Scholarlink

<p>Potential Evapotranspiration (PET) estimation is essential for understanding water requirements in agricultural and hydrological studies, especially in water-scarce regions. This study introduces two novel machine learning (ML) models, Awhari1 and Awhari2, to improve aridity classification across Nigeria. Daily climate data between 1950 and 2022 from fifth European Reanalysis (ERA5) maximum temperature (Tmax), relative humidity (RH), wind speed (ws), actual (ea) and saturated vapour pressure (es) were used for this prupose. The Penman-Monteith (PM) method was utilised as a reference model to assess model accuracy, while Hargreaves and Thornthwaite models were evaluated for comparison. Results reveal that the Awhari2 model outperformed other models, excelling in sensitivity (3.535), specificity (3.916), and balanced accuracy (3.726) across all aridity classes, notably in Humid and Hyper Arid zones. Awhari1 model also showed high precision (3.611), in Semi-arid and arid regions Hargreaves and Thornthwaite underperformed, showing limited daptability to Nigeria’s diverse climatic zones. These findings support evaluating PET models’ performance before being applied in aridity studies. However, limitations in these models, such as their exclusion of solar radiation, suggest that future studies should expand climatic inputs to enhance predictive robustness. This study contributes valuable insights for improved PET modeling, critical for effective water resource management and agricultural planning in arid regions.&#xa0;</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Evaluating evapotranspiration models for precise aridity mapping based on UNEP- aridity classification

  • 期刊论文

  • 发表日期：

    2025年1月20日

查看全文
  • Dauda Pius Awhari,
  • Mohamad Hidayat Bin Jamal,
  • Mohd Khairul Idlan Muhammad,
  • Istifanus Akila Bardey,
  • Shamsuddin Shahid

摘要

Potential Evapotranspiration (PET) estimation is essential for understanding water requirements in agricultural and hydrological studies, especially in water-scarce regions. This study introduces two novel machine learning (ML) models, Awhari1 and Awhari2, to improve aridity classification across Nigeria. Daily climate data between 1950 and 2022 from fifth European Reanalysis (ERA5) maximum temperature (Tmax), relative humidity (RH), wind speed (ws), actual (ea) and saturated vapour pressure (es) were used for this prupose. The Penman-Monteith (PM) method was utilised as a reference model to assess model accuracy, while Hargreaves and Thornthwaite models were evaluated for comparison. Results reveal that the Awhari2 model outperformed other models, excelling in sensitivity (3.535), specificity (3.916), and balanced accuracy (3.726) across all aridity classes, notably in Humid and Hyper Arid zones. Awhari1 model also showed high precision (3.611), in Semi-arid and arid regions Hargreaves and Thornthwaite underperformed, showing limited daptability to Nigeria’s diverse climatic zones. These findings support evaluating PET models’ performance before being applied in aridity studies. However, limitations in these models, such as their exclusion of solar radiation, suggest that future studies should expand climatic inputs to enhance predictive robustness. This study contributes valuable insights for improved PET modeling, critical for effective water resource management and agricultural planning in arid regions.