Benchmarking of hierarchical bayesian model aggregation, xgboost and optimal band ratio analysis (OBRA) models for total suspended sediments retrieval from remote sensing

• Main Authors: Pak, Hui Ying, Law, Adrian Wing-Keung, Lin, Weisi
• Other Authors: Interdisciplinary Graduate School (IGS)
• Format: Conference or Workshop Item
• Language: English
• Published: 2023
• Subjects: Engineering::Civil engineering::Water resources; Engineering::Environmental engineering::Environmental protection; Bayesian Model Aggregation; Remote Sensing; Water Quality Monitoring
• Online Access: https://hdl.handle.net/10356/156772; https://cmswebonline.com/iahr2022/epro/html/stheme-06-09.html
• Institution: Nanyang Technological University

In remote sensing of water quality, the use of data-driven models for the retrieval of water quality parameters has been gaining traction in recent years, especially with the recent advancement in machine learning (ML). The application of deep learning ML models on satellite imageries for retrieving water quality parameters for large areas is also becoming ubiquitous, although in many cases, simpler ML models like band-ratio algorithms can also achieve similar performances because the limited bands available on most satellite imageries restrict high dimensional data for the ML models to exploit. With Unmanned Aerial Vehicles (UAVs) and hyperspectral sensors, however, fine spectral and spatial data can be obtained and there is a need for more advanced ML to be developed to capitalize on the richer information. In this study, a new method called Hierarchical Bayesian Model Aggregation for Optimal Multiple Band Ratio Analysis (HBMA-OMBRA) has been developed as a proof-of-concept for the optimal determination of Total Suspended Solids (TSS) concentrations from the hyperspectral data. A laboratory investigation was also conducted in the present study to measure the hyperspectral reflectance in experiments under various environmental conditions to verify the robustness of HBMA-OMBRA. The performance of the HBMA-OMBRA was benchmarked against the popular XGBoost, as well as a band-ratio algorithm – Optimal Band Ratio Analysis (OBRA). The overall results showed that HBMA-OMBRA performed the best among these models.