Calibration trade-off: good fit on small events, underestimates large flood

[Barbara Popczyk]

Dear WRF-Hydro community,

 I have a calibration problem in my WRF-Hydro setup for the 1742km2 mountain catchment in Poland.

 During the period June–September 2024 (with several small events and one major flood in mid-September), I face a strong trade-off:

 - When parameters are tuned for smaller events, the model significantly underestimates the large September flood peak.

- When tuned for the big flood, it strongly overestimates peaks during smaller events.

 I mainly adjusted REFKDT, OVROUGHRTFAC, MaNN, but the conflict remains.

 Has anyone experienced similar behaviour?

What parameters or approach helped you to find a better compromise?

 Any advice would be greatly appreciated.

Best regards,

Barbara

[Arezoo RafieeiNasab]

Hi Barbara, 

We recommend calibrating the model on a variety of events and multiple years to cover different weather and climate conditions in your study area. Calibrating to a specific type of event will usually not result in a better performance on events the model has never seen before. This holds true regardless of the model or the parameters you choose. A good calibration strategy would be to cover both high and low flow events in the calibration period so the parameters do not adjust toward matching a particular type of events. 

I have seen studies that try calibrating for a short period, however we advise against it. If possible to do multiple year, please consider a longer calibration period. If it is not possible to perform multi year, at least choose a period that has a variety of events. 

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/fc932163-9151-4694-b9e7-5af7ee05acd6n%40ucar.edu.

--

-------------------------------------------------------------------------------------------------

My working day may not be your working day. Please do not feel obliged to reply to this email outside of your normal working hours.

-------------------------------------------------------------------------------------------------

Arezoo Rafieei Nasab, Ph.D.

Scientist V

NSF NCAR Research Applications Laboratory

are...@ucar.edu