Insights into Errors of SMS-Inferred GATE Convective Rainfall

Abstract
In the mean the Griffith/Woodley rain estimation technique underestimated the radar-measured rain of each of the three phases (a total of 56 days) of GATE, to varying degrees, and the satellite-derived isohyets were generally too extensive relative to radar-measured patterns. Three possible error sources are investigated in the present paper: 1) the method of apportionment of satellite-derived rain at the surface; 2) resolution degradation of the digital satellite imagery; and 3) anomalous behavior of convective clouds in the tropical Atlantic relative to those of the Florida derivation data set. To correct the satellite-derived rain patterns, a new method of apportionment was tested by recomputing the GATE satellite rain estimates. Better volumetric comparisons between radar and satellite estimates were observed for 24 h and phase periods, and comparisons of isohyetal patterns improved on all time scales. The relative error caused by resolution degradation was quantified by comparing rain estimates produced from full resolution imagery to estimates derived from degraded imagery for an 8° latitude by 12° longitude area in the eastern tropical Pacific ocean over a 54 h period. Results showed that the volumetric rainfall estimates made at 1/3° spatial and 1 h temporal resolution would be on the order of 10% lower than estimates made with the full resolution data (1/15° and 30 min). The remaining differences between the GATE satellite and radar estimates are attributable to different conditions prevailing in Florida and in GATE. These include significant rain from clouds that do not grow above the −20°C level (“warm rain”) and very long-lived anvils.