Advice on Forcing Data Choice for Precipitation Sensitivity Experiments

[kinguT]

I’m conducting a series of sensitivity experiments using the WRF model and have consistently encountered precipitation underestimation across most configurations when compared against observational+satellite datasets, particularly CHIRPS, TAMSAT, and MSWEP.
Currently, I’m using the NCEP GFS 0.25° data with 3-hourly temporal resolution as the initial and boundary conditions. For reference, I’ve attached the namelist.input file from one of my experiments.
I would greatly appreciate your insights on the following:
- Based on your experience, which forcing dataset has yielded better performance in simulating precipitation, especially in data-scarce or tropical regions?
- Are there specific datasets or configurations you would recommend for improving precipitation accuracy?

Your expert advice and suggestions will be incredibly valuable to fix the issue.
I really appreciate any help you can provide.

 

[Ming Chen]

Your namelist.input looks fine to me except that

(1) increase time step to 36. For dx= 9km, time_step = 20 is way too small.
(2) turn off nudging (i.e., grid_fdda = 0, 0)

As for the foricng data, personally I prefer to use ERA5 pressure level data. We provide a python package that can process ERA5 efficiently. Please see the website: GitHub - NCAR/era5_to_int: A simple Python script for converting ERA5 model-level netCDF files to the WPS intermediate format

You can download the script, follow the instruction to run it and produce intermediate files.

Please try and let me know if you have any issues.

 

[kinguT]

Your namelist.input looks fine to me except that

(1) increase time step to 36. For dx= 9km, time_step = 20 is way too small.
(2) turn off nudging (i.e., grid_fdda = 0, 0)

As for the foricng data, personally I prefer to use ERA5 pressure level data. We provide a python package that can process ERA5 efficiently. Please see the website: GitHub - NCAR/era5_to_int: A simple Python script for converting ERA5 model-level netCDF files to the WPS intermediate format

You can download the script, follow the instruction to run it and produce intermediate files.

Please try and let me know if you have any issues.

Ming Chen, Thank you for your quick and helpful reply.
I will try using ERA5 as the forcing data and I’ll let you know once I generate the intermediate files.
Regarding nudging, I’ve noticed that even when nudging is turned off, some of my experiments still show underestimated precipitation, similar to when nudging is on. I’ll test again with your suggestions.
Also, I wanted to ask: should I only use ERA5 pressure level data? what about surface level data?

 

[Ming Chen]

Yes please try ERA5 pressure level and surface data as the forcing data.

However, note that while the forcing data is important, precipitation simulation is a challenging issue and many factors may affect the simulation.

 

[kinguT]

Yes please try ERA5 pressure level and surface data as the forcing data.

However, note that while the forcing data is important, precipitation simulation is a challenging issue and many factors may affect the simulation.

Many thanks, Ming Chen.
I’ve noted your advice and will use ERA5 data as the forcing for my WRF experiments.

One more thing: previously, I simulated the sensitivity experiments for ten days due to the limitations of time and computational resources. How long should I simulate the sensitivity experiments for a better accuracy of precipitation?