Gui et al. (2026) Climate-adjusted sequence learning reveals atmospheric-demand transitions in agricultural water–heat buffering

Identification

• Journal: Agricultural Water Management
• Year: 2026
• Date: 2026-08-27
• Authors: Qifeng Gui, Pan Li, Jingui Wei, Hailong Qiu, Lianhao Zhao, Pingxing Wan, Haoran Sun, Wen Yin
• DOI: 10.1016/j.agwat.2026.110739

Research Groups

• State Key Laboratory of Aridland Crop Science, College of Agronomy, Gansu Agricultural University, Lanzhou, 730070, China
• College of Communication and Information Engineering, Xi’an University of Science and Technology, Xi’an, China
• Department of Land Surveying and Geo-Informatics, The Hong Kong Polytechnic University, Kowloon, Hung Hom, Hong Kong

Short Summary

This study develops a climate-adjusted sequence-learning framework to identify agricultural water–heat response signatures from seasonal satellite and climate time series. Two dominant response strategies were identified: strong-buffering systems showed greener, wetter, and cooler climate-adjusted response signatures compared to weak-buffering systems.

Objective

• Examine whether agricultural systems across the western United States exhibit distinct climate-adjusted water–heat response strategies.
• Test whether these strategies provide information beyond favourable absolute-state conditions.
• Evaluate whether the identified response signatures are independently consistent with evapotranspiration-based water-use functioning using OpenET evapotranspiration and Cropland Data Layer crop-type information.

Study Configuration

• Spatial Scale: Three major agricultural regions in the western United States: Central Valley, High Plains West, and Snake River.
• Temporal Scale: Multiple years (2019, 2021, and 2023) with a broad range of hydroclimatic conditions.

Methodology and Data

• Models used: Gated Recurrent Unit (GRU)-based sequence model for climate-conditioned expected-response modelling, sequence autoencoder for residual trajectory representation learning.
• Data sources: Landsat 16-day composites, GRIDMET, Cropland Data Layer crop-type information.

Main Results

• Two dominant climate-adjusted water–heat response strategies emerged from residual representation learning: weak-buffering and strong-buffering signatures.
• Strong-buffering systems showed greener, wetter, and cooler climate-adjusted response signatures compared to weak-buffering systems.
• The difference between the two strategies was most evident during the middle to late growing season, when atmospheric water demand was high.

Contributions

• This study provides a practical basis for prioritizing agricultural water-management interventions under rising atmospheric demand.
• The proposed framework can move beyond mapping where croplands look favourable to identifying where they perform better or worse than expected under comparable climate pressure.

Funding

• Not specified in the provided text.

Citation

@article{Gui2026Climateadjusted,
  author = {Gui, Qifeng and Li, Pan and Wei, Jingui and Qiu, Hailong and Zhao, Lianhao and Wan, Pingxing and Sun, Haoran and Yin, Wen},
  title = {Climate-adjusted sequence learning reveals atmospheric-demand transitions in agricultural water–heat buffering},
  journal = {Agricultural Water Management},
  year = {2026},
  doi = {10.1016/j.agwat.2026.110739},
  url = {https://doi.org/10.1016/j.agwat.2026.110739}
}