Spatio–Temporal Modeling and Spatial Inference Using NA-CORDEX Climate Data

Abstract Responding to climate change requires not only modeling changing trends in temperature and precipitation over time but also doing so spatially, identifying geographical regions where such changes are more extreme. This work presents the construction of geographically weighted multivariate time-series models for temperature and precipitation, considering distinct distributional data assumptions in both spatial and temporal domains. To identify regions of significant change in the estimated trend parameters in the proposed models, we employ the spatial inference method of coverage probability excursion (CoPE) sets, providing spatial uncertainty on the regions whose changes exceed a certain threshold. The North American CORDEX (NA-CORDEX) data are used as a case study for the proposed spatio-temporal models and statistical inference method. The proposed models are used to analyze climate changing trends in three distinct geographical areas of the USA (the states of California, Colorado and Kansas), considering both historical and future series. The proposed models can serve as a reference for future studies related to regional climate data modeling and inference.