INTRODUCTION
Nitrogen (N) inputs to agricultural soils contribute to production of the greenhouse gas nitrous oxide (N2O) and animal production accounts for an estimated 1·5 million tonnes N2O-N/year (Oenema et al. Reference Oenema, Wrage, Velthof, van Groenigen, Dolfing and Kuikman2005). In pasture systems, urination by grazing animals causes a mosaic of discrete patches of highly concentrated N loading to soil. Approximately 0·41 of N2O-N emissions from animal production are attributable to urine and dung deposition by grazing animals (Oenema et al. Reference Oenema, Wrage, Velthof, van Groenigen, Dolfing and Kuikman2005). An increasing number of studies have focused on (a) quantifying N2O-N emissions from urine and (b) assessing urine N2O-N emission mitigation strategies in pasture systems. These studies typically use simulated urine patches (Table 1). Natural urine patches are intrinsically heterogeneous in their within-patch N loading and size. Selbie et al. (Reference Selbie, Buckthought and Shepherd2015) summarized the drivers of this variability as urine volume, wind, slope, antecedent soil moisture and soil physical properties. Cattle urine patches were observed to range from 0·16 to 0·49 m2 by Williams & Haynes (Reference Williams and Haynes1994), to have a mean patch area of 0·353 m2 (Saarijarvi & Virkajarvi Reference Saarijarvi and Virkajarvi2009) and to expand naturally over time (Williams & Haynes Reference Williams and Haynes1994). Dairy cow urine patches (4 year mean 0·37 m2) have also been measured using the zone of grass response as a proxy for the urine wetting front (Moir et al. Reference Moir, Cameron, Di and Fertsak2011). Saarijarvi & Virkajarvi (Reference Saarijarvi and Virkajarvi2009) reported that the non-wetted zone of influence extended up to 150 mm from the wetted patch edge. The total area is termed the ‘effective area’ of a urine patch (Selbie et al. Reference Selbie, Buckthought and Shepherd2015). It follows that effective area of the patch would be expected to delineate the zone of increased N2O loss potential associated with a urine deposition.
Table 1. A selection of studies using urine patch simulation for nitrous oxide loss measurement
Values in italics are calculated from information provided in papers.
There is considerable variability in methods used to simulate urine patches for N2O loss estimation (Table 1). The two most common methods are to uniformly apply urine to either (a) a defined area larger than the footprint of the N2O measurement collar and subsequently install the collar or (b) install the collar prior to application to constrain urine (Table 1). These methods, though practical, do not perfectly simulate a naturally occurring urine patch for a number of reasons. Firstly, they create a uniformly wetted area. Secondly, when constrained by a collar, urine infiltration along the horizontal plain in the surface soil, the most active zone of denitrification (Luo et al. Reference Luo, Tillman, White and Ball1998), is restricted. Thirdly, the constraint interferes with the pattern of urine interaction with soil. Fourthly, there are discrepancies between the average footprints of naturally deposited urine patches and the collars used to simulate them (Table 1). In recent work, Rochette et al. (Reference Rochette, Chantigny, Ziadi, Angers, Bélanger, Charbonneau, Pellerin, Liang and Bertrand2014) took an alternative approach by simulating a urine patch with a wetted area, which was 0·33 of the N2O measurement collar area, thus ensuring the zone of influence was accounted for.
The objective of the current work was to summarize patch simulation approaches in the literature and to evaluate the hypothesis that N2O loss induced by a simulated dairy cow urination would be affected by patch simulation and measurement approach. The typical ‘uniform wetted area’ (UWA) method, which artificially limits horizontal movement of urine, is compared with a ‘natural expanding effective area’ (NEEA) method using a collar large enough to allow natural infiltration of urine.
MATERIALS AND METHODS
Site description, experimental design and treatments
Field experiments were conducted under two conditions: (i) ‘high’ N2O loss, which occurred at a moderately drained site in autumn (HL) and (ii) ‘low’ N2O loss, which occurred at a freely draining site in spring (LL). This approach permitted comparison of the methods under contrasting loss conditions and was not designed to explore specific site or seasonal differences, which are heavily influenced by specific soil and environmental factors following treatment application. The HL occurred on a moderately draining Cambisol (58% sand, 30% silt, 12% clay, 73 g organic matter/kg, 32 g total C/kg, 3·0 g total N/kg, pH 5·7 0–10 cm) in autumn 2013 at the Teagasc Johnstown Castle Research Centre, Co. Wexford, Ireland (52°18′N, 6°30′W, 72 m a.s.l). The LL occurred on a free-draining Cambisol (58% sand, 28% silt, 14% clay, 79 g organic matter/kg, 30 g total C/kg, 3·2 g total N/kg, pH 5·8 0–10 cm) in spring 2014 at the Teagasc Moorepark Research Centre, Co. Cork, Ireland (52°09′N, 8°14′W, 35 m a.s.l). Both sites were in long-term grassland dominated by perennial ryegrass (Lolium perenne L.). No organic manures or fertilizers were applied and animals were excluded for a period of at least 6 months in advance of the experiments. Grass was cut to approximately 5 cm before the experiments and allowed to regrow to approximately 8 cm. Stainless steel N2O measurement collars were inserted to 7–10 cm depth at least 4 days prior to treatment application. Soil volumetric moisture (0–10 cm) was measured using a Theta probe soil moisture sensor (Delta-T, Cambridge, UK) in the area surrounding the simulated urine patches. Soil bulk density (0–10 cm) was measured to calculate water-filled pore space (WFPS) following the method of Maljanen et al. (Reference Maljanen, Martikkala, Koponen, Virkajarvi and Martikainen2007). Precipitation, air and soil temperature (0–10 cm) were measured at a nearby (<500 m) meteorological station.
The treatments were: (a) UWA, a patch simulated by uniformly applying 2 litres of urine within 0·156 m2 collars and (b) NEEA, which closely mimicked natural urination by applying 2 litres of urine to a central point within collars of 0·462 m2 and allowing urine to migrate outward as it would naturally. Although the simulated patches originated from the same simulated urination (2 litres) the UWA method resulted in a uniform volume loading of 12·8 litres/m2 and the NEEA a non-uniform urine loading with a mean of 4·33 litres/m2. The urine N loading differed on an area basis but not on a simulated urination basis or on a patch basis. This is an important point because it is the N2O-N emission associated with urination voided by an animal, which represents the unit of interest. The control treatment to measure the soil background N2O emission (control) used a 0·156 m2 collar. Up-scaling N2O emissions from a chamber scale to area scales is a common practice for presenting results, in a similar manner the background emission for a 0·462 m2 area was calculated by up-scaling emissions from 0·156 m2. Treatments were applied on the morning of 14 October 2013 and 8 April 2014 for the HL and LL experiments, respectively. The experimental design was a randomized block design, with three treatments (UWA, NEEA and untreated control) present in each of the five replicate blocks. The experimental unit was the plot, which in all blocks contained one simulated urine patch per urine treatment dedicated to N2O sampling. Additionally, in blocks 1, 3 and 5 each experimental unit contained an additional individual simulated urine patch (HL) or three additional simulated urine patches (LL), which were used solely for soil sampling and mineral N assessment. The dimensions of these experimental units was 4 × 2·5 m. For the experimental units containing one urine patch the plot size was 1·5 × 2·5 m with the treatment located centrally in the plot.
Urine was collected from grazing lactating Holstein Friesian dairy cows less than a week prior to application, homogenized and refrigerated at 4 °C until application. Urine N content was measured using an Aquakem 600 discrete analyser (Thermo Fisher Scientific, Waltham, MA, USA) (Cabrera & Beare Reference Cabrera and Beare1993). Urine N content at application was 8·3 and 5·3 g/l for the HL and LL experiments, respectively.
Nitrous oxide sampling and analysis
Unvented stainless steel covers (10 cm high) were used to form a headspace. Chamber to collar sealing was via a neoprene gasket, compressed by a 6 kg weight. A 10 ml gas sample was taken through a rubber septum after 40 min (Becton Dickinson, Oxford, UK) using a 10 ml polypropylene syringe (BD Plastipak, Becton Dickinson, Oxford, UK) fitted with a hypodermic needle (BD Microlance 3, Becton Dickinson, Oxford, UK) and was injected into pre-evacuated 7 ml screw-cap septum glass vials (Labco, High Wycombe, UK). The N2O sampling procedure of Chadwick et al. (Reference Chadwick, Cardenas, Misselbrook, Smith, Rees, Watson, McGeough, Williams, Cloy, Thorman and Dhanoa2014) was followed. Eight samples of ambient air were collected at each sampling. Their mean N2O concentration was set as a surrogate for N2O concentration at time zero. The assumption of a linear increase in headspace N2O accumulation (Chadwick et al. Reference Chadwick, Cardenas, Misselbrook, Smith, Rees, Watson, McGeough, Williams, Cloy, Thorman and Dhanoa2014) during the 40-min enclosure period was verified on each sampling occasion by collecting five headspace samples per chamber from a random sub-set of urine treated chambers during a 60-min enclosure period. Of the sub-set of chambers, which had a flux, 0·87 were linear according to the criteria of Chadwick et al. (Reference Chadwick, Cardenas, Misselbrook, Smith, Rees, Watson, McGeough, Williams, Cloy, Thorman and Dhanoa2014). At the end of the 60-min enclosure period, the mean N2O concentration inside chambers in the linear group was 3·5 ppm (s.d. 3·96 ppm). For the quadratic group it was 2·62 ppm (s.d. 1·93 ppm). The quadratic group was not dominated by any particular urine treatment. The methodology of Chadwick et al. (Reference Chadwick, Cardenas, Misselbrook, Smith, Rees, Watson, McGeough, Williams, Cloy, Thorman and Dhanoa2014) has been used in the generation of emission factors (e.g. Bell et al. (Reference Bell, Rees, Cloy, Topp, Bagnall and Chadwick2015); Krol et al. (Reference Krol, Carolan, Minet, McGeough, Watson, Forrestal, Lanigan and Richards2016)) and treatment inter-comparison (e.g. Minet et al. (Reference Minet, Jahangir, Krol, Rochford, Fenton, Rooney, Lanigan, Forrestal, Breslin and Richards2016)). Nitrous oxide concentrations were determined using a gas chromatograph (GC) (Varian CP 3800 GC, Varian, USA). Hourly N2O emissions were calculated based on the rate of N2O concentration change during the enclosure period. Flux calculations accounted for air temperature, atmospheric pressure and the ratio of surface area to chamber volume. Sampling took place between 10·00 and 12·00 h and was used to calculate daily emissions (de Klein et al. Reference de Klein, Barton, Sherlock, Li and Littlejohn2003). Cumulative emissions were calculated by integrating the daily fluxes and linear interpolation between measurement points (de Klein & Harvey Reference de Klein and Harvey2012) over 66 and 70 days in HL and LL experiments, respectively. In each experiment, sampling was conducted on 20 occasions with the highest sampling intensity following treatment application.
Soil sampling and analysis
Soil samples (0–10 cm) were collected by sampling at 15 cm intervals across a horizontal cross-section of each patch to obtain a composite sample. In total, there were 12 soil samplings in the HL and 7 in the LL experiment with the highest sampling intensity following treatment application. Samples were fresh-sieved using a 4 mm sieve, and sub-sample gravimetric moisture content and mineral N content was measured. Samples were extracted with 2 m potassium chloride (KCl) and mineral N in the extract was determined using an Aquakem 600 discrete analyser.
Data presentation and statistical analysis
The flux data are presented per simulated urine patch, as previously done by Rochette et al. (Reference Rochette, Chantigny, Ziadi, Angers, Bélanger, Charbonneau, Pellerin, Liang and Bertrand2014). The effect of treatment and time after urine application on the dependent variables of N2O, soil nitrate (NO3-N) and ammonium (NH4-N) were evaluated using the REPEATED statement of the PROC MIXED procedure of SAS 9·3 (© 2002–2010, SAS Institute Inc., Cary, NC, USA). The factors in the model were treatment, time of sampling and block with time of sampling as the repeated factor. The treatment effect on the cumulative mass of N2O-N loss during the measurement period was tested using the PROC GLMMIX procedure of SAS. The analysis included treatment, loss condition, i.e. HL or LL and their interaction as fixed effects and block as a random effect.
RESULTS AND DISCUSSION
Water-filled pore space is an important driver of N2O-N loss (Smith et al. Reference Smith, Thomson, Clayton, McTaggart and Conen1998). Conditions were not favourable for N2O loss under LL due to lower WFPS levels (45–55%). Under LL the urine treatments were not significantly different from the control (Table 2). Consequently, it is not surprising that the patch simulation approach had no effect. In contrast, under HL conditions precipitation occurred almost daily following urine application (Fig. 1(a)) and WFPS exceeded 65% for at least 40 days following urine application. Additionally, soil temperature at patch simulation, a time when N2O-N losses are frequently greatest (Williams et al. Reference Williams, Ineson and Coward1999; Maljanen et al. Reference Maljanen, Martikkala, Koponen, Virkajarvi and Martikainen2007; Krol et al. Reference Krol, Forrestal, Lanigan and Richards2015), was also 3–5 °C higher. Smith et al. (Reference Smith, Thomson, Clayton, McTaggart and Conen1998) reported an exponential increase in N2O production related to temperature. Under these conditions, both urine treatments increased N2O loss significantly compared with the control (P < 0·001). The NEEA, which closely mimics a natural urine deposit, induced a significantly greater loss compared with the UWA method (P < 0·01). The UWA patch had a net relative emission of 64% compared with the NEEA method (Table 2). An important factor explaining the lower loss by the UWA method is thought to be the differential urine-soil interactions between methods. Wachendorf et al. (Reference Wachendorf, Lampe, Taube and Dittert2008) reported that 75% of the urine induced N2O-N loss in their experiment came from native soil N. It is likely that a significant portion of the urine-induced N2O loss under HL in the current work also came from native soil N. A rapid emission peak exceeding 1100 µg N2O-N/patch/h was induced from the NEEA simulated patch on the day of application. The peak in emission occurred at a time when soil total oxidized nitrogen (TON) levels were low (Fig. 2(c)) and was almost three times larger than the initial peak of 398 µg N2O-N/patch/h for the UWA simulated patch (Fig. 3(a)). In the NEEA method, the urine can interact with a greater volume of surface soil as it migrates outwards from the point of application within the collar and tapers off naturally towards the edges. In the current experiments, the NEEA area was approximately three times larger than the UWA. It is suggested that these tapering (Williams & Haynes Reference Williams and Haynes1994) and edge effects could be important because interfaces or edges are often the most active zones of ecosystems. Another factor likely to affect the urine–soil interaction is a degree of transient ponding observed at application in the UWA method. The hydraulic head (Hillel Reference Hillel2004) created by the artificial urine ponding which occurred in the UWA treatment may have promoted deeper infiltration. Deeper infiltration could reduce N2O production because the nitrification rate in the upper soil layer could be at least an order of magnitude higher than in the lower soil layers (Luo et al. Reference Luo, Tillman, White and Ball1998). It is also conceivable that ammonia volatilization loss, an important N loss pathway from urine patches (Fischer et al. Reference Fischer, Burchill, Lanigan, Kaupenjohann, Chambers, Richards and Forrestal2016), could be differentially affected by the patch simulation approach.
Fig. 1. Precipitation, water filled pore space (WFPS), soil and air temperature during the experiment for (a) high loss and (b) low loss conditions.
Fig. 2. Soil NH4-N for (a) high loss and (b) low loss conditions, soil NO2-N and NO3-N (total oxidized N (TON)) for (c) high loss and (d) low loss conditions (0–10 cm) for naturally expanding effective area (NEEA) and uniform wetted area (UWA) methods. Error bars indicate the pooled s.e. of the mean.
Fig. 3. Temporal flux of N2O-N emission for (a) high loss and (b) low loss conditions in response to the uniform wetted area (UWA) and naturally expanding effective area (NEEA) urine patch simulation methods.
Table 2. Effect of the uniform wetted area (UWA) and naturally expanding effective area (NEEA) urine patch simulation methods under high (HL) and low loss (LL) conditions on N2O-N loss
The NEEA method allowed measurement of the naturally occurring patch effective area for the specific soil environmental conditions of the current experiment. The 0·462 m2 collar used in the NEEA method was approximately three times larger than the 0·156 m2 collar used in the UWA method. It was larger than the mean wetted area of 0·353 m2 reported for a 2·37 kg urination by Saarijarvi & Virkajarvi (Reference Saarijarvi and Virkajarvi2009) and the mean zone of grass response of 0·37 m2 reported by Moir et al. (Reference Moir, Cameron, Di and Fertsak2011). It was also larger than any of the collars used in the previous work, listed in Table 1. Anger et al. (Reference Anger, Hoffmann and Kühbauch2003) accounted for the patch zone of influence to a degree by simulating a 0·2 m2 patch in a 0·24 m2 N2O measurement collar (Table 1) and Rochette et al. (Reference Rochette, Chantigny, Ziadi, Angers, Bélanger, Charbonneau, Pellerin, Liang and Bertrand2014) specifically designed their experiment to account for it by simulating 0·1 m2 patches in 0·303 m2 N2O measurement collars.
The method which most closely mimics natural conditions is expected to deliver the most credible quantitative estimate of loss. In the case of these experiments, the NEEA mimicked natural conditions much more closely than the UWA method. Although higher loss was recorded for the NEEA method under HL, this may not always be the outcome, for instance no effect was observed under LL. Under different conditions a concentrated zone of N loading as a result of the UWA method could contribute to an elevated NO3 −-N pool, which persists for longer, favouring denitrification further from the time of urine application. Some evidence of such an effect is present in the LL data. A significant treatment × day of measurement interaction was detected (P < 0·001) with two secondary peaks in emission on days 20 and 30 measured for the UWA method but not for the NEEA method (Fig. 3(b)). The direction of difference in N2O loss between methods cannot be extrapolated from the present study to the diverse soil and environmental conditions in which researchers make N2O loss estimates for urine patches. The present study simply highlights a need for greater attention to the method of urine patch simulation. The nature of urine patches raises practical questions of how to best simulate patches for N2O emission measurements. It is suggested that a representative patch can be achieved, at least in part by the following:
-
(a) Use of a defined urine volume and N content similar to that of the animal of interest, e.g. close to 2·1 litres for dairy cattle (Selbie et al. Reference Selbie, Buckthought and Shepherd2015).
-
(b) Allow natural infiltration of the chosen defined volume of urine for the soil of choice to permit tapering toward the edges as observed in natural patches by Williams & Haynes (Reference Williams and Haynes1994).
-
(c) Measure from the zone of influence (Saarijarvi & Virkajarvi Reference Saarijarvi and Virkajarvi2009) in addition to the wetted area, i.e. the patch effective area (Selbie et al. Reference Selbie, Buckthought and Shepherd2015) or the NEEA.
The authors thank the laboratory and field staff at Teagasc Johnstown Castle and Moorepark. This research was financially supported under the National Development Plan, through the Research Stimulus Fund, administered by the Department of Agriculture, Food and the Marine (Grant numbers RSF10-RD/SC/716 and 11S138).
