Recognizing a spatial extreme dependence structure: A deep learning approach
Manaf Ahmed, Véronique Maume‐Deschamps, Pierre Ribereau
Abstract
Abstract Understanding the behavior of extreme environmental events is crucial for evaluating economic losses, assessing risks, and providing health care, among many other related aspects. In a spatial context, relevant for environmental events, the dependence structure is extremely important, influencing joint extreme events and extrapolating on them. Thus, recognizing or at least having preliminary information on the patterns of these dependence structures is a valuable knowledge for understanding extreme events. In this study, we address the question of automatic recognition of spatial asymptotic dependence versus asymptotic independence, using a convolutional neural network (CNN). We designed a CNN architecture as an efficient classifier of a dependence structure. Extremal dependence measures are used to train the CNN. We tested our methodology on simulated and real datasets: air temperature data at 2 m over Iraq and rainfall data along the east coast of Australia.