Modeling the risk of phosphorus runoff following single and split phosphorus fertilizer applications in two contrasting catchments

The potential loss of P in runoff is a function of the combined effects of fertilizer–soil interactions and climatic characteristics. In this study, we applied a Bayesian approach to experimental data to model the annualized long-term risk of P runoff following single and split P fertilizer applications using two example catchments with contrasting rainfall/runoff patterns. Split P fertilizer strategies are commonly used in intensive pasture production in Australia and our results showed that three applications of 13.3 kg P ha^-1 resulted in a greater risk of P runoff compared with a single application of 40 kg P ha^-1 when long-term surface runoff data were incorporated into a Bayesian P risk model. Splitting P fertilizer applications increased the likelihood of a coincidence of fertilizer application and runoff occurring. We found that the overall risk of P runoff is also increased in catchments where the rainfall/runoff pattern is less predictable, compared with catchments where rainfall/runoff is winter dominant. The findings of our study also question the effectiveness of current recommendations to avoid applying fertilizer if runoff is likely to occur in the next few days, as we found that total P concentrations at the half-life were still very high (18.2 and 8.2 mg P L^-1) following single and split P treatments, respectively. Data from the current study also highlight that omitting P fertilizer on soils that already have adequate soil test P concentrations is an effective method of reducing P loss in surface runoff. If P fertilizer must be applied, we recommend less frequent applications and only during periods of the year when the risk of surface P runoff is low.