Short-term quantitative precipitation forecasting in tropical watersheds

The ability to predict tropical precipitation at short range (0-3 days) and at high spatial resolution is an important capability that has always challenged the limits of mesoscale numerical forecast models. Tropical precipitation is mostly convective in origin, and its forcing is not constrained by strong dynamical balances (like geostrophy in midlatitudes). The tropical environment is often nearly convectively unstable, so variability in precipitation is tied to subtle atmospheric disturbances, superimposed on a very strong diurnal forcing, especially near complex terrain. Yet the need for accurate daily quantitative precipitation forecasts (QPF) is significant, particularly over watersheds in complex terrain. Flash floods are an ever-present risk during the rainy season. Additionally, nearly half of the world's hydroelectric power generation resides in the mountainous tropics and subtropics, and high-resolution QPF is a key input to the hydrologic predictions required for optimal operation of hydroelectric power facilities.

As a first step in evaluating model performance for short-range tropical QPF in mountainous tropical regions, we have performed 48-h 5-km Weather Research and Forecast (WRF) model forecasts of precipitation over central Panama and in the southern foothills of the Himalaya. These regions were chosen because of the availability of special hourly raingauge data for several rainy seasons in recent years. We have compared the performance of the model at three resolutions (45, 15, and 5 km) with raingauge observations, and also are evaluating the climatology of satellite-derived precipitation products (CMORPH and TMPA) versus the model. Results indicate that the model performs well in capturing the diurnal cycle of precipitation in both locations (which are quite different). The model shows limited (but non-zero) skill in capturing the multi-day “envelopes” of heavier or weaker precipitation that are observed, compared to a 24-hour persistence forecast. The model also performs considerably better than the satellite-derived products in terms of capturing the spatial climatology of precipitation, particularly over the Himalayan foothills. The results are promising considering the model configurations tested were relatively simple “out of the box” settings. Additional improvement is expected with upgrades such as high-resolution SST fields, and use of a land-surface model specifically developed for tropical applications, which we will test next.