Technical Abstract: Phosphorus transfer from agricultural soils to surface waters is an important environmental issue. Commonly used models like SWAT have not always been updated to reflect improved understanding of soil P transformations and transfer to runoff. Our objective was to validate the ability of the P routines in both SWAT2000 and SWAT2005 to initialize the size of the soil P pools (and thus, soil total P) and simulate long-term changes in the simulated Solution P pool. We used data from 40 published studies. Results showed that currently published equations to estimate the soil PSP parameter and SWAT’s method of summing four soil P pools under-predict soil total P. An under-prediction of soil total could subsequently result in under-prediction of P loss in runoff with eroded sediment. Conversely, our proposed alternative for estimating soil total P, which includes a new equation to estimate the soil PSP parameter and including the Solution P pool when summing soil P pools, resulted in accurate predictions of soil total P for 484 topsoil samples from 35 published studies. Our results also showed that both the SWAT routines and routines proposed by Vadas et al. (2006) for simulating changes in soil P accurately predict long-term changes in soil Solution P. However, routines in both SWAT 2000 and SWAT 2005 may under-predict Solution P for several weeks after P is added to soils. This could result in under-prediction of dissolved inorganic P loss in runoff soon after a P application to soils. The routines proposed by Vadas et al. (2006) would not result in similar under-predictions.