Understanding the longitudinal dispersion mechanism in natural channels is critical for controlling water pollution and preventing flow stratification, which is dangerous for water resource management for both human and aquatic life. The main objective of the proposed study is to develop an artificial neural network (ANN)-based model for the determination of longitudinal dispersion coefficient (Kx) for natural streams. For training and testing of ANNs of the present work, 71 field datasets (collected from 29 rivers in United States) have been used of which, randomly selected 50 datasets were used for training and 21 were used for testing. In this study, the procedure for the estimation of longitudinal dispersion coefficient (Kx) in rivers has been established in such a way that the error has been reduced, and regression value has been increased when compared with previous studies.Initially, the model developed by Fischer gave very high values (788.58 and 2834.387) of root mean square error (RMSE) in both training and testing data and regression (R2) value as (0.266 and 0.518) were also on lower side. These values were further improved in later studies leading to further improvement in the present study as 4.1032 and 5.9055 of RMSE in both training and testing and R2 as 0.982 and 0.923, respectively.
Elsevier, Water, Land, and Forest Susceptibility and Sustainability: Geospatial Approaches and Modeling, Volume 1, 2023, Pages 87-119