Modelling changes in soil phosphorus when phosphorus fertiliser Is reduced or ceases

In order for land managers and policy makers to manage excessive soil phosphorus (P) concentrations and reduce the risk of this particular source of P from impacting water bodies, models of soil P decline under various scenarios are needed. We modelled the decrease in calcium chloride-extractable P (CaCl2-P), and sodium bicarbonate-extractable P (Olsen-P and Colwell-P) using data from six Australian grazed pasture soils with contrasting P sorption properties, over a period of 4.5 years. Each soil had four initial soil P concentrations (Pinit), each of which received four on-going rates of P fertiliser (Pfert). The model predicts the final P concentration (Pfinal) by taking into account the P concentration previously measured (CaCl2-P, Olsen-P or Colwell-P), Pfert applied since measurement, and time since previous measurement: Final P concentration = (previously measured P concentration + ep x P fertiliser applied) exp (-dp x years since previous P concentration measurement). Where ep is the increase in soil P for each unit of applied P and dp is the decay constant representing how quickly the soil P decreased. The greatest decreases in proportion to Pinit occurred for CaCl2-P, followed by Olsen-P, and then Colwell-P. The model tended to fit the dataset well for Olsen-P and Colwell-P, with mean overestimation (modelled Pfinal concentration greater than actual Pfinal) of the Pfinal concentrations of 6.1 (32%) and 4.3 mg/kg (10%), respectively. Although there was less CaCl2-P data, the model successfully described it, with a mean overestimation of Pfinal CaCl2 of 3.1 mg/kg (26%). The overestimation of Pfinal CaCl2 was possibly due to the high CaCl2-P concentrations of the low P buffering index soils. The model predicted an average of 32 years (ranging from 26 to 49 years) for Olsen-P concentrations of between 55 and 96 mg/kg to decrease to an agronomic optimum of 17 mg/kg. Agronomic optimum was not a reliable indicator of environmental risk as some soils did not exceed the CaCl2-P environmental threshold until Olsen-P concentrations were twice the agronomic optimum, whereas low P sorbing soils tended to exceed the threshold before reaching agronomic optimum. Further work with more soils is required to examine the influence of soil properties – such as P sorption – on decreases in soil P