Diagnostics and Calibration

[ephilip]

Hi Rob,

I calibrated GR4J model using OSTRICH. Uncalibrated model results are low R2 = 0.27. The best results from OSTRICH is not giving any better results than the uncalibrated one. Could you please give me some help on identifying the error?

Thank you
Elizabeth

[rchlumsk]

Hi Elizabeth

My first thought is to double-check that the calibration objective is being read in properly, check the ostOutput.txt and/or ostModel0.txt files to check that the calibration objective is changing (and hopefully improving) with trials. It may be that the wrong file is being referenced, or the wrong metric is being read in somehow.

Cheers,
Rob

[ephilip]

Hi Rob,

Thank you so much. Problem is fixed.
Thank you
Elizabeth

[ephilip]

Hi Rob,

Would you please give me an example/sample on how you make the weighting file (page 128) .

Thank you
Elizabeth

[rchlumsk]

Hi Elizabeth

Here is a sample weighting file for hydrograph observations, the same can be done for other data types. As you mentioned, this format is the same as that on page 128 of the manual

Code: Select all

:ObservationWeights   HYDROGRAPH    72   
2007-10-01   0:00:00   1   1461                  
0
0
0
0
0
...
1
1
1
:EndObservationData


This can be used to weight the observation data as you wish. For example, I typically setup my warm up year(s) in this fashion, using a 0 weight for the warm-up period and a 1 for the remaining data.

Let me know if you have any other questions!
Rob

[ephilip]

I am in the process of setting up separate calibration and validation periods for my model. What is the best/easiest way to setup these runs in Raven for different periods?

There are a few different ways to setup the runs. You could have:
• Separate model files for calibration and validation runs, although not the easiest way. Note that you can use a place holder value of -1.2345 in observed data for periods that should not be evaluated, that is the other way to control what periods the fit metric is evaluated for
• Single file for each gauge data with separate weight files for calibration and validation, so that the metrics are evaluated accordingly to the correct data set
• Have a separate start and duration setup in the rvi file (separate files or commented/uncommented system) to run the model for various times, i.e. starting with calibration period or validation period accordingly. Main trick here is to ensure the solution.rvc from the best calibration run is used to warm start the validation run (true in any validation run). Should be able to have all forcing and observed data in the files in this case, Raven can read the data and only use portion for after the validation start

Hi Rob,
I did calibration using OSTRICH. Now I have to copy the contents of the solution.rvc file to the initial .rvc file for validation. Please let me know whether I am correct or not?

Thank you
Elizabeth

[rchlumsk]

Hi Elizabeth

If you want to run the validation starting at the point where the calibration ends, then yes you can copy+paste the solution.rvc contents into the model.rvc file, which will start the validation with the conditions ended in calibration. If you do this, you should update the beginning date/time of your model and the model duration accordingly in the rvi file in order to start the model run at the validation period.

The alternative is to update the weighting file for diagnostics (or just comment out the original and point to a different file) that will only evaluate diagnostics in the validation period, and run the model for the calibration+validation period. You can also ignore the weightings if you calculate your own diagnostics from the extracted hydrographs or other files. In either case, this will also ensure that the end of calibration conditions are used at the beginning of validation.

It does not really matter which method you use, I prefer the second one personally but whatever you think is easier.

[ephilip]

Hi Rob,

I am getting the warning following message while runnign GR4J model. Would you please help me.

WARNING : maximum state variable limit exceeded in initial conditions for Soil Water[0] (in HRU 1) in .rvc file

Thank you
Elizabeth

[rchlumsk]

Hi Elizabeth

The warning means that the initial storage is greater than the maximum storage for SOIL[0] in HRU 1. The storage is set to the maximum in this case, and if you have a warm-up period you can likely ignore this. If you are worried about it or see that there are effects from initial conditions past the warm-up period, you can adjust the initial storage in the rvc file to some lower value.

Rob

[ephilip]

Hi Rob,

Is it ok to put values like 0.95 in the weighting file other than 0 and 1. The value of 0,95 to account for the effect of ice (B value) during winter.

Thank you
Elizabeth