Abstract. Statistical Postprocessing of medium-range weather forecasts is an important component of modern forecasting systems. Since the beginning of modern data science, numerous new postprocessing methods have been proposed, complementing an already very diverse field. However, one of the questions that frequently arises when considering different methods in the framework of implementing operational postprocessing is the relative performance of the methods for a given specific task. It is particularly challenging to find or construct a common comprehensive dataset that can be used to perform such comparisons. Here, we introduce the first version of EUPPBench, a dataset of time-aligned forecasts and observations, with the aim to facilitate and standardize this process. This dataset is publicly available at https://github.com/EUPP-benchmark/climetlab-eumetnet-postprocessing-benchmark. We provide examples on how to download and use the data, propose a set of evaluation methods, and perform a first benchmark of several methods for the correction of 2-meter temperature forecasts.

Abstract. Statistical postprocessing of medium-range weather forecasts is an important component of modern forecasting systems. Since the beginning of modern data science, numerous new postprocessing methods have been proposed, complementing an already very diverse field. However, one of the questions that frequently arises when considering different methods in the framework of implementing operational postprocessing is the relative performance of the methods for a given specific task. It is particularly challenging to find or construct a common comprehensive dataset that can be used to perform such comparisons. Here, we introduce the first version of EUPPBench (EUMETNET postprocessing benchmark), a dataset of time-aligned forecasts and observations, with the aim to facilitate and standardize this process. This dataset is publicly available at https://github.com/EUPP-benchmark/climetlab-eumetnet-postprocessing-benchmark (31 December 2022) and on Zenodo (https://doi.org/10.5281/zenodo.7429236, Demaeyer, 2022b and https://doi.org/10.5281/zenodo.7708362, Bhend et al., 2023). We provide examples showing how to download and use the data, we propose a set of evaluation methods, and we perform a first benchmark of several methods for the correction of 2 m temperature forecasts.

Current practice in predicting future weather is the use of numerical weather prediction (NWP) models to produce ensemble forecasts. Despite of enormous improvements over the last few decades, they still tend to exhibit bias and dispersion errors and, consequently, lack calibration. Therefore, these forecasts may be improved by statistical postprocessing. In this work, we propose a D‐vine‐copula‐based postprocessing for 10 m surface wind speed ensemble forecasts. This approach makes use of quantile regression related to D‐vine copulas, which is highly data driven and allows one to adopt more general dependence structures as the state‐of‐the‐art zero‐truncated ensemble model output statistic (tEMOS) model. We compare local and global D‐vine copula quantile regression (DVQR) models to the corresponding tEMOS models and their gradient‐boosting extensions (tEMOS‐GB) for different sets of predictor variables using one lead time and 60 surface weather stations in Germany. Furthermore, we investigate which types of training periods can improve the performance of tEMOS and the D‐vine‐copula‐based method for wind speed postprocessing. We observe that the D‐vine‐based postprocessing yields a comparable performance with respect to tEMOS if only wind speed ensemble variables are included and to substantial refinements if additional meteorological and station‐specific weather variables are integrated. As our main result, we note that, in the global setting, DVQR is able to provide better scores than tEMOS‐GB in general, whereas the local DVQR is able to substantially outperform the local tEMOS‐GB at particular stations admitting nonlinear relationships among the variables. In addition, we remark that training periods capturing seasonal patterns perform the best. Last but not least, we adapt a criterion for calculating the variable importance in D‐vine copulas and we outline which predictor variables are due to this approach the most important for the correction of wind speed ensemble forecasts.

<p>Statistical postprocessing of ensemble forecasts has become a common practice in research to correct biases and errors in calibration. Meanwhile, machine learning methods such as quantile regression forests or neural networks are often suggested as promising candidates in literature. However, interpretation of these methods is not always straightforward. <br />Therefore, we propose the D-vine (drawable-vine) copula based postprocessing, where for the construction of a multivariate conditional copula the graphical D-vine model serves as building plan. The conditional copula is based on this tracetable model using bivariate copulas, which allow to describe linear as well as non-linear relationships between the response variable and its covariates. Additionally, our highly data-driven model selects the covariates based on their predictive strength and thus provides a natural variable selection mechanism, facilitating interpretability of the model. Finally, (non-crossing) quantiles from the obtained conditional distribution can be utilized as postprocessed ensemble forecasts. <br />In a case study for the postprocessing of 10 m surface wind speed ensemble forecasts with 24 hour lead time we compare local and global D-vine copula based models to the zero-truncated ensemble model output statistics (tEMOS) for different sets of predictor variables at 60 surface weather stations in Germany. Furthermore, we investigate different types of training periods for both methods. We observe that the D-vine based postprocessing yields a comparable performance with respect to tEMOS models if wind speed ensemble variables are included only and a significant improvement if additional meteorological and station specific weather variables are integrated. The case study indicates that training periods capturing seasonal patterns are performing best for both models. Additionally, we provide a criterion for calculating the variable importance in D-vine copulas and utilize it to outline which predictor variables are the most important for the correction of 10 m surface wind speed ensemble forecasts.</p>

No abstract