Relating land-use/land-cover patterns to water quality in watersheds based on the structural equation modeling

• Source-sink landscape theory was reconstructed by the structural equation modeling. • Natural vegetation has the common characteristics of “source” and “sink” • Watershed fragmentation intensifies the effect of urban areas on organic pollution. • Water quality is mainly affected by point source pollution in the dry season. • The ecological process was reflected. Understanding the relationships between land-use/land-cover (LULC) patterns in watersheds and water quality is of great significance for exploring the causes of non-point source pollution and scientific planning of land use. Multivariate regression methods have often been used to examine the response of specific water contaminants to LULC patterns. However, this also resulted in overlooking the ecological process and biased parameter estimates caused by redundancy where some landscape metrics are strongly correlated. In this study, based on “source-sink” landscape theory, Redundancy Analysis (RDA) and Partial Least Squares Structural Equation Modeling (PLS-SEM) were used to investigate the relationship between LULC patterns and water quality in the surrounding sub-watersheds of the Danjiangkou Reservoir in the dry and rainy seasons. The results showed that water quality parameters were characterized by seasonal differences. In the dry season, only the proportion of the watershed occupied by urban landscapes (P_Urban) had a significant impact on water quality in watersheds ( β = 0.835, p < 0.001), so the water quality was only affected by point source pollution in the dry season. In the rainy season, specific water contaminants were categorized as organic and inorganic pollutants. The “sources” of organic and inorganic pollutants were agricultural and urban land, respectively. However, natural vegetation was not only a “sink” of inorganic pollutants, but also a “source” of organic pollutants. Furthermore, agriculture and natural vegetation mediated the effects of landscape spatial configuration on the water quality, yet landscape aggregation negatively moderated the relationship between urban and organic pollution. Therefore, watersheds with high landscape aggregation have small agricultural landscapes and large natural vegetation landscapes where water quality is affected by low levels of inorganic pollution and high levels of organic pollution. However, organic pollution from urban sources was aggravated by watersheds with low landscape aggregation. This study revealed the relationships between LULC patterns and water quality in watersheds and reflected the ecological process to a certain extent.