Abstract

The breaking of temperature records has become a prominent issue globally. A key question is how a changing climate influences the frequency of these extreme events. This work considers a dataset of over sixty years (1960-2021) of daily maximum temperatures measured at 40 meteorological stations from peninsular Spain. The work introduces the first fully developed model to analyze the occurrence of record-breaking temperatures across years for any given day within the year. The analysis needs detailed modeling of the indicator events that define record-breaking sequences. Building on novel exploratory data analysis, we propose Bayesian hierarchical spatial logistic regression models for these indicator events. The selected model includes an explicit long-term trend, necessary autoregression terms, spatial pattern captured by the distance to the coast, useful interactions, and very strong daily spatial random effects. We also introduce additional model-based inference tools, including maps for the number of records over specific time periods, maps for the ratio of the predicted number of records over the expected number of records under stationary climate conditions, maps of record probabilities on given days, and time series for the extent of record surface. The findings reveal that the number of temperature records has doubled in the last decade compared to expectations without climate change. Notably, deviations from the expected stationary pattern have been observed since the late 20th century, varying both spatially and temporally.