Interannual variability of tropical precipitation as represented in satellite-based data sets

Since the beginning of the World Climate Research Program's Global Precipitation Climatology Project (GPCP) satellite remote sensing of precipitation has made dramatic improvements, particularly for tropical regions. Data from microwave and infrared sensors now form the most critical input to precipitation data sets and can be calibrated with surface gauges to so that the strengths of each data source can be maximized in some statistically optimal sense. It is clear however that there still remain significant uncertainties with satellite precipitation retrievals which limit their usefulness for many purposes. Systematic differences in tropical precipitation estimates have been brought to light in comparison activities such as the GPCP Algorithm Intercomparison Project and more recent Wetnet Precipitation Intercomparison Project 3. These uncertainties are assuming more importance because of the demands for validation associated with global climate modeling and data assimilation methodologies.

This paper examines the ability of present satellite-based data sets to depict interannual variations in tropical precipitation associated primarily with ENSO events. In particular, we show that an index of precipitating ice, DCI, determined from the MSU2 sensors shows remarkable coherence with SST variations when averaged over the tropical oceans. While this data set is corroborated by SSM/I precipitation estimates produced by the Wentz / Spencer algorithm and TRMM precipitation from the combined algorithm, other frequently used data sets (e.g. GPCP and Xie-Arkin) lack this correspondence to tropically-averaged SST. On the other hand, tropical land-averaged precipitation in these data sets shows better agreement, possibly because of the availability of more surface gauge data with which to intercalibrate.

An anti-correlation between tropical land and ocean precipitation is found in these data sets which seems to indicate a perturbing of the global monsoon and associated moisture transport during ENSO events. Consistency of these relationships is assessed by examining parallel diagnostics of moisture budgets variability constructed from the new NCEP Reanalysis-2 which has improved deep convective heating and moistening compared to the original reanalysis model.