Hi Ofir,

Based on my experience, there are a number of factors which will determine the best parameterisation configuration in WRF. In our sensitivity studies, we found that the domain size and topography also impacted sensitivity and combinations that worked best in one region, didn't work well in others. I am attaching a link to the paper we did on this. Interestingly, our best performing parameterisation configuration is similar to yours. The only differences are that we used Dudhia for SW radiation and Monin-Obukhov (Janjic Eta) for the surface layer scheme. Although our study was not in the tropics but in a temperate climate, hopefully some of our findings will help you out. 

Article Sensitivity of WRF to driving data and physics options on a ...

Hi Ofir, 

The simplest answer of your question is that there are no optimal combination of physical parameterization schemes that performs equally everywhere in tropics in all the time. As my friend Julia mentioned above, "there are a number of factors which will determine the best parameterisation configuration in WRF".

Performance of  physical parameterization schemes depends on multiple factors, specially topography if you are planning to run a high resolution simulation. For example, all convective parameterization schemes shows large precipitation bias over eastern India, a region of very complex topography and south Asian monsoon circulation. In your case the south American topography is distinctively different than south Africa. The same thing is true for soil moisture as well.

Over the Amazon rain forest you will find persistent cloud cover, which will be absent over south African continent. So, in your case both short wave and long wave radiation parameterization will play important role.      

In the available literature you will find that YSU PBL, WSM6 microphysics, and KF cumulus cloud parameterization schemes show overall better skill. However, the following study by Evans et al. (2012) strongly adviced not the use the above combination over south-east Australia based on 36 member ensemble experiment.

Instead the above study suggested to use the BMJ cumulus and MYJ PBL schemes combination over their study areas. 

My first suggestion would be to use parameterization schemes which are used on the driving data (in your case CESM1). This will reduce model uncertainty inherited form initial and lateral boundary conditions.  

My second suggestion would be to do a small members ensemble sensitivity experiment to get better credibility of your results.

I hope it helps. 

Thanks

Mostofa Kamal

http://link.springer.com/article/10.1007/s00382-011-1244-5

Hi all, i run the WRF model with different parametrizations everyday 4 times at day for Central America and the caribbean (tropics) here: http://srt.marn.gob.sv/?q=wrfrc  in my experience (empirical for now), to do a correct evaluation you need to consider the synoptic feature, the basic WSM3 microphysics at higher resolutions is ok in most cases but not allways, when you try to use this scheme in lower resolutions (like 15 or 30 km) first you need to add a cumulus scheme (not present in high resolution domain) and you will need to use WSM6 (better than WSM3 in lower resolution but not always) 

Hi Ofir,

For the case of WRF running for Indonesia, after tested several combination schemes to the several case of extreme precipitation events, these are our conclusions in regards:

- All the cases show that KF with the modified trigger function has the tendency of reducing the areal coverage of precipitation while enhancing the precipitation maxima. The convection weakening can lead to the enhancement of the explicit component that causes the intensification of maxima repeatedly seen. Furthermore, the increased contribution of

explicit rain may also explain the higher degree of organization detected in structures simulated by this experiment, inducing a more realistic appearance of the rainfall field.

- When verified against the TMPA analyses, simulations using Tiedtke keep the most resemblance to observations, especially due to its ability to make a clear distinction between rainy and dry areas. In fact, Tiedtke shows a relative low accuracy at predicting the precipitation maxima observed in land-based stations and its maximum values are often below the highest given by TMPA. However, the most serious drawback is its poor performance in surface fields, such as temperature, humidity and wind, with outstanding large biases.

- Grell3D is notably skillful at detecting the precipitation maxima. This circumstance has become evident in both visual and objective verification. This scheme has also shown advantage in surface variables such as nighttime temperatures and moisture. Its weak point, concerning the rain, is the large overestimation shown in many cases: this leads to false alarms of light to moderate rain accumulations in areas where few or no precipitation is observed.