To provide a scientific basis for sustainable land management, a 20-year fertility experiment was conducted in Changwu County, Shaanxi Province, China to investigate the effects of long-term application of chemical fertilizers on wheat grain yield and yield stability on the Loess Plateau using regression and stability analysis. The experiment consisted of 17 fertilizer treatments, containing the combinations of different N and P levels, with three replications arranged in a randomized complete block design. Nitrogen fertilizer was applied as urea, and P was applied as calcium superphosphate. Fertilizer rates had a large effect on the response of wheat yield to fertilization. Phosphorus, combined with N, increased yield significantly （P 〈 0.01）. In the unfertilized control and the N or P sole application treatments, wheat yield had a declining trend although it was not statistically significant. Stability analysis combined with the trend analysis indicated that integrated use of fertilizer N and P was better than their sole application in increasing and sustaining the productivity of rainfed winter wheat.
