COBASE example — A copula-based shuffling for ensemble forecast postprocessing

This repository accompanies the paper:

COBASE: A new copula-based shuffling method for ensemble weather forecast postprocessing
Maurits Flos, Bastien François, Irene Schicker, Kirien Whan, and Elisa Perrone
(Eindhoven University of Technology, KNMI, and GeoSphere Austria)

---

About this project

Weather forecasts are often generated as ensembles, i.e., multiple runs of a numerical weather prediction model, to represent uncertainty.
However, these ensembles usually need statistical postprocessing to correct biases and restore realistic dependence structures across variables and locations.

The COBASE framework introduced in the paper is a copula-based postprocessing approach that combines the strengths of parametric and nonparametric methods through a rank-shuffling mechanism. This design allows COBASE to preserve calibrated univariate forecasts while reconstructing realistic multivariate dependence.

This example uses mock ensemble-style data to demonstrate the workflow.

---

Project Structure

COBASE_github/
├── COBASE_github.Rproj           # R Project file
├── main.r                        # Main script controlling the workflow
├── config.yml                    # Configuration file for directories and model settings
│
├── Creating Plots/               # Scripts for generating figures and tables
│   ├── create_plots_paper_selection.R
│   └── Utilities/
│       ├── dm_plots.R
│       ├── skill_scores.R
│       ├── univariate_pit.R
│       ├── score_boxplots.R
│       └── create_score_tables.R
│
├── Postprocessing/               # Postprocessing routines
│   ├── univariate.R
│   ├── multivariate.R
│   └── Utilities/
│       ├── mvpp_methods.R
│       ├── mvpp.R
│       ├── sim_util.R
│       ├── DM_util.R
│       ├── ensfc_util.R
│       ├── scores_util.R
│       ├── uvpp_util.R
│       └── load_data_util.R
│
├── Data/                         # Input and mock datasets
│   ├── Datasets/
│   │   └── Mock_data.csv         # Mock dataset used for demonstration
│   └── ENS/
│       └── transformed_data_Mock_data.Rdata
│
├── Results/                      # Generated results and figures
│   ├── Figures/
│   │   ├── flos_dmcrps_t2m.pdf
│   │   ├── flos_dmesvs_simssh.pdf
│   │   ├── flos_dmesvs_mvpp.pdf
│   │   ├── flos_dmesvs_parcop.pdf
│   │   ├── crps_table.tex
│   │   └── esvs_table.tex
│   ├── Scores/
│   │   └── score_env_Mock_data_mout_10.RData
│   ├── TestStatistic/
│   │   └── dm_statistics_Mock_data_mout_10.Rdata
│   ├── SimilarityMatrix/
│   │   └── simMatrix_Mock_data.Rdata
│   └── UVPP/
│       └── uvpp_Mock_data.Rdata
│
└── README.md

---

⚙️ Requirements

Tested with R ≥ 4.2.

Required R packages:

install.packages(c("dplyr", "tidyr", "ggplot2", "xtable", "yaml"))

---

▶️ Running the example

1. Open the R Project

   • Open COBASE_github.Rproj in RStudio to automatically set the working directory.

2. Check the configuration

   • Review config.yml to ensure paths and directories match your local setup.

3. Run the workflow

   • Execute main.r, which:
     • Loads the mock dataset.
     • Performs univariate and multivariate postprocessing.
     • Generates the scores, diagnostics, and plots under Results/.

4. Inspect results

   • Figures and tables used in the paper are reproduced under:

     Results/Figures/
     

---

Notes

• The included data are mock examples for demonstration only.
• In the original study, COBASE was applied to real ALADIN-LAEF and ECMWF ensemble data.
• The scripts and structure provided here are designed for clarity and reproducibility.

---

Note on Code Origin

Portions of this repository are adapted from the multiv_pp package by Lerch et al. (2020). These components have been modified and extended to support the specific case studies and methodological developments presented in the accompanying paper.

Reference:
Lerch, S., Baran, S., Möller, A., Groß, J., Schefzik, R., Hemri, S., & Gräter, M. (2020). Simulation-based comparison of multivariate ensemble post-processing methods. Nonlinear Processes in Geophysics Discussions. https://doi.org/10.5194/npg-2019-62

---

License

This repository is released under the MIT License.

---

Contact

For questions or collaboration, please contact:
Elisa Perrone — e.perrone@tue.nl