Intercomparison of Deep Learning Architectures for the Prediction of Precipitation Fields With a Focus on Extremes

Abstract

In recent years, the use of deep learning methods has rapidly increased in many research fields. Similarly, they have become a powerful tool within the climate scientific community. Deep learning methods have been successfully applied for different tasks, such as the identification of atmospheric patterns, weather extreme classification, or weather forecasting. However, due to the inherent complexity of atmospheric processes, the ability of deep learning models to simulate natural processes, particularly in the case of weather extremes, is still challenging. Therefore, a thorough evaluation of their performance and robustness in predicting precipitation fields is still needed, especially for extreme precipitation events, which can have devastating consequences in terms of infrastructure damage, economic losses, and even loss of life. In this study, we present a comprehensive evaluation of a set of deep learning architectures to simulate precipitation, including heavy precipitation events ( > 95th percentile) and extreme events ( > 99th percentile) over the European domain. Among the architectures analyzed here, the U‐Net network was found to be superior and outperformed the other networks in simulating precipitation events. In particular, we found that a simplified version of the original U‐Net with two encoder‐decoder levels generally achieved similar skill scores than deeper versions for predicting precipitation extremes, while significantly reducing the overall complexity and computing resources. We further assess how the model predicts through the attribution heatmaps from a Layer‐wise Relevance Propagation (LRP) explainability method.

Automated Input Variable Selection for Analog Methods Using Genetic Algorithms

“…M. Hamill et al, 2015;Ben Daoud et al, 2016), climate change studies (e.g., Dayon et al, 2015;Raynaud et al, 2016), or past climate reconstruction (Caillouet et al, 2016). AMs are also used for other predictands, such as precipitation radar images (Foresti et al, 2015;Panziera et al, 2011), temperature (Caillouet et al, 2016;Delle Monache et al, 2013;Jézéquel et al, 2017;Raynaud et al, 2016), wind (Delle Monache et al, 2013;Delle Monache et al, 2011;Vanvyve et al, 2015; Alessandrini, Delle Monache, more and more popular in the context of forecasting, postprocessing and downscaling (e.g., Chapman et al, 2022;Leinonen et al, 2020;Miralles et al, 2022;Otero & Horton, 2023), AMs are still relevant and offer the benefit of interpretability. Some AMs combine different predictors together using weights in the calculation of the distances between the target and the analog situations (e.g., Keller et al, 2017;Meech et al, 2020).…”

“…AMs are also used for other predictands, such as precipitation radar images (Foresti et al., 2015; Panziera et al., 2011), temperature (Caillouet et al., 2016; Delle Monache et al., 2013; Jézéquel et al., 2017; Raynaud et al., 2016), wind (Delle Monache et al., 2013; Delle Monache et al., 2011; Vanvyve et al., 2015; Alessandrini, Delle Monache, Sperati, & Nissen, 2015; Junk, Delle Monache, Alessandrini, Cervone, & von Bremen, 2015; Junk, Delle Monache, & Alessandrini, 2015), and solar radiation or power production (Alessandrini, Delle Monache, Sperati, & Cervone, 2015; Bessa et al., 2015; Raynaud et al., 2016). Although deep learning methods nowadays become more and more popular in the context of forecasting, postprocessing and downscaling (e.g., Chapman et al., 2022; Leinonen et al., 2020; Miralles et al., 2022; Otero & Horton, 2023), AMs are still relevant and offer the benefit of interpretability.…”

Automated Input Variable Selection for Analog Methods Using Genetic Algorithms

“…Although deep learning methods nowadays become more and more popular in the context of forecasting, postprocessing and downscaling (e.g. Chapman et al, 2022;Leinonen et al, 2020;Miralles et al, 2022;Otero & Horton, 2023), AMs are still relevant and offer the benefit of interpretability. Some AMs combine different predictors together using weights in the calculation of the distances between the target and the analog situations (e.g.…”