During cultivation, the internal phosphorus cycle of Mollisols (Chernozems) of the Canadian Prairies is perturbed by crop sequences including wheat phases, tillage practices, and regular applications of fertilizers. To monitor these changes, a proximate sequential phosphorus (P) fractionation procedure was developed by Hedley et al. (1982) to extract inorganic and organic P fractions as very labile (resin-P), labile (NaHCO3-P), slowly available (NaOH-P), and very slowly available (HCl-P) pools. Models used so far to monitor P pools do not address the interactive behaviour of P fractions constrained to a closed compositional space. Compositional data analysis using isometric log ratio (ilr) coordinates is appropriate for modelling the interactive P pools using sequential binary partitions of P pools. Our objective was to model changes of P pools in Mollisols in response to management and time using ilr coordinates. We used a dataset with treatments and another where a Mollisol was analyzed at time zero and 4, 65, and 90 yr after sod breakup. Seven P fractions were assigned to P reactivity groups to compute six ilr coordinates. The ilr2 contrasting inorganic (geochemical) and organic (biological) P pools and ilr4 between the most readily available and less P bioavailable pools were the most sensitive to crop sequence and fertilization. Using composition at time zero as reference, the Aitchison distance reached a plateau after the 4th year in the Bm horizon compared to continuous change in the Ah horizon. Time changed the P balance of cultivated Mollisols primarily in the inorganic vs. organic P pools. The risks of yield loss and environmental damage can be minimized using soil tests that quantify the rapidly bioavailable inorganic P pools and crop management strategies that promote biological P pools.