As part of the effort to create a data set of global monthly precipitation for an extended period from 1948 to the present, a new method is developed to produce an analysis of oceanic precipitation anomaly by EOF reconstruction. Applied successfully by Smith et al.(1996) in reconstruction of SST fields, this method defines anomaly fields by interpolating sparsely distributed historical observations constrained by the spatial distributions of EOF patterns of satellite observations.

First, EOF analysis is performed for monthly anomaly fields of the precipitation estimates derived from satellite-observed OLR data (OPI, Xie and Arkin 1998) over global oceanic areas for the 20-year period from 1979 - 1998. The first 20 EOF modes are then used as spatial functions in the reconstruction. The anomaly fields of precipitation are reconstructed by projecting historical rain gauge observations over coastal regions and islands onto these EOF patterns.

Simulation tests and cross validation were conducted. The results show reasonable accuracy of the reconstruction in reproducing ENSO-related anomaly patterns over most of the tropical oceans. To further improve the performance over extra-tropical areas, experiments with regional and seasonal EOF reconstructions were conducted. Based on these results, a test product of the global monthly oceanic precipitation anomaly has been produced for an extended period from 1948 to the present and compared to the precipitation fields produced by numerical models. Details of the rescontruction and its potential applications will be reported at the meeting.