In this study we evaluate three techniques for estimating the rainfall rate using infrared satellite data from the Japanese Geostationary Meteorological Satellite (GMS). Hourly rainfall rates were estimated using the GOES precipitation index technique, a simplified Griffith-Woodley technique, and the infrared power law rain-rate technique, all of which relate the rainfall rate to the cloud-top temperature. The intercomparison also included rainfall estimates from a gridded monthly rainfall climatology and the Australian Regional Assimilation Prognosis (RASP) model forecasts. Validation data consisted of twenty-three months of daily surface rainfall analyses over Australia based on surface precipitation observations from the meteorological synoptic station network. The three infrared algorithms produced similar rainfall estimates and showed limited skill in diagnosing daily rainfall over Australia. Non-precipitating cirrus clouds were frequently diagnosed as raining, while rain occurring in mid and low-level stratus went undetected. Even so, the correlation coefficients with the surface rainfall analyses were much greater for the satellite estimates than for the RASP model or ‘climatology’, indicating that the IR techniques were better able to detect the spatial and temporal variable of rainfall. Tuning of the algorithms significantly reduced the overall bias and root mean square errors.
Ebert et al. (Fri,) studied this question.