Comparison of Precipitation Observed over the Continental United States to That Simulated by a Climate Model

This study compares monthly average frequency, intensity, and amount of hourly precipitation simulated by National Center for Atmospheric Research Community Climate Model version 2/Biosphere–Atmosphere Transfer Scheme to the smoothed estimates of those observed locally over the continental United States. The observed monthly intensities vary from less than 1 to 5 mm h−1, with lowest values in the winter in northern midcontinent and highest around the Gulf Coast in summer. Model intensities are on the average 3–4 times less when drizzle is excluded and an order of magnitude less when drizzle is included. As might be anticipated, relative frequencies are much too high and intensities much too low in the model fields, compared to those observed. The spatial pattern of these quantities and the total precipitation are more realistic. The study also compares extreme events and diurnal and seasonal variations and finds that, in general, the simulated precipitation has larger spatial variability, larger diurnal variation, and longer maximum continuous wet and dry periods than those observed. These discrepancies are largely due to the different nature of model (area averaged) precipitation and observed (local) precipitation. Therefore, the use of GCM output to directly represent local values of intensities and frequencies may lead to large errors in the coupling of the land surface to the atmosphere above it.