The global distribution of phosphorus (P) in agricultural soils is heterogeneous and this heterogeneity results from variability in both virgin soil properties and past agricultural practices. A current picture of this distribution is required to identify the deficit (surplus) areas with negative effect on productivity (respectively environment). However, such a picture is missing due to limitations in soil P measurement techniques and in modeling approaches. Here, we built a simple soil P dynamic model and combined it with global spatially explicit (0.5°lat x 0.5°lon) datasets to reconstruct the temporal evolution of P in agricultural soils from the cultivation starting up to the current time period. Global datasets provided information about land-use, agricultural soil P input/output and current soil P in unmanaged soil. This latter information was used to approach the initial conditions of agricultural soils, i.e. at the time of conversion from non-agricultural to agricultural land. The design of the soil P dynamic model was chosen to be consistent with Hedley fractionation method, on which some available global datasets are based. Maps of current soil P distribution were simulated and evaluated against a database compiling Hedley measurements on cropland and pasture sites from the literature. Our modeling approach allowed to assess the drivers of the current P soil distribution, and through sensitivity tests, to identify the main sources of uncertainty. It is the foundation for assessing how the agricultural soil P would evolve under different future scenarios of P resource management in context of increasing scarcity.