River Bank Erosion Prediction Using Multivariable Linear Regression

This research proposes a new approach using multivariable linear regression to predict the riverbank erosion speed. As a simple and interpretable model, the proposed approach gains two main achievements. First, it can specify the main factors causing riverbank erosion. Notably, the method identifies the river's depth and the water flow's hydraulic gradient, contributing primarily to the erosion speed. Second, multivariable linear regression can be learned from such a small dataset. This aspect makes the range of applications for the method much broader. The Experimental results show that the multivariable linear regression can predict erosion speed well. With a dataset with only 27 records, the method can predict the erosion speed with an error of around 2 meters per year. In the future, a more extensive training dataset or a more complicated regression model is requested to gain a better result.