Wider et al. (2026) Evaluation code for 'unseen-awg: Spatio-temporal weather generation using analogs and unseen data'

⚠️ Warning: This summary was generated from the abstract only, as the full text was not available.

Identification

• Journal: Open MIND
• Year: 2026
• Date: 2026-04-22
• Authors: Jonathan Wider, Jakob Zscheischler
• DOI: 10.5281/zenodo.19698739

Research Groups

[Not specified in the provided text]

Short Summary

The paper introduces unseen-awg, an analog weather generator that utilizes reforecast data to produce high-resolution synthetic weather series capable of simulating unprecedented extremes and maintaining complex spatial and multivariate dependencies.

Objective

• To develop a weather generator that overcomes the limitations of traditional analog generators—specifically the underestimation of temporal correlations and the constraint of historical dataset length—to better simulate unseen weather extremes and dependencies.

Study Configuration

• Spatial Scale: Europe (case study), including grid-cell scale analysis.
• Temporal Scale: Daily timescale, covering the full annual cycle.

Methodology and Data

• Models used: unseen-awg (analog weather generator), incorporating a novel tuning strategy and block sampling.
• Data sources: Reforecasts (weather forecasts initialized with historical conditions).

Main Results

• Temporal & Distributional Accuracy: The generator successfully simulates the full annual cycle and captures the distributional properties of individual weather variables.
• Dependency Capture: It accurately represents the multivariate dependence between summer temperature and precipitation at the grid-cell scale.
• Extreme Event Simulation: unseen-awg can generate "unseen" extremes at a daily timescale, including heatwaves and droughts of unprecedented spatial extent.
• Continuity: Improved temporal continuity in generated time series compared to standard analog weather generators.

Contributions

• Advances the field of weather generation by utilizing reforecast data to break the dependency on the length of historical observation records.
• Provides a tool for assessing risks in water, agriculture, and forestry sectors by enabling the simulation of unprecedented weather hazards with preserved spatial and multivariate dependencies.

Funding

[Not specified in the provided text]

Citation

@article{Wider2026Evaluation,
  author = {Wider, Jonathan and Zscheischler, Jakob},
  title = {Evaluation code for 'unseen-awg: Spatio-temporal weather generation using analogs and unseen data'},
  journal = {Open MIND},
  year = {2026},
  doi = {10.5281/zenodo.19698739},
  url = {https://doi.org/10.5281/zenodo.19698739}
}