Spatial modeling of extreme temperature in the Canadian Prairies using max-stable processes

Abstract

The Prairie region of Canada is an agriculture-intensive area and is strategic to Canada's food security. Moisture deficit due to extreme temperatures leading to high evapotranspiration can have significant impacts on water availability, resulting in poor crop yield. The primary objective of this study is to quantify the spatial structure and dependency of extreme temperatures by employing Max-Stable Process (MSP) modeling on daily annual maximum temperature data spanning 1970–2020. The spatial trend surface of the marginal parameters of the Spatial Generalized Extreme Value (SGEV) shows that geographical coordinates, topography due to the Rocky Mountains, and proximity (Euclidean distance) to Hudson Bay are important covariates in capturing the spatial trend of extreme temperatures. Furthermore, through the SGEV, important products such as the point-wise return periods and levels were derived using the best selected model as determined by Takeuchi's information criteria. The return levels show that the southern portion of the Canadian Prairies shows a consistent increase in extreme temperature for all the return periods. The results show that all the return periods mentioned above have extreme temperatures exceeding 37 °C in the southern portion of the Canadian Prairies. An unconditional simulation using the fitted MSP model provided various realizations of temperature extremes. The results from this study provide important insights into extreme temperatures in this important region, where water resources management is crucial. This study will be beneficial to the hydrologists, water resource specialists, climate change scientists and policy makers involved in the monitoring of extreme events in the Canadian Prairies.