Global Oceanic Precipitation from the MSU during 1979—91 and Comparisons to Other Climatologies

Abstract

Oceanic precipitation is estimated on a 2.5° grid for the period 1979?1991 from Microwave Sounding Unit (MSU) channels 1, 2, and 3 data gathered by seven separate TIROS-N satellites. Precipitation is diagnosed when cloud water and rainwater-induced radiometric warming of the channel 1 brightness temperatures (Tb) exceeds a cumulative frequency distribution threshold of 15% after correction for airmass temperature determined from the channel 2 and 3 measurements. After intercalibration between satellites, the 13-year gridpoint field of average Tb warming is calibrated in precipitation units with data from five to ten years of globally distributed low-elevation island and coastal rain accumulation measurements from 132 gauges. The calibration involves a single scale factor, and has a dependence on air temperature that is estimated from an MSU climatology. Comparisons between the satellite and raingage measurements of the average annual cycle in monthly precipitation are presented for 75 raingage locations from different climatic regimes. At 2.5° gridpoint resolution, peak annual rainfall (5600 rim) occurs in a quasistationary portion of the ITCZ over the eastern Pacific, while peak monthly rainfall (over 900 mm) occurs in the northeastern Bay of Bengal in June. Comparisons between the MSU oceanic rainfall climatology and those of Jaeger (raingage), Legates and Wilmott (mostly ship synoptic code observations), and the GOES precipitation index (GPI) of Janowiak and Arkin (satellite infrared) reveal several important differences. Jaeger largely misses the extratropical storm tracks, as well as the intensity of the intertropical convergence zone in the eastern Pacific and western Atlantic. Legates and Wilmott have the features that Jaeger missed, but without the intensity that the MSU suggests. Two prominent differences in the Pacific ITCZ depiction are probably due to a lack of ship data in the Legates and Wilmott climatology. The GPI shows more rainfall over the eastern Indian Ocean than the MSU, and much less than the MSU over the tropical eastern Pacific, tropical western Atlantic, and in the western Pacific extratropical storm track. These differences are related to regional differences in the amount of cirrus cloud activity versus cloud water activity. The largest consistent discrepancy between the MSU and other climatologies is the eastern Pacific ITCZ, where the MSU indicates up to 8 mm day?1 more rainfall than the GPI. Raingage data from the Line Islands, which protrude into one end of this maximum, suggests that the large MSU amounts could be real. Microwave sounding unit and GPI depictions of seasonal rainfall variability associated with the ENSO warm event of 1991?92 show good agreement in the large-scale patterns. Comparisons of MSU pentad rain estimates for a small region near Sumatra to the GOES precipitation index (GPI) and two numerical weather prediction model forecasts show much better agreement between the two satellite estimates than between either satellite estimate and the model forecasts. Remaining disagreements between the two satellite methods are in the form of numerous GPI-diagnosed light-to-moderate rain amounts for which the MSU shows little or no precipitation. This is consistent with the expectation that cold cirrus cloudiness, on which the GPI estimates are based, often covers larger areas with more persistence than does the rainfall on short time scales. Monthly gridpoint anomaly sampling skill is estimated during 81 months of two-satellite coverage by comparisons between the anomalies measured by the satellites separately. At the 2.5° gridpoint scale, anomaly correlations are generally above 0.8 in the tropics, reaching a peak of 0.99 in the center of the tropical Pacific Ocean. Extratropical gridpoint anomaly correlations are lower, due to smaller interannual variability, ranging from 0.2 to O.8. The corresponding single-satellite gridpoint anomaly sampling error ranges from 10 mm month?1 to 40 mm month?1, the lower errors occurring in climatologically light rainfall areas, and greater errors in the heavy rainfall areas. An example of monthly anomalies in oceanic rainfall is presented for February 1983.