Evaluating optically and non-optically active water quality and its response relationship to hydro-meteorology using multi-source data in Poyang Lake, China. (December 2022)

• Record Type: Journal Article
• Authors: Fu, Bolin; Lao, Zhinan; Liang, Yiyin; Sun, Jun; He, Xu; Deng, Tengfang; He, Wen; Fan, Donglin; Gao, Ertao; Hou, Qinglin
• Is Part Of: Ecological indicators. Volume 145(2023)
• Publication Date: 2022-12
• Subjects: Water quality assessment -- Optically and non-optically active parameters -- SEL algorithms -- Response relationship -- Hydro-meteorological factors -- Hyperspectral and SAR images; Environmental monitoring -- Periodicals; Environmental management -- Periodicals; Environmental impact analysis -- Periodicals; Environmental risk assessment -- Periodicals; Sustainable development -- Periodicals; 333.71405
• DOI: 10.1016/j.ecolind.2022.109675
• Languages: English
• ISSNs: 1470-160X

Abstract

Highlights: Six optically and non-optically active parameters of Poyang Lake, China, were estimated using multi-sensors images. Estimation accuracy (MRE) of water quality parameters was ranged from 2.15% to 44.91%. LOOCV-ML provides better prediction of water quality than ML and SEL models. Daily minimum and mean temperature have a positive influence on optical water quality parameters. Sunshine hours and wind speed positively affected NH4 -N and TP while water temperature and rainfall negatively affected on TP and BOD5 . Abstract: Lakes potentially provide sensitive indicators of global climate change, but water pollution seriously affects the lake ecosystem. Therefore, mapping and monitoring water quality of lake has attracted extensive attention from scholars. However, estimating optically active (OAPs) and non-optically water quality parameters (NOAPs) and clarifying their response relationship with hydro-meteorology still face great challenges. To resolve these issues, this paper presents an approach for assessing OAPs (Chlorophyll a (Chla), Total suspended matter (TSM)) and NOAPs (total phosphorus (TP), total nitrogen (TN), ammonia nitrogen (NH4 -N), five-day biochemical oxygen demand (BOD5 )) in Poyang Lake, the largest freshwater lake in China, by using ensemble learning (SEL), combing leave-one-out validation and machine learning (LOOCV-ML) algorithms with multi-source remote sensing images. We further quantitatively evaluated retrieval accuracy differences between OAPs and NOAPs using multispectral and hyperspectral images. Meanwhile, a novel automatic extraction method was proposed in this paper for mapping water bodies. We systematically explored the response relationship between water quality (OAP and NOAP) and hydro-meteorology. We found that NOAPs achieved higher estimation accuracy than OAPs. The LOOCV-XGBoost and LOOCV-GradientBoost provided higher inversion accuracy of OAP and NOAP than SEL, demonstrating the great potential of combining OHS-2A hyperspectral images with LOOCV-ML to estimate the water quality in inland lakes. We revealed that the estimation accuracy (MRE) of TN and TP was 1.09 %-18.03 % higher than NH4 -N and BOD5 . Finally, we clarified that water temperature had a significantly negative effect on NH4 -N and BOD5 (Projected importance of variables (VIP) > 1, Coefficient < -1), and sunshine hours produced a great positive effect on Chla, NH4 -N and BOD5 (VIP > 1, Coefficient > 1), further revealing that water temperature, water body area, wind speed, rainfall, mean relative humidity and sunshine hours were important divers of OAPs and NOAPs changes. These results provide a scientific basis for restoration and protection of water quality in lake ecosystems.