[Correlating landscape pattern with total nitrogen concentration using a location-weighted sink-source landscape index in the Haihe River Basin, China].

Understanding the effect of land cover pattern on nutrient losses is of great importance in management of water resources. The extensive application of mechanism models is limited in large-scale watersheds owing to the intensive data and calibration requirements. On the other hand, the traditional landscape indexes only take the areas and types of land cover into account, considering less about their topographic features and spatial patterns. We constructed a location-weighted landscape index (LWLI) based on the Lorenz curve, which plots the cumulative proportion of areas for sink and source landscapes respectively against cumulative proportion of their relative location to the outlet in a watershed, including relative elevation, distance and slope. We assessed the effect of land cover pattern on total nitrogen losses in the Haihe River. Firstly, 26 watersheds were derived from 1: 250 000 digital elevation model (DEM), and their "source" and "sink" landscape types were identified from Landsat TM images in 2007. The source" landscapes referred to the paddy land, dry land and residential area, correspondingly the "sink" landscapes referred to the forest and grassland. Secondly, LWLI was calculated according to the landscape types and spatial patterns for each watershed. Thirdly, we accessed the effect of land cover pattern on total nitrogen (TN) flux according to the value of LWLI, comparing with the area proportion of sink-source landscapes. The correlation coefficients were different in three parts of Haihe River, i. e., 0.86, 0.67 and 0.65 in the Yanshan Mts, Taihang Mts and lower Haihe River. The results showed strong correlations between TN and LWLI in contrast to the weak correlations between TN and area proportion of sink and source landscape types. This study indicates the spatial pattern of land cover is essential for accessing the nutrient losses, and the location-weighted landscape pattern analysis may be an alternate to existing water quality models, especially in large watershed scales. The sink-source index is sufficiently simple that it can be compared across watersheds and be easily interpreted, and potentially be used in landscape pattern optimal designing and planning.