Leveraging remote sensing data and machine learning models to estimate suspended sediment concentration (SSC), a vital water quality parameter to assess soil erosion effects
Publication Date
1-1-2024
Document Type
Contribution to a Book
Publication Title
Advanced Tools for Studying Soil Erosion Processes: Erosion Modelling, Soil Redistribution Rates, Advanced Analysis, and Artificial Intelligence
DOI
10.1016/B978-0-443-22262-7.00024-2
First Page
97
Last Page
114
Abstract
Soil erosion has severe environmental consequences as it displaces soil particles into water bodies, leading to increased suspended sediment concentration (SSC) levels. This study utilized Sentinel-2 data from Google Earth Engine (GEE) to estimate SSC. Boosting-based ensemble machine learning (EML) models, including AdaBoost, gradient boosting regression (GBR), light gradient boosting machine (LGBM), and eXtreme Gradient Boosting (XGBoost), were employed to establish the relationship between remote sensing reflectance (Rrs) and SSC along the Missouri River. Bayesian optimization (BO) was utilized to optimize the hyperparameters of each model through five repeats of 5-fold cross-validation. The results indicated that BO improved the performance of EMLs by approximately 5% in R2 and 1.5mgL−1 in mean squared error (MSE). Moreover, optimized XGBoost (B_XGBoost) could outperform other models (R2=77.43%). This research confirmed the effectiveness of EML models in accurately estimating SSC from remote sensing data. Remote sensing-derived estimates of SSC can be utilized to assess the impact of soil erosion in various ways. These include evaluating the extent of soil erosion in a specific area, identifying erosion hotspots characterized by high sediment concentrations, analyzing temporal trends of soil erosion and sediment transport using time-series data, validating and improving erosion models, and integrating SSC estimates with other water quality parameters to obtain a comprehensive assessment of the soil erosion effect.
Keywords
Bayesian optimization, Ensemble machine learning, Google Earth Engine, Machine learning, Missouri river, Water quality
Department
Civil and Environmental Engineering
Recommended Citation
Alireza Taheri Dehkordi, Mohammad Javad Valadan Zoej, Mohsen Jafari, Hani Ghasemi, and Ali Mehran. "Leveraging remote sensing data and machine learning models to estimate suspended sediment concentration (SSC), a vital water quality parameter to assess soil erosion effects" Advanced Tools for Studying Soil Erosion Processes: Erosion Modelling, Soil Redistribution Rates, Advanced Analysis, and Artificial Intelligence (2024): 97-114. https://doi.org/10.1016/B978-0-443-22262-7.00024-2
