Evaluating Various Techniques for Total Sediment Concentration Prediction: Case of Marala Ravi Link Canal in Pakistan

Sediment‐induced issues pose significant threats to the optimal operations of some canal systems globally. Sediment transport, particularly in irrigation canal systems, is crucial for sustainable water management, maintaining water quality, promoting ecosystem health, ensuring infrastructure maintenance, and effective allocation planning. This paper aims to investigate sediment concentration in the Marala Ravi Link Canal (MRLC), Pakistan, employing various techniques, including artificial intelligence (AI) and empirical equations proposed by Ackers and White, as well as those of Brownlie. In artificial neural networks (ANNs), 12 model training functions (MTFs), each with double and triple layers (DL and TL), were tested. The models’ performances were compared with the adaptive neuro‐fuzzy inference system (ANFIS) and particle swarm optimization (PSO) technique. Ten years of sediment data, including discharge, flow velocity, flow depth, total sediment load, and sediment characteristics, were obtained from the relevant agencies for MRLC. A Monte Carlo simulation (MSC) sensitivity analysis was conducted to assess the influence of various flow and sediment parameters on total sediment load prediction. An empirical formula was developed based on MRLC sediment data using an efficient optimization tool, general reduced gradient (GRG). Various performance indicators, including mean square error (MSE), correlation coefficient (R), analysis of variance (ANOVA), t ‐test, and Taylor’s diagram, were used to assess the performance of the employed models. The results show that out of 12 DL‐ANN models, MTF1 performed better than the other models, with a higher R ‐value of 0.95. Similarly, out of 12 TL‐ANN models, MTF13 performed better than the other 11 TL‐ANN models. The results show that the MTF13 model performed better than the remaining AI models. The GRG results were comparatively better, with a higher R ‐value of 0.99 for total sediment load prediction in the MRLC. The superior performance of the AI models provides an efficient framework that is not limited to the MRLC but has applicability on a global scale, especially for river and irrigation networks influenced by sedimentation.