Sweet et al. (2026) A data-driven method for identifying climate drivers of agricultural yield failure from daily weather data

Identification

• Journal: Geoscientific model development
• Year: 2026
• Date: 2026-07-23
• Authors: Lily‐belle Sweet, Christoph Müller, Jonas Jägermeyr, Jakob Zscheischler A
• DOI: 10.5194/gmd-19-6687-2026

Research Groups

• Department of Compound Environmental Risks, Helmholtz Centre for Environmental Research – UFZ, Leipzig, Germany
• Department of Hydro Sciences, TUD Dresden University of Technology, Dresden, Germany
• Potsdam Institute for Climate Impact Research (PIK), Member of the Leibniz Association, Potsdam, Germany
• Columbia University, Climate School, New York, NY, USA
• NASA Goddard Institute for Space Studies (GISS), New York, NY, USA

Short Summary

A data-driven method is proposed to identify climate drivers of agricultural yield failure from daily weather data. The approach systematically generates and tests candidate features using machine learning and then selects a set of robust drivers.

Objective

• Identify the principal hypothesis that causal drivers should generalize across different environments.

Study Configuration

• Spatial Scale: Continental US, with a focus on maize cropping areas.
• Temporal Scale: Daily resolution for climate data, with a time period spanning 1901–2019.

Methodology and Data

• Models used: Process-based global gridded crop models (LPJmL and pDSSAT), machine learning (extra trees) models.
• Data sources: Simulated US maize yield data from LPJmL and pDSSAT, daily multivariate climate data from reanalysis.

Main Results

• The method identifies a set of ten climate drivers that are robust across different environments and time periods.
• The identified drivers align with crop model mechanisms and consistently use only the weather variables taken as input by the respective models.
• Logistic regression models using the identified drivers show strong predictive performance on held-out test sets.

Contributions

• This study proposes a data-driven method for identifying climate drivers of agricultural yield failure from daily weather data, which can be used to improve process understanding and inform crop model development.
• The approach circumvents the limitations of post-hoc interpretability in black-box machine learning models and allows researchers to use parsimonious statistical models to explore relationships between climate and impacts.

Funding

• This research was funded by the Helmholtz Centre for Environmental Research – UFZ, Leipzig, Germany.

Citation

@article{Sweet2026datadriven,
  author = {Sweet, Lily‐belle and Müller, Christoph and Jägermeyr, Jonas and A, Jakob Zscheischler},
  title = {A data-driven method for identifying climate drivers of agricultural yield failure from daily weather data},
  journal = {Geoscientific model development},
  year = {2026},
  doi = {10.5194/gmd-19-6687-2026},
  url = {https://doi.org/10.5194/gmd-19-6687-2026}
}