Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Bivariate Postprocessing of Wind Vectors

Ferdinand Buchner, David Jobst, Annette Möller, Claudia Czado

Abstract

To quantify the uncertainty in numerical weather prediction (NWP) forecasts, ensemble prediction systems are utilized. Although NWP forecasts continuously improve, they suffer from systematic bias and dispersion errors. To obtain well calibrated and sharp predictive probability distributions, statistical postprocessing methods are applied to NWP output. Recent developments focus on multivariate postprocessing models incorporating dependencies directly into the model. We introduce three novel bivariate postprocessing approaches, and analyze their performance for joint postprocessing of bivariate wind vector components for 60 stations in Germany. Bivariate vine copula based models, a bivariate gradient boosted version of ensemble model output statistics (EMOS), and a bivariate distributional regression network (DRN) are compared to bivariate EMOS. The case study indicates that the novel bivariate methods improve over the bivariate EMOS approaches. The bivariate DRN and the most flexible version of the bivariate vine copula approach exhibit the best performance in terms of verification scores and calibration.

Bivariate Postprocessing of Wind Vectors

Abstract

To quantify the uncertainty in numerical weather prediction (NWP) forecasts, ensemble prediction systems are utilized. Although NWP forecasts continuously improve, they suffer from systematic bias and dispersion errors. To obtain well calibrated and sharp predictive probability distributions, statistical postprocessing methods are applied to NWP output. Recent developments focus on multivariate postprocessing models incorporating dependencies directly into the model. We introduce three novel bivariate postprocessing approaches, and analyze their performance for joint postprocessing of bivariate wind vector components for 60 stations in Germany. Bivariate vine copula based models, a bivariate gradient boosted version of ensemble model output statistics (EMOS), and a bivariate distributional regression network (DRN) are compared to bivariate EMOS. The case study indicates that the novel bivariate methods improve over the bivariate EMOS approaches. The bivariate DRN and the most flexible version of the bivariate vine copula approach exhibit the best performance in terms of verification scores and calibration.
Paper Structure (24 sections, 24 equations, 11 figures, 3 tables)

This paper contains 24 sections, 24 equations, 11 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Data
  3. Observation and forecast data.
  4. Model training, validation and testing.
  5. Notation.
  6. Methods
  7. Bivariate Ensemble Model Output Statistics
  8. Gradient-Boosted Bivariate Ensemble Model Output Statistics
  9. Bivariate Distributional Regression Neural Network
  10. Bivariate Vine Copula Regression
  11. Verification Measures
  12. Proper scoring rules.
  13. Sharpness.
  14. Calibration.
  15. Testing for statistical significance.
  16. ...and 9 more sections

Figures (11)

  • Figure 1: Observation stations with $10$ m surface $u$- and $v$-component averaged over all 880 training days.
  • Figure 2: Schematic neural network with one hidden layer adapted from Rasp2018.
  • Figure 3: Y-vine tree structure with 3 covariates.
  • Figure 4: Empirical marginally normalized contour plots for observed and forecast wind vector component variables at station Kandern-Gupf for training data. Non-elliptical contours imply non-Gaussian dependence.
  • Figure 5: ES-based univariate PIT histograms aggregated over 60 stations and 944 days in the test data for six models. RI is the reliability index.
  • ...and 6 more figures