To gain insights into phosphorus (P) dynamics in soils and the ability to predict soil responses to varying fertilizer inputs, mass balance models prove to be valuable tools. In this study, a new dynamic mass balance model, PBalD8, was used to describe the change in extracted P in the A horizon of soils subjected to diverse fertilizer treatments over a period of 50 to 60 years in five soil fertility experiments. The model employed a Freundlich equation to describe soil-solution partitioning of P and assumed that acid-lactate-extractable P represented a labile pool of P in instant equilibrium with soil solution P. Additionally, oxalate-extractable inorganic P was presumed to comprise the sum of the labile and stable pools of P, with mass flux to and from the latter described by Fick’s first law. The model was evaluated using results from extractions and P K-edge XANES spectroscopy. Notably, organic P, as revealed by P K-edge XANES, did not substantially contribute to long-term changes in soil P content and was therefore excluded from consideration. In general, the model offered reasonable fits to the extracted P concentrations. However, for the P-depleted treatments, a prerequisite was that the P removal through harvest was lower compared to measurements. Conversely, in three of the soils, the modelled fertilizer inputs needed to be reduced to 70 % to 85 % of the known additions. These discrepancies may be attributed to the involvement of deeper soil horizons, including deep crop uptake and mixing with lower soil layers, although other factors such as lateral dispersion and inaccuracies in estimating applied fertilizers cannot be discounted. These results underscore the necessity of gaining a more comprehensive understanding of how deeper soil horizons influence P mass balances in agricultural soils. In one of the soils, Fjärdingslöv, P K-edge XANES results demonstrated the formation of calcium phosphate over time in the highest fertilization treatment, consistent with the model. Additionally, in two soils, Kungsängen and the P-depleted Vreta Kloster soil, the model predicted a significant contribution from mineral weathering. However, the PBalD8 model also projected higher P leaching rates than those observed, suggesting that the model may not fully capture this P output term.