Precipitation estimates derived from passive microwave satellite observations are combined with ERA-40 reanalysis precipitation to produce a spatially complete monthly precipitation analysis for 1992-2002. The resulting analysis provides consistent global precipitation fields for this period, without problems associated with changing data input types in time or across land-sea boundaries. Over oceanic regions this reanalysis is similar to the GPCP analysis, which relies heavily on the same microwave-based oceanic estimates.

Using this new monthly 1992-2002 analysis, spatial covariance modes are computed for precipitation anomalies. These covariance modes are used to reconstruct precipitation over an extended historical period. For each month, the historical anomaly reconstruction is a weighted sum of the set of modes. Weights for the set of modes are computed by fitting the available gauge data to the set of modes such that the mean-squared error of the fit is minimized. Cross-validation tests indicate reconstruction skill over the oceans between approximately 30°S and 60°N, with little skill poleward of this region. Skill is almost constant since 1950, and it is only slightly reduced in the first half of the 20th century. Additional testing showed that including ship data along with gauge data can not greatly improve the skill in these reconstructions.