Recently, Researcher He Xianqiang’s team from our laboratory and their collaborators published a research paper titled "Satellite retrievals of water quality for diverse inland waters from Sentinel-2 images: an example from Zhejiang Province, China" in the top remote sensing journal International Journal of Applied Earth Observation and Geoinformation. The first author of the paper is Zhao Yaqi, a direct PhD student jointly trained by our laboratory and Zhejiang University, and the corresponding author is Researcher He Xianqiang. Collaborators include Senior Engineer Pan Shuping from Zhejiang Provincial Ecological and Environmental Monitoring Center, Researcher Bai Yan, Researcher Wang Difeng, Associate Researcher Li Teng, Senior Engineer Gong Fang, and Engineer Zhang Xuan from our laboratory.
Monitoring inland water quality is of great significance for water resource protection and management, water pollution control, and ecological system health protection. However, the sources of water components vary across different water bodies, leading to complex and variable optical properties of water. On the other hand, it is difficult to establish a direct relationship between non-optically active water chemical parameters such as nutrients and remote sensing spectra, requiring the development of empirical statistical or machine learning models based on in-situ measured data. Therefore, constructing remote sensing retrieval models for non-optically active water quality parameters applicable to diverse inland waters (e.g., rivers, lakes, and reservoirs) is a major current challenge.
From a data-driven perspective, this study took Zhejiang Province as an example. Based on the abundant historical water quality data accumulated by Zhejiang’s automatic water quality monitoring stations, matched with 10-meter resolution Sentinel-2 satellite remote sensing images, a large-scale satellite remote sensing-in-situ measurement matched dataset was constructed. The dataset includes 8,760 cases of permanganate index, 7,434 cases of total nitrogen (TN), 8,845 cases of total phosphorus (TP), and 8,642 cases of ammonia nitrogen (NH₃-N), covering various optical types of water bodies and different seasons to ensure extensive and comprehensive sample coverage (Figure 1). On this basis, combined with the eXtreme Gradient Boosting (XGBoost) machine learning method, unified remote sensing retrieval models for each water quality parameter applicable to diverse inland waters were established. Validation results show that the water quality parameters retrieved by remote sensing are in good agreement with in-situ measured data in terms of spatial distribution (Figure 2) and temporal variation (Figure 3). Using this model, water quality monitoring of different water bodies can be achieved every 5 days under clear sky conditions (Figure 4).
Figure 1. Distributions of spectra and concentrations of each water quality parameter in the matched dataset
Figure 2. Variations of satellite-retrieved values and in-situ measured values at stations along the Qiantang River with station locations
Figure 3. Temporal variations of satellite-retrieved values and in-situ measured values at the station in front of the Cao’e River Grand Sluice
Figure 4. Remote sensing monitoring of water quality parameter concentrations in the Qiantang River Basin (2016 ~ 2023 average results). (a) Permanganate index, (b) Ammonia nitrogen, (c) Total nitrogen, (d) Total phosphorus, (e) Turbidity
This study is a new achievement following the team’s previous breakthrough in riverine water quality remote sensing technology, and has been operationally applied to the remote sensing monitoring of marine ecological environment in Zhejiang Province. The research was supported by the Zhejiang Provincial "Jianbing" Program (2023C03011) and the National Key Research and Development Program of China (2023YFC3108101).
Paper Citation: Zhao, Y., He, X.*, Pan, S., Bai, Y., Wang, D., Li, T., Gong, F., Zhang, X., 2024. Satellite retrievals of water quality for diverse inland waters from Sentinel-2 images: An example from Zhejiang Province, China.
Int. J. Appl. Earth Obs. Geoinformation, 132, 104048.
https://doi.org/10.1016/j.jag.2024.104048
