| الملخص: | Reference evapotranspiration (ET<sub>0</sub>) is an essential component in hydrological and ecological processes. The Penman–Monteith (PM) model of Food and Agriculture Organization of the United Nations (FAO) model requires a number of meteorological parameters; it is urgent to develop high-precision and computationally efficient ET<sub>0</sub> models with fewer parameter inputs. This study proposed the genetic algorithm (GA) to optimize extreme learning machine (ELM), and evaluated the performances of ELM, GA-ELM, and empirical models for estimating daily ET<sub>0</sub> in Southwest China. Daily meteorological data including maximum temperature (<i>T</i><sub>max</sub>), minimum temperature (<i>T</i><sub>min</sub>), wind speed (<i>u</i><sub>2</sub>), relative humidity (RH), net radiation (<i>R</i><sub>n</sub>), and global solar radiation (<i>R</i><sub>s</sub>) during 1992–2016 from meteorological stations were used for model training and testing. The results from the FAO-56 Penman–Monteith formula were used as a control group. The results showed that GA-ELM models (with R<sup>2</sup> ranging 0.71–0.99, RMSE ranging 0.036–0.77 mm·d<sup>−1</sup>) outperformed the standalone ELM models (with R<sup>2</sup> ranging 0.716–0.99, RMSE ranging 0.08–0.77 mm·d<sup>−1</sup>) during training and testing, both of which were superior to empirical models (with R<sup>2</sup> ranging 0.36–0.91, RMSE ranging 0.69–2.64 mm·d<sup>−1</sup>). ET<sub>0</sub> prediction accuracy varies with different input combination models. The machine learning models using <i>T</i><sub>max</sub>, <i>T</i><sub>min</sub>, <i>u</i><sub>2</sub>, RH, and <i>R</i><sub>n</sub>/<i>R</i><sub>s</sub> (GA-ELM5/GA-ELM4 and ELM5/ELM4) obtained the best ET<sub>0</sub> estimates, with R<sup>2</sup> ranging 0.98–0.99, RMSE ranging 0.03–0.21 mm·d<sup>−1</sup>, followed by models with <i>T</i><sub>max</sub>, <i>T</i><sub>min</sub>, and <i>R</i><sub>n</sub>/<i>R</i><sub>s</sub> (GA-ELM3/GA-ELM2 and ELM3/ELM2) as inputs. The machine learning models involved with <i>R</i><sub>n</sub> outperformed those with <i>R</i><sub>s</sub> when the quantity of input parameters was the same. Overall, GA-ELM5 (<i>T</i><sub>max</sub>, <i>T</i><sub>min</sub>, <i>u</i><sub>2</sub>, RH and <i>R</i><sub>n</sub> as inputs) outperformed the other models during training and testing, and was thus recommended for daily ET<sub>0</sub> estimation. With the estimation accuracy, computational costs, and availability of input parameters accounted, GA-ELM2 (<i>T</i><sub>max</sub>, <i>T</i><sub>min</sub>, and <i>R</i><sub>s</sub> as inputs) was determined to be the most effective model for estimating daily ET<sub>0</sub> with limited meteorological data in Southwest China.
|
|---|
