A SPARROW (Spatially Referenced Regression on Watershed attributes) model was established to estimate total nitrogen (TN) in Xin'anjiang River Basin. The land-to-water delivery factor (LDF) denoted the influences of three land-water variables of rainfall, slope and air temperature, combined with discharge coefficients of three pollutant sources, i.e., agricultural source, forestland-grassland source and life source, the delivery ratio (LDR) of total nitrogen after transported by land-water and reached river channel was analyzed, thus to reveal the nonpoint source pollution of the total nitrogen. The results showed that, the influences of slope differed greatly in the whole basin, LDF value ranged between 0.86 and 1.06, therefore, the transport ratios of three pollution sources into the stream channels had large differences. Considering the overall effects of three land-to-water delivery variables, the maximum and minimum LDF occurred in the 60th and 225th subbasins, respectively. Thus the areas where the 60th subbasin located was considered prior to the other subbasins in the management if the emission reduction measures required the same reductions into the river. The influences of hydrological and non-hydrological factors were described by using a improved stream attenuation equation. Instead of the first-order reaction coefficient, the mass transfer coefficient was taken as the model simulation parameter. The reduction rate was negatively correlated with the flow rate, and most of reach reduction rates were within the value scopes of the previous literatures. It showed that the improved mass transfer rate was feasible for the TN model of Xin'anjiang River Basin.