关键词: 景观格局, 氮污染, 景观空间负荷对比指数, 偏最小二乘法, 非参数突变点分析

Abstract: Understanding the relationship between landscape patterns and water quality in river headwater watersheds is essential for developing sustainable landscape policies to protect water quality in water source areas. With the Qingshan Lake headwater watershed as the research object and based on the data of 25 water sampling sites between 2023 and 2024, we used the partial least squares regression (PLSR), non-parametric change-point analysis and bootstrap methods to quantitatively assess the impacts of landscape patterns on riverine nitrogen concentration during high-flow, normal-flow, and low-flow periods. The results showed that there were significant differences in landscape dominance and fragmentation among different sub-watersheds. High landscape weighted load index (LWLI) values (>0.50) were predominantly observed in low-altitude, gently sloping areas were characterized by extensive “source” landscapes, whereas low LWLI value (<0.10) were mainly distributed in mid-altitude regions dominated by forests. The optimal PLSR model accounted for 60.6%, 69.7%, and 78.3% of the variance in total nitrogen (TN) concentrations during the high-flow, normal-flow, and low-flow periods, respectively. Variable importance in projection (VIP) analysis revealed that LWLI was the dominant landscape factor driving TN concentrations throughout the year. The proportion of build-up land primarily affected TN concentrations during the high-flow period, while the proportion of grassland and the largest patch index had more substantial effects during the normal-flow period. During the low-flow period, the proportion of forest land emerged as the most dominant factor. LWLI and the proportion of construction land exerted positive effects on TN concentrations, whereas the proportion of grassland, the largest patch index, and the proportion of forest land exhibited negative effects. When the LWLI value exceeded 0.35, the cumulative probability of abrupt changes in TN concentration during the high-flow period exceeded 95.0%, thereby elevating the risk of water quality degradation. Optimizing landscape patterns could effectively control non-point source pollution and improve water quality in headwater watersheds.

Key words: landscape pattern, nitrogen pollution, landscape weighted load index, partial least squares regression, nonparametric change-point analysis