J. Castillo-Mateo, A. E. Gelfand, A. C. Cebrián Guajardo, J. Asín Lafuente, J. Abaurrea
This work proposes a Bayesian hierarchical spatio-temporal model for daily maximum temperature which introduces several innovations. The model adopts two temporal scales, year and day within year. It captures temporal dependence through autoregression on days within year and on years. In addition to the fixed effects (linear trend, seasonality and elevation), the complex spatio-temporal structure of temperature requires four spatial Gaussian processes to model intercept, slope, autocorrelation and residual variance parameters, and three pure error terms (years, sites within years and sites for days within years). The model is fitted with a MCMC algorithm. This enables inference for parameters and provides spatio-temporal predictions at unobserved locations. These predictions from the model are used to analyze climate change in a region of the Ebro basin. In particular, an approach to compute the extent, percentage of area under certain conditions, is proposed.
Keywords: autorregressive model, Gaussian process, hierarchical modeling, long-term trend, MCMC, spatial extent
Scheduled
Spatial Statistics and Temporal Space I
June 7, 2022 3:30 PM
Conference hall
