Abstract Identifying rainfall regions associated with specific modes of variability is of practical interest for water resources management, seasonal forecasting, and mitigation of weather‐related risks. This study aims to identify homogeneous rainfall regions within the ~1 million km 2 Upper Madeira River basin—southwestern Amazon—by their interannual and decadal variability and relates this variability to ocean indices. An observed dataset of 146 ground‐based rainfall stations, distributed throughout the Andes and the Amazon, and homogenized at the monthly time‐step for the period 1980–2016, was used for the analysis. With no spatial constraints, hierarchical cluster analysis and principal component analysis (PCA) optimally grouped stations into 10 rainfall homogenous regions. The value of the regionalization for interpreting the rainfall variability was evaluated by relating the seasonal rainfall time series of the regions with ocean indices. Then, by applying PCA to seasonal rainfall series and linking the principal components to sea surface temperature and ocean indices, an insight into the main large‐scale drivers of the rainfall spatio‐temporal variability in this basin at interannual and decadal scales is provided. This analysis identified differences in the year‐round influences of the tropical Pacific and/or Atlantic oceans on the 10 homogenous regions.