A0285
Title: Spatiotemporal modeling of record-breaking daily maximum and minimum temperatures
Authors: Jorge Castillo-Mateo - University of Zaragoza (Spain) [presenting]
Zeus Gracia - University of Zaragoza (Spain)
Jesus Asin - University of Zaragoza (Spain)
Ana C Cebrian - University of Zaragoza (Spain)
Alan Gelfand - Duke University (United States)
Abstract: The increasing occurrence of record-breaking temperatures presents a critical global challenge. To investigate the influence of climate change on these extreme events, daily maximum and minimum temperature data from 40 meteorological stations across peninsular Spain were analyzed over the period 1960-2023. The dataset is encoded as a daily pair of binary events, indicators, for that day, of whether a yearly upper record was broken for the daily maximum temperature and/or for the daily minimum temperature. A Bayesian hierarchical spatiotemporal bivariate probit regression model was developed to study the probability of such events. The model incorporates an explicit long-term trend, vector autoregressive components, spatial effects based on distance to the coast, relevant interactions, and strong daily spatial random effects. The spatial random effects are modeled using coregionalized Gaussian processes, where covariates are incorporated both in the coregionalization structure and in the covariance function to account for non-stationarity and anisotropy. Model-based inference reveals a strong correlation between record-breaking maxima and minima, with distinct spatial and temporal patterns of climate change signals, as well as differing degrees of persistence and spatial dependence in the two processes.
