Agricultural land is widely acknowledged as a primary non-point source of nutrients that affect the health of freshwater ecosystems throughout many parts of the world. Simulated rainfall-induced runoff trials were conducted at four P amended primary field sites under annual crop cultivation and three un-amended secondary field sites during the fall of 2009 and 2010. Highly significant (P <0.001) linear relationships were observed between measures of soil P and runoff TDP (r2 = 0.20–0.43). Olsen P was as accurate as other soil test methods for predicting runoff P losses. However, site, sampling year, runoff interval and crop residue management practices also affected these relationships. In particular, the predictability of these relationships on a site by site basis, as well as from one year to the next was extremely variable (r2 = 0.00-0.86), and declined significantly as the range of STP values decreased below approximately 80-100 mg P kg-1.