The performance of the MM5 and the RegCM3 models in simulating total monthly rainfall has been evaluated. Total rainfall for the two months of June 2000 and June 2001 are simulated. These two months have been selected because they are characterized by two different rainfall regimes; hence, they should provide a good test for the models. June 2000 is a relatively dry month while June 2001 is a wet month. Simulations are made with different schemes for parameterizing cumulus convective rainfall for both models. Two schemes are tested for the MM5 simulations while four schemes are tested for the RegCM3 model. It is found that the rainfall simulations are highly sensitive to the parameterization scheme. The results of the evaluation show that the MM5 model performs much better than the RegCM3 model.
The performance of any model configuration also depends on the characteristics of the weather situation which prevails during the month. For example, in the case of the MM5 model with the
Betts-Miller convective scheme, this particular configuration performs well in June 2000 which is a relatively dry month. The correlation coefficient between simulated and observed rainfall is 0.62. On
the other hand, it performs badly for June 2001, which is relatively a wet month. The corresponding correlation coefficient is only 0.02. The reason for this poor performance for June 2001 is presumably due to the fact the model has been unable to predict accurately the weather disturbances which occurred during the month.