Unraveling the effects of hydrological connectivity and landscape characteristics on reservoir water quality

• Published in: Journal of hydrology (Amsterdam) 2022-10, Vol.613, p.128410, Article 128410
• Main Authors: Wang, J., Yin, W., Wang, L.X., Li, N.X., Hao, R., Xu, J.F., Xiao, H.B., Zhong, Z.M., Zhang, Y.P., Shi, Z.H.
• Format: Article
• Language: English
• Subjects: Hydrological connectivity; Landscape characteristics; Machine learning; Reservoir bay; Water quality index
• ISSN: 0022-1694; 1879-2707
• DOI: 10.1016/j.jhydrol.2022.128410

[Display omitted] •Machine learning approaches quantified the effects of hydrological and landscape metrics on water quality.•Impacts of hydrological and landscape metrics on water quality varied with hydrological regulations.•Landscape composition exhibited the largest impact on water quality during the FDPs.•Landscape configuration had the largest impact on water quality during the WSPs. Dam construction and reservoir operation altered the landscape and hydrological process of reservoir bays, affecting reservoir water quality. However, many landscape and hydrological connectivity metrics are highly correlated and may introduce redundancies and misleading results when use conventional multivariate regression techniques. Knowledge concerning the pure effects of landscape and hydrological connectivity metrics are crucial for understanding nonpoint pollution processes and guiding water quality protection strategies in reservoirs. Based on water quality monitoring data for six years (2015–2020) from 66 reservoir bays of the Danjiangkou Reservoir in China during both flood discharge periods (FDPs) and water storage periods (WSPs), machine learning approaches (boosted regression trees and random forest) were conducted to decipher the effects of hydrological connectivity and landscape characteristics on water quality. The results showed that landscape composition, landscape configuration, and topography had the combined importance of 46.69%, 31.48%, and 10.14% on the overall water quality changes during the FDPs, respectively. However, landscape configuration was the largest importance factor controlling overall water quality for the WSPs with the combined importance of 38.57%. For the FDPs, the top two importance variables of overall water quality variation were the proportions of shrub and agricultural land in the reservoir bay. For the WSPs, the top two importance variables were flow length and index of connectivity. For specific water quality parameters, the highest importance factor controlling the variation in total nitrogen, total phosphorus, ammonia nitrogen, nitrate and chlorophyll a were landscape configuration, while the landscape composition had the highest importance on the variation in permanganate index, suggesting landscape characteristics affected water quality with specific responses to distinct water quality parameters. These findings emphasize the distinct roles of landscape and hydrological characteristics on water quality and provide import