Spatial calibration using multiple gauges

[Ameer]

Good morning,
Suppose, I discretized the watershed into 6 subbasin but have data for only 3 gauges. I am, however, interested in all 6 points (both gauged and ungauged subbasins outlet). Can I use the 3 monitoring points for which i have observed data to calibrate the model but extract data for all 6 points? Its possible in other models but not sure about HBV-EC.
Thank you,
Ameer

[rchlumsk]

Hi Ameer

Yes, Raven is well setup to do this. As long as your three gauge data sets are read into Raven and associated with the right subbasins (in the ObservdationData .rvt files), and the appropriate subbasins with data are flagged as Gauged in the Subbasins table of the .rvh file, then Raven will generate performance metrics for those gauged basins that can be used in calibration.

If you wish for Raven to output the hydrographs for additional basins, you can mark them as gauged as well, even if there is no observed data for those basins. In your case, you would mark all six basins as gauged to extract the simulated flows for all six basins. This is true in Raven regardless of the choice of process algorithms.

Hope that helps!

Cheers,
Rob

[Ameer]

Thanks Rob, that helps.
Ameer

[Ameer]

Hi Ameer

Yes, Raven is well setup to do this. As long as your three gauge data sets are read into Raven and associated with the right subbasins (in the ObservdationData .rvt files), and the appropriate subbasins with data are flagged as Gauged in the Subbasins table of the .rvh file, then Raven will generate performance metrics for those gauged basins that can be used in calibration.

If you wish for Raven to output the hydrographs for additional basins, you can mark them as gauged as well, even if there is no observed data for those basins. In your case, you would mark all six basins as gauged to extract the simulated flows for all six basins. This is true in Raven regardless of the choice of process algorithms.

Hope that helps!

Cheers,
Rob

Thanks Rob- Following is some snap shot of my .rvh file. I have 18 subbasin and I am interested to get hydrograph for all of them, even though i have observed data for only 10 points. When i run the model, the simulated flow is autogenerated for the points that i flagged gauged, however, the actual gauged data is not read by the model. That means the hydrograph CSV file in my output folder has only simulated flow data but no observed data.

Code: Select all

:Attributes,  NAME, DOWNSTREAM_ID, PROFILE, REACH_LENGTH, GAUGED
 :Units,      none,  none,          none,    km,           none
   1   Sub1   7   Reach1   10.37   1
   2   Sub2   7   Reach2   18.95   1
   3   Sub3   7   Reach3   41.39   1
   4   Sub4   6   Reach4   1.63   1
   5   Sub5   6   Reach5   138.25   0
   6   Sub6   10   Reach6   71.71   0

Subbasin properties are defined as

Code: Select all

:SubBasinProperties
  :Parameters, TIME_CONC, TIME_TO_PEAK, TIME_LAG,
  :Units     ,      d,      d,     d,
   1   3.0   1.5   3.0
   2   3.0   1.5   3.0
   3   3.0   1.5   3.0
   4   3.0   1.5   3.0
   5   3.0   1.5   3.0
   6   3.0   1.5   3.0

The channels profile is read via .rvp file as

Code: Select all

:RedirectToFile channel_sections.rvp

Here is a screen shot of the channel_section.rvp - i am showing the profile for the first two channels only but have data prepared for other sections in similar fashion.

Code: Select all

ChannelProfile   Reach1
:Bedslope   0.066801
:SurveyPoints   
0   0
100   0.5
300   2.2
500   0.8
600   0
:EndSurveyPoints   
:RoughnessZones   
0   0.035
50   0.045
500   0.045
600   0.035
:EndRoughnessZones   
:EndChannelProfile   
   
:ChannelProfile   Reach2
:Bedslope   0.083663
:SurveyPoints   
0   0
100   1.5
300   4
500   0
:EndSurveyPoints   
:RoughnessZones   
0   0.035
100   0.045
400   0.045
500   0.035
:EndRoughnessZones   
:EndChannelProfile   

A screen shot of the hydrograph.csv file in the output folder after running the model is attached. It seems the model is not reading the observed streamflow data.

Hydrograph.PNG

Not sure what is wrong? will appreciated any help.
Thanks,
Ameer

[Ameer]

Hi Ameer

Yes, Raven is well setup to do this. As long as your three gauge data sets are read into Raven and associated with the right subbasins (in the ObservdationData .rvt files), and the appropriate subbasins with data are flagged as Gauged in the Subbasins table of the .rvh file, then Raven will generate performance metrics for those gauged basins that can be used in calibration.

If you wish for Raven to output the hydrographs for additional basins, you can mark them as gauged as well, even if there is no observed data for those basins. In your case, you would mark all six basins as gauged to extract the simulated flows for all six basins. This is true in Raven regardless of the choice of process algorithms.

Hope that helps!

Cheers,
Rob

Thanks Rob- Following is some snap shot of my .rvh file. I have 18 subbasin and I am interested to get hydrograph for all of them, even though i have observed data for only 10 points. When i run the model, the simulated flow is autogenerated for the points that i flagged gauged, however, the actual gauged data is not read by the model. That means the hydrograph CSV file in my output folder has only simulated flow data but no observed data.

Code: Select all

:Attributes,  NAME, DOWNSTREAM_ID, PROFILE, REACH_LENGTH, GAUGED
 :Units,      none,  none,          none,    km,           none
   1   Sub1   7   Reach1   10.37   1
   2   Sub2   7   Reach2   18.95   1
   3   Sub3   7   Reach3   41.39   1
   4   Sub4   6   Reach4   1.63   1
   5   Sub5   6   Reach5   138.25   0
   6   Sub6   10   Reach6   71.71   0

Subbasin properties are defined as

Code: Select all

:SubBasinProperties
  :Parameters, TIME_CONC, TIME_TO_PEAK, TIME_LAG,
  :Units     ,      d,      d,     d,
   1   3.0   1.5   3.0
   2   3.0   1.5   3.0
   3   3.0   1.5   3.0
   4   3.0   1.5   3.0
   5   3.0   1.5   3.0
   6   3.0   1.5   3.0

The channels profile is read via .rvp file as

Code: Select all

:RedirectToFile channel_sections.rvp

Here is a screen shot of the channel_section.rvp - i am showing the profile for the first two channels only but have data prepared for other sections in similar fashion.

Code: Select all

ChannelProfile   Reach1
:Bedslope   0.066801
:SurveyPoints   
0   0
100   0.5
300   2.2
500   0.8
600   0
:EndSurveyPoints   
:RoughnessZones   
0   0.035
50   0.045
500   0.045
600   0.035
:EndRoughnessZones   
:EndChannelProfile   
   
:ChannelProfile   Reach2
:Bedslope   0.083663
:SurveyPoints   
0   0
100   1.5
300   4
500   0
:EndSurveyPoints   
:RoughnessZones   
0   0.035
100   0.045
400   0.045
500   0.035
:EndRoughnessZones   
:EndChannelProfile   

A screen shot of the hydrograph.csv file in the output folder after running the model is attached. It seems the model is not reading the observed streamflow data. Hydrograph.PNG

Not sure what is wrong? will appreciated any help.
Thanks,
Ameer

I think the project is correctly constructed. I was looking at the very first few lines of Hydrograph.csv (output files). Since i do not have data for winter months... the model do not read that portion of the record (so the observed column is empty). The observed data start from the month March of each year all the way to October. I, however, have pretty bad performance metrics values- Do you think that the missing winter months record is a factor to bad model performance?

[rchlumsk]

Hi Ameer

This depends on how the rvt observation files are setup. If you have missing values in the appropriate period flagged as -1.2345 in your observation rvt files, that period/timestep should not count against your metrics. If you want better control over period of metric evaluation, you can also supply an rvt format observation weights file (see the manual) that would basically have a 0 or 1 time series to indicate whether the time step should be counted in the metrics (there is a function in the RavenR package now to produce this rvt file for you).

If the observations are appearing for the appropriate subbasins (as they appear to be), and the metrics are simply poor, I would check the missing values and the hydrograph performance with a plot. It may also just be that the model is performing poorly in that period, so a visual check would be helpful.

Cheers
Rob