Seasonal prediction of droughts over European Russia using artificial intelligence methods

Among natural disasters, droughts cause the most damage and losses to the agricultural sector in Russia. Measures to minimise the damage caused by droughts depend on timely and accurate prediction of droughts. Seasonal prediction of droughts is challenging due to the non-linearity of processes in the coupled atmosphere-land system. For the same reason, it is expected that Artificial Intelligence (AI) methods may be superior to traditionally used statistical methods for drought prediction. The aim of this work is to apply and evaluate different methods of AI for seasonal drought prediction over European Russia. Several indices are used to identify droughts, such as the Standardised Precipitation Index (SPI), the Standardised Precipitation Evapotranspiration Index (SPEI) and the Selyaninov hydro-thermal coefficient (HTC). The latter index is traditionally used in agricultural meteorology in Russia and its relation to the production of various crops is well established. The indices are calculated for the period 1958-2022 using ERA5 reanalysis data. A set of predictors includes several indices characterising the large-scale atmospheric circulation over northern Eurasia (e.g. NAO, AO and others), as well as the ENSO index and the sea surface temperature anomaly over the North Atlantic. Several AI methods are used to predict the values of the drought indices based on the set of predictors for different lead times ranging from two weeks to several months. Prediction is made separately for each of the administrative regions in the European Russia with a special focus on those regions where droughts cause most damage. Another focus of the study is the relationship between droughts and atmospheric blocking events and heat waves. The application of the used AI methods for heat wave prediction is discussed.