How do we reduce the precipitation bias over mountains

[Liliputian_123]

Hello,

This is just a general question.

I have been simulating typhoon induced precipitation over an area that includes a mountain (bottom figure). This was run at 5x5 km with IC from ERA5.
Unfortunately, despite trying all combinations of the cumulus, microphysics, and PBL schemes, the best performing combination still show positive rainfall bias over mountains.

I have noticed a similar result even for non typhoon simulations. WRF tends to overestimate rainfall over mountains.
Do you have any suggestions on how I can further reduce this bias? I am not sure which physics scheme has more impact on orographic precipitation.


Will really appreciate any help on this.

 

[Ming Chen]

Hi,
A few research papers indicate that WSM5 microphysics, scale-aware GF cumulus scheme, and higher resolution generally yield better results for precipitation simulation. However, note that the model performance is often case-dependent and these options may not be the best options for East Asia.
Another issue is that, the 5-km resolution is within the typical grey zone for cumulus scheme. Have you run a test case with cumulus scheme off?

 

[Liliputian_123]

Hi,
A few research papers indicate that WSM5 microphysics, scale-aware GF cumulus scheme, and higher resolution generally yield better results for precipitation simulation. However, note that the model performance is often case-dependent and these options may not be the best options for East Asia.
Another issue is that, the 5-km resolution is within the typical grey zone for cumulus scheme. Have you run a test case with cumulus scheme off?

Hi Dr. Ming Chen,

Thank you for this!
I just checked and found that the cumulus is turned off in the above example.
The best peroming MP-PBL combo is Goddard-ACM2.
You are correct that this is case dependent, as we have some runs where YSU PBL performs better than ACM2.
The overestimation occurs on the wind ward side of the mountain (blue shaded regions in the figure below).
I'll need to do more research on this. Perhaps we need to change the parameters (i.e., some constants) inside the schemes to further reduce the bias.


This figure shows dh/dx (a) and dh/dy (d). h is height in meters. The topography map is shown below.