Predictions of nitrogen leaching from a well-drained soil under dryland and irrigated dairy farming using APSIM and OVERSEER

Abstract

As urine patches from grazing livestock are the single largest source of leached N, biophysical models that incorporate the effect of N leaching from urine patches are required to assess the impact of changes in farm management on N leaching. The objective of this study was to compare predictions of N leaching from a dairy farm as made by two contrasting tools; APSIM (Agricultural Production Systems Simulator) and the OVERSEER Nutrient Budget Model. Typically, these models are used for different purposes, require different input data, and produce outputs that are not readily comparable. APSIM is a process-based model that works on a fine scale and daily time-step whereas OVERSEER calculates N leaching on a monthly time-step and reports annual averages of N leaching for relatively large areas. As OVERSEER has been calibrated for New Zealand's farming systems and uses inputs that are readily accessible by farmers, it is the favoured tool for assessing compliance of dairy farms to proposed new regulations. However, APSIM is increasingly being used to analyse existing and proposed farming scenarios in greater detail. The dairy system modelled in this study is located on a well-drained Manawatu silt loam soil at Massey No. 1 Dairy Farm, Palmerston North, New Zealand. Dryland and irrigated scenarios were defined to cover a range of possible management options. Results from simulations over 25 years using APSIM were analysed to obtain long-term estimates of N leaching to compare with OVERSEER predictions. Although the model inputs were set such that both models were under similar weather and management conditions, there were differences between the models regarding the effect of irrigation on N leaching (Table 1). These differences were attributed to a) differences in how the two models describe irrigation and its effect on drainage, b) uncertainties in the calculations of urinary N load, and c) differences in simulations of other components of the N cycle. Our modelling exercise revealed those N transformation processes that behave similarly from those that differ between the two models, highlighting potential knowledge gaps. The exercise also highlighted the difficulties of comparing models, and great care should be taken when comparing outputs from different sources.