic export results. Calibration is discussed in the userās guide section describing the IC_0 and K_b parameters.
Please let us know if you have any further questions,
James
Dear @jdouglass
Thanks for your prompt reply. Pls lt me know if there is any maximum cap for Kb parameter for calibration. I am getting close values with my observed data for KB = 6 -7.
Thanks again
While Iām not the model author, I am not aware of a maximum value for Kb. Iāve looked through Hamel 2015 and Vigiak et al., 2012 (as referenced in the User Guide), and donāt see specific mention of a maximum value. Hamel uses a max value of 3 for sensitivity analysis, but this is just based on altering the input values by +/-50%.
That said, have you only used Kb for calibration, or have you looked at other input parameters as well? Thereās a fair amount of uncertainty associated with the USLE parameters in particular.
Also, if version 3.8.0 of SDR worked well for you, you are welcome to continue using it. While we feel that the LS updates in 3.14 are generally calculating more realistic USLE values (which are usually lower than the excessively high values sometimes reported with the previous LS method), we have done plenty of analyses with the older version, and still feel that itās fine for capturing the patterns of high-low erosion and deposition on the landscape.