"There is not yet enough real-world data to train the forecast sufficiently, so the ML models actually learned about heat wave drivers in a model world but successfully applied the training to the ...