Nutrient loading into rivers is generally increased by human-induced land-use changes and can lead to increased surface water pollution. Understanding the extent to which land-use patterns influence nutrient loading is critical to the development of best-management practices aimed at water-quality improvement. In this study, we investigated total nitrogen (total N) concentration as a function of land-use patterns and compared the relative significance of the identified land-use variables for 26 upstream watersheds of the Haihe River basin. Seven land-use intensity and nine landscape complexity variables were selected to form the land-use pattern metrics on the landscape scale. After analyzing the significance of the land-use pattern metrics, we obtained five dominant principal components: human-induced land-use intensity, landscape patch-area complexity, area-weighted landscape patch-shape complexity, forest and grassland area, and landscape patch-shape complexity. A linear regression model with a stepwise selection protocol was used to identify an optimal set of land-use pattern predictors. The resulting contributions to the total N concentration were 50% (human-induced land-use intensity), 23.13% (landscape patch-shape complexity), 14.38% (forest and grassland area), and 12.50% (landscape patch-area complexity), respectively. The regression model using land-use measurements can explain 87% of total N variability in the upstream regions of Haihe River. The results indicated that human-related land-use factors, such as residential areas, population, and road density, had the most significant effect on N concentration. The agricultural area (30.1% of the study region) was not found to be significantly correlated with total N concentration due to little irrigative farmland and rainfall. Results of the study could help us understand the implications of potential land-use changes that often occur as a result of the rapid development in China.