Studying long-term variations and trends in grid datasets is often complicated by variations in the station network used for spatial interpolation. The increase in station density over time gives rise to inhomogeneities in the datasets. Here we introduce a gridded dataset of monthly precipitation for the entire Alpine region with a grid-spacing of 5 km that extends over more than 100 years (1901–2008), is largely unaffected by network variations and still exploits information from recent dense observations. It is derived by a reconstruction procedure that combines data from several thousand stations over recent decades with long-term and almost continuous station series at much coarser resolution (120 stations). Our dataset is a recalculation and update of a previous and similarly constructed dataset that ended in 1990. The reconstruction method is modified for the present application and an entirely new high-resolution reference was used for calibration. The reconstruction explains more than 80% of the month-by-month variance in Switzerland and Austria where the long-term station network is comparatively denser. We also demonstrate that the reconstruction can effectively filter out artifacts in trend patterns due to local inhomogeneities in the data and the station network. Precipitation trends calculated with the updated grid dataset are qualitatively similar with those found previously, but statistical significance has partly changed. A likely reason for these discrepancies is decadal variations, notably in winter, and this calls for prudence in the interpretation of trend results from conventional tests.