Extreme weather prediction
Extreme weather events can have huge impact on society. Wind gusts of more than 100km/h can overthrow trees, snowfall can disrupt the train system and heavy precipitation can lead to flooding. Therefore, it is crucial to take precautions to reduce the societal impact of extreme weather events.
The big problem is, when do we warn the society to take preventative action? False alarms lead to a big waste of social resource. Moreover, warning society for extreme weather when it is not extreme can have dulling effect to the warning itself. On the other hand, missing an extreme event would result in a lot of damages.
A decent solution to the following question enables to build a reliable alarm system:
“What is, given the information we have today, the probability distribution of the extreme weather in the coming days?”
In our research, we address this problem by combining techniques from extreme value theory and quantile regression among others to develop new statistical tools for this application.
Applications of the probability distributions of the weather appear in many places. For example, think of the weather alarms that KNMI issues, code yellow, orange and red. The waterboards in the Netherlands need to decide based on the precipitation prediction whether or not to pump out water to prevent flooding. Finally, based on the probability of snowfall, Pro-Rail makes the decision to change the timetable in order to make sure the train system will not be shut down.