Printer Friendly

Field-scale verification of nitrous oxide emission reduction with DCD in dairy-grazed pasture using measurements and modelling.


Nitrous oxide ([N.sub.2]O) from agricultural soils is a major source of greenhouse gas emissions in New Zealand, accounting for 15.2% of total greenhouse gas emissions on a C[O.sub.2]a-equivalent basis (MfE 2010). Nitrous oxide is a by-product of the microbial breakdown of N compounds applied to soil (typically as animal excreta or N fertiliser). In New Zealand, [N.sub.2]O emissions from agricultural soils rose from 30.3 to 35.8 Gg/year between 1990 and 2008 due to increased fertiliser and excretal N inputs (MfE 2010). Nitrification inhibitors such as dicyandiamide (DCD) inhibit the nitrification process where soil microbes convert ammonium (N[H.sup.4+]) to nitrate (N[O.sup.-3]) ions. Inhibiting the nitrification process can potentially reduce the [N.sub.2]O emitted both by the nitrification process itself, and from subsequent denitrification of N[O.sup.3](Abbasi and Adams 2000; Cookson and Cornforth 2002). A review of lysimeter and field studies using DCD in New Zealand reported an average reduction of 67 [+ or -] 6% in [N.sub.2]O emissions from animal urine (Kelliher et al. 2007). The effects of using DCD have been incorporated into the national inventory assuming a reduction of 67% in direct [N.sub.2]O emissions from animal excreta when DCD is applied (MfE 2010). In many of the previous studies, DCD has been applied directly to the urine patch. However, farmers apply DCD to grazed pastures shortly before or after grazing rather than applying it specifically to the urine patches. In this study, we attempted to measure the effectiveness of DCD in reducing [N.sub.2]O emissions under grazing conditions using both gas chambers and micrometeorology, and modelled the results using the processbased NZ-DNDC model. The micrometeorological results will be presented separately (M. Harvey, A. McMillan, J. Laubach, S. Saggar, D. Giltrap, J. Singh, R. Martin, T. Bromley, M. Evans, unpubl, data). The current paper reports on the gas chamber and modelling approaches.

The process-based model DNDC (DeNitrification-DeComposition) (Li et al. 1992) simulates the soil physical, chemical, and biological processes that produce greenhouse gas emissions. NZ-DNDC is the New Zealand specific model that has been adapted for use in New Zealand grazed pasture conditions and tested on dairy and sheep-grazed pastures (Saggar et al. 2004, 2007). Giltrap et al. (2010) modelled the effect of DCD on N20 emissions from a urine patch, assuming that the nitrification inhibitor caused a constant percentage reduction in the nitrification rate. Reasonable agreement between measured and modelled results was found when 70% reduction in nitrification rate with DCD was used.

Methodology Study site

The study site was Massey University Dairy Farm 4 in Palmerston North, New Zealand. The soil was a poorly drained Tokomaru silt loam (Argillic-fragic Perch-gley pallic, Hewitt 1998). Two circular plots of 40 m diameter (0.126 ha) were each grazed by 20 cows for 5 h on 11 June 2009. The following day, DCD was applied in 800 L of water to one of the plots at 10 kg/ha. [N.sub.2]O emissions were measured periodically for 20 days following the grazing event, using 20 soil chambers in each plot. The period of 20 days was chosen to align with a concurrent micrometeorology study (described in M. Harvey, A. McMillan, J. Laubach, S. Saggar, D. Giltrap, J. Singh, R. Martin, T. Bromley, M. Evans, unpubl, data). Chambers were placed in a regular circular pattern in either plot, which constitutes random placement with respect to the distribution of cattle excreta. Soil and environmental variables were also monitored. Daily temperature, rainfall, and solar radiation data were downloaded from a nearby climate station.

Chamber measurements

Fluxes of [N.sub.2]O were measured using 20 gas flux chambers (diameter 250mm, 300mm high) in each plot. On each sampling day, the chamber was closed with a lid for 1 h, and the air above the soil was sampled through a three-way tap on the chamber lid using a 60-mL syringe at 0, 30, and 60 min. Gas samples were then analysed using a Shimadzu GC-2010 gas chromatograph and flux rates calculated based on a linear regression of the three sampled concentrations using the method described in Saggar et al. (2010).

NZ-DNDC model

The NZ-DNDC model was used to simulate [N.sub.2]O emissions from both treatments (with and without DCD). Correct initialisation of the parameter values is important for accurate simulation. The NZ-DNDC model has previously been used to model [N.sub.2]O emissions from this soil type (Saggar et al. 2004, 2007), and these previous simulations provided several of the parameters. Other parameters were set so to ensure that the measured soil moisture content and [N.sub.2]O emissions from the grazing-only treatment were well simulated. No fitting was done to the grazing+DCD treatment. Thc model was run using the parameter values listed in Table 1.

The NZ-DNDC model has an optional water retention layer that can be used to model soil texture heterogeneity caused by compaction. The soil hydraulic conductivity drops beneath the water retention layer, allowing water accumulation above the water retention layer following rainfall events. The soil waterfilled pore-space (WFPS) at field capacity and depth of water retention layer were set to minimise the root mean square error (RMSE) of the prediction for soil WFPS from 0 to 0.10m (Fig. la). In the absence of measurements, the initial soil N[H.sub.4.sup.+] and N[O.sub.3.sup.-] concentrations were selected to minimise the RMSE in the [N.sub.2]O emissions for the urine-only treatment (Fig. lb). The same soil conditions were then used for the urine+DCD treatment. A previous field study on the same soil had found that DCD reduced the nitrification rate by 60-80% (Giltrap et al. 2010), so the effect of DCD was simulated by reducing the nitrification rate by 70%. Note that as [N.sub.2]O is produced by both nitrification and denitrification processes, the reduction in the [N.sub.2]O emissions will not be the same as the reduction in the nitrification rate. It was assumed that the inhibitor effectiveness remained constant over the 20-day measurement period.

The measured and modelled results were compared using the model mean error (ME) and RMSE. The ME is the average difference between the predicted and observed values. The model RMSE is defined as:

where [P.sub.i] is the predicted value, [O.sub.i] is the observed value, and n is the number of observations. Note that both the ME and RMSE are measures of the model deviation from observation on a point-by-point basis. It is possible that the model could have high errors for predicting the emissions on a given day but still produce accurate emissions estimates over a longer time period.


Sensitivity analyses

The [N.sub.2]O emissions predicted by the NZ-DNDC model are sensitive to the choice of initial soil parameter values. For the sensitivity analysis, we examined the effects of uncertainty in soil N[H.sub.4.sup.+], N[O.sub.3.sup.-] , initial soil WFPS, soil organic carbon (SOC), bulk density, clay content, pH, and WFPS at field capacity on the simulated [N.sub.2]O emissions. The effect on [N.sub.2]O emissions of varying each parameter was examined individually, then high and low scenarios were generated using the extreme values of all eight parameters. Table 2 shows the parameter ranges considered. These ranges represent the potential spatial variability of the soil properties within the plot.

The tested parameters are not independent of each other (e.g. clay content will affect WFPS at field capacity, bulk density will depend upon SOC and clay content). In addition to the direct effect that bulk density has in NZ-DNDC, it is also used to convert the measured soil moisture to WFPS. As the WFPS at field capacity and water retention layer depth were adjusted to provide a good model simulation of WFPS, the uncertainty in the bulk density would also affect these parameters. However, these secondary effects have not been quantified.

Sensitivities of individual parameters are quoted as (% change in [N.sub.2]O emission)/(% change in parameter).


Chamber measurements

Box-plots are presented for the chamber emissions from the grazing-only plots (Fig. 2a) and grazing + DCD plots (Fig. 2b). The measured emissions were positively skewed, as was expected given the highly patchy nature of urine deposition. In particular, for the grazing-only treatment, the highest emissions come from a single chamber and are likely to be the result of a urine patch.


To calculate total emissions over the 20-day period, the fluxes from each chamber were integrated using linear interpolation between measurement days, and then the arithmetic mean and standard error were calculated. Taking the mean and standard error of all 20 chambers, for the grazing-only treatment the total [N.sub.2]O emission was 220 [+ or -] 90g [N.sub.2]O-N/ha, and for the DCD treatment it was 110 [+ or -] 20 g [N.sub.2]O-N/ha. From these values, the calculated reduction due to DCD application was 50 [+ or -] 40%.

Modelled emissions

The [N.sub.2]O emissions are shown for the grazing-only treatment (which was used to establish some initial parameter values) (Fig. 3a) and the grazing+DCD treatment (Fig. 3b). Table 3 compares the measured and modelled [N.sub.2]O emissions for the two treatments. The modelled [N.sub.2]O emissions for the grazing-only treatment were within the uncertainty range of the measured emissions. The RMSE was larger than the ME, indicating that the differences between measured and modelled predictions can be quite high on a daily basis, but that under-predictions and over-predictions tend to cancel out. The ME was positive, so overall the model may slightly overestimate [N.sub.2]O emissions. However, it should be noted that the initial soil N[O.sub.3.sup.-] and N[H.sub.4.sup.+] were selected to optimise the model agreement with the measured [N.sub.2]O emissions for this treatment.

For the grazing + DCD treatment, the model slightly underpredicted [N.sub.2]O emissions. The most likely reason is that the assumed 70% reduction in nitrification was slightly too high.

The measured and modelled reduction in [N.sub.2]O emissions using DCD agreed within the large uncertainty limits.

Sensitivity analysis

Figure 4a-h shows the range of modelled [N.sub.2]O emissions as initial soil N[O.sub.3.sup.-] , N[H.sub.4.sup.+], WFPS, SOC, bulk density, clay content, pH, and WFPS at field capacity are individually varied within the ranges listed in Table 2, and Fig. 4i presents the effect of varying all eight parameters simultaneously. The range in the modelled [N.sub.2]O emissions for the combined effect of variability in the input parameters was noticeably larger than the range produced by varying any one parameter.

Table 4 shows the sensitivity of the total modelled [N.sub.2]O emissions to each of the input parameters. The SOC, pH, and bulk density had sensitivities >l, meaning that uncertainty in these parameters produced an even larger relative uncertainty in the modelled [N.sub.2]O emissions. In contrast, soil N[O.sub.3.sup.-] and N[H.sub.4.sup.+] and initial WFPS parameters had sensitivities <1, so the modelled [N.sub.2]O emissions were relatively insensitive to uncertainties in these parameters. The WFPS at field capacity and clay content both had large negative sensitivities, indicating that an increase in these parameters results in a decrease in [N.sub.2]O emissions. In this case, increasing clay content or WFPS at field capacity reduced the amount of N[O.sub.3.sup.-] lost via leaching and therefore increased the amount of soil N available for nitrification and denitrification. However, both of these parameters affected [N.sub.2]O emissions primarily by modifying soil water flow, and our initial model parameterisation no longer produced a good fit with measured soil moisture when these parameters were changed.



The combined effect of the uncertainty in all eight parameters produced an uncertainty range from -87% to +150% in the total [N.sub.2]O emissions modelled. While this is quite high, it is of a similar magnitude to the variability found in the field. This suggests that high variability in measured [N.sub.2]O emissions could be the result of spatial variability in the soil properties.


In a grazed pasture, [N.sub.2]O emissions tend to be highly variable and positively skewed due to the uneven deposition of animal excreta creating emission 'hot spots' around urine patches. As many statistical tests require normally distributed data, a common technique is log-transformation of results (with the mean of the log-transformed data being equivalent to the geometric mean). However, in this study our aim was to estimate the total [N.sub.2]O emissions from the field. In a grazed pasture, high-emitting hot spots are responsible for a large proportion of the total [N.sub.2]O emissions. Measures such as the geometric mean and median, which lessen the influence of extreme values, will tend to underestimate the total field emissions. Therefore, we have used untransformed data to calculate the arithmetic mean emission and standard error. Accordingly, it was not possible to apply many common statistical probabilities, which assume a normal distribution, to our data (e.g. t-statistics to calculate 95% confidence intervals).

The accuracy of the field-scale emission rate estimated from chamber measurements depends on the distribution of urine patches among the chambers being representative of the distribution of urine patches across the field. We are unable to ascertain whether this was the case for our experiment. In the grazing-only trial, there was one very highly emitting chamber, which was most likely due to a urine patch. However, there was no correspondingly high emitting chamber in the grazing + DCD treatment. This could either be because the DCD reduced the emissions from urine patches, or because the random sample did not include any urine patches. As we were unable to distinguish between these explanations, the possibility exists that the differences observed may be an artethct of sampling error.

However, the fact that the level of reduction measured agreed with that measured in a urine patch trial (50 [+ or -] 20%, Giltrap et al. 2010) and the NZ-DNDC simulation is encouraging.

Previous studies have attempted to determine empirically the number of chambers required for adequate sampling of a grazed pasture. The results from two intensive 6-week measurements using 20 small chambers (diameter 250 mm, 300 mm high, used in this study) and 6 large chambers (1 m by 0.5 m, 300 mm high) reported by Saggar et al. (2008) showed no significant differences in gaseous emissions between the two types of chambers, but the spatial variability was higher from small chambers than large chambers. Subsequently, Saggar et al. (2010) performed another grazing experiment using 40 chambers in a 0.48-ha plot grazed by 205 cows/ha for a total of 12 h; based on the comparison of average emissions from the even- and odd-numbered chambers, they found that 20 chambers appeared to produce a representative estimate of emissions from dairy grazed pasture.

The methodology for using chambers to estimate [N.sub.2]O emissions at field scale is continuously undergoing refinement. A simple test for sampling errors in future experiments would be to conduct simultaneous emissions measurements from chambers on known urine patches for both treatments. If the [N.sub.2]O emissions from the urine patches were substantially higher than any individual chamber in the corresponding plot, this could indicate that the sample did not include any urine patches. Another possible approach for future trials is a targeted placement, with a defined number of chambers representing urine patches and another group of chambers representing urine-free areas, combined with an attempt to estimate the area fraction in the plot covered by urine. Such an approach allows weighted upscaling of the chamber fluxes to a paddock flux estimate.

The mean daily [N.sub.2]O emission rates obtained here from chambers are small: 11 [+ or -] 4.5 g [N.sub.2]O-N/ for grazing-only and 5.5 [+ or -] 1 g [N.sub.2]O-N/ for grazing+ DCD. Simultaneously, M. Harvey et al. (M. Harvey, A. McMillan, J. Laubach, S. Saggar, D. Giltrap, J. Singh, R. Martin, T. Bromley, M. Evans, unpubl. data) attempted to measure these emission rates with a micrometeorological flux-gradient technique. They obtained a mean daily emission of 13 [+ or -] 30g [N.sub.2]O-N/ for grazingonly and 12 [+ or -] 27g [N.sub.2]O-N/ for grazing+DCD. The magnitude of the error shows that, in the present experiment, the flux-gradient method was unable to resolve a difference between the two plots; but that does not mean no difference existed.

This study looked only at the effects of DCD on [N.sub.2]O emissions from a single grazing event, with the DCD applied the following day. However, DCD has a limited lifetime in soil, as it is subject to microbial decay. The half-life of DCD in soil is temperature-dependent with decreasing lifetimes at higher temperatures (Kelliher et al. 2008). Therefore, the benefit of a single DCD application on emissions from subsequent grazing events would be reduced.

The NZ-DNDC is sensitive to variability in the soil input parameters, and the uncertainty in input parameters leads to large errors in the modelled [N.sub.2]O emissions. This uncertainty reflects an underlying spatial variability in the soil properties. Reducing the uncertainty in the model estimates may require characterising the soil properties in terms of a probability distribution function rather than using average values. A better knowledge of the underlying distribution of urine patches would also be useful both for modelling and for experimental design.

Conclusion Application of DCD to a grazing pasture appeared to reduce the N[O.sub.2] emissions from a single grazing event by 50 [+ or -] 40% over a 20-day period. However, this result should be treated with caution, because the possibility of sampling error due to the chamber distribution cannot be excluded.

The effectiveness of DCD in reducing N[O.sub.2] emissions over the longer term is a matter of ongoing research. NZ-DNDC simulated a reduction of a similar magnitude (60%) in N[O.sub.2] emissions due to DCD. This level of reduction is consistent with that found in experiments with individual urine patches.

The NZ-DNDC model was sensitive to uncertainty in the input parameters. The combined effect of typical uncertainty in soil N[O.sub.3.sup.-], N[H.sub.4.sup.+], SOC, bulk density, clay, pH, and WFPS at field capacity was an overall uncertainty of -87% to +150% in the cumulative N[O.sub.2] emissions.


This research was funded by the Foundation for Research, Science and Technology. Thanks to Massey University for access to the farm and management of cattle. Thanks also to Peter Berben, John Dando, and Steve McGill at Landcare Research, and Tony Bromley, Ross Martin, and Matthew Evans at NIWA for technical support. </DO>/10.1071/SR11090</DO>

Received 28 April 2011, accepted 24 October 2011, published online 6 January 2012


Abbasi MK, Adams WA (2000) Estimation of simultaneous nitrification and denitrification in grassland soil associated with urea-N using 15N and nitrification inhibitor. Biology and Fertility of Soils 31, 38-44. doi:l0.1007/s003740050621

Cookson WR, Comforth IS (2002) Dicyandiamide slows nitrification in dairy cattle urine patches: effects on soil solution composition, soil pH and pasture yield. Soil Biology & Biochemistry 34, 1461-1465. doi:10.1016/S0038-0717(02)00090-1

Giltrap DL, Saggar S, Singh J, Zaman M (2010) A preliminary study to model the effects of a nitrification inhibitor on nitrous oxide emissions from urine amended pastures. Agriculture. Ecosystems & Environment 136, 310-317. doi:10.1016/j.agee.2009.08.007

Hewitt AE (1998) 'New Zealand Soil Classification.' 2nd edn. (Manaaki Whenua Press: Lincoln, New Zealand)

Kelliher I:M, Clough TJ, Clark H (2007) Developing revised emission factors for nitrous oxide emissions from agricultural pasture treated with nitrification inhibitors. Report prepared for Ministry of Agriculture and Forestry, Wellington, New Zealand. Available at: news-resources/publications

Kelliher FM, Clough T J, Clark H, Rys G, Sedcole JR (2008) The temperature dependence of dicyandiamide (DCD) degradation in soils: A data synthesis. Soil Biology & Biochemistry 40, 1878-1882. doi:10.1016/j. soilbio.2008.03.013

Li CS, Frolking S, Frolking TA (1992) A model of nitrous-oxide evolution from soil driven by rainfall events. 1. Model structure and sensitivity. Journal of Geophysical Research Atmospheres 97, 9759-9776.

MfE (2010) 'New Zealand's greenhouse gas inventory 199-2008. (Ministry for the Environment: Wellington, New Zealand)

Saggar S, Andrew RM, Tate KR, Rodda N J, Townsend JA (2004) Modelling nitrous oxide emissions from dairy-grazed pastures. Nutrient Cycling in Agroecosystems 68, 243-255. doi:10.1023/B: FRES.0000019463.92440.a3

Saggar S, Hedley CB, Giltrap DL, Lambie SM (2007) Measured and modelled estimates of nitrous oxide emission and methane consumption from a sheep-grazed pasture. Agriculture. Ecosystems & Environment 122, 357-365. doi:10.1016/j.agee.2007.02.006

Saggar S, Tate KR, Giltrap DL, Singh J (2008) Soil-atmosphere exchange of nitrous oxide and methane in New Zealand terrestrial ecosystems and their mitigation options: a review. Plant and Soil 309, 25-42. doi:10.1007/s11104-007-9421-3

Saggar S, Harvey M, Singh J, Giltrap D, Pattey E, Bromley T, Martin R, Dow D, Moss R, McMillan A (2010) Chambers, micrometeorological measurements, and the New Zealand Denitrification--Decomposition model for nitrous oxide emission estimates from an irrigated dairygrazed pasture. Journal of Integrative, Environmental Sciences 7, 61-70. doi:10.1080/19438151003621433

Donna L. Giltrap (A,E), Surinder Saggar (A), Jagrati Singh (A,B), Mike Harvey (C), Andrew McMillan (C), and Johannes Laubach (D)

(A) Landcare Research, Private Bag 11052, Palmerston North 4442, New Zealand.

(B) University of Melbourne, Melbourne, Vic. 3010, Australia.

(C) NIWA, PO Box 14-901, Wellington, New Zealand.

(D) Landcare Research, PO Box 40, Lincoln 7640, New Zealand.

(E) Corresponding author. Email:
Table 1. Parameters used in the NZ-DNDC simulations
WFPS, Water-filled pore-space
Parameter                               Value

Bulk density                            1.3 g/[cm.sup.-3]
Clay content                            23%
Initial soil N[H.sub.4.sup.+]-N         10 mgN/kg soil
Initial soil N[O.sub.2]-N               9 mg N/kg soil
Initial WFPS                            85%
pH                                      6
Soil organic carbon at surface          0.047 kg C/kg
Soil texture                            Silt loam
Soil WFPS at field capacity             80%
Soil WFPS at wilting point              28%
Depth of water retention layer          50 mm
Nitrification inhibitor effectiveness   70%

Table 2. Parameter ranges considered in uncertainty analysis
WFPS, Water-filled pore-space

Parameter                Range

Soil N[0.sub.3]          5-15 mg N[0.sub.3] -N/kg soil
Soil N[H.sub.4.sup.+]    5-15 mg N[H.sub.4.sup.+]-N/kg soil
Initial WFPS             80-90%
Soil organic carbon      0.043-0.051 g C/g soil
Bulk density             1.20-1.35 g/[cm.sup.3]
Clay content             19-27%
pH                       5.2-6.2
WPFS at field capacity   75-85%

Table 3. Measured and modelled [N.sub.2]O emissions between 12 June
and 1 July 2009 for grazing-only and grazing + DCD treatments

ME, Mean error; RMSE, root mean square error

                          Measured ([+ or -] s.e.)   Modelled
                          (g [N.sub.2]0-N/ha)

Grazing only              220 [+ or -] 90            169
Grazing +DCD              110 [+ or -] 20             68
Reduction in [N.sub.2]O   50 [+ or -] 40%             60
emissions using DCD

                          Model ME   Model RMSE
                          (g [N.sub.2]O-N/ha * day)

Grazing only                1.5      9
Grazing +DCD               -1.9      2
Reduction in [N.sub.2]O
emissions using DCD

Table 4. Sensitivity of modelled 20-day [N.sub.2]0 emissions to changes
in initial soil N[0.sub.3.sup.-], N[H.sub.4.sup.+], water-filled
pore-space (WFPS), soil organic carbon (SOC), and bulk density
(relative to baseline values)

Sensitivity =(% change in [N.sub.2]O emission)/(% change in parameter)

Parameter                NO emission      Parameter      Sensitivity
                         (t %             change)

Soil N03                 + 13             -44 to +67     0.29-0.19
Soil NH4'                23 to +24        + 50           0.47 0.49
Initial WFPS             -4.0 to +3.7     + 6.3          0.63-0.59
SOC                      -17 to +19       + 8.5          2.0-2.2
Bulk density             11.6 to +7.4     -7.7 to 13.8   1.5-1.9
Clay content             +41.6 to -37.5   + 22           -2.4 to -2.2
pH                       -25.7 to +4.2    -13 to +3      1.9-1.3
WFPS at field capacity   +14.7 to -18.1   + 6.3          -2.4 to -2.9
All parameters           -87 to +150
Measured uncertainty     40
COPYRIGHT 2011 CSIRO Publishing
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2011 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Giltrap, Donna L.; Saggar, Surinder; Singh, Jagrati; Harvey, Mike; McMillan, Andrew; Laubach, Johann
Publication:Soil Research
Article Type:Report
Geographic Code:8NEWZ
Date:Nov 1, 2011
Previous Article:Meeting bulk density sampling requirements efficiently to estimate soil carbon stocks.
Next Article:Effects of dairy cow treading pressures and food resources on invertebrates in two contrasting and co-occurring soils.

Terms of use | Privacy policy | Copyright © 2019 Farlex, Inc. | Feedback | For webmasters