University of Wollongong
Research Online
Faculty of Science, Medicine and Health - Papers Faculty of Science, Medicine and Health
2013
Accuracy of micrometeorological techniques for detecting a change in methane emissions from a herd of cattle
Johannes Laubach
Landcare Research NZ
Mei Bai
University of Wollongong, [email protected]
Cesar S. Pinares-Patino
Agresearch Grasslands Research Centre
Frances A. Phillips
University of Wollongong, [email protected]
Travis A. Naylor
University of Wollongong, [email protected]
See next page for additional authors
Research Online is the open access institutional repository for the University of Wollongong. For further information contact the UOW Library:
Publication Details
Laubach, J., Bai, M., Pinares-Patino, C. S., Phillips, F. A., Naylor, T. A., Molano, G., Cardenas Rocha, E. A. & Griffith, D. W. T. (2013).
Accuracy of micrometeorological techniques for detecting a change in methane emissions from a herd of cattle. Agricultural and Forest Meteorology, 176 50-63.
Accuracy of micrometeorological techniques for detecting a change in methane emissions from a herd of cattle
Abstract
Micrometeorological techniques are effective in measuring methane (CH4) emission rates at the herd scale, but their suitability as verification tools for emissions mitigation depends on the uncertainty with which they can detect a treatment difference. An experiment was designed to test for a range of techniques whether they could detect a change in weekly mean emission rate from a herd of cattle, in response to a controlled change in feed supply. The cattle were kept in an enclosure and fed pasture baleage, of amounts increasing from one week to the next. Methane emission rates were measured at the herd scale by the following techniques: (1) an external tracer-ratio technique, releasing nitrous oxide (N2O) from canisters on the animals’ necks and measuring line-averaged CH4 and N2O mole fractions with Fourier-transform infra-red (FTIR) spectrometers deployed upwind and downwind of the cattle, (2) a mass-budget technique using vertical profiles of wind speed and CH4 mole fraction, (3) a dispersion model, applied separately to CH4 mole fraction data from the FTIR spectrometers, the vertical profile, and a laser system measuring along four paths surrounding the enclosure. For reference, enteric CH4 emissions were also measured at the animal scale on a daily basis, using an enteric tracer-ratio technique (with SF6 as the tracer). The animal-scale technique showed that mean CH4 emissions increased less than linearly with increasing feed intake. The herd-scale techniques showed that the emission rates followed a diurnal pattern, with the maximum about 2 h after the feed was offered. The herd-scale techniques could detect the weekly changes in emission levels, except that the two vertical-profile techniques (mass-budget technique and dispersion model applied to profile) failed to resolve the first step change. The weekly emission rates from the external tracer-ratio technique and the dispersion model, applied to data from either the two FTIR paths or the four laser paths, agreed within ±10%
with the enteric tracer-ratio technique. By contrast, the two vertical-profile techniques gave 33–68% higher weekly emission rates. It is shown with a sensitivity study that systematically uneven animal distribution within the enclosure could explain some of this discrepancy. Another cause for bias was the data yield of the vertical-profile techniques being higher at day-time than at night-time, thus giving more weight to times of larger emission rates. The techniques using line-averaged mole fractions were less sensitive to the exact locations of emission sources and less prone to data loss from unsuitable wind directions; these advantages outweighed the lack of a method to calibrate CH4 mole fractions in situ.
Keywords
herd, emissions, methane, cattle, change, accuracy, detecting, techniques, micrometeorological, GeoQuest
Disciplines
Medicine and Health Sciences | Social and Behavioral Sciences
Publication Details
Laubach, J., Bai, M., Pinares-Patino, C. S., Phillips, F. A., Naylor, T. A., Molano, G., Cardenas Rocha, E. A. &
Griffith, D. W. T. (2013). Accuracy of micrometeorological techniques for detecting a change in methane emissions from a herd of cattle. Agricultural and Forest Meteorology, 176 50-63.
Authors
Johannes Laubach, Mei Bai, Cesar S. Pinares-Patino, Frances A. Phillips, Travis A. Naylor, German Molano, Edgar A. Cardenas Rocha, and David W. T Griffith
Accuracy of micrometeorological techniques for detecting a change in methane emissions from a herd of cattle
Johannes LaubachA,*, Mei BaiB,**, Cesar S. Pinares-PatiñoC, Frances A. PhillipsB, Travis A. NaylorB, German MolanoC, Edgar A. Cárdenas RochaC,D, David W. T.
GriffithB
ALandcare Research, P.O. Box 40, Lincoln 7640, New Zealand
BCentre for Atmospheric Chemistry, University of Wollongong, Wollongong NSW 2522, Australia
CAgResearch, Grasslands Research Centre, Tennent Drive, Palmerston North 4442, New Zealand
DFacultad de Medicina Veterinaria y Zootecnia, Universidad Nacional de Colombia, Carrera 45 No. 26-85, Bogotá, Colombia
*Corresponding author:
Tel.: +64-3-321 9999 Fax: +64-3-321 9998
email: [email protected]
**present address: CSIRO Animal, Food & Health Sciences, Townsville QLD 4811, Australia
Submitted to Agricultural and Forest Meteorology 15 November 2012 Revised version, submitted 23 February 2013
Abstract 1
Micrometeorological techniques are effective in measuring methane (CH4) emission 2
rates at the herd scale, but their suitability as verification tools for emissions mitigation 3
depends on the uncertainty with which they can detect a treatment difference. An 4
experiment was designed to test for a range of techniques whether they could detect a 5
change in weekly mean emission rate from a herd of cattle, in response to a controlled 6
change in feed supply. The cattle were kept in an enclosure and fed pasture baleage, of 7
amounts increasing from one week to the next. Methane emission rates were measured 8
at the herd scale by the following techniques: 1) an external tracer-ratio technique, 9
releasing nitrous oxide (N2O) from canisters on the animals’ necks and measuring line- 10
averaged CH4 and N2O mole fractions with Fourier-transform infra-red (FTIR) 11
spectrometers deployed upwind and downwind of the cattle, 2) a mass-budget 12
technique using vertical profiles of wind speed and CH4 mole fraction, 3) a dispersion 13
model, applied separately to CH4 mole fraction data from the FTIR spectrometers, the 14
vertical profile, and a laser system measuring along four paths surrounding the 15
enclosure. For reference, enteric CH4 emissions were also measured at the animal scale 16
on a daily basis, using an enteric tracer-ratio technique (with SF6 as the tracer). The 17
animal-scale technique showed that mean CH4 emissions increased less than linearly 18
with increasing feed intake. The herd-scale techniques showed that the emission rates 19
followed a diurnal pattern, with the maximum about two hours after the feed was 20
offered. The herd-scale techniques could detect the weekly changes in emission levels, 21
except that the two vertical-profile techniques (mass-budget technique and dispersion 22
model applied to profile) failed to resolve the first step change. The weekly emission 23
rates from the external tracer-ratio technique and the dispersion model, applied to data 24
from either the two FTIR paths or the four laser paths, agreed within ±10 % with the 25
enteric tracer-ratio technique. By contrast, the two vertical-profile techniques gave 33 26
to 68 % higher weekly emission rates. It is shown with a sensitivity study that 27
systematically uneven animal distribution within the enclosure could explain some of 28
this discrepancy. Another cause for bias was the data yield of the vertical-profile 29
techniques being higher at day-time than at night-time, thus giving more weight to 30
times of larger emission rates. The techniques using line-averaged mole fractions were 31
less sensitive to the exact locations of emission sources and less prone to data loss 32
from unsuitable wind directions; these advantages outweighed the lack of a method to 33
calibrate CH4 mole fractions in situ.
34 35
Keywords: Cattle CH4 emissions, Gas dispersion, Atmospheric surface layer, Tracer-ratio 36
techniques, Mass-budget technique, Backward-Lagrangian stochastic model 37
38
1. Introduction
39
Methane (CH4) emissions from ruminant livestock constitute about 30 % of New 40
Zealand's and 12 % of Australia's greenhouse gas emissions. For any future practice or 41
technology to mitigate these emissions, it must be verified at the “herd scale” (or 42
“paddock scale”), under representative farming conditions, that the expected emissions 43
reduction is achieved. In New Zealand (NZ) and Australia, cattle and sheep are farmed 44
outdoors year-round. To measure CH4 emissions outdoors, micrometeorological 45
techniques are effective at the herd scale (Laubach et al., 2008). These can potentially 46
verify small changes in emission rates, provided that the uncertainty with which such 47
changes are detected is accurately known. Here, we report an experiment measuring 48
the emissions from a herd of beef cattle that was designed to specify this uncertainty 49
for a suite of micrometeorological techniques. These include: a mass-budget technique 50
using vertical profiles of wind speed and CH4 mole fraction ([CH4]), a backward- 51
Lagrangian stochastic (BLS) dispersion model using the same [CH4] profiles, the same 52
BLS model using line-averaged [CH4] data gathered with two types of instruments, 53
and an external tracer-ratio technique, releasing nitrous oxide (N2O) co-located with 54
the CH4 emission sources and measuring line-averaged N2O and CH4 mole fractions 55
upwind and downwind of the sources. The last technique was classified by Harper et 56
al. (2011) as “non-micrometeorogical”, since it does not require any meteorological 57
information to compute emission rates; however, as its feasibility relies on atmospheric 58
transport, we consider it more appropriate to include it among the 59
“micrometeorological” techniques. With these, it also shares the spatial and temporal 60
scales at which it operates. To obtain non-micrometeorological reference values of 61
CH4 emissions on a daily basis, an enteric tracer-ratio technique was employed, 62
commonly known as the SF6 tracer-ratio technique. This technique operates at the 63
“animal scale”, i.e. individual animals. Details and references for all techniques are 64
given in later sections.
65
Laubach et al. (2008) already reported an experiment at the same site, with 66
similar animals, and using the same techniques except for the external tracer-ratio 67
technique. In that experiment, the animals were freely grazing in rectangular strips that 68
were changed daily. This represented usual farming practice in NZ but had two 69
disadvantages: the instrumentation had to be moved frequently to be in suitable 70
locations for capturing the emissions, and the feed intake of the animals could not be 71
measured. Knowing the feed intake is desirable because it is a major factor 72
determining CH4 emissions. The “methane yield” Ym, defined as the ratio of CH4 73
emissions per dry-matter intake (DMI) where both variables are expressed in units of 74
combustion energy, is recommended for inventory purposes (IPCC, 2006; Lassey, 75
2007). To overcome the two disadvantages in the present experiment, the cattle were 76
held in a grass-free area and were fed known quantities of pasture baleage. The 77
experiment was run for three weeks. Within each week, daily feed rations were held 78
constant, then increased for the following week in order to produce a measurable step 79
change in the herd’s CH4 emissions. The objective was to test for each technique 80
whether it was possible to detect this step change on the basis of weekly averages, 81
which in turn required quantification of the uncertainty of these averages. Factors 82
determining this uncertainty are not only measurement accuracy, but also for each 83
technique its data yield, i.e. the number of runs meeting specific quality criteria, and its 84
sensitivity to the spatial distribution of sources (which is always assumed 85
homogeneous across a prescribed area, except for the external tracer-ratio technique, 86
where no such assumption is needed).
87
A side issue, inadvertently discovered from consistency checks between the 88
different CH4 instruments, is a strong temperature dependence of the “GasFinder” CH4 89
laser, previously unreported in the micrometeorological literature. This result is 90
presented and empirically corrected for in the main text; the causes are discussed in the 91
Appendix.
92 93 94
2. Experimental design
95
The experiment was conducted in November 2008. The site (40.336° S, 175.465° E) 96
was located on the Aorangi Research Farm, ca. 20 km inland from the west coast of 97
the North Island of NZ, near the city of Palmerston North. It is ideally suited for 98
micrometeorological techniques and tracer dispersion studies because the surrounding 99
terrain is flat for several km in all directions and there is a predominant wind direction, 100
from W (which includes frequent afternoon sea breezes when synoptic winds are 101
weak). The cattle were managed in two groups, of 30 and 31 animals, respectively, 102
with identical treatments as described in 2.2.
103
104
2.1 Site preparation and setup geometry 105
A paddock area of about 200 m by 200 m had been sprayed with herbicide prior to the 106
experiment, so that the ground was initially bare; by the end, a thin cover of herbage 107
had grown back. In the NW quarter of this area, a rectangle of 80 m by 55 m was 108
fenced to contain the cattle herd. A 7 m tall mast to collect vertical profiles of wind 109
speed and CH4 mole fraction was erected at the midpoint of the E boundary of the 110
fenced area, and with additional fencing of a semicircle with 20 m radius, the cattle 111
were kept at a minimum distance of 20 m from the mast. The nominal source area 112
(rectangle minus semicircle) thus covered 3772 m2. This area was subdivided into two 113
equal-size enclosures (Fig. 1) to allow handling of the cattle in two separate groups and 114
to limit clustering of the animals at feeding time.
115
The surrounding terrain consisted mainly of flat paddocks, with no significant 116
flow obstacles to the W, S and E for at least 500 m. To the N, there was a water ditch 117
dropping 2 m below ground level at 50 m distance from the profile mast, and a 118
shelterbelt at ca. 150 m distance. The bare ground extended ca. 100 m from the mast to 119
the W, S and E, and 45 m to the N.
120
Two types of line-averaging optical sensors (described below) were employed 121
(Fig. 1). One was a CH4-specific laser system with four paths that were mounted 122
outside the cattle area, with one path along each side of the rectangle, at 5 to 10 m 123
distance from the fence. Path lengths were between 53 and 57 m, measured with 0.1 m 124
accuracy, and path heights above ground were 1.85, 1.37, 1.38 and 1.07 m (±0.02) m 125
for the W, S, E and N path, respectively. The other instrument type was a Fourier- 126
transform infra-red (FTIR) spectrometer, measuring mole fractions of multiple gas 127
species simultaneously along an open path. Two identical FTIRs were set up parallel to 128
the W and E sides of the rectangle, at about 5 and 12 m distance to W and E, 129
respectively. The path lengths were 88.9 and 89.5 m and the heights above ground 130
were 1.39 (±0.05) m.
131
Wind direction, atmospheric stability and velocity statistics (required as inputs 132
for the dispersion model) were measured with a sonic anemometer (81000V, RM 133
Young, Traverse City, Michigan, USA), mounted on top of a telescope mast, 3.85 m 134
above ground and 13 m NNW of the profile mast.
135 136
2.2 Animals and feed 137
The experiment involved 61 one-year-old Friesian-Hereford crossbred steers. They 138
were fed perennial ryegrass/white clover baleage that had been prepared on-farm, with 139
an approximate content of white clover of 20 % (DM basis). The chemical 140
composition of the feed was determined by near-infrared spectroscopy.
141
Prior to the start of measurements, the animals were acclimatised to the 142
management conditions over a period of 10 d by gradually increasing the baleage 143
fraction in the diet while decreasing herbage allowance. At the same time, the animals 144
were accustomed to wear the gas collection gear required for the SF6 tracer-ratio 145
technique. Starting on the 6th day of acclimatisation, baleage constituted 100 % of the 146
diet. The cattle were fed at three increasing feeding levels (low, medium and high) 147
over three consecutive periods of one week duration each. The first three days of each 148
week were considered as adjustment periods to the new feeding level, and the last four 149
days were used to conduct CH4 emissions measurements with the animal-scale 150
technique. Feeding levels were set with the intention of making 1.0, 1.5, and 2.0 times 151
maintenance energy requirements available to the steers during Weeks 1, 2 and 3, 152
respectively. The maintenance energy requirement was estimated following CSIRO 153
(2007). In order to allow for increasing fractions of feed wastage with increasing 154
feeding level, the feed amounts offered were set to 1.10, 1.73 and 2.60 times 155
maintenance requirements. The animals were weighed at the start of the experiment 156
and at the end of each one-week feeding period. Drinking water was freely available to 157
them, from troughs placed centrally in each enclosure.
158
The feed bales weighed around 450 kg. Every morning at 09:00 h the feed was 159
distributed on the ground inside the two enclosures, using a forage mixer feed-out 160
wagon equipped with a scale. The accuracy of the wagon’s scale was calibrated against 161
a load cell at the start of the experiment and found to agree within ±10 kg for a bale. In 162
each enclosure, the feed was laid out approximately in a “B” shape, which maximised 163
the areal spread of the feed while avoiding that the feed-out wagon had to cross the 164
supply pipes to the water troughs (Fig. 1).
165
Estimation of feed intake on the basis of metabolisable energy algorithms was 166
not considered reliable, because the 7 d long feeding period at each level was too short 167
to accurately account for liveweight change due to the potentially overwhelming 168
influence of variable gut fill when the animals were weighed. Alternatively, feed 169
intake estimates were to be based on the difference between feed offered and refused.
170
However, refusals were trampled and mixed with faeces and soil, and attempts to 171
collect representative samples proved impractical. Consequently, the proportion of 172
refused feed was estimated daily by visual assessment.
173 174
2.3 Profile mast instrumentation 175
The vertical profile of [CH4] was measured with an analyser using off-axis integrated 176
cavity output spectroscopy (DLT-100, Los Gatos Research, Mountain View, 177
California, USA). Air was drawn continuously from seven intakes via tubes into 178
ballast volumes. A datalogger-controlled valve-switching system sequentially selected 179
each of the intake lines for sampling by the analyser. Each switching cycle lasted 180
20 min, allocating 171 s of sampling time to each intake, of which the first 140 s were 181
discarded as flushing time to purge the sample cell of air from the previous intake.
182
Five of the intakes were mounted on the mast, at heights of 0.62, 1.22, 2.24, 4.13 and 183
7.18 m (±0.02) m. The other two were placed to the W and SE away from the cattle 184
area (Fig. 1), at 1.94 and 1.77 m height, respectively, so that for any wind direction one 185
of them would be suitably located to provide the local background mole fraction 186
(upwind of the herd).
187
Wind speed, relative humidity (RH) and temperature were measured at the same 188
five heights as [CH4]: wind speed by cup anemometers (A101M, Vector Instruments, 189
Rhyl, Clwyd, UK) with matched calibrations, humidity by capacitive RH sensors 190
(MP100A, Rotronic, Bassersdorf, Switzerland) and temperature by thermocouples 191
inside the same aspirated double radiation shields as the RH sensors.
192 193
2.4 Open-path instrumentation 194
The four-path CH4 laser system (GasFinder MC, Boreal Laser, Edmonton, Alberta, 195
Canada) consists of a central control unit, four remote heads (Fig. 2a) and four retro- 196
reflectors. The central unit houses the laser source, of wavelength 1653 nm, a sealed 197
reference cell filled with a known amount of CH4, an optical multiplexer and the 198
controlling and processing electronics. The wavelength is modulated across a narrow 199
band in order to sample the shape of the absorption line. The IR light is ducted via 200
fibre-optical cables to the remote heads, switching between them in a cycle of ca. 18 s 201
duration. The head emits the light along the open path towards the retro-reflector and 202
collects the returning light on a photodiode. The electrical output signal of the 203
photodiode is transmitted back to the central unit via a coaxial cable. There, the light 204
intensity measurements, as scanned across the waveband, are converted to CH4 mole 205
fractions with a regression algorithm that compares the measured line shape to that of 206
the reference cell (Boreal Laser, 2005). These mole fraction data were collected on a 207
laptop computer and subjected to post-processing to remove dubious data (guided by 208
quality flags and light level data provided by the instrument), apply correction factors 209
for signal dampening along the coaxial cable length, and form 20-min averages aligned 210
with the switching cycles of the vertical profile system.
211
The two identical open-path multi-gas FTIR spectrometer systems were 212
constructed at the University of Wollongong (Bai, 2010). The spectrometer is a 213
Matrix-M IRcube (Bruker Optik, Ettlingen, Germany). Light from an air-cooled SiC 214
infrared source is modulated by the interferometer and passes via a beam splitter (ZnSe 215
window) into a modified 10” Schmidt-Cassegrain telescope (LX200R, Meade 216
Instrument Corp., Irvine, California, USA). From there, the primary beam is 217
transmitted across the measurement path, returned by a 0.3 m diameter retro-reflector 218
array (PLX, Deer Park, New York, USA) and received by a HgCdTe detector (Infrared 219
Associates, Stuart, Florida, USA) which sits on the same tripod-mounted optical bench 220
as the source, interferometer and telescope (Fig. 2b). A Zener-diode thermometer (type 221
LM335) and a barometer (PTB110, Vaisala, Helsinki, Finland) are integrated in the 222
source/detector array to provide real-time air temperature and pressure data for the 223
analysis of the measured spectra. The spectrometer is controlled by software developed 224
at the University of Wollongong. The interferometer performs about 80 scans min−1 225
across the waveband of interest (700 – 5000 cm−1), which were in this experiment 226
averaged to 2-min spectra. From the spectra, the mole fractions of CH4, N2O, CO2, CO 227
and H2O are retrieved by a non-linear optimisation algorithm (MALT software), 228
developed by Griffith (1996). The algorithm uses absorption line strengths of these 229
gases as listed in the HITRAN database (Rothman et al., 2005) and constructs the 230
combination of mole fractions that provides the best match between the transmission 231
spectrum expected from this combination and the observed spectrum (Smith et al., 232
2011; Griffith et al., 2012). Only the CH4 and N2O mole fractions are used in this 233
experiment.
234 235 236
3. Techniques to determine CH
4emission rates
237
The various techniques to determine the CH4 emission rates are listed in Table 1. The 238
first technique (Section 3.1) was applied to individual animals, providing daily 239
averages. The others (Sections 3.2 to 3.4) all integrated over emissions from the herd, 240
as the emitted CH4 was transported and dispersed by the wind. They were applied 241
using 20-min averages of the CH4 mole fraction, relative to moist air, and of the 242
meteorological variables.
243 244
3.1 Enteric tracer-ratio technique (SF6) 245
The SF6 tracer-ratio technique (Johnson et al., 1994) was employed to estimate CH4
246
emissions from individual animals. For this purpose, on the first day of the pre-trial 247
acclimatisation period of 10 d, individually calibrated brass permeation tubes 248
containing the SF6 tracer were dosed per os into the reticulo-rumen of each 249
participating animal. For the last four days of each feeding level period, daily breath 250
and background air samples were collected from each steer, using a yoke fitted around 251
his neck (Fig. 3). Although the animals were out of the paddock for 2 to 3 h every 252
morning (time spent in transit to and from the management yards and for handling), 253
changing the sampling yoke took only about 1 min per animal, thus the collection 254
period was effectively 24 h. Samples were transported to the laboratory for analysis of 255
SF6 and CH4 mole fractions using a gas chromatograph (GC-2010, Shimadzu, Kyoto, 256
Japan), as described by Pinares-Patiño et al. (2007). Daily CH4 emissions from 257
individual animals were calculated from the ratio of mole fraction elevations of CH4
258
and SF6 (above the respective background-air mole fractions) and the known 259
permeation rate of SF6 from the permeation tubes (Johnson et al., 1994).
260 261
3.2 External tracer-ratio technique (N2O) 262
The basic approach of the N2O tracer-ratio technique is the same as for the SF6
263
technique: a tracer gas is released at a known rate co-located with the CH4 emission 264
from the animals, the mole fractions of both gases are measured post-emission in the 265
same volume, and then the ratio of the two mole fractions (after subtracting 266
atmospheric background from each) is equated with the ratio of their emission rates.
267
The main difference is that with the N2O technique the gases are not collected in a 268
container, but measured in situ, some distance downwind, by the open-path FTIR. The 269
path length is long enough to capture the emissions from many, if not all, animals 270
simultaneously, which makes the technique effective at the herd scale. Along with this 271
spatial integration comes high temporal resolution, not achievable with the animal- 272
scale technique.
273
Nitrous oxide was chosen because it is inert, non-toxic, commercially available 274
and its spectrum documented in HITRAN, which makes it easily measurable by the 275
FTIR in the same wavebands as CH4. Also, it could be released at rates much larger 276
than from any environmental sources (soil and animal excreta), so that the measured 277
downwind-upwind [N2O] differences were clearly attributable to the manufactured 278
release. In a previous experiment, Griffith et al. (2008) released the N2O from a 279
perforated pipe along the upwind fence line. Here, the release location was much more 280
closely matched to that of the CH4 by using pressurised canisters carried by the 281
animals near their mouths (Fig 3). These were 12 oz paintball gas canisters (Catalina 282
cylinders, Garden Grove, California, USA), which were filled with ca. 0.3 kg liquid 283
compressed N2O and fitted to the animals on a daily schedule, along with the SF6 284
collection yokes to which they were attached. Due to the long preparation time 285
required to fill the N2O canisters, only half of the cattle (15 in each group) were 286
equipped with them. This was unlikely to introduce additional error because animals 287
with and without release canister distributed themselves randomly across the same 288
area. Each canister was fitted with a tap and restricting orifice to regulate the release 289
rate to about 10 g h−1. The release rate’s time average was determined for each canister 290
by weighing it, before and after it was carried by the animal, and the release rate’s 291
temporal evolution was constructed by taking into account its systematic decrease with 292
canister pressure as well as its temperature dependence. Technical details of the 293
canister filling procedure and the computation of the N2O release rate are given in Bai 294
(2010). The CH4 emission rate was then calculated from the total N2O release rate and 295
the path-integrated CH4 and N2O mole fractions measured downwind from the herd.
296 297
3.3 Micrometeorological mass-budget technique 298
The mass-budget technique, sometimes also named integrated horizontal-flux 299
technique, was first applied to animal emissions by Harper et al. (1999) and Leuning et 300
al. (1999). Here, it was implemented using the vertical 5-point profiles of CH4 mole 301
fraction, measured with the Los Gatos analyser, and of wind speed, measured with the 302
cup anemometers. From these profiles and the background mole fraction, measured 303
upwind of the cattle, the integrated horizontal flux of CH4 was computed. To convert 304
this flux to an areal emission rate, it was divided by the distance across the cattle area, 305
from the profile mast in the upwind direction. Corrections for cross-wind variation of 306
this distance, for horizontal flux contributions above the top measurement height, and 307
for turbulent backflow were applied as described in Laubach and Kelliher (2004).
308 309
3.4 BLS technique 310
The principle of the backwards-Lagrangian stochastic (BLS) technique is to employ a 311
dispersion model to simulate trajectories of air parcels backwards in time from a 312
location where a gas concentration was measured, and to analyse statistically where 313
these trajectories intersect with a source volume (or area) in which the concentration of 314
the simulated air parcels would have been altered. For a given flow field (mean wind, 315
direction, and certain turbulent parameters), this trajectory-mapping delivers an 316
unambiguous linear relationship between measured concentration and emission rate.
317
The BLS model used here, developed by Flesch et al. (1995; 2004), is distributed 318
as a user-friendly software (WindTrax 2.0, www.thunderbeachscientific.com). The 319
model can accommodate concentration measurements at a point, such as realised by an 320
intake to a gas analyser, or along a line, such as the path of an optical sensor measuring 321
absorption. It is here applied separately to the three types of CH4-measuring 322
instruments: the closed-path analyser (vertical profile), the open-path laser system 323
(4 paths) and the open-path FTIR instruments (2 paths). While the three simulations 324
differ in their concentration input data, they assume identical CH4 source distributions 325
and identical flow fields. The CH4 source is defined as a ground-level area source with 326
the dimensions of the cattle enclosure. The flow field is specified, as in Laubach 327
(2010), by wind speed from the highest cup anemometer, the stability parameter and 328
the standard deviations of the three wind components from the sonic anemometer (at 329
3.85 m), and roughness length. Roughness length was not allowed to vary randomly 330
from run to run; rather, its evolution for the whole experiment was determined as a 331
function of time and wind direction. Over time, the roughness length increased slowly 332
as vegetation grew back, from 0.8 cm (bare soil) to 1.7 cm. The directional analysis 333
showed systematically larger values from the N sector than from the other directions, 334
consistent with the presence of ditch and shelterbelt to the N. The directional 335
dependence was fitted with a squared cosine function for directions within ±90 ° from 336
N, added to the time-dependent roughness length for the other sectors. The maximum 337
roughness length obtained for N winds was 9.1 cm.
338
339 340
4. Calibration checks and corrections to CH
4mole fractions
341
4.1 Closed-path analyser 342
Every 6 h, sampling of vertical [CH4] profiles with the closed-path analyser (Los 343
Gatos) was interrupted for 20 min to perform an automated calibration check, using 344
two gas cylinders that provided air with near-zero and near-ambient CH4 mole 345
fractions. These checks showed that throughout the campaign, [CH4] was reproducible 346
within ±3 ppb standard deviation (±0.18 % of ambient). Testing for temperature 347
dependence, a linear regression slope of 0.07 ppb K−1 was found, which is a factor 12 348
smaller than reported by Tuzson et al. (2010) for the same type of instrument. With 349
R2 = 0.03, this temperature dependence was not significantly different from zero and 350
thus neglected. Instead, the observed differences between subsequent calibration 351
checks were used to correct the mole fraction data in post-processing, assuming a 352
linear drift in time. Within a 20-min run, this drift correction had negligible effect on 353
mole fraction differences between intakes, but it provided minor adjustments to the 354
diurnal courses.
355 356
4.2 Open-path FTIR instruments 357
For the open-path instruments, in-situ calibration checks were not possible. Instead, a 358
series of consistency checks was conducted, first between paths of the same system, 359
then between these and the closed-path instrument. For these checks, only the periods 360
of cattle absence were used, expecting that then all instruments would effectively 361
sample the local background CH4 mole fraction – not necessarily identical to 362
hemispheric background, but influenced only by sources (or sinks) far enough away 363
that the position differences between the compared instruments did not matter.
364
As the periods of cattle absence were in the morning, strongly stable 365
stratification from the residual nocturnal boundary could occur on some days, with 366
elevated CH4 concentrations that potentially showed a vertical structure. To be able to 367
identify the influence of stable versus well-mixed conditions, the [CH4] data from the 368
two FTIR instruments are compared by plotting their ratio as a function of the wind 369
speed at 7.18 m (Fig. 4). As a change in alignment can change the instrument 370
calibration, the data are split into two groups, before and after a realignment of the W 371
path that was necessary following disturbance to the instrument while diagnosing an 372
electronic failure. Before the realignment, the ratio was consistently near 1, except for 373
some runs of low wind speed which are considered as stably stratified and not well- 374
mixed. The mean (±SD) for wind speed > 2 m s−1 was 1.007 (±0.009, n = 67). After 375
the realignment, the ratio was 1.034 (±0.006, n = 30). The SD values indicate a random 376
error of order 5 to 10 ppb at ambient [CH4]. This estimate would include an error 377
contribution from the horizontal distance between the instruments; it is thus 378
compatible with several independent instrument precision checks that yielded 379
estimates between 2 and 4 ppb (Bai, 2010).
380
Similar tests were done for the N2O mole fractions, required for the external 381
tracer-ratio technique (not shown). These yielded W/E ratios of 1.02 (±0.01) before 382
and 1.03 (±0.01) after realignment, respectively. The SD values indicate a random 383
error of 2 ppb at ambient [N2O], which again – due to the error from the horizontal 384
separation – is compatible with independent precision checks, yielding 0.4 ppb (Bai, 385
2010).
386
To compare the two FTIR to the closed-path analyser, [CH4] from each of the 387
former was divided by [CH4] from the nearest intake of the latter, selecting the same 388
cattle-free periods as before. The nearest path for the W path FTIR was the background 389
intake at 1.94 m height, and for the E path, the second-lowest intake from the profile 390
mast, at 1.22 m height. Since the open-path and closed-path data differ in three 391
respects: height, averaging volume (“line” vs. “point”), and sampling duration (shorter 392
for closed-path due to the switching cycle), one should not expect perfect agreement.
393
However the ratios showed remarkable consistency, at about 0.97 for the E path and 394
about 0.98 and 1.01 for the W path before and after the realignment, respectively 395
(±0.01 for each ratio). The ratios were independent of temperature and humidity. Any 396
ratios less than 0.94 were associated with wind speed < 2 m s−1. This indicates that the 397
absolute mole fractions of the two FTIR instruments were 2 to 3 % too low, equivalent 398
to an absolute error of 40 to 60 ppb at background, and they did not drift.
399
After these checks, the mole fractions from the W path were corrected against 400
those from the E path, by factors determined as the slopes of period-wise linear 401
regressions (of runs without cattle presence). This ensured matched calibrations 402
between the two FTIRs, which are crucial for accurate determination of upwind- 403
downwind differences, and in turn, emission rates. The FTIR data were not corrected 404
against the closed-path data. Thus, it is expected that the 3 % difference between the E 405
FTIR and the closed-path analyser carries through to the emission rate computations 406
with the BLS technique.
407 408
4.3 Four-path laser system 409
For the open-path laser (GasFinder MC), the same consistency checks were carried out 410
as for the FTIR. In Fig. 5, consistency between paths is assessed, by showing the ratios 411
of [CH4] from the W, S and N path, respectively, to [CH4] from the E path, against 412
wind speed. For each ratio, the spread is considerably larger than for the W/E FTIR 413
ratio in Fig. 4, and wind speed does not affect the spread, indicating that the variability 414
reflects instrument precision, not true [CH4] variations. The means (±SD), for 126 runs 415
of cattle absence, were 1.015 (±0.039) for W/E, 0.963 (±0.038) for S/E, and 416
1.033 (±0.041) for N/E, respectively. The SD values indicate a random error of order 417
48 ppb, at ambient [CH4], for each path. This is compatible with the manufacturer’s 418
specification of 2 ppm m precision for the path-integrated mole fraction, equivalent to 419
36 ppb for a 55 m long path. The laser system is thus one magnitude less precise than 420
the FTIR instruments and the closed-path analyser.
421
In Fig. 5, the differences of the mean ratios from 1 cannot be explained by height 422
differences (the W path was the highest, N the lowest, and S and E were at equal 423
height). Like for the FTIR, the E path was selected as the reference path, and the [CH4] 424
data from the other three paths were corrected against it. This was done separately for 425
the three feeding-level periods, to ensure the [CH4] differences between paths were 426
bias-free within each such period.
427
As for the FTIR, mole fractions from the W and E paths of the laser system were 428
compared to those from the nearest intakes of the closed-path analyser. Both horizontal 429
and vertical distances between path and intake location were smaller than for the FTIR.
430
Therefore, it was considered acceptable to include runs with and without cattle 431
presence in the comparison. To ensure sufficient atmospheric mixing, only runs with 432
wind speed > 2 m s−1 were selected. The path/intake [CH4] ratios reveal a significant 433
temperature dependence (Fig. 6), with a slope of −9.4 x 10−3 K−1 for both. The ratio 434
equals 1 at 23 (±2) °C (equal to the reference temperature, 296 K, of the HITRAN 435
database – see Appendix), while at 0 °C the open-path laser would overestimate [CH4] 436
by about 20 %. It was further tested whether the [CH4] ratios depended on specific 437
humidity or pressure, with negative result. This test was repeated with the data binned 438
into 3 K wide temperature classes, to prevent that the temperature dependence would 439
overwhelm any trends with the other variables, but no dependence on humidity or 440
pressure could be identified. By contrast, when the data were binned into five classes 441
of specific humidity, or pressure, then within each class there was still a temperature 442
dependence, represented by a slope that agreed within ±30 % with that in Fig. 6. For 443
all subsequent analyses, the [CH4] data from the laser system are therefore linearly 444
corrected for temperature, using the slope from Fig. 6 as the adjustment factor.
445
To our knowledge, such a temperature dependence of the GasFinder has not been 446
reported in the micrometeorological literature before. Possible physical causes are 447
discussed in the Appendix.
448 449 450
5. Results and Discussion
451
5.1 Animal behaviour 452
On the majority of mornings, the cattle were taken off-site between 7:15 and 8:00 h, to 453
have their gas collection yokes changed. During their absence, the feed was laid out in 454
the enclosures. When the cattle returned, between 10:00 and 10:30 h, they began 455
feeding immediately, and used up the available rations within a few hours. The rest of 456
the day they spent mostly drinking, resting and ruminating, moving around less than 457
grazing animals typically would. It was clear from observations, of the cattle 458
themselves as well as the distribution of their excreta, that they stayed preferably in the 459
W parts of the enclosures, frequently bunching near the W fence line (and 460
occasionally, around the water troughs). This behaviour was very different from the 461
freely-grazing behaviour observed by Laubach and Kelliher (2004; 2005b) and 462
Laubach et al. (2008), when the paddock area was by and large evenly covered by the 463
cattle, for most of the time. The uneven spatial animal distribution in the present 464
experiment had implications for the emission rate determinations. These are discussed 465
in Section 5.5.
466 467
5.2 Feed intake and emission rates at the animal scale 468
On average, the baleage contained 399 (±26) g dry matter (DM) per kg wet weight. On 469
a DM basis, it contained 46.5 (±5.5) % neutral detergent fibre, 19.0 (±3.4) % crude 470
protein, 9.4 (±2.3) % soluble sugars, 3.3 (±0.5) % lipids and 13.8 (±3.0) % ash. The 471
estimated metabolisable energy content of the feed was 10.5 (±0.4) MJ (kg DM)−1. 472
The average liveweight (LW) of the cattle increased with level of feeding from 473
331 kg at low level to 352 kg at the high level (Table 2). On a DM basis, feed on offer 474
increased by 58 % from the low to the medium level of feeding and by 52 % from the 475
medium to the high level. Visual assessment of feed refused showed negligible 476
amounts (< 1 %) for Weeks 1 and 2. In Week 3, the soiled feed residues were 477
considerably larger and estimated at 10 %. The resulting estimates of dry matter intake 478
(DMI) were 4.3, 6.7 and 9.3 kg DM d−1 animal−1 for the low, medium and high feeding 479
level, respectively (Table 2), and hence the relative increases from one feeding level to 480
the next were 58 and 38 %, respectively.
481
The CH4 emission rates were obtained with the SF6 tracer-ratio technique, for the 482
last 4 d of each week. The first 3 d were considered as an adjustment period to the new 483
feeding level (as diet composition did not change and feeding level increased, rather 484
than decreased, a period of 3 d sufficed). The average emission rates were 485
70.8 (±13.5), 89.7 (±11.1) and 119.1 (±16.4) g CH4 d−1 animal−1 for Weeks 1, 2 and 3, 486
respectively (Table 2), where the numbers are means (± SD) of the 61 animals. Even in 487
Week 3, when DMI was relatively generous, the CH4 emission rate was low in 488
comparison to those observed previously for freely-grazing steers at the same location 489
(Laubach et al., 2008). The standard deviations in Table 2 convert to standard errors of 490
the mean ≤ 2 g CH4 d−1 animal−1. We assume the weekly mean CH4 emission rates to 491
be “true” within this limit and use them as reference values for the other techniques.
492
The accuracy of the SF6 technique for estimating mean emissions of CH4 has been 493
proved by comparison to animal chamber measurements, both for sheep (Hammond et 494
al., 2009) and cattle (Grainger et al., 2007).
495
The CH4 emission rate increased with increase in feeding level, by 27 % from 496
the low to the medium feeding level and by 33 % from medium to high. Similarly, the 497
estimates of CH4 emissions expressed per unit of LW increased with increasing 498
feeding levels, by 24% between low and medium feeding levels and by 28% between 499
medium and high feeding levels. These increases were significant (P < 0.01), but less 500
than a proportional increase with DMI. Hence, the estimated CH4 yield per unit of feed 501
intake decreased, by 19 % from the low to the medium feeding level and by 4 % from 502
medium to high (Table 2). Decreasing CH4 yields at increasing feeding levels were 503
already reported by Blaxter and Clapperton (1965), and have been attributed to a faster 504
rate of passage of feed through the alimentary tract (Benchaar et al., 2001; Pinares- 505
Patiño et al., 2003).
506
The CH4 emissions per feed intake are unusually low, compared to the mean for 507
ryegrass-fed cattle in NZ, of 19.1 (±3.70) g CH4 (kg DMI)−1 (Hammond et al., 2009).
508
The observed CH4 yields, of 4.6, 3.7 and 3.6 % of gross energy intake (GEI), are closer 509
to the IPCC default value for concentrate-fed cattle, 3.0 %, than to the default value for 510
grazing cattle, 6.5 % (IPCC, 2006, p. 10-30). This is likely to be due to the baleage 511
used in this study, in contrast to fresh grass being the usual cattle diet in New Zealand.
512
It is well established (McDonald et al., 1991; Huhtanen and Jaakkola, 1993; González 513
et al., 2007) that ensiling reduces the ruminal degradability of dry matter. A computer 514
simulation by Benchaar et al. (2001) took this effect into account, and also that CH4 515
production decreases as pH and acetate/propionate ratio decrease. It predicted that CH4 516
emission from feeding lucerne silage would be lower than that from hay (3.7 vs. 5.4 % 517
of GEI). In light of this result, the observed CH4 yields appear plausible. Yet, a deeper 518
discussion of this finding is beyond the scope of this study.
519 520
5.3 Emission rates at the herd scale 521
For any of the herd-scale techniques, the run-to-run variations of the obtained CH4 522
emission rate, Qc, were considerable. This was the compounded effect of several 523
sources of variability: true emission rate changes in response to the animals’ digestion 524
processes, variability of source locations relative to the instruments as the animals 525
moved around, influence of variations in wind speed and direction on the effective 526
footprint of the various techniques, influence of wind speed and direction on the 527
magnitude of the concentration differences to be resolved, and random instrument 528
error (precision). Data availability also differed between techniques. This was partly 529
due to scheduled calibration checks and occasional malfunctions, e.g. dew on optical 530
components, failure of electronic components (resolved by replacement). In addition, 531
wind direction affected availability of the different techniques in different ways. The 532
techniques involving the vertical profile mast (mass-budget and BLS-profile) required 533
the mast to be downwind of at least a few steers, so gave meaningful emission rates 534
only for directions between 210 and 330 °. The techniques involving the two FTIR 535
instruments required one of them to be up- and the other downwind, so data were 536
accepted only if wind direction was within ±50 ° of either W or E. By contrast, the 537
four-path laser system could be used for all wind directions. Periods of low wind speed 538
were excluded for all techniques, when friction velocity was less than 0.1 m s−1. 539
Given the different operational constraints for each technique, and the overall 540
goal to assess the suitability of each to detect a change between weekly emission rates, 541
the results were first compared on the basis of weekly diurnal courses. To construct 542
these, for each week and technique the 20-min emission rates were sorted by time of 543
day, into 1-h wide bins, and then bin-averaged. The results are discussed in the 544
following subsections.
545 546
5.3.1 Diurnal pattern of emissions 547
Despite obvious discrepancies between techniques, a few features of the emission 548
pattern appear robust (Fig. 7). From about 10:00 h, when the cattle entered their 549
enclosures, Qc increased steeply until it reached a maximum typically around 13:00 h.
550
In Weeks 1 and 2, Qc began declining after that, according to all techniques, while in 551
Week 3 the techniques disagreed as to whether Qc further increased or decreased until 552
18:00 h. This pattern suggests that the maximum CH4 emissions typically occurred 553
within 2 h of the time of maximum feeding activity. Feeding always started soon after 554
10:00 h and in Weeks 1 and 2 was practically finished by noon, while it was stretched 555
out longer in Week 3, when food amounts on offer exceeded the animals’
556
requirements. Similar phase relationships between feeding time and maximum- 557
emission time were observed for feedlot cattle (Loh et al., 2008; Gao et al., 2011) and 558
sheep in test chambers (Lassey et al., 2011). From late afternoon throughout the night, 559
Qc generally decreased, reaching the lowest levels from about 7:00 to 9:00 h, before 560
the new day’s feed ration was offered.
561
With all herd-scale techniques, Qc from 10:00 to about 22:00 h (±3 h) was 562
usually larger than the daily average obtained from the SF6 tracer-ratio technique, and 563
during the other half of the diurnal cycle, Qc was smaller than the average. This 564
reflected that most of the feeding and ruminating occurred during the daylight hours.
565
Similar activity patterns are common for free-ranging animals.
566 567
5.3.2 Profile-based techniques 568
The mass-budget and BLS technique that used the vertical [CH4] profile from the 569
closed-path analyser detected virtually the same temporal pattern of emission rates 570
(Fig. 7). On a run-to-run basis, the two techniques agreed typically within 10 %, as 571
was found in other experiments (Laubach and Kelliher, 2005a; Gao et al., 2009;
572
Laubach, 2010).
573
From one week to the next, the daytime observations (10:00 to 18:00 h) clearly 574
showed increasing Qc. For the night-time hours, this cannot be stated. On many days 575
the wind direction turned from W in the afternoon, via N in the evening, to NE at 576
night, and then wind speed usually dropped, with friction velocity falling below the 577
acceptance threshold. Since the profile-based techniques relied on westerly wind 578
directions, night-time data availability was poor, and the available data suffered from 579
low replication rates. This is particularly evident in Week 1 (Fig. 7, left panel), where 580
lack of points for some hours, and a lack of error bars for others, indicated that none or 581
only one run per bin, respectively, was available.
582
Consistently for all three weeks, and almost all times of day, these techniques 583
gave higher Qc than the other techniques. The likely cause for this is that for most of 584
the time, the animals were not spread evenly across the enclosures. Rather, they 585
preferred to stay near the W fence line. Consequently, for the wind directions most 586
frequent and best-suited for the profile-based techniques, the actual animal density in 587
the effective source area tended to be larger than the nominal mean animal density 588
across the enclosures, which was used for converting emission rates per area to 589
emission rates per animal. Hence, the latter were frequently overestimated. This point 590
is further elaborated in Section 5.5.
591 592
5.3.3 BLS with open-path instruments 593
During the nights of Week 1, the laser system showed the highest data availability of 594
all herd-scale techniques, while no data were available from the FTIR either due to 595
unsuitable wind direction or, during one night, failure of the instrument W of the 596
cattle. During day-time in Week 1, and throughout Weeks 2 and 3, there was generally 597
reasonable agreement in the temporal pattern of Qc retrieved by BLS from the FTIR 598
and from the laser system. The laser system gave larger variability throughout, 599
indicated both by the error bars in Fig. 7 and the variations from hour to hour. This 600
may appear counter-intuitive, given that the laser system consisted of four paths almost 601
completely surrounding the emission sources, so it should have provided the most 602
representative coverage of the emissions plume for any wind direction. However, the 603
more erratic behaviour of Qc from the laser system can be explained by its poorer 604
measurement precision.
605
The BLS results from both open-path systems showed an increase of Qc from 606
one week to the next. Each week, the hourly averages from the laser system were 607
spread roughly from 0.3 times to 1.7 times the average Qc from the SF6 tracer-ratio 608
technique. The corresponding FTIR data spanned roughly between night-time minima 609
(when available) of 0.5 times Qc and day-time maxima of 1.5 times Qc from the SF6 610
technique.
611 612
5.3.4 External tracer-ratio technique 613
During Week 1, Qc from the N2O tracer-ratio technique agreed closely with that from 614
BLS using the same [CH4] input data from the FTIR (Fig. 7). During Week 2, the N2O 615
technique gave systematically lower Qc than BLS, and for many hourly bins the lowest 616
of all herd-scale techniques. The same is true at day-time in Week 3, but not at night.
617
In effect, this technique gave the smallest variations with time of day, tracking the 618
average Qc from the SF6 technique more closely than the other herd-scale techniques.
619
This can be explained by the fact that the tracer ratio is independent of the flow field 620
parameters, hence these parameters do not contribute to its error. The sampling errors 621
of the external tracer-ratio technique are those of the mole fractions and the mean N2O 622
release rate, while for BLS they are those of the mole fractions and a number of wind 623
and turbulence parameters.
624 625
5.4 Weekly mean emission rates 626
To obtain mean CH4 emission rates for each feeding level, with each technique, the 627
mean diurnal courses were averaged. This gave more representative estimates than 628
simply averaging all available runs, which would have weighted day-time more 629
strongly than night-time (because of the more frequent occurrence of unsuitable calm 630
periods at night). In some instances, especially in Week 1, some night-time hours were 631
not covered by the data, which made some residual bias towards day-time inevitable.
632
Random errors of the weekly means were obtained by propagating the standard errors 633
of the mean for the hourly bins (root-mean-squares estimate). Fig. 8 displays the 634
results, discussed in the following two subsections.
635 636
5.4.1 Comparison of absolute mean emission rates 637
For all three weeks, the techniques using vertical profile data gave the largest mean 638
emission rates. In Week 1, these were 68 and 56 % higher than Qc from the SF6 tracer- 639
ratio technique, for mass-budget and BLS, respectively. In Weeks 2 and 3, they 640
exceeded the SF6 technique consistently by 35 (±2) %. As noted in 5.3.2, the likely 641
main cause for these biases was the systematically uneven animal distribution, 642
resulting in a discrepancy between the actual animal density (in the effective source 643
area of these techniques) and the nominal mean animal density. Another contributing 644
factor was probably the low data yield at night, which could have led to potentially 645
erratic results for some night-time hours. In Week 1, when several night-time hours 646
had no data coverage at all, that would also have led to bias in the diurnal average.
647
The BLS technique with the four-path laser gave weekly mean Qc that were 648
18 % lower than from the SF6 technique in Week 1, and 8 and 6 % higher (not 649
significantly different) in Weeks 2 and 3, respectively. The low average in Week 1 650
appears to be caused mostly by low hourly values between 13:00 and 15:00, which do 651
not fit with the diurnal patterns observed in the other weeks and by the other 652
techniques (Fig. 7). Some of the contributing runs with low Qc had unstable 653
stratification and SE to S winds; it is possible that the N laser path, at only 1.07 m 654
height, was too close to the ground to accurately sample the emissions plume in 655
convective conditions.
656
The BLS technique with FTIR data exceeded Qc from the SF6 technique by 25 657
and 21 % in Weeks 1 and 2, and agreed within < 1 % (no significant difference) in 658
Week 3. The N2O tracer-ratio technique exceeded Qc from the SF6 technique by 18 % 659
in Week 1, by 3 % (not significant) in Week 2, and fell short of it by 10 % in Week 3 660
(different at the 95 % confidence level). For both techniques, the overestimate in 661
Week 1 is explained by the lack of night-time data, causing the weekly mean to over- 662
represent the times of feeding activity. For the BLS technique in Week 2, there is no 663
obvious cause for an overestimate. Valid FTIR data were obtained for either W or E 664
winds; the fraction of W winds was 90, 80 and 74 % for Week 1, 2 and 3, respectively, 665
so any effects related to wind direction are unlikely to explain differences between 666
weeks in the BLS technique’s performance. For the N2O tracer-ratio technique in 667
Week 3, it can be seen in Fig. 7 that the day-time emission rates were substantially 668
lower than those recorded by the other techniques. A possible cause for low emission 669
rate estimates is an underestimate of the N2O release rate, obtained from weighing of 670
the canisters and using the time and temperature dependence determined by Bai 671
(2010). However, the release rates during Week 3 showed very consistent diurnal 672
courses. It thus remains unclear whether the N2O tracer-ratio technique underestimated 673
the day-time emissions for this week. If it did, that would have caused the weekly 674
average to be an underestimate.
675
With the exceptions as discussed, it appears that all techniques using open-path 676
[CH4] measurements delivered weekly mean Qc that agreed with the enteric tracer- 677
ratio technique within 10 %, which is the commonly assumed magnitude of 678
uncertainty for micrometeorological flux measurement techniques. It should be noted 679
that the herd-scale techniques include CH4 emissions from the rectum, while the SF6 680
tracer-ratio technique does not. Comparisons of the SF6 technique to animal-chamber 681
measurements by McGinn et al. (2006) and Grainger et al. (2007) indicated lower 682
emission estimates for the SF6 technique of 4 % and 6 %, respectively, which could in 683
part be attributed to rectal emissions missed by this technique. Accounting for a small 684
underestimate by the SF6 technique would not change the finding that, for weekly 685
averages, the three open-path techniques agreed well with it. For the two vertical- 686
profile techniques, the differences to the SF6 technique would be somewhat reduced.
687 688
5.4.2 Suitability to resolve a change in mean emission rate 689
According to the SF6 tracer-ratio technique, the mean CH4 emission rate increased by 690
27 % from Week 1 to 2, and by 33 % from Week 2 to 3. The mass-budget technique 691
failed to detect the first change, but identified an increase of the right magnitude 692
(34 %) for the second. The BLS technique using vertical profile data recorded CH4
693
emission increases for both feed level changes, of 10 % and 34 % respectively, but the 694
change from Week 1 to 2 was not significant. For both these techniques, the failure to 695
detect the first step change must be attributed to sparse data coverage and large 696
variability during night-time hours, since Fig. 7 shows that the day-time emission rates 697
increased for both techniques.
698
The other techniques (N2O tracer-ratio and BLS with FTIR or four-path laser) 699
detected both changes in weekly emission rates at > 99 % confidence levels. The N2O 700
tracer-ratio technique recorded the lowest week-to-week increases of these techniques, 701
of 11 and 16 %, respectively, which were just under half of the increases shown by the 702
SF6 technique. For the first step change, this can be explained by the overestimate in 703
Week 1 due to a lack of night-time data. For the second step change, this is 704
computationally caused by low day-time emission rates in Week 3, yet the ultimate 705
cause for these is unclear.
706
Since the 95 % confidence intervals are about twice and 99 % confidence 707
intervals roughly 3 times as large as the standard errors in Fig. 8, one may infer that (in 708
all instances where data availability was sufficient) all techniques would probably 709
