Abstract: Non-point source pollution (NPS) has become a serious problem worldwide. Currently the methods to evaluate N loads of NPS in the large scale are limited to nitrogen budgets and export coefficient of N fertilizer application, which are highly uncertain. On the basis of 187 data across China's croplands, a stepwise regression model including key factors of N application, rainfall, and soil clay content was proposed to simulate N runoff and leaching of cropland of China. Using the statistic yearbook of 2011, the simulations show that the total N runoff from cropland was 0.96 Tg, accounting for 6.0% of the total N fertilizer input, in which dryland and paddy field released 0.76 and 0.20 Tg, respectively. While the total leached N from cropland was 1.01 Tg, accounting for 6.3% of the total N fertilizer input, in which dryland and paddy field released 0.87 and 0.14 Tg, respectively. The results have highlighted the importance of non-point source N from the middle and low reaches of Yangzi River, hilly areas of Southwest China, Shandong Province, and North China Plain. The model has not only provided the approach of estimating the distribution of non-point source N from cropland, but also greatly reduced the uncertainty of non-point source N estimation at large scale, compared to the traditional export coefficient model.