TOWARDS AN IMPROVEMENT OF THE FORECAST OF WIND RESOURCES IN EUROPE: APPLICATION OF UNSUPERVISED MACHINE LEARNING ON FUTURE PROJECTIONS OF POLAR VORTEX

摘要

Studying the impact of climate change on wintertime polar stratosphere is of particular relevance not only for climate knowledge but also for tropospheric projections. Machine learning provides a way to extract information from different climate models and combine the data in a such way that patterns are clearer, so predictions can be inducted. The methods used in this study have been region growing algorithm, K-means and combination of predictions. The final results show three clusters of response trends for the intensity and location of the stratospheric polar night jet. The prediction shows an increase in zonal wind intensity over the 2xC02 and 4xC02 concentration points and a decrease in the related latitude. Our methods can be extended for more climate models and simulation periods, and will allow not only to map the behavior of the polar night jet, but other stratospheric and tropospheric features and interactions between them.