Detail Abstract

Using a Neural Network Based Model to Retrieve the Inland Water Bodies\' Water Quality Parameters from Remote Sensing Data in Thailand

Abstract

Freshwater is vital for human wellbeing and the ecosystem in general. Inland water bodies are the primary sources of freshwater, and they are both directly and indirectly impacted by climate change and anthropogenic activities. Constant monitoring of water bodies is essential in understanding the dynamic nature of the water column as well as its properties for the long-term goal of fresh water and sanitization in such a situation. There are conventional methods for retrieving the water quality indicators, but in the monitoring of large surfaces waterbodies, combining remote sensing data with its high temporal, geographical, and spectral characteristic capabilities will provide a wide range of information to comprehend and estimate the patterns of the waterbodies. Several techniques for extracting active water quality indicators from images, as a proxy for phytoplankton and sediments in the water column, have become available with the development of machine learning and deep learning over the last few decades.</br> Considering the above-mentioned advancements, the main aim of this study was to apply spatially and temporally validated NN-based Case-2 Regional Coast Color (C2RCC) processor and to extract Water Quality Parameter (WPQ) in tropical inland and riverine waters of Thailand using in situ observation and permanent water quality datasets. C2RCC was indeed used to retrieve and estimate the concentration of chlorophyll, total suspended matter and turbidity using high spatial and temporal Earth observation data from Sentinel-2 MSI and Landsat 8 OLI. The satellite-based estimates have been compared to conventional field observations and the accuracy assessed.