179 Additive Logistic Regression Models for Convective Hazards and their Application to Reanalyses and Climate Scenarios

Thursday, 10 November 2016
Broadway Rooms (Hilton Portland )
Anja T. Raedler, Munich Re, Munich, Germany; and P. Groenemeijer, T. Pucik, R. Sausen, and E. Faust

We developed statistical models for large hail and severe wind gusts that take a range of atmospheric predictors as input parameters. In order to model these hazards, the occurrence of each of them was regarded as the product of the probability of a thunderstorm P(storm), and the probability of a hazard given that a storm occurs.

First, a proxy parameter for the occurrence of electrified convection P(storm) was developed as a function of several predictors derived from the ERA-Interim global atmospheric reanalysis. Lightning data from European Cooperation for Lighting Detection served as evidence for thunderstorm occurrence. Subsequently, the probabilities of the following hazards were modelled: hail exceeding 2 cm in diameter, very large hail exceeding 5 cm in diameter and severe wind gusts ≥ 25 m/s given that a convective storm occurred in the first place.

The probability functions were computed by an additive logistic regression using lightning observations and severe weather reports from the European Severe Weather Database for Central Europe in the period 2008-2013.

The statistical models for thunderstorms, large and very large hail and severe wind gusts have been applied to ERA-Interim reanalysis data for the period 1979-2014. The models are capable  of modelling the annual cycle of thunderstorms and the hazards with reasonable accuracy.

The frequency of severe weather events in future climate scenarios was calculated by applying the models to an ensemble of EuroCORDEX regional climate simulations. We will present the resulting trends and changes in severe weather variability.

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner