Weather forecasting is challenging because of the complex interplay between local conditions and regional atmospheric forcings. In this article, we analyse the relationships between local daily rainfall and large‐scale synoptic patterns in the geographical context of Réunion Island, a high volcanic island in the southwestern Indian Ocean basin. Given the critical role of trade winds on weather conditions at island scale, we analyse those relationships across seasons defined with respect to yearly trade‐wind regimes. The analysis of the distribution of inversion events' elevation and frequency allows us to characterize the trade‐wind inversion layer (TWIL) and identify four seasons with homogeneous distributions. We characterize the spatio‐temporal variability of rainfall measured at island scale by a dense network of rain‐gauges over 37 years and relate it to large‐scale weather regimes identified using geopotential height meteorological data. After seasonal signal removal using Fourier transforms and dimension reduction via Principal Component Analysis, we perform Canonical Correlation Analysis to identify canonical variables relating rainfall and geopotential height patterns at the two different studied spatial scales. We then combine Ascending Hierarchical Classification and partitioning k‐means methods to identify homogeneous large‐scale synoptic conditions within each season. From this information, we build composite maps of geopotential height that characterize weather regimes and further analyse rainfall patterns at island scale in each regime. Elevation and orientation are used as descriptive variables at island scale as they strongly structure the patterns. Overall, four homogeneous seasons were identified based on trade‐wind regimes, within which six large‐scale weather regimes were identified for summer, the most variable season, and four for the others. Finally, rainfall patterns at island scale are described in relation to the highlighted synoptic regimes and local descriptive variables.