Development of a deep learning based system for heatwave detection using seasonal forecast data
- 1Center for International Development and Environmental Research, Justus Liebig University Giessen, Senckenbergstrasse 3, 35390 Giessen, Germany
- 2Department of Geography, Climatology, Climate Dynamics and Climate Change, Justus Liebig University Giessen, Senckenbergstrasse 1, 35390 Giessen, Germany
Heatwaves have been widely studied in recent years because of their major impact on human health, mortality, ecosystems, agriculture, and the economy. Globally, heatwaves are becoming more severe, longer, and recurrent with global temperature rise. Therefore, the study of heat waves and the development of an early warning system for prediction of regional heatwaves help climate preparedness and decision-making. In this research, we propose a heatwave prediction algorithm based on a deep learning model, a convolutional neural network (CNN). This CNN model is trained with reanalysis data ERA5 and real heatwave events from EMO observation data for years from 1993 to 2021. We illustrate the relationship between the patterns in geopotential height at 500 hpa (GPH), sea surface temperature (SST), and the real heatwaves that happened in the last 20 years. This study employs the hindcast data from SEAS5.1 with 25 ensemble members, available at C3S. GPH and SST from observation data are input to the model and the heatwave magnitude at every single grid point is the output. The heatwave is defined as a period of three or more consecutive hot days and nights when the daily maximum and minimum temperature (TX/TN) exceeds the long‐term (1993–2022) daily 90th percentile. For estimating the heat wave magnitude we accumulated TX exceedance the local 90th percentile for all heat wave days over a user-defined interval (monthly, seasonal, etc.) as in Zampieri et al. (2017), Toreti et al. (2019). The results show the CNN model using atmospheric circulation fields (SST and GPH) with adjusted parameters is able to forecast extreme events in Europe, and it can potentially enhance the AI-based early warning systems for extreme weather.
Zampieri, M., Ceglar, A., Dentener, F., and Toreti, A. (2017). Wheat yield loss attributable to heat waves, drought and water excess at the global, national and subnational scales. Environmental Research Letters, 12 (6), 064008. doi:10.1088/1748-9326/aa723b
Toreti, A., Cronie, O., and Zampieri, M. (2019). Concurrent climate extremes in the key wheat producing regions of the world. Scientific Reports, 9(1), 5493. doi:10.1038/s41598-019-41932-5
How to cite: Heidari, F., lin, Q., Díaz Esteban, Y., Espitia Sarmiento1, E., and Xoplaki, E.: Development of a deep learning based system for heatwave detection using seasonal forecast data, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-17545, https://doi.org/10.5194/egusphere-egu24-17545, 2024.
