TY - JOUR ID - 73660 TI - Estimation of Transverse Dispersion Coefficient of Pollutant Transport in Rivers Using Evolutionary Computations JO - Iranian journal of Ecohydrology JA - IJE LA - en SN - 2423-6098 AU - Barahouie Nezhad, Asma AU - Ghaemi, Alireza AU - Hashemi Monfared, Seyyed Arman AU - Azizyan, Gholamreza AU - Dehghani Darmian, Mohsen AD - MSc. Student, Department of Civil Engineering, University of Sistan and Baluchestan, Zahedan, Iran AD - Ph.D. Student, Department of Civil Engineering, University of Sistan and Baluchestan, Zahedan, Iran AD - Associate Professor, Department of Civil Engineering, University of Sistan and Baluchestan, Zahedan, Iran Y1 - 2019 PY - 2019 VL - 6 IS - 4 SP - 901 EP - 912 KW - Rivers KW - Pollution transport KW - Transverse dispersion coefficient KW - Support vector machine KW - Tree model DO - 10.22059/ije.2019.281787.1113 N2 - Surface water is taken into account as one of the most important water resources available to mankind, which is used for various purposes, such as drinking and agriculture. Recently, with the growing urban population, there are many problems associated with the pollution and quality of water resources. Therefore, recognizing and studying the process of mixing and conveying materials in rivers is one of the prominent activities in water resource management programs. In the process of mixing, after the longitudinal dispersion coefficient, the transverse dispersion coefficient is considered as the most effective parameter. According to the importance of dispersion and distribution of pollution in rivers, in order to estimate the transverse dispersion coefficient of pollutants in surface flows, MT and SVM using two Kernels including radial basis function (RBF) and polynomial (Poly) are applied. To achieve this aim, 187 dataset including flow depth (H), flow velocity (U), shear rate (U*) and channel width (W) are used. The results of the evaluation criteria showed that the SVM-Poly model had higher accuracy (R = 0.992 = 0.92 OI =) compared to the SVM-RBF (R = 0.968 and O = 950) and MT (R = 0.966 and OI =0.946) in the training phase for DT estimation. The DT values ​​obtained by proposed models were also evaluated for testing dataset. Based on the result, it was found that SVM-RBF had the best ability to estimate DT with the lowest error (RMSE = 0.029). In addition, comparing the performance of intelligent methods with empirical relationships suggests that empirical relationships failed to show acceptable accuracy. UR - https://ije.ut.ac.ir/article_73660.html L1 - https://ije.ut.ac.ir/article_73660_b1c8879ba0738472e30d8be34770dd8b.pdf ER -