New_pars ranges in swat cup - Need clarification

[Navs Kumar]

Dear all, 

After the 1000 iterations, i got my result, which seems to be good, but it is recommended that to use the New_par ranges (result from 1st iteration with 1000 simulations).

Here i found that the range suggested in New_pars.txt for some of the parameters is out of range (as mentioned in Absolute_SWAT_values.txt) e.g., SURLAG is having min value of -13.00 but it cant be negative as the minimum range is 0. (In this case i took the minimum range as 0 and maximum as 18.0 as suggested by max range in New_pars.txt. Am i right?

Why the range of some parameters are falling out of range in New_pars.txt?

 I think New_Pars.txt is the final range, which is used as final calibrated range further used for validation and in  in SWAT model to re-estimate the other variables, like recharge.

Dr. Karim or anyone of you, please shed some light on my doubts!!!!!!!!!!!!!!!

Best Regards

Navs

[Karim Abbaspour]

.... Absolute_SWAT_values.txt) e.g., SURLAG is having min value of -13.00 but it cant be negative as the minimum range is 0. (In this case i took the minimum range as 0 and maximum as 18.0 as suggested by max range in New_pars.txt. Am i right?

You are correct. If you don't do that, the swat_edit program will do it automatically, but it is better that you do it.

Why the range of some parameters are falling out of range in New_pars.txt?

 These are simulated, like any other simulation you may get things that are out of range. The new parameter values are meant to serve as a suggestion

 I think New_Pars.txt is the final range, which is used as final calibrated range further used for validation and in  in SWAT model to re-estimate the other variables, like recharge.

No, the final range are those in par_inf at the last iteration. You should use those for validation

Dr. Karim or anyone of you, please shed some light on my doubts!!!!!!!!!!!!!!!

Best Regards

Navs

--
You received this message because you are subscribed to the Google Groups "SWAT-CUP" group.
To unsubscribe from this group and stop receiving emails from it, send an email to swat-cup+u...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Navs Kumar]

Dear Dr. Karim,

Thanks a lot for yours reply and the fruitful information. SWAT CUP program is excellent and the manual is well explained but I still have some doubts, please clarify the below listed points:

1. New_pars. txt parameter ranges:

After yours clarification, till now my understanding about the new_pars.txt file range states that it is the range suggested by the SWAT CUP program for the next iteration in calibration process? And not the final calibrated range for validating the model. Am I right Dr. Karim??

I already have 1000 simulations for 1^st Iteration, now I got New_pars.txt range, which I should copy and paste in Par_inf file and again run the model for 2^nd iteration (with 1000 or any number of simulations), then again I will get the New_pars.txt range which I may edit, copy & paste again to the Par_inf file for the 3^rd Iteration (depending on the simulation result, likewise I can opt for another 4^th iteration), am I right?

Do you recommend a minimum number of iterations with New_par.txt file ranges out of each iteration before reaching to the final conclusion? Or it depends on the summary results only??

2. You have suggested that for validation, the final ranges are those in par_inf at the last iteration. But when I check my Par_inf file for my 1^st iteration with 1000 simulations, I have found no change in the ranges, why is this so? Where shall I found the final ranges in par_inf? The ranges were the same as I edited and put in Par_inf file before the start of iteration

Again with new _pars. txt file range of 1^st iteration (1000 simulations), when I ran the SWAT CUP, the result of 2^nd iteration gives me the same ranges as I feed from new_pars.txt values of 1^st iteration.

There is no change in the Par_inf file? Where the change does occurs?? As you suggested using the final ranges in par_inf file of the last iteration for validation but it is exactly the same range of new_pars.txt input file before the iteration.

3. Objective function:

Here in my case objective function plays a big role: for example if I choose R2 as the objective function my result of calibration is not good with R sq below 0.5, NS is negative, R factor 1.82. But If I Choose NS as the objective function the result improve significantly, as R sq .85, NS 0.35 and R factor 1.02. Also with different objectives functions, the simulated discharge shows bit under (NS) and over estimation (R2 & PBIAS) compared to observed discharge. What is yours remarks on this? Is the selection of Objective function can differ the results quite significantly??

4. Recharge estimation:

After the calibration and validation of model, shall I select the final ranges and edit in my buildup ArcSWAT model for simulating the recharge? Or shall I make active the no observation functions in SWAT CUP to see the results of simulated recharge by calibrated SWAT CUP model? Is both the procedures yields the same result?

Best Regards

Navs

[Karim Abbaspour]

1. New_pars. txt parameter ranges:

After yours clarification, till now my understanding about the new_pars.txt file range states that it is the range suggested by the SWAT CUP program for the next iteration in calibration process? And not the final calibrated range for validating the model. Am I right Dr. Karim??

Yes, this is correct. Those are suggested parameters for the next iteration. You need to look through them and make sure they are not out of range. You can also adjust the suggested ranges if you have any knowledge of the parameters.

I already have 1000 simulations for 1^st Iteration, now I got New_pars.txt range, which I should copy and paste in Par_inf file and again run the model for 2^nd iteration (with 1000 or any number of simulations), then again I will get the New_pars.txt range which I may edit, copy & paste again to the Par_inf file for the 3^rd Iteration (depending on the simulation result, likewise I can opt for another 4^th iteration), am I right?

Yes, correct. Each set of new parameters have smaller ranges than the previous set. Therefore, you should expect that at each iteration the 95PPU band to get smaller, and   P-factor and R-factor to get smaller. You do not want the P-factor to get too small, while you do not want the R-factor to remain too large. So this is a decision you, as the analyst, has to make as to when to stop the iterations.  

Do you recommend a minimum number of iterations with New_par.txt file ranges out of each iteration before reaching to the final conclusion? Or it depends on the summary results only??

It depends on the results as discussed above. Usually it takes around 4-5 iteration at the most. If you find yourself doing many more than this, then something is probably wrong.

2. You have suggested that for validation, the final ranges are those in par_inf at the last iteration. But when I check my Par_inf file for my 1^st iteration with 1000 simulations, I have found no change in the ranges, why is this so? Where shall I found the final ranges in par_inf? The ranges were the same as I edited and put in Par_inf file before the start of iteration

Don’t understand this question. Of course ranges in the par_inf do not change by themselves. If the results of the set parameter ranges in par_inf are satisfactory, then those are your calibrated parameter ranges. You should then use those for validation (or even a better term verification).

 

Again with new _pars. txt file range of 1^st iteration (1000 simulations), when I ran the SWAT CUP, the result of 2^nd iteration gives me the same ranges as I feed from new_pars.txt values of 1^st iteration.

Yes, of course! See above

 

There is no change in the Par_inf file? Where the change does occurs?? As you suggested using the final ranges in par_inf file of the last iteration for validation but it is exactly the same range of new_pars.txt input file before the iteration.

See above

3. Objective function:

Here in my case objective function plays a big role: for example if I choose R2 as the objective function my result of calibration is not good with R sq below 0.5, NS is negative, R factor 1.82. But If I Choose NS as the objective function the result improve significantly, as R sq .85, NS 0.35 and R factor 1.02. Also with different objectives functions, the simulated discharge shows bit under (NS) and over estimation (R2 & PBIAS) compared to observed discharge. What is yours remarks on this? Is the selection of Objective function can differ the results quite significantly??

Yes, exactly. This is why I included different objective functions. You as researchers should start thinking about this issue when doing calibration. An important issue to address is that using different objective functions leads to different calibrated parameter ranges (See Yang, J., Reichert P., Abbaspour K. C., and Yang H. 2008. Comparing uncertainty analysis techniques for a SWAT application to Chaohe Basin in China. Journal of Hydrology. 358, 1-23). These ranges may be very different indicating completely different processes in the watershed.

4. Recharge estimation:

After the calibration and validation of model, shall I select the final ranges and edit in my buildup ArcSWAT model for simulating the recharge? Or shall I make active the no observation functions in SWAT CUP to see the results of simulated recharge by calibrated SWAT CUP model? Is both the procedures yields the same result?

I don’t understand this question either! Didn’t you use discharges to calibrate your model? If you used discharges from only some subbasins to calibrate the model and now want to see discharges from all subbasins, then you can use  no_observation extract programs to see any variables for all subbasins. For this purpose you don’t need to put these parameters back in the ArcSWAT.  

 

-------------------------------------------------
Dr. K.C. Abbaspour
Eawag, Swiss Federal Institute for Aquatic Science and Technology
Ueberlandstr. 133, P.O. Box 611, 8600 Duebendorf, Switzerland
email: abba...@eawag.ch
phone: +41 44 823 5359
fax: +41 44 823 5375
http://www.eawag.ch/index_EN

[Navs Kumar]

Dear Dr. Karim,

Thanks a ton for the great suggestions!!

Just to clear my final understanding: 

1. The final par inf ranges (for a calibrated model and can be use for verification) are the ones which are the new pars.txt ranges of the last iteration and was feed in in the model as Par_inf before the start of final iteration?

2. Shall i only change the new_pars.txt ranges in the Par_inf file ?? Or the whole original files in the swat cup project folder should also replaced by backup files after each iterations??

3. Also the Sensitive parameters after each iterations should be judged by global sensitivity analysis and the least sensitive parameters should be removed in each iteration, so we will be left with the final best sensitive parameter in the last iteration??

Am i correct Dr. Karim?? any final note from yours side!!

Thank you very much yours time and kind consideration!!!!!!!!!!!!!!!!!!!

Best Regards

Navneet