A spatio-temporal model for temporal evolution of spatial extremal dependence

par Manal ABDULKAREEM

vendredi 11 avr. 2025, 09:00 → 10:00 Europe/Paris

Salle 112, bâtiment Braconnier (La doua)

In recent years, climate change has significantly impacted the nature of extreme environmental events,  highlighting the need for statistical models that accurately capture their evolution with time. In this study, we address the issue of the temporal non-stationarity of spatial extremal dependence, challenging the common assumption of stationarity when modeling long-term extreme environmental events. We propose a space-time model named the time-varying max-stable mixture model. It consists of a mixture of max-stable spatial processes, with time-dependent mixing proportions modeled as a linear trend. For statistical inference, the maximum composite likelihood estimator (MCLE) is used to estimate the parameters of the proposed model. Additionally, we propose a test statistic on the values of time-varying mixing proportion parameters of the proposed model, intending to validate the existence of a temporal trend. Finally, we use the proposed model to investigate how the rainfall in southern France evolves with time.