Ability of a dynamic model of phosphorus to account for observed patterns of response in a series of phosphate fertilizer trials on pasture

Abstract

A national series of rates × forms of phosphate fertilizer trials has provided a large amount of data on patterns of response of pasture production and soil P tests to forms, rates and frequencies of P fertilizer application over time. A simple mechanistic model of P in pastoral systems has been developed in an attempt to account for the observed patterns. The model considers active P in the soil to be in one of two compartments: undissolved fertilizer P (PF) and a pool of labile, plant-available P (PA). P dissolving from PF enters PA, and P is lost from PA by immobilisation in the soil and by net removal above ground through non-recycled uptake by plants. Both these transfers are taken to follow first order kinetics, with rate constants designated K1 and K2 respectively. Pasture dry matter (DM) production in any year is related to the mid-year value of PA by the diminishing-returns Mitscherlich equation, with nil pasture yield being associated with nil PA. Data from different sites with the same design were combined to give generalised response patterns with which to assess the model. Only the control treatment and treatments receiving triple superphosphate (TSP) and Sechura phosphate rock (SPR) were considered. With a value for K1 corresponding to almost immediate dissolution, the model accounted well for the Olsen soil P test patterns from TSP when K2 corresponded to an annual loss of 16.3% of P in PA and a factor of 10 was used to convert Olsen P to kg P/ha in PA. Pasture DM yield response curves to rates of TSP in individual years were well accounted for when the mid-year value of PA required for 90% relative yield was taken as 130 kg P/ha. The model also accounted well for the difference between response patterns to annual and triennial TSP applications for both soil P tests and DM production. The declines in soil P tests and DM production which occurred in nil P controls were less than model predictions: this could be due to unfertilised pasture making more use of P from greater soil depth or less available forms. Olsen P tests with the higher rates of SPR were always lower than model predictions using a value for K1 derived from chemical measurement of SPR dissolution rates in the trials. However, DM yields from SPR were generally close to model predictions. These divergent results support independent data suggesting that available soil P from reactive phosphate rocks (RPRs) is underestimated by the Olsen P test. The relative success of this simple compartmental model in accounting for generalised response patterns in such a large body of data suggests that it could be a useful starting point for a dynamic P fertilizer recommendation model.