Modelling nitrogen leaching from overlapping urine patches

Abstract

Urine depositions have been shown to be the main source of N leaching from grazing systems and thus it is important to consider them in simulation models. The inclusion of urine patches considerably increases the complexity of the model and this can be further aggravated if the overlaps of urine patches are also considered. Overlapping urine patches are potentially important sources of N loss because the N load in these areas can be very high. In this work, we investigate a methodology to simplify the process of accounting for overlapping urine patches. We tested a two-stage approach, where, on one hand, the urine of two consecutive depositions could be aggregated and deposited at the time of the second deposition. This was called the Delayed Representation (DR) and would be useful when the delay (T_d) between overlaps is short. On the other hand, if T_d is sufficiently large, the depositions would become functionally independent and the urine patches could be considered separately. We called this the Independent Representation (IR). We tested this methodology by comparing simulations where the overlapping urine patches were considered explicitly and using the DR or IR in several combination of climates, soils and management options, chosen to span the likely range in New Zealand, using the Agricultural Production Systems Simulator (APSIM) model.

The results from the simulations indicated that when T_d < 20 days, the DR introduced only a low, acceptable, error in simulated N leaching. When T_d > 180 days, the IR was found to be acceptable in all the climates, soils and management options simulated. This left a generic window, 20 < T_d < 180 days, where explicit simulations would be required. In some conditions, that window was considerably shorter. In a dry climate and shallow soil, the window was found to be 75 < T_d < 180 days, while for the simulations with a wetter climate and deep soils, that window was about 70 < T_d < 90 days, except for depositions in the middle of winter. These limits should apply only to long-term simulations where the user is interested in the average behaviour of the system, as considerable year-to-year variability was observed. We suggest that these windows can be used to guide the development of simulation models that include representations of urine patches under many conditions but where the soil and/or climate conditions vary markedly from those used here the analysis should be repeated.

Introduction

Urine deposited by grazing ruminants is a major contributor to the high heterogeneity in soil nitrogen conditions found in pastoral systems (Haynes and Williams, 1993; Bogaert et al., 2000; McGechan and Topp, 2004; Hutchings et al., 2007). The areas affected by urine depositions receive much larger quantities of nitrogen (N) than the remaining area, typically from 500 to 1000 kg N/ha for dairy cows (Haynes and Williams, 1993). Such quantities are far in excess of the pasture's ability to take N up before leaching occurs and thus urine patches are the major sources for N loss in grazing systems (Ball and Ryden, 1984; Di and Cameron, 2002; van Groenigen et al., 2005). The N load and the timing of deposition have been shown to be important factors defining the fate of N in urine patches (Cuttle and Bourne, 1993; Shepherd et al., 2011; Snow et al., 2011).

The heterogeneity arising from the deposition of urine patches complicates the measurement of N loss from a grazed paddock (Lilburne et al., 2012). The high spatial variability makes it difficult and costly to quantify N loss from pastoral systems and means that estimates of N loss from grazed paddocks are often associated with large uncertainties. These large uncertainties are a cause for controversy when defining environmental policies and evaluating the efficacy of mitigation actions. Simulation modelling can be an effective adjunct to experimental methods for monitoring N cycling and estimating N losses. Computer models of various levels of complexity have increasingly been used to estimate N losses worldwide. Most of these models, however, do not account for field heterogeneity. This is mainly because the complexity of the modelling setup required to simulate heterogeneity presents a technical challenge in terms of computing resources (Addiscott, 1995; Hutchings et al., 2007; Wang, 2008; Romera et al., 2012). Furthermore, the importance of heterogeneity and methodologies for the use of information with uncertainty for decision making has not been well established (Beven, 2002; Lowell, 2007).

It has been shown that for estimating N leaching from systems with grazing animals it is necessary to account explicitly for the effects of urine patches (Ryden et al., 1984; Haynes and Williams, 1993; Snow et al., 2009). However, accounting for the heterogeneity created by urine depositions is not trivial. Describing such a system within models places a high demand on computing resources and has been attempted only with considerable simplifications (McGechan and Topp, 2004; Hutchings et al., 2007; Snow et al., 2009), Apart from the variability created by single urine depositions, urine overlaps can further alter the load of N deposited onto the soil and so are potentially important for defining the amount of N leached (Pleasants et al., 2007). Overlapping urine depositions on the same grazing day are likely to be minor because these overlaps typically represent a very small fraction of the grazing area (Shorten and Pleasants, 2007). However, a urine deposition affects the soil and the plants for some length of time, and leaching from a second, overlapping, urine patch deposited some weeks or months after the first deposition will be affected by the first deposition. These temporal overlaps are likely to be more significant than those occurring within the same grazing day because the probability of the overlap and therefore the area affected greatly increases. For example, Pleasants et al. (2007) using a statistical approach estimated that, at a typical 24-h grazing density of 100 cows per ha, less than 1% of the area grazed would be affected by overlapping urine patches. Considering that there would typically be about 14 grazings per year per paddock, and that in winter the animals could be mobbed with stocking densities of up to 1000 cows per ha, the proportion of a paddock that will experience delayed overlaps rises considerably. Over one year, urine depositions affect about a quarter of the area for a typical dairy farm (Pleasants et al., 2007; Moir et al., 2011) and overlaps can represent 20% of this, according to the statistical approach of Pleasants et al. (2007). Accounting for all temporal overlaps would result in a rapid increase in modelling complexity and thus simplifications are needed. While these issues are of most interest for intensively grazed systems where the animals graze the pastures year round, the principles are applicable to grazing situations in general.

The purpose of this study was to investigate options for simplifying the representation of overlapping urine patches in the simulation modelling of grazed systems. While the work here is presented in the context of New Zealand soils and climates, the concepts are generally applicable to any grazed system. We used outputs from the simulation model Agricultural Production Systems Simulator (APSIM) to identify the nature of the interactions between successive urine depositions and how this interaction is affected by the interval between depositions. We hypothesise that when the delay (T_d) between overlaps is short, the first deposition can be delayed and aggregated with the second deposition without significant error; we call this the Delayed Representation (DR). At higher values of T_d, urine patches become functionally independent because the N from the first urine patch is depleted before the second is deposited and in this case, the urine patches can be considered separately with an Independent Representation (IR). We test these simplifications by comparing simulations employing DR and IR against simulations where the overlapping urine patches are considered explicitly (Explicit Representation, ER) in the same simulation. We analyse whether it is possible to define general rules for simplifying the description of urine patch overlaps using DR and IR.