Statistical postprocessing of ensemble forecasts for severe weather at Deutscher Wetterdienst

Abstract

Abstract. This paper gives an overview of Deutscher Wetterdienst's (DWD's) postprocessing system called Ensemble-MOS together with its motivation and the design consequences for probabilistic forecasts of extreme events based on ensemble data. Forecasts of the ensemble systems COSMO-D2-EPS and ECMWF-ENS are statistically optimised and calibrated by Ensemble-MOS with a focus on severe weather in order to support the warning decision management at DWD. Ensemble mean and spread are used as predictors for linear and logistic multiple regressions to correct for conditional biases. The predictands are derived from synoptic observations and include temperature, precipitation amounts, wind gusts and many more and are statistically estimated in a comprehensive model output statistics (MOS) approach. Long time series and collections of stations are used as training data that capture a sufficient number of observed events, as required for robust statistical modelling. Logistic regressions are applied to probabilities that predefined meteorological events occur. Details of the implementation including the selection of predictors with testing for significance are presented. For probabilities of severe wind gusts global logistic parameterisations are developed that depend on local estimations of wind speed. In this way, robust probability forecasts for extreme events are obtained while local characteristics are preserved. The problems of Ensemble-MOS, such as model changes and consistency requirements, which occur with the operative MOS systems of the DWD are addressed.

Calibrated ensemble forecasts of the height of new snow using quantile regression forests and ensemble model output statistics

“…8b). The sensitivity of snow cover models to errors in precipitation phase was already illustrated by Jennings and Molotch (2019), with a meteorological forcing built from weather stations. The magnitude of errors is expected to be much higher when the forcing comes from NWP forecasts.…”

A comparison of statistical and dynamical downscaling methods for short‐term weather forecasts in the US Northeast

“…D01 has resolution of 9 km, while D02 and D03 have 3 km resolutions outputs predictands (e.g. T) that are significantly more accurate forecasts than the raw model data (Hess, 2020). While the MOS is advantageous to raw model output, the forecasts are still deterministic for most variables and are therefore disseminated in a non-probabilistic format, though the regression could provide a distribution of predictions based on the regression uncertainty.…”

“…The sharper the forecasts, the better, subject to calibration. Statistical post-processing of ensemble forecasts has been applied to operational forecasts at the world's top meteorological centers, including the National Oceanic and Atmospheric Administration (NOAA), the ECMWF, the Deustcher Wetterdienst (DWD) (Hess, 2020), the UK Met Office (Ayliffe and Roberts, 2019), and Météo-France (Taillardat and Mestre, 2020), as well as to forecasts by private companies for renewable energy and in hydrological centers (e.g. Parks et al, 2011).…”

Preface: Advances in post-processing and blending of deterministic and ensemble forecasts

“…The post-processing techniques should be able to cope with new types of high-resolution products and, at the same time, be highly efficient. These challenges are discussed by Hess (2020) in which the post-processing suite of the German national weather service (Deutscher Wetterdienst) is described. This suite uses relatively simple post-processing techniques, with a specific focus on the correction of extreme events.…”

“…Usually, univariate metrics are used. Jacobson et al (2020) propose use of metrics that take into account the spatial properties of the fields of interest. They propose a new diagnostic tool called the fraction of threshold exceedance, which allows projection of possibly high-dimensional multivariate fields onto a single univariate quantity, which can then be analyzed with standard metrics.…”