MACHINE LEARNING MODELS (MLP, RANDOM FOREST, LIGHTGBM) - FOR DAILY ET₀ ESTIMATION WITH LIMITED DATA IN HUMID MEDITERRANEAN REGION (JIJEL), ALGERIA

Aug 5, 2026

Upload by Ophelia

Accurate estimation of the reference evapotranspiration (ET₀) is essential for effective irrigation planning and sustainable water resource management, especially in climatically heterogeneous regions. This study evaluates three machine learning models—MLP, Random Forest, and LightGBM—for predicting daily FAO-56 Penman–Monteith ET₀ in the humid Mediterranean region of Jijel, northeastern Algeria. Daily meteorological data from ten stations during 2000–2024 are used with six input variables, and performance is tested using multiple statistical metrics across training, validation, and independent test sets.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages30

Reading Time1 h 16 min

AccessFree to download

1 INTRODUCTION

Reference evapotranspiration and its role

Conventional ET₀ estimation approaches

Motivation for machine learning alternatives

Model types: MLP, Random Forest, and gradient boosting

MACHINE LEARNING MODELS (MLP, RANDOM FOREST, LIGHTGBM) FOR DAILY ET₀ ESTIMATION WITH LIMITED DATA IN HUMID MEDITERRANEAN REGION (JIJEL) ALGERIA MODELOS DE APRENDIZAGEMAUTOMÁTICA (MLP, RANDOM FOREST, LIGHTGBM) PARA AESTIMATIVADIÁRIA DE ET₀ COM DADOSLIMITADOS NA REGIÃOMEDITERRÂNICA HÚMIDA (JIJEL), ARGÉLIA Article received on: 8/29/2025 Article accepted on: 11/28/2025 Assia Meziani* *Department of Hydraulic and Civil Engineering, New Technology and Local Development Laboratory, Faculty of Technology, University of El-Oued, El-Oued, Algeria. Orcid: [https://orcid.org/0000-0002-2569-1552](https://orcid.org/0000-0002-2569-1552)[ ](https://orcid.org/0000-0002-2569-1552)[assia-meziani@univ-eloued.dz](assia-meziani@univ-eloued.dz) The authors declare that there is no conflict of interest Abstract Accurate estimation of the reference evapotranspiration (ET₀) is essential for effective irrigation planning and sustainable water resource management, particularly in climatically heterogeneous regions. This study evaluated the performance of three machine learning models—multilayer perceptron (MLP), Random Forest (RF), and Light Gradient Boosting Machine (LightGBM)—for predicting daily FAO-56 Penman–Monteith ET₀ in the humid Mediterranean region of Jijel, northeastern Algeria. Daily meteorological data from ten stations for the period 2000–2024 were used, incorporating six input variables: air temperature, relative humidity, wind speed, sunshine duration, solar radiation, and vapor pressure deficit. The model performance was assessed using multiple statistical metrics across the training, validation, and independent testing datasets. All models achieved high predictive accuracy, with R² values exceeding 0.97. RF exhibited the highest training performance (R² = 0.997, RMSE ≈ 0.09 mm day⁻¹) but showed signs of mild overfitting on test data. In contrast, MLP demonstrated the best generalization capability (test R² = 0.983, RMSE = 0.21 mm day⁻¹, NSE = 0.983), closely followed by LightGBM (test R² ≈ 0.980) . Trend analysis revealed no significant long-term change in annual ET₀ (p = 0.907) . The results confirm the robustness of machine learning approaches, particularly MLP and LightGBM, for reliable ET₀ estimation in humid Mediterranean environments. Keywords: Reference Evapotranspiration. FAO- 56. LightGBM. Multi-Layer Perceptron. Random Forest. Jijel. Algeria. Resumo A estimativa precisa da evapotranspiração de referência (ET₀) é essencial para o planejamento eficiente da irrigação e a gestãosustentável dos recursos hídricos, especialmenteem regiões climaticamente heterogêneas. Este estudo avaliou três modelos de aprendizado automático —Perceptrão Multicamada (MLP), Random Forest (RF) e LightGBM—paraprevera ET₀ diária FAO-56 Penman–Monteith na região mediterrânea húmida de Jijel (nordeste da Argélia). Foram utilizados dados meteorológicos diários de dez estações (2000– 2024) com seis variáveis: temperatura do ar, humidade relativa, velocidade do vento, duração da insolação, radiação solar e défice depressão de vapor. Odesempenhofoi medido com métricas estatísticas em conjuntos de treino, validação e teste. Todos os modelos alcançaramalta precisão (R² > 0,97). O RF obteve o melhordesempenho no treino (R² = 0,997; RMSE ≈ 0,09 mm dia ⁻¹), mas mostrou leve sobreajuste noteste. O MLP apresentou a melhorgeneralização (R² teste = 0,983; RMSE = 0,21 mm dia ⁻¹; NSE = 0,983), seguido de perto pelo LightGBM (R² teste ≈ 0,980). A análise de tendências não detectou alterações significativas na ET₀ anual (p = 0,907). Os resultados confirmam a robustez dos enfoques de machine learning, especialmente MLP e LightGBM, para estimarET₀ deformafiável em ambientes mediterrâneos húmidos. Palavras-chave: Evapotranspiração de Referência. FAO-56. LightGBM. Perceptrão Multicamada. Random Forest. Jijel. Argélia. 1 INTRODUCTION Reference evapotranspiration (ET₀) is a fundamental parameter in hydrology and agriculture and plays a pivotal role i

FAQs

What ET₀ estimation problem does this study address?

The study targets accurate daily reference evapotranspiration (ET₀) estimation for irrigation and water management using a standardized FAO-56 Penman–Monteith reference.

Which machine learning models are evaluated and how are they compared?

Three models—MLP, Random Forest, and LightGBM—are trained and evaluated using statistical metrics on training, validation, and independent test datasets to compare predictive performance and generalization.

What input data and time period are used to train the models?

The models use daily meteorological observations from ten stations spanning 2000–2024, based on six variables: air temperature, relative humidity, wind speed, sunshine duration, solar radiation, and vapor pressure deficit.

Aug 5, 2026

Upload by Ophelia