please help me understand this for SUFI2 validation

[Sheila]

Still trying to get the answer to this question...

When I run validation in SUFI2, do I import the txtinout data from my
UNCALIBRATED SWAT model (of course with a different set of dates than
my calibration dates).

Then I indicate the calibration parameter values in SWAT-Cup and SWAT-
Sup calculates the validation stats based on those calibration
parameter values?

[David Coffman]

Shiela,

 

SWAT-CUP is not set up to run SUFI2 Validations.  This is what I have done, and you may want to try it. 

 

1.  Run SWAT-CUP on your uncalibrated model to get calibration input parameters. 

2.  Calibrate your SWAT model.

3.  Use Excel for validation.  Set up a spreadsheet with your calibrated model results and your observation values.  Look through the literature and find the formula(s) for your preferred validation statistic(s) and program it into Excel.  (http://naldc.nal.usda.gov/download/9298/PDF)

 

Good Luck!

Dave

[Karim]

I don't understand all the confusion when clear instruction is given for validation!

When you calibrate your model and obtain your calibrated parameter ranges then you need to run the model with these ranges but with your validation data set. You edit the files given in the instruction and do one iteration as before and you have your validation results with 95ppu, statistics and all.... You don't need any Excel file or anything like that...

[David Coffman]

[David Coffman]

I guess I am not understanding how that works.  I do not want to run only one iteration if I give SWAT-CUP a range for my input parameters.  I would understand running it with the best_par.txt output from my multiple (500+) calibration iterations, but not a range of parameter values.  How can I be sure SWAT-CUP is choosing the best value for each parameter if it is running a range of input values for one iteration.  If I am mis-understanding (which is likely), please explain how this works.  Thanks!

On Wed, May 23, 2012 at 1:18 PM, Karim <k_abb...@yahoo.com> wrote:

[Karim]

... Two things:

1- forget about the "best" parameters. In fact it would be better to remove it from the program all together! I left it there only for historic reasons and some especial purposes. As the whole process is stochastic, there is no best solution. Your solution is the the range of parameters you obtain in the last calibration iteration. You must propagate this range for whatever you want to do with the calibrated model, including validation.

2- an iteration may consist of several hundred simulations. For validation, it is best to do the same number of simulations as you used in the final step of the calibration. Sufi2 propagates the calibrated "range" of parameters for validation, unlike other programs that use the behavioral "set" of parameters. For this reason the validation icon in Sufi2 gives you only the instruction, because the steps are exactly the same as the calibration steps. Once you edited the input files with validation data, the steps are the same: use commands under the calibration icon, pre process, run, and post process. Here, of course, you only need to do one iteration of several hundred simulations to propagate the uncertainty in the parameters without further changes to the parameter ranges.

I hope this is clear and sorry about the confusion! 

I suppose we should have more workshops to explain the program. Any suggestion is appreciated...

Best wishes,

Karim

[David Coffman]

Thank you Karim.  That information is extremely helpful.  My next question...After I calibrate my model with SWAT-CUP, is there a way to transfer the new calibrated input parameters into my model in ArcSWAT?  What files need to be copied/replaced?  I will be running a number of different management scenarios, so I need my calibrated inputs values to be available in ArcSWAT, because that is the easiest way to manipulate other inputs (not related to calibration) to run my scenarios.  Thanks again for your help.  SWAT-CUP is a great program, and I will be more than happy to pass along a few suggestions.

 

Dave

[Karim Abbaspour]

 

... when you parameterize the model and, for example use reltive changes, then each hru may have different parameters...

it is not possible to put all these in the soil or landuse or other swat databases. Furthermore, you have a range for the parameters and it is not possible to pass these ranges to ArcSwat. Again, remember that the "best" parameter set is really meaningless! Because of the inherent non-uniqueness problem you have thousands of parameter sets that are all acceptable and the "best" one does not mean very much.

 

Best, Karim 

[Sheila]

Karim,
In which outout file do you get parameter "ranges" for calibrated
parameters? I was under the impression that you were only able to
obtain the best "fitted value" in the Best_Par.sf2 file for each
parameter. You then use this value for your validation such that the
minimum and maximum value is identifcal in your validation model set-
up (in the Par_inf.sf2 input file).

I ask this because, when I run validation, I do not get a 95PPU
graph. Nor do I get an r-factor or a p-factor.

Thanks.
Sheila

On May 23, 2:18 pm, Karim <k_abbasp...@yahoo.com> wrote:
> I don't understand all the confusion when clear instruction is given for validation!
> When you calibrate your model and obtain your calibrated parameter ranges then you need to run the model with these ranges but with your validation data set. You edit the files given in the instruction and do one iteration as before and you have your validation results with 95ppu, statistics and all.... You don't need any Excel file or anything like that...
>

> On May 23, 2012, at 18:51, David Coffman <dkcoff...@gmail.com> wrote:
>
>
>
> > Shiela,
>
> > SWAT-CUP is not set up to run SUFI2 Validations.  This is what I have done, and you may want to try it.
>
> > 1.  Run SWAT-CUP on your uncalibrated model to get calibration input parameters.
> > 2.  Calibrate your SWAT model.
> > 3.  Use Excel for validation.  Set up a spreadsheet with your calibrated model results and your observation values.  Look through the literature and find the formula(s) for your preferred validation statistic(s) and program it into Excel.  (http://naldc.nal.usda.gov/download/9298/PDF)
>
> > Good Luck!
> > Dave
>

[yudan1...@gmail.com]

After calibrated the model,where can I find the  calibrated parameter ranges? In whicn file? In the Sufi2.Out/best_par.txt I just find the  Fitted_Value,can't find the ranges.

在 2012年5月24日星期四UTC+8上午2时18分59秒，k_abb...@yahoo.com写道：

[Abbaspour, Karim]

Parameter ranges are in par_inf file, parameters for the next iteration are in new_par file. If there is no need for the next iteration, the parameters in par_inf are your solution.

Karim

On Sep 18, 2017, at 19:38, Brian Hall <brianhall...@gmail.com> wrote:

I know this is an old thread, but the way I understand it the parameter ranges are in the new_pars.txt file.  Please someone correct me if I am wrong.

-Brian Hall

--
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.

[Brian Hall]

Thank you, I understand it now.

-Brian Hall