Scholarlink

<p>Empirical models can estimate reference evapotranspiration (ET₀) with reasonable accuracy using fewer input variables. Satellite data can be used as efficient inputs in ET₀ estimation models. Therefore, the main goal of this study was to develop an empirical model to estimate and predict ET₀ values using European Centre for Medium-Range Weather Forecasts Reanalysis v5 (ECMWF-ERA5) data. Data from three watersheds in Iran were collected between years 2000 and 2023. Initially, 32 well-known empirical models for estimating evapotranspiration were examined using two different datasets, namely ground-based data and ERA5 data, and evaluated against evaporation pan data. To improve the results, a hybrid model was developed with the linear regression method. Moreover, the proposed linear ET₀-Predict model was investigated against four Machine Learning (ML) algorithms, including Artificial Neural Network (ANN), Random Forest (RF), Deep Neural Network (DNN), and Least Squares Boosting (LSBoost) using the same input variables. The results showed that the Jensen–Haise and Caprio models were the most accurate model based on ground data and ERA5-Land data, respectively. The hybrid model results in the root mean square error of 0.39&#xa0;mm and the correlation coefficient to 0.695. Moreover, the correlation coefficient of the proposed hybrid model was comparable with the values of 0.577 to 0.661 for the ML models. The findings emphasize that the use of hybrid data combined with advanced statistical methods can significantly improve the accuracy of ET₀ prediction. Additionally, the capability of ERA5-Land satellite data to replace ground data in empirical modeling was confirmed.</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

A new empirical model to estimate and predict reference evapotranspiration using ERA5 data

  • 期刊论文

  • 发表日期：

    2026年7月21日

查看全文
  • Seyed Reza Fatemi,
  • Majid Rahimzadegan

摘要

Empirical models can estimate reference evapotranspiration (ET₀) with reasonable accuracy using fewer input variables. Satellite data can be used as efficient inputs in ET₀ estimation models. Therefore, the main goal of this study was to develop an empirical model to estimate and predict ET₀ values using European Centre for Medium-Range Weather Forecasts Reanalysis v5 (ECMWF-ERA5) data. Data from three watersheds in Iran were collected between years 2000 and 2023. Initially, 32 well-known empirical models for estimating evapotranspiration were examined using two different datasets, namely ground-based data and ERA5 data, and evaluated against evaporation pan data. To improve the results, a hybrid model was developed with the linear regression method. Moreover, the proposed linear ET₀-Predict model was investigated against four Machine Learning (ML) algorithms, including Artificial Neural Network (ANN), Random Forest (RF), Deep Neural Network (DNN), and Least Squares Boosting (LSBoost) using the same input variables. The results showed that the Jensen–Haise and Caprio models were the most accurate model based on ground data and ERA5-Land data, respectively. The hybrid model results in the root mean square error of 0.39 mm and the correlation coefficient to 0.695. Moreover, the correlation coefficient of the proposed hybrid model was comparable with the values of 0.577 to 0.661 for the ML models. The findings emphasize that the use of hybrid data combined with advanced statistical methods can significantly improve the accuracy of ET₀ prediction. Additionally, the capability of ERA5-Land satellite data to replace ground data in empirical modeling was confirmed.