Previous studies show that the seasonal precipitation over land may have strong nonlinear relationships with the concurrent sea surface temperature (SST). In this study, we demonstrate that summer precipitation also has a strong nonlinear relationship with preceding SSTs, which are more robust than the linear relationships. A strategy is employed to demonstrate the robustness of the nonlinear relation. With the 60 year observed precipitation and SST, we use data of 50 years to fit the linear and nonlinear rainfall‐preceding SST relations. The SSTs used include all the samples that are from each of the global ocean grids over each of the preceding time periods considered. The fitted relations are then used to predict the rainfall of the remaining 10 years with the SST data. From all the predictions that use a large number of SST samples, we choose the ocean grid cell and the season that give the minimum prediction error. Results indicate that for an individual station, it may well be that the linear relation is better than the nonlinear relation. However, overall, the nonlinear method is better than the linear method as assessed from all stations in the country. This statistical analysis demonstrates that, when treated nonlinearly, the rainfall‐preceding SST relationship can be much strengthened.

Institute of Mathematics of the Czech Academy of Sciences provides access to digitized documents strictly for personal use. Each copy of any part of this document must contain these Terms of use.

No abstract