Abstract: The phosphorus extraction coefficient (ECp) of surface soil runoff is an important parameter to simulate the release and transport of dissolved phosphorus from upper soil layer to surface runoff, and to estimate the phosphorus output of agricultural non-point source (NPS) pollution. However, the vague connotation and diverse determination methods lead to varied ECps, and the variation range and influence mechanisms on ECps remains unclear. We reviewed the relative researches at home and abroad in aim to provide relatively accurate and comparable ECps for evaluation of agricultural NPS simulation. ECp is defined as the ratio of dissolved phosphorus concentration (DP) in surface runoff to that in surface soil. ECps are usually tested using rainfall experiments, in which the linear regression equations of DP between surface runoff and surface soil are established, and the slope of the regression is defined as ECp. ECps are usually affected by such factors as precipitation, soil property, vegetation, and land use patterns. ECp is inversely proportional to soil clay content, adsorptivity, vegetation coverage, etc., and is directly proportional to the initial water content of soil. The ECps in non-calcareous soil are higher than those in calcareous soil. For calculation convenience, ECp is usually taken as a constant in some NPS pollution models, normally 7.5 in the chemicals, runoff, erosion from agriculture management models (CREAMS) and in agricultural nonpoint source pollution models (AGNPS), and 5.7 in the erosion/productivity impact calculator (EPIC) and soil and water assessment tools (SWAT) models. In the future, the physical connotation of ECp still needs studied and defined; the ECps in each rainfall event and on average need monitored in accuracy. The nonlinear correlation between soluble phosphorous concentration in surface runoff and surface soil should be studied to improve the applicability of the determination method. Specific or regional ECps require systemically monitoring, especially in slope land with frequent fertilization and complex hydraulic erosion, so as to propose suitable calibration coefficients for ECps under different conditions. We strongly recommend to replace the constant ECp value in some NPS pollution models with the calibrated or observed ECps for precise simulation of P load from surface soil.