The aim of this study is to build a high-resolution mapping model for Réunion, a mountainous island with highly complex terrain. The dataset used here, which consists of solar irradiation, is not available from the regular weather station network over the island. This network is relatively dense and includes quality-monitoring stations, thus providing enough information to tackle the problem of climate data interpolation over the complex terrain. A model for mapping the monthly means of such variables is presented. It combines Partial Least Squares (PLS) regression with kriging interpolation of residuals. For all the variables, the same set of nine predictors, including altitude, geographical and topographical features, was selected for PLS regression. The regression model gives statistically good estimates of monthly solar irradiation. Accuracy improves significantly using solar radiation mapping built with regression+kriging than for mapping built with regression only.