Precipitation from a 10-yr regional climate simulation is evaluated using three complementary analyses: self-organizing maps, bias scores, and arithmetic bias. Collectively, the three reveal a precipitation deficit in the south-central United States that emerges in September and lingers through February. Deficient precipitation for this region and time of year is also evident in other simulations, indicating a generic problem in climate simulation.

Analysis of terrestrial and atmospheric water balances shows that the 10-yr average precipitation error for the region results primarily from a deficit in horizontal water vapor convergence. However, the 10-yr average for fall only suggests that the primary contributor is a deficit in evapotranspiration. Evaluation of simulated temperature and soil moisture suggests the model has insufficient terrestrial water for evaporation during fall. Results for winter are mixed; errors in both evapotranspiration and lateral moisture convergence may contribute substantially to the precipitation deficit. The model reproduces well both the time-average and time-filtered large-scale circulation, implying that the moisture convergence error arises from an error in simulating mesoscale circulation.