Abstract
A two component mixed log-normal distribution effectively models annual precipitation totals at two stations in Peru characterized by widely differing interannual patterns of precipitation. Physical evidence supports the division of station records into two subsamples. Years with ENSO events and years without ENSO events identify the components of a mixed probability model. The mixed model produces a superior fit to the two parameter log-normal distribution. Model application provides a reliable means of precipitation prediction and also quantitatively describes the highly variable temporal and spatial pattern of annual precipitation in western Peru.
Notes
∗ This research was funded in part by National Science Foundation Grant SES 8713738. We are grateful for the constructive comments of Cesar Caviedes, anonymous reviewers and the editors, and for the cartographic assistance of Mr. Steve Rogers.