Gyamfi et al. (2026) Integrating ENSO dynamics into physics-informed Graph Neural Networks for extreme evapotranspiration prediction under climate change

Identification

• Journal: Agricultural and Forest Meteorology
• Year: 2026
• Date: 2026-08-25
• Authors: Kwame Adutwum Gyamfi, Eun-Sung Chung
• DOI: 10.1016/j.agrformet.2026.111427

Research Groups

Department of Civil Engineering, Seoul National University of Science and Technology

Short Summary

This study develops an ENSO-aware spatio-temporal Graph Neural Network (ST-GNN) for extreme evapotranspiration prediction across the Korean Peninsula, improving predictions during ENSO phase transitions.

Objective

• Investigate the impact of El Niño-Southern Oscillation (ENSO) on extreme evapotranspiration events in the Korean Peninsula.
• Develop a physics-informed Graph Neural Network framework that captures both local hydroclimatic controls and nonlocal ENSO-driven variability.

Study Configuration

• Spatial Scale: Korean Peninsula
• Temporal Scale: Seasonal to interannual timescales

Methodology and Data

• Models used: Physics-informed Graph Neural Networks (ST-GNN)
• Data sources: GLEAM-based evapotranspiration, CMIP6 predictors under SSP2-4.5 and SSP5-8.5, Niño 3.4 anomalies, temperature, humidity, radiation, soil moisture, wind

Main Results

• The ENSO-aware ST-GNN framework improves extreme evapotranspiration prediction, particularly during ENSO phase transitions.
• Future projections indicate shifts in ET extremes associated with ENSO variability.

Contributions

• This study provides a physically interpretable machine-learning approach for assessing teleconnection-modulated hydroclimatic risk.
• The developed framework captures both local and nonlocal controls on extreme evapotranspiration, improving predictions under climate change.

Funding

• This research was funded by [project/program name] with reference code [code].

Citation

@article{Gyamfi2026Integrating,
  author = {Gyamfi, Kwame Adutwum and Chung, Eun-Sung},
  title = {Integrating ENSO dynamics into physics-informed Graph Neural Networks for extreme evapotranspiration prediction under climate change},
  journal = {Agricultural and Forest Meteorology},
  year = {2026},
  doi = {10.1016/j.agrformet.2026.111427},
  url = {https://doi.org/10.1016/j.agrformet.2026.111427}
}