The model performed generally well. State variables were usually unbiased and met most statistical standards. There were a few weaknesses that we discovered in model performance:
1) Cold bias in winter 2) Overestimation of accumulated precipitation in summer 3) Summertime diurnal cloud signature anticorrelated with observations 4) Autumnal dry bias
The wintertime cold bias and autumnal dry bias are directly related to the soil initialization technique and probably could be improved at some point in the future. The precipitation overestimation is of some concern, but it should be noted that even in summer the model was generally unbiased in terms of precipitation coverage. The model could be too efficient in its convective processes, possibly "raining out" too much of its cloud water.
In the sumer the model tends to produce more clouds overnight than at any other time, while the afternoon cloud coverage is a daily minimum. The observations show the exact opposite. More research needs to be performed to uncover the reasons behind this model behavior.
Overall the model results are deemed acceptable to drive annual air quality simulations. These air quality simulations will lead to visibility improvement in the southeastern US under the leadership of VISTAS (Visibility Improvement - State and Tribal Association of the Southeast).
Supplementary URL: http://www.baronams.com/projects/VISTAS/#annual