Comparing Simulated Size Distributions of Precipitation Systems at Different Model Resolution

摘要

Size distributions of tropical convective systems in regional numerical atmospheric models are analyzed over a 2.5 × 105 km2 domain using different model grid spacing and parameterization schemes. The 5- and 20-km-resolution experiments are configured with a cumulus parameterization scheme, whereas the 2- and 4-km-resolution experiments are not. Precipitation systems are defined by either synthetic satellite infrared images, surface rain rates, or vertical winds at 600 hPa. The size distributions of systems defined by shallower clouds, lower rain rates, and weaker updrafts follow power laws, whereas those defined by deep clouds, higher rain rates, and stronger updrafts show lognormality. The cloud size distribution of the 5-km-resolution experiment is most similar to that of the real geostationary satellite observations. Generally, the largest system size becomes larger in the 5- and 20-km-resolution experiments, implying that the cumulus parameterization may have an impact on that scale. Exceptionally, all the model-simulated size distributions of heavy rain areas agree well at the largest scale. Lower-resolution experiments tend to underestimate the number of small-scale systems when compared with higher-resolution experiments. The size distributions also capture a temporal modulation of precipitation during the 2007 Jakarta flood event; small-scale intense precipitation systems increase during the period.