[Objective] Studies on accumulation and potential loss risk of soil phosphorus in non-wood forest soils are essential to management of soil phosphorus and control of non-point source P pollution in watersheds. So a research project was carried out in the Xitiaoxi watershed.[Method] In the project, a total of 105 soil samples were collected in the main non-wood forests, such as moso bamboo (Phyllostachys heterocycla (Carr.) Mitford cv. Pubescens Mazel ex H.de leh.), white tea (Camellia sinensis (L.) O. Ktze.), chestnut (Castanea mollissima Bl.) in the watershed for lab analysis of concentrations of water-extractable phosphorus (WEP) and simulated-acid-rain-extractable phosphorus (SARP) and some main soil physico-chemical properties such as total nitrogen (TN), total phosphorus (TP), total potassium (TK), alkalyzable nitrogen (AN), available phosphorus (Bray 1-P), readily available potassium (AK), pH, soil organic carbon (SOC). Acid rain set at 4.75 in pH was simulated and prepared with sulphate acid and nitrate acid. Correlation analysis and regression analysis was performed of the obtained data to determine influences of soil physico-chemical properties on the risk of phosphorus runoff loss and threshold of soil available phosphorus.[Result] In the survey area, soil TP content varied in the range of 0.22-0.73 g·kg^-1 (0.42 g·kg^-1 on average) and Bray 1-P content in the range of 0.93-313.2 mg·kg^-1 (30.87 mg·kg^-1 on average) with coefficient of variation reaching up to 204.7% and the soil samples over 40 mg·kg^-1 and below 5 mg·kg^-1 in Bray-P content accounted for 17.14% and 38.01% of all the analyzed ones, respectively; soil WEP content varied in the range of 0.03-38.15 mg·kg^-1 (2.64 mg·kg^-1 on average) with coefficient of variation reaching up to 267.5%; and soil SARP content varied in the range of 0.03-42.91 mg·kg^-1 (2.86 mg·kg^-1 on average) with coefficient of variation reaching up to 268.6%. It was found that soil WEP and SARP were significantly and positively related to soil organic carbon (SOC) and TP (P<0.01), and negatively to soil pH, with determination coefficient (R²), however, being only 0.266, 0.251, 0.280, 0.262, 0.187 and 0.190, respectively, which indicates that they are relatively not much affected by these soil properties. The relationship between WEP, SARP and Bray 1-P could be described by the piecewise linear regression equation, which was y=0.148 3x-8.601, x>93.63;0.056 9x-0.043, x<93.63. for WEP with R² being 0.992(r=0.996), and y=0.057 1x-0.021 6, x<87.68;0.160-4x 9.079, x>87.68. for SARP with R² being 0.991 (r=0.995). Hence, the thresholds of soil Bray 1-P was reckoned to be 93.63 and 87.68 mg·kg^-1, respectively, based on WEP and SARP and the latter was 5.95 mg·kg^-1 lower than the former.[Conclusion] All the findings in the study demonstrate that both the phenomena of phosphorus deficiency and excessive accumulation exist in the non-wood forest soils of Xitiaoxi watershed. Soil Bray 1-P, TP, organic matter and pH are the main factors affecting the potential risk of phosphorus runoff loss and Bray 1-P is the most significant one. Acid rain enhances the potential risk of phosphorus runoff loss. As the studied region is one that suffers from frequent acid rain, the content of soil available phosphorus should be more strictly controlled.