MOPREDAScentury: a long-term monthly precipitation grid for the Spanish mainland

Abstract. This article describes the development of a monthly precipitation dataset for the Spanish mainland (western Mediterranean basin), covering the period between December 1915 and December 2020. The dataset combines ground observational data from the National Climate Data Bank (NCDB) of the Spanish national climate and weather service (AEMET) and new data rescued from meteorological yearbooks published prior to 1951 that was never incorporated into the NCDB. The yearbooks data represented a significant improvement of the dataset, as it almost doubled the number of weather stations available during the first decades of the 20^th century, the period when the dataset was more scarce. The final dataset contains records from 11,312 stations, although the number of stations with data in a given month varies largely between 674 in 1939 and a maximum of 5,234 in 1975. Spatial interpolation was used on the resulting dataset to create monthly precipitation grids. The process involved a two-stage process: estimation of the probability of zero-precipitation (dry month), and estimation of precipitation magnitude. Interpolation was carried out using universal kriging, using anomalies (ratios with respect to the 1961–2000 monthly climatology) as dependent variable and several geographic variates as independent variables. Cross-validation results showed that the resulting grids are spatially and temporally unbiased, although the mean error and the variance deflation effect are highest during the first decades of the 20^th century, when the observational dataset was more scarce. The dataset is available at https://doi.org/10.20350/digitalCSIC/15136 under an open license, and can be cited as Beguería et al. (2023).