There is growing evidence that the release of phosphorus (P) from "legacy" stores can frustrate efforts to reduce P loading to surface water from sources such as agriculture and human sewage. Less is known, however, about the magnitude and residence times of these legacy pools. Here we constructed a budget of net anthropogenic P inputs to the Baltic Sea drainage basin and developed a three-parameter, two-box model to describe the movement of anthropogenic P though temporary (mobile) and long-term (stable) storage pools. Phosphorus entered the sea as direct coastal effluent discharge and via rapid transport and slow, legacy pathways. The model reproduced past waterborne P loads and suggested an~30-year residence time in the mobile pool. Between 1900 and 2013, 17 and 27 Mt P has accumulated in the mobile and stable pools, respectively. Phosphorus inputs to the sea have halved since the 1980s due to improvements in coastal sewage treatment and reductions associated with the rapid transport pathway. After decades of accumulation, the system appears to have shifted to a depletion phase; absent further reductions in net anthropogenic P input, future waterborne loads could decrease. Presently, losses from the mobile pool contribute nearly half of P loads, suggesting that it will be difficult to achieve substantial near-term reductions. However, there is still potential to make progress toward eutrophication management goals by addressing rapid transport pathways, such as overland flow, as well as mobile stores, such as cropland with large soil-P reserves.Plain Language Summary All life depends on phosphorus (P), which is why it is an important crop fertilizer. Humans generally consume more P than needed and the excess ends up in sewage systems. Past management of P in fertilizer and human sewage has led to the accumulation of P in soils and sediments of lakes and streams. This accumulation is called "legacy" P because it can leak for decades to downstream lakes and coastal areas where it contributes to environmental problems. We developed a model to understand P dynamics for the entire drainage basin of the Baltic Sea since 1900. This model included a rapid transport pathway that represented sources such as runoff from cropland and a slow pathway that represented leakage from mobile legacy sources. The model suggests that loss from the mobile pool contributes about half of current waterborne inputs to the sea; as a result, it could be difficult to make substantial near-term reductions. However, there are opportunities to meet environmental goals by slowing the accumulation of P in the landscape and by implementing measures that address the rapid transport pathway, such runoff from cropland, and the mobile stores, such as cropland with large soil-P reserves.