University of Tehran, College of Aburaihan Water and Irrigation Management 2251-6298 4 2 2014 09 23 Comparison of Regression tree, artificial neural network and Hargrives-Samani in estimation of reference evapotranspiration in semi region Comparison of Regression tree, artificial neural network and Hargrives-Samani in estimation of reference evapotranspiration in semi region 149 160 53031 10.22059/jwim.2014.53031 FA Mitra Bakhshoodeh MSc., Former Graduate of Irrigation and Drainage Engineering Department, Aburaihan Campus, University of Tehran, Tehran, Iran Ali Rahimikhoob Professor, Irrigation and Drainage Engineering Department, Aburaihan Campus, University of Tehran, Tehran, Iran Journal Article 2013 06 28 The purpose of this study was to evaluate three models of artificial neural networks (ANN), regression trees (M5) and Hargrives-Samani (HG) in estimation of reference evapotranspiration. For this purpose was used climate information of Sistan va Baloochestan, Kerman, Yazd and Khorasan Jonoobi from 1998 to 2008. In addition to effect of wind (U) on evapotranspiration (ET<sub>0</sub>), estimation of ET<sub>0</sub> was done based on wind change in tree groups including U<2.48 m/s (U<sub>1</sub>), 2.48<U<3.67 m/s (U<sub>2</sub>) and U>3.67 m/s (U<sub>3</sub>). The results showed that optimum result of each tree methods was in U<sub>1</sub> group. The amount of RMSE and R<sup>2</sup> in ANN were 1.41 mm/day and 0.84 respectively, in MS were 1.46 mm/day and 0.83 and in HG were 2.02 mm/day and 0.69. These results showed that both ANN and MS methods are better than HG model. Besides, the run of MS to ANN is easy. The purpose of this study was to evaluate three models of artificial neural networks (ANN), regression trees (M5) and Hargrives-Samani (HG) in estimation of reference evapotranspiration. For this purpose was used climate information of Sistan va Baloochestan, Kerman, Yazd and Khorasan Jonoobi from 1998 to 2008. In addition to effect of wind (U) on evapotranspiration (ET<sub>0</sub>), estimation of ET<sub>0</sub> was done based on wind change in tree groups including U<2.48 m/s (U<sub>1</sub>), 2.48<U<3.67 m/s (U<sub>2</sub>) and U>3.67 m/s (U<sub>3</sub>). The results showed that optimum result of each tree methods was in U<sub>1</sub> group. The amount of RMSE and R<sup>2</sup> in ANN were 1.41 mm/day and 0.84 respectively, in MS were 1.46 mm/day and 0.83 and in HG were 2.02 mm/day and 0.69. These results showed that both ANN and MS methods are better than HG model. Besides, the run of MS to ANN is easy. Artificial Neural Network evapotranspiration Hargrives-Samani Regression tree Wind https://jwim.ut.ac.ir/article_53031_9455c945be124efd91e67c600e92bdd7.pdf