A1265
Title: Improving probabilistic forecasts of extreme winds by training post-processing models with weighted scoring rules
Authors: Jakob Wessel - University of Exeter (United Kingdom) [presenting]
Christopher Ferro - University of Exeter (United Kingdom)
Gavin Evans - Met Office (United Kingdom)
Frank Kwasniok - University of Exeter (United Kingdom)
Abstract: Accurate forecasts of extreme wind speeds are of high importance for many applications. Such forecasts are usually generated by ensembles of numerical weather prediction (NWP) models, which, however, can be biased and have errors in dispersion, thus necessitating the application of statistical post-processing techniques. The aim is to improve statistical post-processing models for probabilistic predictions of extreme wind speeds. It is done by adjusting the training procedure used to fit ensemble model output statistics (EMOS) models - a commonly applied post-processing technique - and propose estimating parameters using the so-called threshold-weighted continuous ranked probability score (twCRPS), a proper scoring rule that places special emphasis on predictions over a threshold. It is shown that training using the twCRPS leads to improved extreme event performance of post-processing models for a variety of thresholds. A distribution body-tail trade-off is found where improved performance for probabilistic predictions of extreme events comes with worse performance for predictions of the distribution body. However, strategies to mitigate this trade-off are introduced based on weighted training and linear pooling. Finally, some synthetic experiments are considered to explain the impact of training on the twCRPS. The results enables researchers and practitioners alike to improve the performance of probabilistic forecasting models for extremes and other events of interest.
