Basic SWAT-CUP questions

[Evan Ross]

Hey all,

I am an arcswat novice and have never used swat-cup. I'm trying to use it to calibrate for a large watershed. I want to make sure I'm doing it correctly. The manual leaves me with some lingering confusion.

I chose a number of parameters and ran the calibration for 10 simulations. I then used the new parameters given in the output to run another 10 simulations. Then another 10 with the new parameter values. My R2 has been steadily improving. Do I just keep doing this, replacing parameter min/max values and running simulations until my R2 is at a level I deem acceptable? I'm worried it will soon plateau around 0.52, at which point I could add more parameters?

Once I have a R2 that is high enough, how do I get the new parameter values (I assume the fitted value from the best parameters output file?) into my default arc swat model so that I can run it there?

Thank you for any and all guidance.

[dirombihon]

Hi Ross,

assuming you are using SUFI2 program for your calibration, you should first increase the number of simulations to a much higher number, like 500 or even 1000 if you have a fast computer system. In doing that you will give SUFI2 the possibility to find the most fit parameter values for your watershed. 

Moreover, I would recommend to add as many relevant parameters as possible at first. Since SUFI2 does a sensitivity analysis for you, you can use the result of the sensitivity analysis from the first few iterations as a guidance to exclude some of the parameters. Before you decide to exclude certain parameters, make sure you know your watershed well. Sometimes you may need to keep certain parameters that are practically relevant for your watershed even if they don't seem sensitive enough during the analysis.

Then you can go ahead and do the iteration to the point the change in R2 and/or NSe between iterations is small. Once you are satisfied with your R2/NSe values, you can use SWAT-CUP to set your parameters to the "fitted values" . To do that, assuming your are in your last iteration you are satisfied with, go to "SUFI2_swEdit.def" and set "starting simulation number " and "ending simulation number" to the "Best_sim_no" in "Best_Par.txt" or "Summary_Stat.txt". Then run only the "SUFI2_run.bat". Your model parameters should now be set to the "fitted" parameter values of your last iteration.

Best