regarding calibration

[Aarti Soni]

Hi,

I have few questions regarding model calibration.

1. I was preparing csv file obtained from obsStrData.Rdata. In csv file there are number of columns (site_no, POSIXct, quality_flag, obs, threshold, basinTypeName, basinType). I want to understand how the threshold value, basin number and basin type can be assigned?

2. I have streamflow for 6-stations. Can I do multiple station calibrations at one time  or I have to do it at single station? 

Thank you.

[Arezoo RafieeiNasab]

Hi Aatri, 

Here is my attempt to answer your questions. 

1. Here are some description : 

• Quality flag: that is not used, so you could ignore it, and set it something. 
• threshold: This is a value that is used later for calculation of the categorial metrics such as POD, FAR. It is up to you how you would like to define it, you could use 90% or 95% depends on what is of interest for you. 
• basinTypeName = could be anything, this is more for information and not used. We usually categorize basins into flashy, regular, slow, and snowy. 
• basinType = a number that shows what basin type we have.  0: snowy; 1: slow; 2: flashy and 3: regular. 
• The basinType is used in setting up the parameters that are used to identify events for event based error metrics. 

2. You could use all gauges at the same time meaning to calibrate to several gauges instead of 1 at once. Below is the option in the setup file for such setup and explained below. 

enableMultiSites

Specify whether to calibrate to more than one streamflow gages or not if enableMultiSite set to 1, a file containing the list of the gages of interest and the corresponding weights in the calculation of the objective function is required. Name of the file is hardcoded to "calib_sites.csv" that is placed under the domain dir and contains the following fields: FID : Feature id of the gage as it appears in the CHANOBS site_no : site identifier as it appears in the obsStrDate.Rdata file weight : weight to be given to site when calculating the objective function

This is a sample file that I had for the calib_sites.csv: 

"FID","site_no","weight"
2040783,"02330450",0.5
2041579,"02331000",0.5

This means I am giving 50% weight to gage 02330450 and 50% to the gage 02331000. 

You could play with this as you wish. 

Hopefully this helps! Thanks!

Arezoo

--
You received this message because you are subscribed to the Google Groups "wrf-hydro_users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to wrf-hydro_use...@ucar.edu.
To view this discussion visit https://groups.google.com/a/ucar.edu/d/msgid/wrf-hydro_users/20ce34d3-c66f-4f2a-a815-79af1973e419n%40ucar.edu.

--

Arezoo Rafieei Nasab, Ph.D.

NCAR/RAL Project Scientist II

office: 303-497-2888

[Aarti Soni]

Yes, it is clear. 

Thanks Arezoo for the clarification. 

I have one more query, in ground-based streamflow observations, there are many zero values, which likely indicate missing data where no measurements were taken. Should I retain these zeros, or replace them with NaN to better represent missing observations?

[Arezoo RafieeiNasab]

Hi Aarti, 

If those are missing values, you just leave them out of the Rdataset. It does not matter if there is missing data in the Rdataset, just leave it out. Or set it to negative  value, I think we remove any negative values so that should do it.  However, if they are real zero values I would leave them as zero. 

Thanks!

Arezoo

[Aarti Soni]

Hi Arezoo,

Thank you for the clear explanation.