Introduction
The natural gas industry has undergone a transformation in recent years,
largely due to technological advancements such as hydraulic fracturing
and horizontal drilling. These advances have led to increases in domestic
natural gas production (EPA, 2014b), although concomitant with this
increase has been a rising concern over methane emissions from the entire
natural gas system from the perspective of both environmental impact and a
loss of resources or product. Over the past decade, many studies have aimed
at quantifying these emissions using a variety of methods, yielding a wide
range of methane loss rate assessments for various sectors and basins from
< 0.5 % to greater than 10 % (Pétron et al., 2012a and b; Allen
et al., 2013; Karion et al., 2013; Bullock and
Nettles, 2014; Subramanian et al., 2014; Zimmerle et al., 2014; Harrison et
al., 2011; Zavala-Araiza et al., 2014).
The path of natural gas from well to the consumer can be considered in terms
of five possible steps: production, gathering, processing, transmission and
storage, and distribution. A recent series of studies have investigated
CH4 emissions from each of these activities (Subramanian et al., 2014;
Zimmerle et al., 2014; Allen et al., 2013). Presented here is a discussion of
the methods used during one such investigation in which tracer release
techniques were used to study emissions from gathering and processing
(G&P) facilities (Mitchell et al., 2015; Marchese et al., 2015). This
approach is similar to that employed in previous field measurements of
distribution, production, transmission, and storage facilities (Allen et al.,
2013; Subramanian et al., 2014; Lamb et al., 2015). Of particular emphasis in
this report are the measurement approach to the field campaign and the unique
emission profiles associated with gathering and processing, illustrating the
wide variety of handling, treating, and processing tools at the disposal of
the natural gas industry. The G&P field campaign was executed by Aerodyne
Research, Inc. (ARI), Carnegie Mellon University (CMU), and Colorado State
University from October 2013 through April 2014. Mobile laboratories operated
by ARI and CMU sampled emissions from a total of 130 G&P facilities across
20 natural gas basins in 13 states, using tracer release methodology as
discussed below. The measurements were performed with cooperation from
industry partners, who provided site access and detailed facility data, such
as natural gas throughput, gas type, gas composition, equipment inventories,
compressor power, age, and inlet/outlet pressures. Efforts were made by the
study participants to ensure that the facilities were sampled as found, and
the resulting data were assigned random numbers such that they cannot be
traced back to a specific facility or partner company.
The inherent chemical profile of natural gas from different sources can
significantly affect the technological approach that G&P facilities use
to prepare the gas for delivery into the transmission pipeline system. In
order to sample from the wide range of equipment employed during gathering
and processing, the campaign measured emissions from facilities associated
with a variety of types of gas, such as gas with low- and high-C2+
hydrocarbon content (here referred to as dry and wet gas, respectively), as
well as sour (high sulfur and/or CO2 content) and sweet gas sources
(low sulfur and/or CO2 content). More detailed information about site
selection is presented by Mitchell et al. in the companion paper,
“Measurement Results” (Mitchell et al., 2015). These facilities handled
natural gas derived from a variety of origins, including shale, coal-bed,
and conventional wells. In many cases, the emission profiles associated with
these facilities reflect the equipment used to prepare the natural
gas (EIA, 2006; Kidnay et al., 2011). For example, the first step
during gathering is often passage through gathering lines and a compressor
(gathering) station. One of the primary purposes of gathering facilities is
to collect and compress the input stream of gas to pipeline pressures,
usually ∼ 800 psi (∼ 55 bar). This requires the
use of compressors and associated equipment, for which there are multiple
possible emission sources such as compressor seals, natural-gas-driven
pneumatic devices, and engine exhaust. Frequently gathering facilities will
also remove water from the gas stream using dehydration trains, which
provide more possible emissions points. Following gathering, sweet, dry gas
can typically be easily conditioned and sent to the distribution network.
However, gas that is sour, wet, or with a high water content requires
significant subsequent processing, such as the removal of natural gas
liquids (NGLs) using forced extraction, and sometimes a dehydration step to further
remove water (Kidnay et al., 2011; Jumonville, 2010). These
relatively complex structures can involve distillation columns,
turboexpanders, separators, compressors, pneumatic devices, and heat
exchangers, all of which can emit CH4 either through minor fugitive
components or venting. Finally, extracted natural gas can have high CO2
and/or H2S content (i.e., sour, especially in coal-bed methane and some
shale-gas regions), which requires amine treating (frequently collocated
with other gas processing or compression facilities) to make it
distribution-ready (Kidnay et al., 2011). Again, this equipment and
additional processing adds to the number of possible emission sources.
Presented in the second half of this paper are examples of the unique
chemical profiles associated with the gathering, treatment, and processing
systems utilized by the natural gas industry. In the process of measuring
CH4 emission rates, these signatures can provide important information
about contributions from specific methane sources on
site.
Challenges in measuring emissions from natural gas facilities
The necessity for emissions measurements at natural gas facilities is
two-fold: (i) as an assessment of the impact of facility operation upon
regional and national air quality and climate (EPA, 2014a) and (ii) to
quantify losses due to normal operation or identify large emission sources.
In the case of (i), measured emissions provide an opportunity to compare to
national estimates and assess the overall impact of the natural gas supply
chain on CH4 emissions in the US. (Marchese et al., 2015;
Subramanian et al., 2014). In the case of (ii), these measurements aid the
natural gas industry in minimizing product losses.
Bottom-up approaches
Several approaches have been utilized to observe emissions at industrial
facilities. In some cases, a bottom-up approach is employed, wherein the
magnitudes of emissions from individual components are directly measured and
then added together to estimate the facility-level emission rate
(FLER) (Subramanian et al., 2014; Harrison et al.,
2011). This makes use of stack test data, manufacturer data, emission
factors, engineering estimates, activity factors, and on-site measurements.
These on-site measurements can take many forms, such as acoustic emission
detection, which quantifies leaks through suspected leak points such as
valves, and Hi-Flow® sampling, which can accurately determine
emission rates from a variety of fixtures. While these methods are widely
used and are capable of many measurements in a short time, they are not
applicable to all possible emission sources due to the number and
accessibility of fixtures within facilities (Subramanian et
al., 2014). This issue is particularly relevant at large processing and
treating plants, where the inability to measure emissions from a large
number of components could lead to an asymmetric bias in the reported FLER.
In addition, in order to accurately scale bottom-up studies to nationwide
(or even regional) estimates, care must be taken to ensure that the sampled
population, which is typically small, accurately represents the national or
regional inventory of facilities.
Optical gas imaging (e.g., infrared cameras such as FLIR®) is
a method by which leaks can be identified by using real-time infrared
imaging. This method provides a high duty cycle – dozens of fixtures within
a facility can be investigated per hour – and large emitters can be readily
identified. It is often used in conjunction with the above methods to locate
possible leak sources. However, because the method does not measure CH4
concentrations or flow rates, it does not quantify the emission magnitudes.
It nonetheless serves as a powerful qualitative tool in leak detection and
is therefore leveraged in this study to identify suspected emission points
at each G&P facility.
Top-down approaches
Top-down estimates aim to quantify methane emissions from a particular
geographic region. These results can then be compared to inventories
constructed from bottom-up measurements. Two top-down approaches are
commonly used for determining regional methane emissions: mass-balance
flights and fixed sensors fields (Zavala-Araiza et al., 2014). The mass-balance flight method, exemplified in several recent oil and gas
basin studies (Karion et al., 2013; Pétron et al., 2012b, 2013), uses upwind and downwind transects to capture emissions from
a bounded region. This area can be as small as an individual facility or as
large as an entire basin. Under favorable meteorological conditions, such
measurements can potentially estimate emissions from a large area with a
single flight, but these techniques are costly and provide little to no
source-specificity. This lack of source-specificity makes it especially
difficult for top-down studies to determine the relative emissions from
various activities within the industry (i.e., from gathering, processing,
transmission, or production) or even differentiate between emissions from
different industries, such as natural gas vs. feedlots vs. farming
operations vs. natural emissions. In addition, due to costs, these
studies have a limited number of samples over a short duration (hours) and
therefore may not be representative of actual emissions when extrapolated
and compared with annual nationwide inventories.
Top-down estimates of regional emissions are also commonly performed using
meteorological transport simulations in combination with a network of fixed
sensors or using inverse modeling coupled with dispersion or advection
models (Wofsy, 2013; Bullock and Nettles, 2014;
Zavala-Araiza et al., 2014). Such methods can leverage preexisting sensor
networks with data available 24 h day-1. However, the interpretation
of sensor data for emissions measurements is highly dependent upon
atmospheric modeling, with large uncertainties (Nehrkorn et al., 2010;
Draxler and Hess, 1997, 1998).
Tracer release approach
Because the goal of this study was to develop an understanding of the total
emissions from individual G&P facilities and to use these measurements
to estimate total national emissions from natural gas gathering and
processing (Marchese et al., 2015), the measurement approach
described here uses an established measurement technique called tracer flux
ratio (or tracer ratio). It has previously been demonstrated that the tracer
ratio method can quantify the total emissions from industrial sites (Lamb
et al., 1995; Allen et al., 2013) and
landfills (Czepiel et al., 1996;
Mosher et al., 1999). The strengths of the method are that it does not
require theoretical modeling, can measure facility-wide emissions, and under
the proper conditions can be useful in identifying large sources within a
facility. The tracer ratio method has been shown to effectively and
accurately yield the total emissions from many small sources within a large
area, where measurements of individual leak rates would be
challenging (Shorter et al., 1997; Mosher et al., 1999; Subramanian et
al., 2014; Lamb et al., 1995). It therefore allows for FLERs to be determined for large facilities such as
processing and treatment plants, where a multitude of possible emissions
sources exist that may not be accessible or quantifiable using bottom-up
approaches. For this study, the method is applied to quantify total
facility-level methane emission rates (fugitive, venting, and combustion) at
natural gas processing plants, treatment facilities, and midstream compressor
stations.
Conceptually, the tracer release method is based upon the simple relation
that the downwind concentration enhancement of gas X above ambient
background, Δ[X], is directly related to the flow rate at its
source, FX:
Δ[X]=α⋅FX.
The relation between these two quantities is determined by α. The
coefficient α is a complicated function of meteorological
information, such as wind speed, wind history, turbulence, solar irradiance,
temperature, boundary layer height, local topography, and downwind distance.
In principle this information can be estimated using, for example, a
Gaussian dispersion model (Beychok, 2005). Such models have had success
in qualitatively reproducing measured plume data but frequently lack the
precision and accuracy required for this study, especially in areas with
complex terrain and meteorology.
Schematic of dual-tracer release technique. At distances far
downwind (top), both tracers and CH4 are spatiotemporally overlapped.
At distances closer to the facility, the spatial position of the CH4
plume relative to the two tracer plumes can indicate the location of an
emission vector on-site with sub-facility resolution.
The tracer release method provides an empirical means to bypass the need for
determining α (Lamb et al., 1986, 1995). By deploying a known flow of tracer gas located physically
near a CH4 emission source, the downwind tracer concentration
enhancement (above background), Δ[T], downwind CH4
concentration enhancement (above background), Δ[CH4], and
tracer flow rate, FT, become measurable quantities. The ratio of the
two downwind concentrations is then equal to the ratio of flow rates:
Δ[CH4]Δ[T]=ααFCH4FT=FCH4FT,
where FCH4 refers to the flow of CH4 from the facility.
Because concentrations Δ[CH4] and Δ[T] are measured
and FT is known, FCH4 can be determined without the need for
detailed information about α.
The underlying assumption in this technique is that the tracer release point
is located close enough to the unknown emission source that both gases
experience the same dilution factor α. This separation distance
becomes less important as the concentration measurement (aboard a mobile
platform) moves further downwind. However, when the separation distance is
of the same order as the downwind distance, the α values associated
with CH4 and T are expected to be significantly different. Under ideal
circumstances, the tracer is collocated with the emission source, and their
concentrations are measured far downwind in stable meteorological
conditions. In practice this is not always possible due to facility size,
interfering methane sources, road access, or varying winds.
To mitigate these issues, this study made use of a dual-tracer release
technique (Allen et al., 2013) in which two different
tracer gases, in this case N2O and C2H2, are released from
different locations within the facility, bracketing the on-site equipment, as
shown in Fig. 1. The use of a second tracer has two important advantages
over single-tracer measurements. First, closer downwind measurements (50–200 m downwind) afford a refined assessment of an emission source location based
upon the position of its CH4 plume relative to each tracer plume.
Second, when conducting mixed plume characterization in the far-field
(downwind), where αN2O∼αC2H2∼αCH4, the second tracer becomes an internal standard to
the measurement. This capability mitigates the need for a calibration or
for benchmarking against other measurements. Emissions rates determined by
tracer release have, however, been compared to detailed on-site leak
measurements in Subramanian et al. (2014). That study found that these two techniques usually agreed to within
experimental uncertainty. The use of two known tracer gas flow rates and an
observed downwind molar ratio also provides an empirical measure of the
uncertainty for every plume. This error will be further described below, in
the Supplement, and in the associated Measurements
report (Mitchell et al., 2015).
Understanding and optimizing data quality
In the context of the two possible transect scenarios depicted in Fig. 1
(spatially overlapping plumes vs. spatially separated plumes), it is
important to qualitatively understand what measurement conditions (tracer
separation, transect distance, meteorology) yield these two results. This
can be developed using Gaussian dispersion modeling as a
guide (Beychok, 2005). As a rule of thumb, for typical mid-day
atmospheric conditions (stability classes A, B, or C, as described in
the Supplement) and downwind distances (100–3000 m), the
horizontal width of a plume that is propagating according to Gaussian
dispersion is ∼ 20–50 % of the distance that it has traveled
from its source. That is, the ratio of plume width to downwind distance is
0.2–0.5, where low wind conditions yield wider plumes (∼ 0.5) and high wind conditions yield narrower plumes (∼ 0.2).
A plume observed 1000 m downwind of its origin, for example, is typically
200–500 m wide.
If the plume widths of two gases being measured downwind (e.g., CH4 and
N2O) are much larger than the separation of their sources, the plumes
will generally be co-dispersed or spatially overlapping. Therefore the
ratio of the distance between emission sources to the downwind transect
distance must be less than 0.2–0.5 in order to achieve co-dispersion. When,
for example, the separation between an N2O tracer and a CH4 source
is 100 m, the downwind distance required to observe the onset of
co-dispersion is > 500 m in high winds and > 200 m in
low winds. Alternatively, when local road access limits the downwind distance
to 200–500 m, the N2O tracer must be placed within 100 m of the
suspected CH4 emission source.
This same rule-of-thumb approach can be applied to cases where a nearby
CH4 source, such as a wellhead, may interfere with the FLER measurement
at a G&P facility. In these cases, the downwind transect must be close enough
that the interfering plume width is smaller than its separation from the
G&P facility. For example, if the distance between a wellhead and
facility is 50 m, downwind transects must be less than 100–250 m in order
isolate and exclude the wellhead plume from the FLER estimate.
When the second tracer is used as an internal standard, it can serve to
quantify the uncertainty of the measurement. As will be shown below and in
the Supplement, this uncertainty decreases when the two tracer
plumes are spatially overlapping compared to cases where the plumes are
separated. Because this precision reflects the uncertainty in the FLER,
efforts are made by the study team to maximize the co-dispersion of methane
and tracer plumes. In light of the above discussion, this can be achieved by
attempting to place one or both tracers near the dominant suspected emission
source at a facility, when one exists. When these conditions are met, the
downwind distance required to observe co-dispersion is reduced, thereby
increasing the instrumental signal-to-noise and further separating any
possible interfering sources.
Initial placement of the tracers at opposite ends of the facility allows for
early transects to identify suspected methane emission locations. In some
cases, the observed methane plume will appear covariant with one of the two
tracers, indicating that the dominant methane emitter is in the vicinity of
that tracer. In many cases, however, the methane plume is observed between
the two tracer plumes. In this scenario, one (or both) of the tracers is
typically moved such that its plume is spatially overlapping the methane
plume. This process is iterated multiple times over the course of the
measurement in order to yield plumes that exhibit high degrees of
CH4-to-tracer correlation.
While two tracers act as an internal standard in the horizontal plane, a
complicating factor unique to some large facilities (e.g., processing plants
and larger gathering facilities) studied here is the presence of flares
and/or engine exhaust stacks, some of which can be over 20 m tall. Presented
in the Supplement is a Gaussian plume and Brigg's equation
analysis of the effect of a possible elevated CH4 source on the
measured emission rate (Beychok, 2005). A simple rule-of-thumb approach
as used above is hampered here by both buoyant plume rise effects and plume
reflection off of the ground. These calculations indicate that in strong
wind conditions (i.e., high atmospheric stability classes, such as in winds
above 5 m s-1), the measured emission rate determined from close transects can
be biased considerably low, depending upon the fraction of the emission
coming from elevated positions. In wind conditions below 5 m s-1, the
dispersion is large enough that the bias is lessened to 0–50 %. To
minimize this bias, plumes were obtained as far downwind as possible, and at
several processing plants a tracer was emitted at an elevated position such
as the side of a demethanizer column or stack. The impact of the bias upon
the overall data set and resulting conclusions is discussed in more detail
in the accompanying Measurements paper (Mitchell et al., 2015).
Auxiliary species
The study team also used measurements of other species, CO, CO2, and
C2H6, to aid in identifying and attributing methane emissions to
targeted G&P facilities. For example, engine exhaust from reciprocating
engines and turbines that power compressors at many natural gas facilities
will contain CO and CO2. This enables potential differentiation between
emissions of G&P equipment and those emanating from nearby well pads
(which typically do not include combustion sources or emit much smaller
amounts of CO and CO2). Similarly, amine treatment systems serve as
non-combustion sources of CO2 and are easily distinguishable from other
facilities (Rochelle, 2009; EIA, 2006; Kidnay et al., 2011).
Ethane measurements serve multiple purposes within the context of this
study. First, the presence of ethane associated with methane in downwind
plumes indicates that some fraction of the methane is of thermogenic, rather
than biogenic, origin. The ability to distinguish between these sources is
especially important in farming and ranching regions, where livestock
emissions can be a substantial source of CH4. Second, the observed
ethane-to-methane ratio (E / M ratio) in a downwind plume can serve as a
unique identifier of a facility of interest. It can therefore be used to
differentiate a particular emission source from others in the area. Finally,
variations in ethane content over close transects can indicate active
distillation or other processing present on-site. The utility of these
measurements will be explicitly illustrated via examples in the Results
section.
Instruments and sensitivities for measured species on Aerodyne and CMU mobile laboratories.
Instrument
Species detected
Sensitivity
Aerodyne mobile laboratory
Aerodyne dual QCL
CH4
1 ppb
C2H2
200 ppt
Aerodyne mini QCL
C2H6
100 ppt
Aerodyne mini QCL
N2O
100 ppt
CO
100 ppt
Li-Cor NDIR
CO2
500 ppb
Carnegie Mellon mobile laboratory
Picarro CRDS
CH4
3 ppb
C2H2
600 ppt
Aerodyne dual QCL
C2H6
100 ppt
N2O
100 ppt
CO
100 ppt
Laboratory and instrument details
The two mobile laboratories used in this study were operated by Aerodyne
Research, Inc. (Herndon et al., 2005) and Carnegie Mellon
University (Subramanian et al., 2014). Both mobile
laboratories contain a variety of spectroscopy-based gas-detection
instruments, which sample the ambient air from an inlet mounted on the front
of the vehicle. In the case of the Aerodyne mobile laboratory, three ARI
direct-absorption quantum cascade laser (QCL) spectrometers (Jiménez
et al., 2005; Yacovitch et al., 2014; McManus et al., 2005) operating at
20–40 Torr are employed in series to detect CH4, C2H6, CO,
N2O, and C2H2. To detect CO2, a non-dispersive infrared (NDIR)
LiCOR® instrument is used. In this work, the QCL spectrometers
are operated in series, with flow rates through the instruments of
∼ 10 SLPM. This flow rate afforded a time response of
< 1 s. The NDIR instrument draws a small flow from the inlet line
before the air sample entered the QCLs. The QCL spectrometers report mixing
ratios of all species in parts per billion by volume (ppbv), while the NDIR
instrument reports CO2 in parts per million by volume (ppmv). In the
Carnegie Mellon mobile laboratory, CH4 and C2H2 are measured
using a Picarro cavity ring-down spectrometer (Crosson, 2008;
Rella et al., 2009) running at 4–5 Hz, while C2H6, N2O, and
CO are measured using an ARI Dual QCL spectrometer operating at 1 Hz.
Detection limits of all instruments are listed in Table 1. Except for
practically limiting the minimum detectable concentration of certain
species, the differences in equipment manufacturer and sensitivity do not
affect the results of the measurements. In addition to the concentration
information, both mobile laboratories record their location, bearing, and
heading using Global Positioning System (GPS; Garmin® 76 and
Hemisphere GPS Compass® for the ARI laboratory,
Airmar® for the CMU laboratory). A small meteorological
station (Airmar® 200WX or LB150) is also mounted on a boom at
the front of the vehicle to record true wind speed (speed corrected for
vehicle velocity), true wind direction (wind direction relative to true
north), and GPS location. Along with the mixing ratios, this information is
recorded at 1 s intervals on a main onboard acquisition computer, where all
of the acquired data are visualized in real time and can be overlaid on
maps.
Both laboratories are accompanied by a tracer release vehicle (i.e., pickup
truck) to facilitate the storage, setup, and release of the N2O and
C2H2 tracers. Tracer gas bottles are stored on the bed of the
truck, along with flow control systems and associated valves, tubing, and
telemetry systems. Polyethylene tubing for each tracer is rolled out from
the pickup truck up to 200 m to the intended release location, where the end
of the tube is attached to on-site equipment or placed on a tripod. For
both laboratories, tracer flow rates are controlled by Alicat®
MC-series mass flow controllers. The mass flow rates are recorded via RS232
to an onboard computer in the vehicle.
In addition to the tracer gas flow systems, three portable meteorological
stations (Airmar® 200WX) are deployed on tripods, sometimes
serving as physical supports for the tracer release tubing. They are capable
of recording GPS, true wind direction, and wind speed with 1 s resolution.
Each unit broadcasts that information wirelessly or via an RS232 cable at 1 Hz to a computer onboard the tracer release vehicle, where it is recorded
and displayed for observation by the tracer release personnel to advise the
mobile laboratory as needed. When considered in the context of tracer
placement, the wind data can immediately inform mobile laboratory personnel
whether a tracer is being deployed in an area on-site that is not
well ventilated. If this is the case (frequently due to the local wind
currents near buildings) the tracer can then be moved to allow it to be
carried downwind by the larger regional wind mass. This information also
provides a crude wind field for later analysis to better understand the
sources of error and uncertainty in tracer release methods.
Calibrations and ranges
In both laboratories, the inlet was periodically overblown (injected with a
flow larger than the intake flow) with ultra-zero air (AirGas®
or Praxair®) to zero the instruments, typically every 15 min for 30 s. Because CH4 and N2O are present in
background ambient air (1900 and 325 ppbv, respectively), zeroing events
also serve as an approximate check of those instrument calibrations. Full
instrument calibrations were performed several (4–5) times over the course
of the measurement campaign using calibration standards. For these dilution
calibrations, a controlled mass flow of calibration gas is released into a
known zero-air flow, and the resulting mixture is overblown into the inlet.
The mixture is changed by varying the calibration gas flow using either a
series of critical orifices or mass flow controllers (Alicat®
MC Series). Typical calibration ranges were 0–10 ppm for CH4, 0–500 ppb
for C2H2, and 0–1000 ppb for N2O. The calibrations were
linear, with typical R2 > 0.99. The results of these
calibrations changed less than 5 % over the course of the campaign. The
mass flow controllers onboard the tracer release vehicle are also
periodically calibrated using a NIST-traceable Dry-Cal® flow
meter.
Field implementation
In practice, when the mobile laboratory arrived at a facility a safety
meeting was conducted with the facility supervisor, after which the tracer
release apparatuses were set up. The tracer positions were decided upon
after discussion with the supervisor regarding likely emission sources (near
compressors, dehydrators, tanks, etc.), a cursory survey with infrared
imaging, consideration of the current wind conditions, site size and safety
issues, and sometimes after performing an initial drive within facility
boundaries. After setup, the tracer gases were released and the mobile
laboratory was deployed downwind. Constant communication was maintained
either over CB radio or cellular phones. During this period, an additional
study team member (“the on-site observer”) surveyed the facility with an
infrared camera, inventoried facility components, and recorded relevant
information such as facility throughput, equipment counts, and motor, engine,
or turbine horsepower. In many cases the identification of emission sources
by survey of the facility using infrared imaging agreed with or informed the
results of close-pass plume transects. If the mobile laboratory detected
CH4 plumes that were spatially separated from the tracer plumes, one or
both tracers were moved to maximize co-dispersion with CH4. When
possible, on-site ethane-to-methane ratios were measured by driving the
mobile laboratory within fence line immediately downwind (< 25 m) of
on-site equipment, for future comparison with partner company gas
chromatograph (GC) data.
After acquiring enough downwind plumes (a target of 10) to provide a
statistically meaningful time-averaged FLER and uncertainty, the mobile
laboratory returned on-site, and the tracer release hardware was packed.
Usually at least two facilities were surveyed daily and sometimes as many
as four, depending upon wind conditions, time, and the locations of nearby
facilities. Because of their size and scale, a full day was reserved to
sample emissions from processing facilities.
Plume types and analysis methods
There are multiple ways in which downwind tracer plumes can be analyzed,
depending upon the plume intensity and spatial overlap between the tracer
and CH4 plumes (Subramanian et al., 2014). Figures 2–5 show the four possible plume types observed during the G&P
campaign.
Example dual-correlation plume from a natural gas facility. Top
panel: time trace of CH4, C2H6, N2O, and C2H2
concentrations, showing high temporal correlation. Center left panel: map of
tracer location (right side) and transect location (left side) during the
course of the plume. Red, blue, and green weighted lines correspond to
CH4, C2H2, and N2O intensities during the
transect, spatially offset for clarity. Thin lines point into the wind at
the mobile laboratory (red) and at the facility (light blue, pink, and
yellow). Blue square and green triangle indicate C2H2 and N2O
release locations, respectively. Lower panels: Correlation analysis of
C2H6 vs. CH4, N2O vs. C2H2, CH4 vs.
C2H2, and CH4 vs. N2O. The measured emission rate from
this plume was found to be 3.4 SCFM.
Similar to Fig. 2, illustrating dual-area-type plumes. Top
panel: time trace of CH4, C2H6, N2O, and C2H2
concentrations, showing high temporal correlation. Center left panel: map of
tracer location (right side) and transect location (left side) during the
course of the plume. Red, blue, and green weighted lines correspond to
CH4, C2H2, and N2O intensities during the
transect, spatially offset for clarity. Thin lines point into the wind at
the mobile laboratory (red) and at the facility (light blue, pink, and
yellow). Blue square and green triangle indicate C2H2and N2O
release locations, respectively. Lower panels: correlation analysis of
C2H6 vs. CH4, N2O vs. C2H2, CH4 vs.
C2H2, and CH4 vs. N2O. Note the lack of correlation in
lower left and center panels, indicating that the analysis must rely on an
area method. Note, however, the strong correlation between C2H6
and CH4 (bottom left), indicating that the observed methane is derived
from natural gas. The emission rate determined from this plume was found to
be 3.1 SCFM.
Example of a single-correlation plume (CH4 correlation with
C2H2). Top panel: time trace of CH4, C2H6,
N2O, and C2H2 concentrations, showing high temporal
correlation. Center left panel: map of tracer location (right side) and
transect location (left side) during the course of the plume. Red, blue, and
green weighted lines correspond to CH4, C2H2, and N2O
intensities during the transect, spatially offset for
clarity. Thin lines point into the wind at the mobile laboratory (red) and
at the facility (light blue, pink, and yellow). Blue square and green
triangle indicate C2H2 and N2O release locations,
respectively. Lower panels: correlation analysis of C2H6 vs.
CH4, N2O vs. C2H2, CH4 vs. C2H2, and
CH4 vs. N2O. The emission rate determined for this plume was found
to be 8.1 SCFM.
Example of analysis using a linear combination of tracer plumes.
Note that N2O and C2H2 are associated with different sections
of the CH4 plume (top). Adding the two tracer plumes in an
81/19 % combination yields a correlation diagram (below) with high
R2 value (0.87). The emission rate determined from this plume is 56.1
SCFM.
Dual correlation
The ideal scenario occurs when the measurement transect is far enough
downwind of the facility that the CH4, N2O, and C2H2
plumes are spatially overlapping. The resulting measurements of
concentration vs. time exhibit a high degree of covariance between species,
as shown in the top panel of Fig. 2. Analysis of these “dual-correlation”
plumes consists of plotting the concentration of one species vs. another
and performing a linear orthogonal distance regression fit as shown in
the bottom panels of Fig. 2. This regression analysis is performed for
CH4 vs. N2O, CH4 vs. C2H2, N2O vs.
C2H2, and C2H6 vs. CH4. From these linear
regressions, the slope indicates the ratio of concentrations of the two gas
species (for use in Eq. ), and R2 indicates the degree of
correlation. These values are recorded for use in determining whether the
plume meets the acceptance criteria for the CH4 emission rate to be
considered valid. If the R2 values derived from fits of CH4 vs.
N2O, CH4 vs. C2H2, and N2O vs. C2H2 are
all greater than 0.75, and the tracer ratio ([C2H2] / [N2O]) is
within a factor of 1.5 of the known tracer flow rate, the plume is a
candidate for dual-correlation analysis. The choice of acceptable R2
and tracer ratio were based upon values at which further relaxation of the
criteria would alter the uncertainty and accuracy of the FLER
measurement (Mitchell et al., 2015). A discussion of the
use of a factor for the tracer ratio criterion, as opposed to a deviation
such as ±50 %, is presented in the Supplement.
Dual area
In certain circumstances, wind conditions along with local road access and
intervening CH4 sources prevent the ability to get far enough downwind
for the tracer gas and CH4 plumes to become spatially overlapped.
However, transects may still be performed closer to the facility
(∼ 50–500 m) such that all three species will be observed.
As illustrated in the example shown in Fig. 3, under these circumstances
correlation diagrams do not provide useful information about the ratio of
species (bottom panels). In these cases a “dual-area” technique is used,
in which the analysis must rely on the integrated area of each species' plume
over the time of the transect. Here, the deviation of the species' mixing
ratios from ambient conditions must be considered, rather than the raw
integrated intensity. This point is particularly relevant for CH4 and
N2O, whose ambient concentrations are ∼ 1900 and
∼ 325 ppb, respectively. In the analysis of the data, the
baseline (non-plume) mixing ratio was determined by fitting a line through
the average of several data points immediately before the plume transect
began and the average immediately after the transect ended. The fit line was
then subtracted from the data to yield a baseline-corrected plume. This
accounted not only for background concentrations (e.g., 1900 or 325 ppb) but
also any minor baseline drift that may have occurred over the course of the
transect. The quality of the baseline fit was visually confirmed and
corrected if it did not accurately represent the true baseline. For the
plume to be considered a candidate for dual-area analysis, the ratio of
areas of the C2H2 and N2O plumes must be within a factor of 2
of the known tracer flow rates.
Single correlation
In scenarios where the CH4 mixing ratio was highly correlated with only
one of the two tracers, a “single-correlation” analysis was performed, as
shown in Fig. 4. This approach corresponds to that originally used by Lamb et
al. in early demonstrations of the tracer release method (Lamb et al., 1995).
The need to use the single-correlation technique can be the consequence of
several possible measurement conditions: (i) one of the tracers is placed
geographically close to the dominant emitter within the facility (e.g., a
compressor or large fugitive source), (ii) the site is emitting a tracer
species (i.e., C2H2 during certain combustion processes), forcing
the measurement to become single-tracer only, or (iii) the plume transect is
far enough downwind (frequently > 2 km) that one of the tracer
species' mixing ratio is at or below the instrumental detection limit. In
single-correlation cases, correlation analysis is performed for both tracers
but only the well-correlated tracer serves to provide the true CH4
emission rate. For a plume to be a candidate for single-correlation analysis,
the R2 value derived from the linear regression fit of CH4 to one
of the two tracers must be greater than 0.75.
Linear combination of tracer plumes
In certain circumstances, unique tracer placement, road access, and wind
conditions allow for intermediate-distance transects where the CH4
plume profile is not well correlated with either individual tracer but is
well correlated with a linear combination of the tracer plumes, i.e.,
ΔCH4=a⋅ΔN2O+b⋅Δ[C2H2],
where a and b are multiplicative coefficients of the N2O and
C2H2 plumes, respectively. Such an example is shown in Fig. 5.
This scenario is equivalent to performing two independent single-tracer
measurements, where the plumes are overlapping in time. In these cases
facility emission rates can be determined by performing a correlation
analysis of CH4 vs. (a ⋅ Δ[N2O] + b ⋅ Δ[C2H2]) while adjusting the values of a and b
in Eq. (). The a and b values that provide the largest possible
R2 value in the fit are used to determine the CH4 emission rate
associated with each tracer. While the sum of these values serves as the
FLER, the individual emission rates contain
information at sub-facility-level resolution, such as leak or vent
magnitudes associated with condensate tanks, compressors, or dehydrators.
This analysis method has also been applied in cases where equipment not
associated with the G&P (e.g., a natural gas production well) is present
within a facility boundary. In such a case, one tracer is placed at or near
the non-associated equipment while the other is placed near a suspected
emitter that is part of G&P facility. If the plume from the former tracer
is well correlated with the non-associated equipment emission and the plume
from the latter tracer is well correlated with the rest of the CH4 from
the facility of interest, then the facility level emission rate can be
estimated, even when the CH4 from the non-associated equipment is
overlapping with the facility plume.
Implementation of plume analysis
Table 2 summarizes the preference of the four analysis methods, their
acceptance criteria, the number of accepted plumes that were analyzed using
each method, and the measurement variance associated with each plume type.
The determination of the variance for each plume type is discussed in detail
in
the Supplement.
The large number of plumes observed during the measurement campaign allows
for extensive statistical analysis of dual correlation, dual area, and
single-correlation plumes. As is discussed in the Supplement and
the associated Measurements report (Mitchell et al., 2015),
this statistical analysis yields variances for each plume type, the
inverses of which are used as weighting factors for determining the weighted-average
FLER. Not surprisingly, the dual-correlation method exhibits the
lowest variance of all plume types and is therefore the most preferred.
This is likely due to the fact that these plumes correspond to a limit where
full co-dispersion of the tracers has been achieved, i.e., both tracer plumes
are experiencing the same local turbulence by the time they are measured by
the mobile laboratory. In addition, no baseline subtraction is required in
the dual-correlation method, which can be a source of uncertainty depending
upon the signal-to-noise exhibited by the plume. The larger variance of the
dual-area method is likely derived from the lack of co-dispersion of the
tracers. In these scenarios, one tracer concentration may be enhanced
relative to the other due to the fact that each tracer plume is experiencing
different local turbulence en route to the mobile laboratory.
Plume analysis types, preference, criteria, prevalence, and variance.
Analysis type
Preference
Criteria
# of
Variance
plumes
(variance)
Dual correlation
1
250
0.04 (0.2)
R2>0.75: N2O vs. C2H2,
N2O vs. CH4, C2H2 vs. CH4, C2H6
vs. CH4
Tracer ratio error <1.5
E / M ratio error <1.5
Dual area
2/3
441
0.14 (0.37)
R2>0.75: C2H6 vs. CH4
Tracer ratio error <2
E / M ratio error <1.5
Single correlation
3/2
728
0.09/0.22(0.3/0.47)
R2>0.75: C2H6 vs. CH4,Tracer vs. CH4
E / M ratio error <1.5
Linear combination
4
16
–
R2>0.75: C2H6 vs. CH4
In the case of single-correlation plumes, the observed variance is found to
be relatively small when the downwind tracer ratio (determined using
integrated areas) is within a factor of 1.5 of the tracer flow rates
(variance of 0.09 in Table 2). Because this variance is less than that for
dual area (0.09 vs. 0.14), single-correlation analysis is preferred over
dual-area analysis for these plumes. Notably, the variance increases
significantly from 0.09 to 0.22 when including all single-correlation plumes
(i.e., with no tracer ratio filter). When the tracer ratio is more than a
factor of 1.5 different than the tracer flow rates, the dual-area method is
then preferred over single-correlation analysis. This indicates that
although the both tracers are not being used to determine the FLER
associated with that plume, filtering by their ratio can still yield more
precise results. The decision tree employed during the analysis of this
data set is presented in the Supplement.
Ethane-to-methane ratio
Finally, the ratio of ethane to methane in the measured downwind plume can
also serve as an acceptance criterion regardless of plume classification.
The amount of ethane in a natural gas mixture can vary from well to well and
from one gathering facility to another (Kidnay et al., 2011). As
such, the ethane content represents a unique “fingerprint” of a facility,
providing a means to identify whether the CH4 measured in a plume is
coming from the facility of interest. In this study, the ethane-to-methane
ratio (E / M ratio) associated with a given facility was determined in one of
two ways: from partner company GC analysis of the inlet/outlet gas or from
C2H6 vs. CH4 correlation analysis of plumes when the mobile
laboratory was on-site (and thus only observing emissions from the facility).
While GC analysis data are preferred since they provide a completely
independent (and external) check of the methodology, they were not always
available on the date of the measurement. When possible, observed E / M ratios
of plumes obtained when the mobile laboratory was on-site were compared to
the GC data to confirm (or disprove) that the emission composition was in
agreement with the GC data.
Both mobile laboratories measured ethane and methane at a 1 Hz sampling rate
or faster, allowing for an accurate determination of the E / M ratio of
individual plumes. The E / M ratio for every downwind plume obtained in the
campaign (determined using correlation analysis) was measured and compared
to the known ratio from GC analysis (or measured on-site E / M ratio in cases
where the GC data were unreliable). A detailed comparison between the
observed E / M ratio and that from the inlet GC analysis is presented in the
results section. A plume was only accepted for further analysis when the
observed ratio was within a factor of 1.5 of the known value. This criterion
was suspended in cases where the facility itself was actively changing the
ethane content (e.g., from a demethanizer), where the E / M ratio was varying
across the facility, or when the downwind C2H6 mixing ratio was
below the detection sensitivity limit.
Finally, under certain scenarios, a small number of plumes that would be
rejected as described above are manually accepted during analysis. These
exceptions are possible for one of several reasons. One is that the plume
transect is far enough downwind that the tracer or CH4 plume
concentrations are near the detection limit of the onboard instruments.
Under such a scenario the correlation analysis may reveal R2 < 0.75 despite the plume being legitimate. Another possible reason for
manually accepting a plume is when the E / M ratio is variable across the
facility, which is frequently due to the presence of a high emission point source
such as a venting condensate tank. Because condensate tank emissions may
exhibit an E / M ratio larger than that of the remainder of the facility, the
observed downwind ratio may be variable, even on the timescale of a single
plume.
Results
In this section, we present results from a number of case studies that
illustrate the capabilities of the dual-tracer release method.
Gathering facilities
A gathering station serves as a point where multiple natural gas sources
(wells) are combined to produce a high-pressure stream of gas. These
facilities typically include equipment such as inlet separators to remove
liquid phase water and condensate (C5+), when present, and systems for
pipeline maintenance activities (e.g., “pigging”). Compression at these
facilities is accomplished by a series of 1 to 20 individual compressors
powered by electric motors, reciprocating engines, or gas turbines with total
engine powers ranging from 500 to 25 000 HP depending on the inlet gas
pressure and total gas throughput (Mitchell et al., 2015).
Gathering stations also typically contain condensate storage tanks, produced
water storage tanks, and other gas handling equipment including pneumatic
valves (often powered by natural gas) and gas metering systems. If the gas
has a high water content, glycol dehydration systems are also frequently
present to dry the gas (Goetz et al.,
2014; Kidnay et al., 2011).
There are three main sources of continuous emissions from these facilities.
First, compressors can serve as significant sources of CH4 via both
fugitive leaks as well as through seals in the compressor housing. In the
case of wet compressor seals, it should be noted that the primary emission
route is due to absorption of methane into the seal fluid at high pressure,
followed by exposure of the fluid to ambient pressure, where the methane is
routed through a vent to atmosphere (EPA, 2006). Second, because the
natural gas is typically under high pressure, fugitive and vented emissions
may occur at the facility, including from continuous-bleed natural gas
pneumatic devices, dehydration units, and a variety of flanges and valves.
Third, methane slip (i.e., unburned methane in engine exhaust gases) through
on-site combustion sources such as engines and turbines can be a source of
CH4, depending upon a wide variety of combustion characteristics. The
relative importance of this emission source to the FLER is discussed in the
associated Measurements report (Mitchell et al., 2015) and
in previous studies of combustion emissions in natural gas transmission and
storage (Subramanian et al., 2014). Similarly, methane and
other unburned hydrocarbons are present in flare emissions and may vary
greatly depending upon the flare combustion
efficiency (Torres et al., 2012).
Some intermittent methane emission sources may also be found at gathering
facilities, such as intermittent-bleed natural gas-driven pneumatic
controllers, produced water tanks, and condensate tanks. Of particular
importance to the associated Measurements paper (Mitchell
et al., 2015), produced water and condensate tanks may transiently emit CH4,
C2H6, and higher hydrocarbons from thief hatches or other pressure
relief valves attached to the tank. Because of the nature of the liquids
stored in them, i.e., long-chain hydrocarbons, the ethane-to-methane ratio
observed from a condensate tank can be much higher than the natural gas
composition entering or exiting the facility. However, these units may
sometimes also serve as venting release points for equipment on-site, in
which case the E / M ratio will be very similar to that of the inlet stream.
Three exemplary plumes from a gathering station: (a) far-field
plume (1.6 km) showing strong correlation between CH4, C2H6,
N2O, C2H2, CO2, and CO; (b) close plume transect (100 m away) of same facility, showing loss of correlation and isolation of
CO2 and CO combustion products to a section of the facility;
(c) example of a close plume transect (200 m away) showing CO and CO2
correlation with a component of the CH4 trace.
Example of varying E / M ratio during a close transect due to the
presence of a condensate tank battery on-site. Note the ∼ 2×
decrease in the E / M ratio toward the end of the plume.
An example of an emission rate measurement from a compressor station (C station)
is shown in Fig. 6a. Similar to the example plume shown in
Fig. 2, this plume as accepted as dual correlation (R2= 0.998,
tracer ratio error = 1.05, E / M ratio error = 1.4). The average emission
rate from this facility was found to be 43.8 ± 8.4 kg h-1. In this case,
the methane and ethane signals are strongly correlated with both tracers at
a distance of 1600 m downwind of the facility. Note that inclusion of the CO
and CO2 in the analysis indicates that both of these gases are also
being emitted from the facility, likely due to combustion. While this plume
alone can provide an accurate determination of the FLER from the facility,
even more information can be extracted by also investigating transects from
only 100 m away, shown in Fig. 6b (a dual-area plume, with tracer ratio
error = 0.7, E / M ratio error = 1.5). While such a close transect may not
provide as precise a FLER, we see from the figure that the CO and
CO2 signatures are coincident with only a fraction of the methane being
emitted and are not well correlated with it. This indicates that some, but
not all, CH4 emitted at the facility may be associated with combustion.
In this case, the remaining CH4 emission is likely from other
non-combustion sources on-site. At some facilities, such as that shown in
Fig. 6c, CO and CO2 are correlated with a distinct part of the
CH4 plume, indicating the presence of a combustion source that is
emitting CH4 or co-located with one that is and clearly associated
with one section of the facility. Because the goals of the G&P study are
to understand both overall emissions and their origins, this type of
analysis can aid in understanding the relative role of combustion sources
and methane slip in G&P CH4 emissions. In the case of the compressor
station associated with the plume in Fig. 6c, the area of the facility
with CO, CO2, and CH4 emissions is the compressor/engine section,
while the area with no CO / CO2 corresponds to other non-combustion
sources on-site. Thus, Fig. 6 illustrates the important role that the
auxiliary gas measurements (in this case CO and/or CO2) can play in
identifying sources of emissions.
Because they are ubiquitous at both production and gathering facilities, it
is of interest to this study to understand, and quantify when possible, what
fraction of emitted methane is coming from condensate and produced water
tanks. Shown in Fig. 7 is an example of the emission profile observed at a
compressor facility containing a condensate tank, illustrating another
example of the utility of close (< 200 m) transects. In this case,
one tracer (N2O) was placed next to the compressors, while another
(C2H2) was placed near a battery of three condensate tanks. As shown
in the transect trace, both of these sources (compressors and tanks) are
correlated with their respective tracers but have very different E / M
ratios. Here the relative intensities of the CH4 plumes associated with
the different E / M ratios indicate comparable emission rates between the two
sources. As discussed in the associated Measurements
paper (Mitchell et al., 2015), the sub-facility spatial
resolution afforded by tracer release, along with the measurement of
auxiliary species such as ethane, provides the ability to address the
contributions of particular equipment, especially condensate tanks, to
emissions from G&P facilities. Here, for example, analysis using a linear
combination of tracers as described above reveals that the CH4 emission
from the condensate tank represents 50 % of the overall CH4 emission
rate from the facility. The average total emission rate from this facility
was found to be 48 ± 22 kg h-1.
Example of differing CO2 plume profiles as a function of gas
play: (a) emissions from a plant in a coal-gas region, with an amine
scrubbing unit, showing significant CO2 emissions; and (b) emissions
from a gathering facility with no treatment in a shale-gas region.
While not always the case, it is common to find a larger ethane content in
emissions from condensate tanks relative to the inlet gas composition due
to the larger fraction of ethane in the condensate itself. It should be
noted that daily temperature variations (producing “breathing” emissions)
may change the relative vapor pressures of ethane and methane in the
condensate tank, and the filling/emptying schedule of the condensate tank
(producing “working” emissions) may alter condensate composition. Both of
these activities can therefore change the E / M ratio of the tank emissions
over the course of the day.
Amine treatment
The composition of natural gas often depends upon its geologic origin (or
play). To illustrate this effect, we compare emissions from facilities
associated with different gas sources: shale and coal bed
methane (Whiticar, 1994; Kidnay et al., 2011). Shale gas,
tight gas, and conventional gas contain varying amounts of ethane and higher
hydrocarbons, typically with low levels of CO2. Coal bed methane, however, typically contains little ethane and up to 40 %
CO2 (Kidnay et al., 2011). This carbon dioxide is particularly
interesting since in this case it is not an indicator of combustion. Other
combustion sources within the facility can be distinguished by the presence
of CO.
If CO2 is present in high amounts (> 3 %), it must be
removed from the natural gas prior to transmission and storage. It can be
removed from a gas stream by passing the natural gas through a vapor of
monoethanolamine or other related amine compounds. This process is called
“amine treatment” or “amine scrubbing” (Kidnay et al.,
2011; Rochelle, 2009; Bottoms, 1930). The amine binds to the CO2 and is
then regenerated through heating. CO2 is thus evolved from this
process, so the facility's CO2 emissions relative to CH4 will be
higher than would be expected for a direct leak of the untreated gas.
Heating is applied through combustion of excess fuel (natural gas or other
easily available source) so CO2 may sometimes be present along with
small amounts of combustion products such as CO and NOx. Amine
treatment is also used for the removal of hydrogen sulfide (H2S), with
the main difference being that the H2S is highly toxic and must be
captured or combusted.
Figure 8 contrasts emissions from facilities associated with coal bed
methane and shale gas. The facility in Fig. 8a is a coal bed methane
treatment plant without compression. The average emission rate from this
facility was 142 ± 50 kg h-1. The compressor/dehydration facility shown
in Fig. 8b (the same compressor facility discussed above) had four
compressors and is in a shale region with characteristically high ethane
content in the gas. The ethane content of the coal bed methane is observed
at a molar ratio C2H6 / CH4= 0.0215 (Fig. 8a), while the
shale-gas facility emissions have a much higher measured ratio,
C2H6 / CH4= 0.164 (Fig. 8b). The CO2 emissions vary
even more greatly between the facilities, at CO2 / CH4= 165 vs.
CO2 / CH4= 3.3. The molar ratio of CO2 to CH4 in the
former facility's emissions (CO2 / CH4= 165) is 4 orders of
magnitude higher than the operator data for the inlet gas (CO2 / CH4
= 0.106). For Fig. 8a, at the distances sampled no other significant
combustion products (such as CO) were observed, indicating that the primary
source of CO2 is the amine treatment process. This information,
along with the observed high degree of correlation between CO2 and
CH4 at intermediate distances (∼ 500 m), suggests that
the primary CH4 emission source is located within or near the amine
treatment area of the facility.
Natural gas processing
Natural gas processing plants are large, complex facilities that remove
unwanted compounds in the incoming gas stock (e.g., H2S, CO2,
H2O) and separate other high-value compounds (i.e., natural gas liquids,
as discussed below) from the gas to produce pipeline-quality natural gas.
Physically, processing plants often serve as the nexus between the gathering
networks in the area and a transmission system working to serve longer-range
transport. They are typically characterized by capacity throughputs of
3–1500 million standard cubic feet per day (MMscfd; equivalent to 2400–1 200 000 kg h-1). The types of equipment
and the processes that are undertaken at a gas-processing plant depend on
the composition of the gas in the region. Many plants utilize multiple
processing “trains” to enable flexible operation. The equipment and steps
in each train can vary depending again on the region and the engineering
decisions made by the operator of the plant (Kidnay et al., 2011).
It should also be noted that not all natural gas in the US supply chain is
processed. Rather, in cases where natural gas composition does not contain
substantial levels of natural gas liquids or H2S / CO2 (i.e., is dry
and sweet), the natural gas flows directly from gathering facilities into
transmission pipelines (and sometimes directly into distribution networks).
The initial process that is typically found at a gas-processing plant
involves a continuation of the treatment types found in the gathering system
of the region. At some facilities, the initial product will be a first cut
at collecting natural gas condensate, which is typically comprised of
functionalized hydrocarbons above C5, using an inlet separator (when they
have not been collected further upstream in the gathering network). Water
may also be removed using glycol dehydration. Other trace contaminants are
often filtered using a series of molecular sieve apparatus that are
staggered for effective continuous regeneration. As discussed below, natural
gas liquids are removed from the gas stream using either a cryogenic
separation or separation based on solubility in lean oil (Kidnay
et al., 2011). Additional details of this class of compounds and specific
equipment used are discussed in the next section.
Due to the nature of the various processing steps and types of equipment
found at processing plants, as well as the somewhat larger geographic scale
they typically occupy, there are typically multiple methane emission points
with various co-emitted compounds. On the surface, this type of source is a
direct challenge to the tracer release methodology given the constraint for
the controlled tracer release to be as close to the emission source as
possible. The following examples and discussion describe how these types of
facility are quantified using the dual-tracer methodology as well as using
the nature of the co-emitted compounds to deduce the dominant emission
sources.
In the left-hand panel, the time series for methane, ethane,
nitrous oxide, and acetylene are depicted for two transects, (a) and (b). In
the right-hand panel, the geographic location is portrayed for the
processing plant (grey) and the two transects (a) and (b). See text for
additional discussion.
The geographic scale of processing plants presents a challenge to the dual-tracer flux ratio quantification given the constraints of wind direction and
roadway access. Figure 9 depicts a pair of transects from a processing
plant. The average emission rate measured at this plant was found to be 128 ± 66 kg h-1. Each transect was collected with the mobile lab maneuvering
from north to south. This is depicted by the rainbow bar in each of the two split
time series (a) and (b) in the left hand panel and portrayed on the right-hand panel with the relative distance (north vs. east). In the case where
the transect was captured at the facility fence line (a), we see relatively
high spikes in plume mixing ratios with three different quantifiable E / M
ratios. Note that the tracer release locations were relatively close to one
another and this is reflected in the spatial coherence in both of the
transects.
In the case of the more distant (∼ 1.2 km) transect, the
mixing ratios of ethane and methane are significantly less spiked. Careful
analysis of the time and space dependence of the E / M ratio suggests that
even at this distance the ratio in the northern sector of the facility is
different than that in middle and southern sections. This observation is
corroborated anecdotally by the physical location of the liquids storage and
natural gas transmission hardware on-site. In this facility the recompression of
pipeline-grade natural gas takes place in the southern third of the
facility. This corresponds to the lowest E / M ratio (red-purple in the time
series) but is a significant source of CH4 emissions (∼ 50 %) from the facility. The liquids storage and handling takes place at
the northern section of the facility. The effective leak rate of methane is
less than in other sections of the facility because the methane is at
residual levels in the liquids headspace. The E / M ratio in the green and
yellow section of the time series is greater because this is where the NGL
stock is being processed.
To quantify the FLER from processing facilities, frequently the dual-area
analysis method is used. In the case of the close transects, the measured
methane emission rates often exhibit substantial variance. The average of
multiple close transects typically was found to be comparable to values
determined by more distant, better-mixed plume intercepts, when such a
comparison was available.
Natural gas liquids and condensates
NGL is an umbrella term (EIA, 2013) for the many
different chemicals and blends extracted in the liquid form from natural
gas. Depending on the equipment available and the demand for the various
products, the amount of processing of natural gas can vary greatly. At the
lower end of the spectrum, the gas may undergo dehydration and just enough
removal of C2+ to meet pipeline specifications, such that no liquids
condense at pipeline pressures. Removal of other impurities such as CO2
and H2S may also be required to meet pipeline specifications. At the
highest end of the processing spectrum, cryogenic distillation will be
employed to sequentially extract methane (demethanizer), ethane
(deethanizer), propane, iso- and n-butane, and higher hydrocarbons. This
processing can occur at a single facility or can be performed in several
steps between different facilities. The net result is to separate the
methane (and/or ethane) from other condensable compounds that may still be
present in the feed stock after the various upstream treatments. The liquid
product at this stage is referred to as “x” or “y” grade liquid depending on
the cut temperature and ethane content in the liquid. In some of the
processing plants in this study, this liquid stream is stored in this state
and shipped off-site via an NGL pipeline or tanker truck. In other
facilities studied, the liquid is further fractionated, sequentially
removing ethane, then propane, then butane (Kidnay et al., 2011). Because of the low methane content within the liquid, this further
processing of the NGL is not expected to significantly contribute to the
FLER but may play a role in the E / M ratio that is observed downwind.
Many of the facilities visited in this study were in so-called “ethane
rejection” mode, meaning that distillation towers were operated at lower
liquids recovery levels and purified ethane is treated as a byproduct of the
C3+ extraction. As a byproduct, it frequently was re-injected into the
natural gas stream. This occurs when there is less demand for purified
ethane as a feedstock for ethylene, a process that occurs at an extremely
limited number of locations in the US.
As in the case of identifying condensate tank emissions, the E / M ratio can
inform the attribution of a methane emission source to individual pieces of
NGL equipment. A striking example is shown in Fig. 10. This facility has two
compressors, dehydrators, condensate tanks, and processing equipment. The
measured CH4 emission rate from this facility is 58 ± 22 kg h-1. The
nitrous oxide tracer (green marker) was placed near the condensate tanks and
the acetylene tracer (blue marker) near the compressors. Northeast of the
acetylene tracer, above-ground piping marks the facility's inlet and outlet
(natural gas) as well as a liquids pipeline carrying a mixture of ethane and
propane produced at the facility. The E / M ratio for the mixed facility plume
was 0.0576, while the ratio for the liquids pipeline and inlet/outlet region
was 14.58, i.e., nearly entirely ethane. Therefore, this transect indicates
that the pipeline is not a significant source of CH4
emissions.
Measured E / M ratios as a function of gas type at gathering and processing
facilities. Minimum, median, and maximum average measured ratios are noted. Offshore gas is not
included here due to the small number of offshore facilities measured.
Gas type
Measured E / M ratio
min
median
max
count
Coal bed methane
0.00
0.014
0.045
8
Coal bed methane
0.0057
0.018
0.031
4
and conventional
Shale
0.0055
0.051
0.24
64
Conventional
0.012
0.068
0.22
37
Downwind plume transect showing mixing ratio as a function of
time (top) and a map (bottom). Tracer release locations are shown as a green
triangle (nitrous oxide) and a blue square (acetylene). The plume transect
is colored by methane mixing ratio (black to yellow). Ethane mixing ratio is
also shown with a geographic offset. Wind vectors (pink, red, and yellow)
point into the wind.
Comparison between measured ethane/methane ratio and operator
data on gas composition. Error bars correspond to the 95 % confidence
limits from the replicate experimental plumes. Points are also colored by
the type of gas at each site. A line to guide the eye is drawn at a 1 : 1
correspondence between measured and operator data.
Comparison of C2 content with operator data
In this study, the E / M ratio serves several purposes: (i) confirmation that a
plume is from a target facility, (ii) elimination of plumes from neighboring
facilities or biogenic sources, and (iii) distinguishing between different
emission sources within a given facility. The quantification of a facility's
methane emissions leverages (i) and (ii) above. Figure 11 shows a comparison
between the measured E / M ratios at each facility and the operator-provided
data on gas composition. Agreement is good overall, with a few outliers.
Also shown in the figure are 95 % confidence limits on the measured E / M
ratios. Large error bars in the facility average for E / M ratios are usually
due to variations in the emission composition, since the error for any
individual ratio measurement is low. The operator gas composition
information was not always measured on the same day as the field testing.
For gathering facilities, gas composition is periodically measured by gas
sampling and subsequent third-party analysis. For processing plants, gas
composition data are typically acquired in real time at multiple locations at
the facility. In either case, the gas composition exiting the gathering
facility or processing plant may not always reflect the gas composition of
the emission sources. This can be due to the E / M ratio changing as the gas
moves through the facility or from emissions from condensate/produced water
tanks. This variety of equipment and processes at gathering facilities and
processing plants explains much of the discrepancy between measured and
operator E / M ratios compared to the transmission and storage study,
where the composition of the gas does not vary during
handling (Subramanian et al., 2014; Yacovitch et al., 2014). Table 3
outlines the minimum, median, and maximum facility average E / M ratios divided
by primary gas type. It should be noted that the classification by gas type
is not rigid. That is, there may be multiple gas types other than the
primary present at these facilities. The points in Fig. 11 are colored
based on this gas classification. As noted above, coal bed methane
facilities typically have the lowest E / M ratios. Conventional facilities sit
somewhere in the middle, with the shale-gas facilities split into several
clusters. The shale gas is scattered about the plot, with some clustering
associated with various geographic basins. The three main shale clusters
observed in Fig. 11 (green points) correspond loosely to: the Denver
(Denver–Julesburg), Permian (Eagle Ford and Delaware), and Appalachian
basins (∼ 12–23 %); the Anadarko (Mississippian Lime gas
play), Uinta (Natural Buttes), and Piceance basins (∼ 4–6 %); and the Arkoma basin (∼ 1 %). Other shale basins
were also visited but the number of facilities for each of these basins is
low.