We combined meteorological data with additional information (soil, rock, land cover and elevation) on a single grid that spans five river basins in central Europe over 45 years. Data was not aggregated within river catchments so as to retain the full spatial covariance. It can be efficiently processed in parallel and has been used in our recent study on end-to-end rainfall streamflow modeling with neural networks. We hope to promote further development of spatially resolved modeling approaches.