水生态学杂志

探究中国四级子流域景观格局对河流氮磷空间分异的影响及丰枯水期差异，为跨流域面源污染分区分类管控与“源-汇”景观空间优化提供量化依据。以中国境内110个四级子流域为基本单元，基于2021年河流水质监测数据和30 m分辨率土地利用数据，结合全局/局部Moran’s I、冗余分析（RDA）、弹性网惩罚回归、bioenv分析及变差分解等方法，系统识别丰、枯水期下的关键景观因子及“源-汇”景观贡献。结果表明：TN、TP在两水期均呈显著正空间自相关，具有明显区域聚集特征；景观破碎化与边缘效应指标（PD、ED、LSI）与营养盐浓度呈正向关联，聚集连通性指标（PLADJ、CONTAG、LPI）总体呈负向关联；边缘密度（ED）在4个惩罚回归模型中具有一致正向系数与较高入模频率，是跨水文期稳定的关键景观因子；“源-汇”景观对水质空间分异具有显著耦合性，丰水期总解释率（65.57%）高于枯水期（61.54%），且共享贡献均超过单一景观组独立贡献。研究表明河流氮磷空间分异具有明显水文期依赖性和“源-汇”协同调控特征，可为高风险子流域识别、“源-汇”景观空间优化及分水文期营养盐管控提供依据。

The effects of landscape patterns on the spatial differentiation of riverine nitrogen and phosphorus are a key prerequisite for basin-scale water environment management; however， most existing studies rely on a single statistical method and rarely compare landscape-level and class-level effects within a unified framework. In this study， 110 fourth-order sub-basins in China were used as basic units. Based on total nitrogen (TN) and total phosphorus (TP) monitoring data from 2021 and 30 m resolution land-use data， global and local Moran's I tests were applied to examine spatial autocorrelation during the wet and dry seasons. Correlation analysis， redundancy analysis (RDA)， elastic net penalized regression， bioenv analysis， and variation partitioning were integrated to identify key landscape factors and the contributions of “source–sink” landscapes at both the landscape and class levels. The results showed that TN and TP concentrations exhibited significant positive spatial autocorrelation in both hydrological periods. Landscape fragmentation and edge-effect indices (PD， ED， LSI) were positively associated with nutrient concentrations， whereas aggregation and connectivity indices (PLADJ， CONTAG， LPI) showed overall negative associations. Edge density (ED) consistently exhibited positive coefficients and high model-entry frequency across all four penalized regression models， making it the key landscape factor stable across hydrological periods. Source and sink landscapes showed significant coupling effects on spatial water quality differentiation， with the total explanatory rate being higher in the wet season (65.57%) than in the dry season (61.54%). This study demonstrates that edge expansion and landscape fragmentation are the key pattern characteristics driving elevated riverine nitrogen and phosphorus levels， and that the influence of source–sink landscape patterns on water quality is modulated by hydrological periods， providing quantitative indicators for the zoned and classified management of non-point source pollution and landscape pattern optimization across basins. Landscape patterns influence the spatial differentiation of riverine total nitrogen (TN) and total phosphorus (TP) by regulating pollutant generation， hillslope runoff， river-network connectivity， and nutrient retention and transformation. To address the limited understanding of multidimensional landscape metrics， multi-method validation， and source–sink contribution partitioning at the national sub-watershed scale， this study selected 110 HydroSHEDS Level-4 sub-watersheds in China as analytical units. River water-quality monitoring data from 2021 and 30 m land-use data were integrated， and dry and wet seasons were defined using the 25th and 75th percentiles of monthly precipitation in each sub-watershed. Spatial autocorrelation analysis， directional correlation analysis， redundancy analysis (RDA)， Elastic Net regression， bioenv analysis， and variation partitioning were used to quantify the effects of landscape patterns on the spatial differentiation of TN and TP and their hydrological-period differences. The results showed that TN and TP exhibited significant positive spatial autocorrelation in both seasons， indicating clear regional clustering. Fragmentation and edge-effect metrics， including patch density (PD)， edge density (ED)， and landscape shape index (LSI)， were generally positively associated with TN and TP， whereas aggregation and connectivity metrics， including percentage of like adjacencies (PLADJ)， contagion index (CONTAG)， and patch cohesion index (COHESION)， were generally negatively associated with nutrient concentrations. These results suggest that landscape fragmentation and edge complexity may increase nutrient export risk， while continuous sink landscapes can help reduce pollutant delivery to rivers. Landscape-level RDA explained 38.92% and 37.17% of nutrient variation in the dry and wet seasons， respectively. At the class level， cropland had a higher explanatory rate in the dry season (37.8%)， whereas built-up land showed a stronger role in the wet season (36.9%). The optimal bioenv indicator shifted from PD in the dry season (r = 0.299) to ED in the wet season (r = 0.308)， indicating the enhanced role of edge structure in characterizing nutrient spatial differences during the wet season. Variation partitioning showed that source–sink landscapes explained more variation in the wet season (65.6%) than in the dry season (61.5%)， and their shared contribution exceeded the independent contribution of either landscape group. These findings indicate that riverine nitrogen and phosphorus differentiation is characterized by hydrological-period dependence and source–sink coupling， providing support for identifying high-risk sub-watersheds， optimizing source–sink landscape patterns， and implementing season-specific nutrient management.