Developing a New Artificial Intelligence Framework to Estimate the Thalweg of Rivers

Hydrographic operations to investigate the riverbed form throughout the entire length of a river are costly and time-consuming. This has made scholars use a wide range of alternative methods to address the issue. In the present study, however, a new framework using Artificial Intelligence (AI) based models is introduced to identify the thalweg of rivers, which provides an accurate estimate of a river thalweg via linking coordinates of their left and right banks. In this regard, we trained and tested the performance of two AI-based models, including Artificial Neural Network (ANN) and Adaptive Neuro-Fuzzy Inference System (ANFIS) models. The database of two rivers, namely the Qinhe River in China and the Gaz River in Iran was used to help evaluate the developed model. Outcomes of the two investigated case studies demonstrated that the values of the statistical error estimators, including the Root Mean Square Error (RMSE) of the ANFIS model were less than those of the ANN model. As a result, the ANFIS model can lead to more accurate results than the ANN model, and it is suitable for cases with less available data. Moreover, comparing the results from the developed models with those of the River Channel Morphology Model (RCMM) showed that AI-based models outdo numerical approaches in the identification of the thalweg of rivers. All in all, it is inferred that the proposed approach not only helps us achieve an accurate geometry of rivers but reduces the side costs and can be used as an effective alternative to field operations. The applicability of the proposed models to different river systems is also discussed as a potential direction for future research.