Remote monitoring of water clarity in coastal oceans of the Guangdong-Hong Kong-Macao Greater Bay Area, China based on machine learning

• Published in: Ecological indicators Vol. 160; p. 111789
• Main Authors: Lu, Xinyi, Mo, Zifeng, Zhao, Jun, Ma, Chunlei
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.03.2024; Elsevier
• Subjects: algorithms; China; Empirical model; estuaries; governance; Guangdong-Hong Kong-Macao Greater Bay Area; Landsat; Machine learning; Remote sensing; rivers; sediment yield; time series analysis; turbidity; Water clarity; water quality; wind speed; China; Water clarity; Guangdong-Hong Kong-Macao Greater Bay Area; Empirical model; Remote sensing; Machine learning
• ISSN: 1470-160X; 1872-7034; 1872-7034
• DOI: 10.1016/j.ecolind.2024.111789

•A machine learning based algorithm for retrieving Secchi depth (ZSD) was developed.•ZSD were shallowest in the Pearl River Estuary, followed by Daya Bay and Mirs Bay.•ZSD increased in the Pearl River Estuary and Daya Bay, decreased in the Mirs Bay.•Sediment discharge, wind speed, and precipitation regulated ZSD in the study area. The development of the Guangdong-Hong Kong-Macao Greater Bay Area (GBA), one of the most developed and densely populated regions in China, has posed an increasing threat to the health of the water environment in adjacent coastal oceans. However, the spatiotemporal variations of water clarity in the coastal oceans of the GBA (COGBA) have not been well-documented. Therefore, this study aims to develop a remote sensing retrieval algorithm for assessing water clarity, represented by Secchi disk depth (ZSD), based on Landsat-8 OLI imagery for the COGBA using a machine learning method. Cross-validation demonstrated that the algorithm performed exceptionally well, with an R2 value greater than 0.8. By applying the algorithm to 263 Landsat-8 OLI images, we obtained a time series of ZSD for the period of 2013–2021 over the COGBA. Results indicate that the Pearl River Estuary (PRE) exhibited the highest turbidity, followed by the Daya Bay and the Mirs Bay. Generally, ZSD increased from the northwest to the southeast of the COGBA. Seasonal, interannual, and long-term variations in ZSD were also observed with long-term increases in the PRE and the Daya Bay. The interannual variations in ZSD during fall were primarily regulated by different factors in each region. In the PRE, the negative effects of sediment discharge and wind speed played a significant role. In Mirs Bay, wind speed and sediment discharge had a negative impact on ZSD. In Daya Bay, precipitation and wind speed were the key factors influencing ZSD. The development and findings of our algorithm contribute to the protection and management of the water environment in the COGBA, facilitating effective governance measures.