Improving Subseasonal and Seasonal sUmmer forecast over southern Europe through machine Learning

In the recent years, the continual improvements of weather forecasting models and the sustained need for reliable weather predictions beyond the weekly timescale resulted in the development of subseasonal to seasonal (S2S) forecast models and an intense research work from the scientific community. Despite the large number of research studies, S2S forecast models still show a limited skill in summer over Europe.

In addition, southern Europe, has received much less attention, even though it is highly vulnerable to high-impact summer heatwaves, and very sensitive to climate change.

The aim of this project, ISSUL, is to better understand and improve the S2S prediction of heatwave frequency and intensity and their associated weather patterns over southern Europe.

To do this, a combination of two machine learning algorithms, an optimisation algorithm, to identify the best set of predictors, and a neural network, to provide non-linear predictions will be used. This approach has never been attempted before for these timescales.

It is expected to perform better than standard S2S forecast models in predicting heatwave frequency and intensity and associated weather patterns and to bring larger improvements compared with traditional statistical forecasts that do not identify all the predictors and cannot represent non-linear complex interactions.ISSUL is divided into three parts.

The first part aims at identifying the best set of predictors, using the optimisation algorithm, at evaluating it and understanding it is related to heatwaves over southern Europe via a dynamical analysis.

The second part aims a predicting the frequency and intensity of heatwaves and associated weather patterns using a neural network.

The third part aims at evaluating the performance of this combined machine learning approach compared with standard S2S forecasting model.