Assessment of the TRMM 3B42 Precipitation Product in Southern Brazil

Abstract

AbstractThe increase in the use of satellite‐derived precipitation products generated by different methods and algorithms emphasizes the need for a deeper analysis of their quality and accuracy. Using the contingency table method, we evaluated the accuracy of versions 6 and 7 of the Tropical Rainfall Measuring Mission Precipitation (TRMM) 3B42 product in southern Brazil by comparing daily precipitation over 13 years (V6 was tested for historical context). The interpolated data from 25 rain gauges were compared with both versions of TRMM. The V7 product tended to produce a slight increase in PC (proportion correct). V7 also showed a slight increase in the correlation coefficient (CC) and a significant increase in the H (hit rate) and CSI (critical success) indexes. However, the upgraded version shows an undesirable increase in the false alarm ratio. When the rainfall volumes were compared, V6 clearly underestimated the total rainfall over the entire period, but the V7 product slightly overestimated the cumulative volume (11%) which still represented a more reliable estimate than from V6. Furthermore, the main improvement in V7 was a large increase in the quantitative recognition of extreme precipitation events: V6 detected only 1% of the daily rainfalls above 60 mm, whereas V7 detected 57% of the events.