Saminathan et al. (2026) A data driven approach for reference evapotranspiration forecasting: merging weather forecasts with persistence based Approaches

Identification

• Journal: Agricultural and Forest Meteorology
• Year: 2026
• Date: 2026-08-12
• Authors: Sakila Saminathan, Subhasis Mitra, Lahari Chowdary Talasila
• DOI: 10.1016/j.agrformet.2026.111414

Research Groups

Department of Civil Engineering, Indian Institute of Technology Palakkad, Kerala, India

Short Summary

A data-driven approach is proposed to forecast reference evapotranspiration (ETo) by merging weather forecasts with persistence-based approaches, resulting in improved ETo forecasts compared to traditional Numerical Weather Prediction (NWP) models.

Objective

• Investigate the potential of combining weather forecasts and persistence-based approaches for ETo forecasting at short and medium range time scales.

Study Configuration

• Spatial Scale: Indian region
• Temporal Scale: Short and medium range time scales

Methodology and Data

• Models used: Support Vector Regression (SVR), Random Forest (RF), Extreme Gradient Boosting (XGB), Long Short Term Memory (LSTM), Convolutional Neural Network (CNN)
• Data sources: Weather hindcasts from European Centre for Medium-Range Weather Forecasts (ECMWF) and Global Ensemble Forecast System (GEFS)

Main Results

• The proposed framework provides better ETo forecasts than NWP models, with XGB requiring the least time for training.
• Performance of the proposed approach is lower in mountainous regions due to low skill of NWP forecasts.

Contributions

• Original contribution lies in merging weather forecasts with persistence-based approaches for improved ETo forecasting.
• The proposed framework outperforms traditional NWP bias-correction methods and persistence-based approaches in literature.

Funding

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

Citation

@article{Saminathan2026data,
  author = {Saminathan, Sakila and Mitra, Subhasis and Talasila, Lahari Chowdary},
  title = {A data driven approach for reference evapotranspiration forecasting: merging weather forecasts with persistence based Approaches},
  journal = {Agricultural and Forest Meteorology},
  year = {2026},
  doi = {10.1016/j.agrformet.2026.111414},
  url = {https://doi.org/10.1016/j.agrformet.2026.111414}
}