Estimation of Suspended Sediment Concentration in Rivers using Sentinel-2 Imagery: A Case Study of Lam Pao Basin

Abstract

Monitoring and quantifying suspended sediment concentration (SSC) along rivers provides important information for reservoir management. Traditional monitoring based on in-situ sampling and measurement of SSC in rivers is expensive and time-consuming to perform. The objective of this study was to use spectral information provided by remote sensing from Sentinel-2 images in combination with machine learning to estimate the SSC of a river in the Lam Pao basin. Three machine-learning regression algorithms (Multiple Linear Regression, Deep Learning, and Support Vector Machine: SVM) were evaluated and a suitable model was created to estimate the SSC of the river. The results show that the Support Vector Machine model gave the most balanced results, with the lowest RMSE values and a high statistical correlation (R2=0.863; RMSE=11.9) for the whole range of SSC (0 to 90 mg/l) measured at this station during the studied period. The methodology presented in this study can be used as a guideline for the combination of machine learning with Sentinel-2 images for estimation of the SSC other rivers, although some factors require further study to improve the accuracy of SSC estimation.

Keywords

• remote sensing
• sentinel-2 data
• suspended sediment concentration (ssc)
• machine learning