Simplifying pastoral systems modelling - Accounting for the effect of urine deposition on N leaching

Abstract

Models are increasingly being used to assess the performance of farming systems. Computer modelling is a cost-effective way to study a wide range of factors and the only way to consider future scenarios. In pastoral systems, it is important to account for urine patches deposited by grazing animals as they are a major source of nitrogen (N) leaching. Explicitly accounting for urine depositions is very challenging because the introduced spatial and temporal variability may require thousands of simulations depending on the system modelled. In this case the computing time becomes a significant limitation so there is a need for methods to reduce computing time of pastoral systems simulations. Here we present an approach that can be used in the APSIM modelling framework to account for the effects of urine patches in a simplified way. Using a meta-model coded as an APSIM module, a fraction of the urine N deposited is removed from the simulation as direct urinary N leaching (Figure 1), the rest is retained in the simulation and is returned to the paddock evenly distributed as mineral N. This approach allows the user to model the whole paddock with relatively simple simulations, enabling to study a wide range of scenarios faster while still appropriately considering the impacts of urine depositions. The proposed meta-model relates the fraction of N leached (fLEACH) from urine patches to environmental and management factors. It was constructed using empirical functions derived from a comprehensive N leaching dataset, which was produced using simulations that explicitly described the urine patches. We present in this paper the results using data from the Canterbury region of New Zealand. Stepwise regression and principal components analysis were used to identify the most important factors related to fLEACH and then empirical functions were fitted to the dataset using regression analysis. This was performed in three steps to account for the major static factors (soil plant available water, urine N load, annual rainfall and day of urine deposition) and one step accounting for dynamic factors (average temperature on the week prior to urine depositions and soil moisture at the time of deposition). Most of the variation was accounted for by the static factors, and the improvement by using dynamic factors was barely significant. This means that the meta-model lacks sensitivity to year-to-year variations, therefore it is better suited for long-term simulations. However, overall agreement was good (R 2=0.73), with little overall bias. The approach is thus promising, capturing the effects from the largest drivers of N leaching while considerably reducing the simulation time.