Extended Abstract
Introduction: Sediment transport by surface flows is one of the primary processes responsible for reshaping the Earth's surface. This process is widely expressed in the shape of river drainage systems and their alluvial deposits, which dominate the geomorphology of large areas of the Earth's surface and are responsible for the majority of sediment from land to oceans. One of the most important problems threatening dams is the sediment inputs to the dam reservoir. Due to various problems, estimating the volume of sediments is a complicated process. So, some methods have been created by researchers to overcome these problems. Modeling of suspended sediment load (SSL) is an important subject for decision-makers in the catchment. Accurate and reliable modeling of SSL is one of the important subjects for planning, managing, and designing soil and water structures in the drainage networks. The purpose of this study was to compare the efficiency of Sediment Rating Curve (SRC) and Machine Learning Algorithms (MLA) for estimating SSL in karst rivers in Bahram Joo, Cham Anjir, Sarab Seyd Ali, and Kakareza hydrometric stations in Lorestan province, Iran.
Materials and Methods: In this study, a rating curve and five soft computing techniques, Support Vector Machine with RBF kernel (SVM-RBF), Support Vector Machine with PUK kernel (SVM-PUK), Gaussian processes with PUK kernel (GP-PUK), Gaussian processes with RBF kernel (GP-RBF), M5P, reduced error pruning tree (REPTree), and Random Forest (RF) were used and evaluated. They were used to predict SSL in the Kashkan watershed, Iran. Cham Anjir, Bahram Joo, Sarb Seyd Ali, and Kakareza stations were selected for this study. The data of temperature, rain, discharge, and SSL of 20 years (2001–2021) were utilized as input and output parameters. Thus, four stations with a long-term data were selected. The total dataset consists of temperature, rain, discharge, and SSL of watersheds, of which 70% of the data were used for training and 30% for the testing phase. Finally, the models’ accuracy was assessed using three performance evaluation parameters: Correlation Coefficient (C.C.), Root Mean Square Error (RMSE), and Maximum Absolute Error (MAE).
Results and Discussion: Results showed that the soft computing methods (SVM-PUK, GP-PUK, GP-RBF, M5P, REPTree, and RF) performed better than the traditional technique (SRC), as they made use of non-linear techniques for data reconstruction. It can be concluded that, among all the models, the M5P model, which used decomposed data that captured the dynamic features of the non-linear and non-stationary SSL time series data, performed better than other models. The SRC performed with a C.C of 0.5941. The best M5P model (best among soft computing methods) scored a mean C.C of about 0.89, surpassing the best SRC results. Although it captured the peaks better than SRC, it still overestimated the sediment load and was unable to capture the peak sediment rates, which are of great importance for design purposes.
Conclusion: Sediments carried by water are a serious problem, as they shorten the life of a reservoir, reduce the channel discharge-carrying capacity, especially to tail-end users, etc. Therefore, sediment management is the golden rule in river engineering, to which much effort and energy are directed. An important aspect of sediment management is sediment estimation, which is mostly found in a suspended form in rivers and other water bodies. This research focused on a comparison of the different methods of suspended sediment estimation in rivers. This includes the traditional method, i.e., SRC, and soft computing techniques, i.e., SVM-RBF, SVM-PUK, GP-PUK, GP-RBF, M5P, REPTree, and RF. The results of this study provide scientific information to predict SSL, and Soft Computing Techniques could be an efficient technique to simulate the SSL time series, because they extract key features embedded in the SSL signal. Finally, the results showed that the M5P model is effective in predicting suspended sediment content in the KhorammAbad, Biranshahr, and Alashter watersheds.