There is a general assumption that using “all” available precipitation estimates from passive microwave data will likely result in an improved multi-satellite estimate of global precipitation. In the course of developing the Multi-satellite Precipitation Analysis (MPA) system, the authors discovered that achieving a uniform, high-quality analysis using multiple passive microwave estimates as input raised several very important issues. The MPA aims to provide 0.25°x0.25° 3-hourly precipitation estimates for the global latitude belt 50°N-50°S. In the MPA, Tropical Rainfall Measuring Mission (TRMM) observations are used to calibrate precipitation estimates from SSM/I, AMSR-E, and AMSU-B. All these are merged, and the results are used to provide calibration parameters for a variable-rainrate, threshold-matched infrared (IR) technique. The final MPA product is computed as the merged microwave estimate, when available, and the calibrated IR otherwise.

The primary issue when incorporating a new microwave estimate is coping with systematic differences in the fraction of the time each sensor detects precipitation. For example, the currently available AMSU-B estimates have a distinct shortfall in detecting light precipitation over oceans, resulting in underestimates in the subtropical highs. SSM/I estimates suffer this problem to a lesser extent. A second issue is the degree to which the calibrations can be done once, either in space or time, as opposed to being computed for different regions and/or months. Once these and other issues are addressed, we need to assess whether the additional microwave data improves or degrades the final analysis compared to using the calibrated IR. In part, this is done by analyzing parallel data sets created by withholding various microwave data sources. The quality issue is considered both at the full-resolution scale, as well as at monthly climate scales.

Illustrations will be drawn from both the real-time MPA (RT-MPA) product that has been computed and publicly posted on an on-going basis starting in February 2002, and a post-real-time research version that was implemented in Version 6 of the TRMM data product 3B-42, which will cover the period from January 1998 to the (delayed) present. Even after the end of the TRMM era a revised RT-MPA will continue to be computed as an early test of the Global Precipitation Measurement (GPM) project’s Precipitation Processing System (PPS).