Küçüktopçu et al. (2026) Assessing the Impact of Geographical and Meteorological Information on Machine Learning-Based Reproduction of FAO Penman–Monteith Reference Evapotranspiration

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

Identification

• Journal: Agronomy
• Year: 2026
• Date: 2026-07-25
• Authors: Erdem Küçüktopçu, Petr Šařec, Václav Novák, Emre Tunca, Martin Procházka
• DOI: 10.3390/agronomy16151409

Research Groups

Not specified in the provided text.

Short Summary

This study evaluates four machine learning algorithms for estimating reference evapotranspiration (ETo) in the Czech Republic, concluding that the availability of meteorological predictors is more critical for model accuracy than the specific algorithm selected.

Objective

• To evaluate the performance of four machine learning (ML) algorithms in reproducing FAO Penman–Monteith (FAO-PM) ETo under different levels of geographical and meteorological information availability.

Study Configuration

• Spatial Scale: Czech Republic (59 meteorological stations).
• Temporal Scale: 1980–2024 (daily observations).

Methodology and Data

• Models used: Kernel Approximation Regression (KAR), Multilayer Perceptron (MLP), Extreme Gradient Boosting (XGB), and Random Forest (RF).
• Data sources: Daily observations from 59 meteorological stations.

Main Results

• Predictor availability influenced model performance more significantly than the choice of ML algorithm.
• The lowest performance was observed in the geographical-information-only scenario, while the inclusion of meteorological variables led to substantial improvements.
• Among single-variable meteorological additions, relative humidity provided the greatest improvement in agreement with the FAO-PM ETo benchmark.
• Overall testing performance across all scenarios and algorithms ranged from $R^2 = 0.683$ to $0.998$ and $\text{RMSE} = 0.076$ to $0.939\text{ mm d}^{-1}$.

Contributions

• Provides a practical framework for approximating FAO-PM ETo at stations where some meteorological inputs are unavailable by utilizing reduced-input ML scenarios.

Funding

Not specified in the provided text.

Citation

@article{Küçüktopçu2026Assessing,
  author = {Küçüktopçu, Erdem and Šařec, Petr and Novák, Václav and Tunca, Emre and Procházka, Martin},
  title = {Assessing the Impact of Geographical and Meteorological Information on Machine Learning-Based Reproduction of FAO Penman–Monteith Reference Evapotranspiration},
  journal = {Agronomy},
  year = {2026},
  doi = {10.3390/agronomy16151409},
  url = {https://doi.org/10.3390/agronomy16151409}
}