Compositions and Greenhouse Gas Emission Factors of Flared and Vented Gas in the Western Canadian Sedimentary Basin

Matthew R. Johnson* and Adam R. Coderre Energy & Emissions Research Laboratory, Mechanical & Aerospace Engineering, Carleton University, Ottawa, ON, Canada, K1S 5B6

The article should be cited as: M.R. Johnson and A.R. Coderre (2012) Compositions and Greenhouse Gas Emission Factors of Flared and Vented Gas in the Western Canadian Sedimentary Basin, Journal of the Air & Waste Management Association, 62(9):992-1002 (doi: 10.1080/10962247.2012.676954).

* Corresponding author: Email: [email protected]; Office: (613) 520 2600 ext. 4039; Fax: (613) 520 5715

ABSTRACT A significant obstacle in evaluating mitigation strategies for flaring and venting in the upstream oil and gas industry is the lack of publicly available data on the chemical composition of the gas. This information is required to determine the economic value of the gas, infrastructure and processing requirements, and potential emissions or emissions credits, all of which have significant impact on the economics of such strategies. This paper describes a method for estimating the composition of solution gas being flared and vented at individual facilities, and presents results derived for Alberta, Canada, which sits at the heart of the Western Canadian Sedimentary Basin. Using large amounts of raw data obtained through the Alberta Energy Resources Conservation Board, a relational database was created and specialized queries were developed to link production stream data, raw gas samples, and geography to create production-linked gas composition profiles for approximately half of the currently active facilities. These were used to create composition maps for the entire region, to which the remaining facilities with unknown compositions were geographically linked. The derived data were used to compute a range of solution gas composition profiles and greenhouse gas emission factors, providing new insight into flaring and venting in the region and enabling informed analysis of future management and mitigation strategies.

IMPLICATIONS Accurate and transparent determination of environmental impacts of flaring and venting of gas associated with oil production, and potential benefits of mitigation, are severely hampered by the lack of publically available gas composition data. In jurisdictions within the Western Canadian Sedimentary Basin, frameworks exist for regulating and trading carbon offset credits but current potential for mitigation is limited by a lack of standardized methods for calculating CO2 equivalent emissions. The composition and emission factor data derived in this paper will be useful to industry, regulators, policy researchers, and entrepreneurs seeking statistically significant and openly available data necessary to manage and mitigate upstream flaring and venting activity and estimate greenhouse gas impacts.

When these liquids are produced and brought to INTRODUCTION the surface, the pressure acting on them is In the energy and petrochemical industries, reduced from formation to atmospheric, causing excess or unwanted flammable gases are often these dissolved gases to come out of solution. disposed of by flaring or venting. Flaring is the These evolved gases are commonly referred to process of combusting the gases in an open- as solution gas. The term associated gas, is atmosphere flame, and provides a means of perhaps even more commonly used, although in disposing of flammable gases in a cost-effective general associated gas is understood to refer to manner. If stable combustion of surplus gas is the combination of solution gas and gas that not possible, for instance if the flow rates are too exists separate from the oil at reservoir low or too intermittent, or if the heating value of conditions. In the upstream oil and gas industry, the gas is too low to sustain combustion, or if the solution gas is the source for the majority of all gases are deemed uneconomic to recover and flaring and venting activity that takes place. regulations permit, the gases are instead vented, From an air emissions management perspective, meaning they are simply released to atmosphere. the practice of flaring and venting is a concern The U.S. Energy Information Administration, due to the scale at which it takes place. In based on reports from individual countries, addition to carbon dioxide (CO2), an important estimates that global flaring and venting totalled greenhouse gas, flares can produce airborne 122 billion m3 in 20081. By contrast, pollutants such as particulate matter in the form examination of visible light images captured by of soot 4,5, unburned fuel and carbon monoxide orbiting satellite suggests that global flaring 6,7 (especially if the heating value of the flare gas alone exceeds 139 billion m3 annually, and that is low 8), and potentially other by-products of these volumes have been relatively stable over incomplete combustion 9. When the raw flare 2 the past fifteen years . Vented gas is not as gas contains hydrogen sulphide (H2S), the major readily detected and to the authors’ knowledge, pollutant sulphur dioxide (SO2) is also produced. accurate estimates of global venting volumes do Although direct venting of gas precludes not exist. However, for the case of Alberta, combustion related emissions, from a Canada, a mature oil and gas producing region greenhouse gas (GHG) perspective, venting of with extensive pipeline infrastructure, a recent high-methane content gas associated with analysis of production data shows venting petroleum production is even worse. This is volumes similar to flared volumes as well as a because methane (CH4) has a 100-year global trend toward proportionally greater amounts of warming potential that on a mass basis is 3 10 venting as more heavier oils are produced . twenty-five times more potent than CO2 . The majority of global flaring and venting Predicting impacts of flaring and venting on a occurs during upstream production of oil and gas broader scale requires knowledge of gas resources. The production of conventional oil is compositions being flared and vented. As well, nearly always accompanied by the production of the viability of any potential mitigation flammable gases, even when no gas is initially strategies such as collection of gas into pipelines present in the reservoir. This is because the or the use of the gas to generate heat and hydrocarbons are contained in sub-surface electricity, are highly dependent on chemical geological formations under high pressure, composition of the gas, especially in terms of which allows for volatile chemical species to energy and H2S content. The lack of statistically equilibrate and dissolve in the formation liquids. significant, published data on compositions of

2 flared and vented associated gas is thus a production and 76% of gross natural gas significant impediment to engineering analysis production 15. of impacts and mitigation options. Successful Upstream Flaring and Venting in Alberta, regulation and trading of carbon offset credits Canada. Much of the conventional oil in from flaring and venting mitigation projects are Alberta is produced from smaller-volume wells further hampered by a lack of consistently connected to “battery” sites, i.e. surface facilities applied and transparently derived greenhouse in which reservoir fluids, including solution gas, gas (GHG) emission factors. The objective of are separated and measured. Oil and bitumen this paper is to address this gap in knowledge batteries in Alberta produced nearly 15 billion through comprehensive analysis of available m3 of solution gas in 2008 16, the latest year for production and reservoir data for a significant which data were available. The large majority petroleum production region of the world. The (95.3%) was conserved, meaning that it was derived results are subsequently used to estimate either used onsite as fuel or directed into natural a range of gas composition-based emission gas pipelines for processing and sale. The factors to predict greenhouse gas emissions from remainder was disposed of by flaring or venting. flaring and venting activities. Although 4.7% is a relatively small fraction of the total amount of solution gas produced, it still Petroleum Production in the Western represents a significant volume of gas which Canadian Sedimentary Basin totalled 687 million m3 in 2008 3. Upstream The Western Canadian Sedimentary Basin flaring and venting from all sources in Alberta (WCSB) is a vast geological formation of 3 16 totalled 1.11 billion m in 2008 , or sedimentary rock that spans several western 3 approximately 0.9% of the 122 billion m global Canadian Provinces, bordered by the Rocky flaring and venting estimate from the U.S. Mountains to the west and the Canadian Shield 1 Energy Information Administration . to the east. The bulk of Canada’s oil and gas resources lie within this basin, including the vast The body that regulates the upstream oil and gas quantities of oil sands that place Canada’s industry in Alberta is the Energy Resources proved oil reserves third highest in the world, Conservation Board (ERCB). ERCB’s Directive behind Saudi Arabia and Venezuela 12. More 60 contains guidelines for the decision-making than 97% of Canada’s proven reserves are in the process pertaining to solution gas conservation form of oil sands deposits, while conventional options that industry operators are required to 17 reserves in the WCSB account for nearly 2% 13. follow . Whereas ERCB Directive 007 However, conventional sources in the WCSB mandates that operators submit monthly account for a much greater fraction of current production reports through the Petroleum 18 production. In 2009, roughly 3% of global oil Registry of Alberta (PRA) , Directive 60 further production was sourced from the WCSB, of specifies that “volumes of gas greater than or 3 3 which approximately half (55%) originated from equal to 0.1·10 m /month (adjusted to 101.325 oil sands deposits 14,12. The province of Alberta kPa(a) and 15°C) that is flared, incinerated, or 17 sits at the heart of the WCSB, and is a mature vented” are to be included . However, the and very active oil and gas production region. composition of the gas being flared or vented is Alberta is by a wide margin the largest producer not included in these reports. of oil in gas in Canada, accounting for roughly 68% of Canada’s 2008 crude oil and equivalent

Origins and Production of Solution Gas pipelines and processing facilities before A discussion of solution gas composition begins entering a sales gas line. It should be noted that with some background and basic hydrocarbon in practice, several different H2S content reservoir terminology. According to the thresholds for defining sour gas are also generally accepted organic theory, hydrocarbons common. For example, the Alberta Oil and Gas 21 17 were formed when sediments including organic Conservation Act and Directive 60 define matter were buried by geological shifts, and sour gas simply as gas “containing” H2S. Other subjected to intense temperatures and pressures ERCB guidelines tend to differentiate sour sites over periods of geological time 19. These based on potential release rates of sulphur, rather combined factors converted the organic matter to than by raw volume concentrations of H2S in the 22,23 the fluids found in reservoirs today, and gas stream . converted the sand, mud, and silt sediments to Therefore, to evaluate the economics of any rock. Hydrocarbon fluids, being less dense than potential mitigation strategies and to determine water, were displaced upwards through porous the GHG contributions of solution gas flaring and permeable rock, until they either breached and venting, the composition of the gas must the surface or were trapped by an impermeable first be determined. This is problematic in that layer of rock (called a cap rock) that prevented the composition often goes unmeasured, further upward migration. However, the pores particularly at smaller production facilities. This in the rock are small enough that surface paper presents a strategy for assigning estimated wettability and capillary forces prevent complete solution gas compositions to production segregation of the fluids. Thus, hydrocarbon facilities in the WCSB. The derived results were reservoirs consist of porous and permeable rock, used to determine volume- and site-weighted the pore space of which is filled with mixtures of solution gas composition ranges, and to calculate water, oil, and gas phases, partially segregated to GHG emission factors for flaring and venting form a gradient of fluid saturations (i.e. the activities under a range of scenarios. In fraction of pore space filled with each fluid addition, separate maps for flared and vented phase) from primarily gas just underneath the solution gas in the Province of Alberta were cap stone, to a primarily oil further down, to developed to assess the geographic distribution primarily water toward the bottom. of gas compositions within the Province. These The composition of solution gas can vary new data enable proper estimation of considerably, comprising differing mixtures of environmental impacts and to support light hydrocarbon species (primarily alkanes quantitative evaluation of mitigation strategies such as methane), non-flammable gases such as for upstream flaring and venting activities in the nitrogen and carbon dioxide, and toxic WCSB. Finally, the methodology developed herein could be usefully applied to other mature impurities such as hydrogen sulphide (H2S). oil and gas producing regions of the world. H2S content has a particularly significant impact on the economics of flaring and venting mitigation. As specified in the Alberta Pipeline METHODOLOGY Act20, gas with more than 10 mol/kmol (i.e. 1% Figure 1 defines key terms used in oil production by volume) H S is designated “sour” (as 2 at batteries in Alberta. An oil field, or simply a opposed to “sweet”). Sour gas is handled and field, refers to the surface area above an regulated separately from sweet gas and has underground hydrocarbon reservoir. A pool different infrastructure requirements, meaning refers to the hydrocarbon reserve itself, whether that it must be directed to specialized sour gas 4 a geological pool, which refers to an actual work. Also obtained through the ERCB were a underground geological hydrocarbon-containing large number (60,000+) of gas samples from formation as discussed above, or a comingled wells attached to non-confidential pools, which pool (sometimes also called an administrative contained molar fractions of 13 chemical groups pool) which describes some combination of including hydrocarbons by carbon content (C1, geological pools that are tapped by an individual C2, C3, IC4, NC4, C5, C6, C7+), combustible well. The term ‘administrative pool’ is primarily non-hydrocarbon species (H2, H2S), as well as used for production accounting purposes. non-combustibles (He, N2, CO2). Similarly,

these data are a raw form of the “Individual Well Gas Analysis Data” files available for public purchase24. These gas analyses originate from industry supplied reports to the ERCB, and although they do not contain details of the specific analysis procedure used, it is understood that they are almost exclusively obtained using gas chromatography with extracted samples, and are reported with a mole fraction precision of 0.0001 as per ERCB Directive 01722. Given the reported precision of the gas samples, overall uncertainties in the present analysis will be dominated by site to site variability which is presented in terms of percentile limits and considered in more detail in the results section. Finally, the ERCB provided production stream data for nearly 9,000 non-confidential batteries, Figure 1: An illustration of hydrocarbon reservoir which link those batteries to the reservoirs from terminology which they produce via numeric field and pool identification codes, including the proportions of In the simplistic representation seen in Figure 1, production attributed to each pool for multi-well a battery is present in a field, and produces from batteries (see Figure 1). These linkages are three wells. Well 1 and Well 2 produce from continually updated as production patterns Pool 1 and Pool 2, respectively, which are each change, and the data considered in the present distinct geological pools. Well 3 produces from analysis reflect linkages in place in June 2008. both Pool 1 and Pool 2; this combination would All data were merged into a large relational then be assigned a separate administrative pool database created to analyze results using scripted code. queries. Industry-reported monthly production data from To meet the key objectives of estimating battery- more than 18,000 oil and bitumen batteries in specific and volume-weighted average the Province of Alberta, spanning the years compositions of gas flared and/or vented in the 2002-2008, were obtained in collaboration with province, several pieces of information needed the ERCB. These data are a raw form of the to be connected. Since the available gas analysis ERCB ST-60 series reports available for public 24 data contained only limited location information purchase and are the basis of the volumetric (i.e. each gas analysis identified a pool code and production and battery location data used in this sometimes a field code, but was otherwise not

5 attached to a specific location), the most direct While this would be a relevant source of way to link a gas sample to a specific battery uncertainty in attempting to assign gas (and its known geographic location) was through compositions to individual batteries, the data the production stream data. These production further suggest that uncertainties in C1 mole stream data contained a numeric identification fraction of less than 0.05 would be typical. code for the field in which each battery was To further verify the validity of this approach, located (rather than specific location information statistical analysis using Levene’s test for such as latitude and longitude) and one or more equality of variances was completed to compare code numbers for the pool(s) (i.e. hydrocarbon the variations among gas samples from a single reservoir(s)) from which it produced. Coupled pool with variations among all gas samples. with battery specific production data (which Calculations were performed for each of the included location data and monthly volumes of three pools with the most available data (i.e. two gas flared and/or vented), composition of gases geological pools with 323 and 337 available gas being flared and/or vented could be determined samples, and one administrative pool with 2529 and directly linked with gas volumes. available gas samples), and results easily Preliminary analysis revealed that of the showed that the differences among the variances 60,000+ gas samples available, only a small were statistically significant. Thus, multiple subset could be matched for both the field and samples within a common pool had statistically pool codes associated with any particular less variation than the set of all gas samples, and battery. However, given the geological time it was reasonable to combine them for the scales the fluids had to equilibrate within a pool, purpose of the aggregate analysis. it was deemed reasonable to assume that Under this assumption, two batteries located in solution gas compositions (as opposed to liquid different fields but producing from the same compositions) would not vary significantly pool would be expected to have similar solution within a single geological pool, and available gas compositions, and production stream data gas samples for a given pool could be averaged. could then be linked to gas samples by matching Indeed, for geological pools with multiple only the pool codes within the database. samples, half had standard deviations in C1 Because most pools had multiple available concentration of less than 5.2% of the mean corresponding gas samples, the impacts of any value within the pool, and 75% of the pools had solution gas variability within the pools was standard deviations in C1 of less than 8.3% of further minimized in the aggregate analysis. the mean. Similarly, half of all administrative/comingled pools with multiple Of the 60,000+ gas samples obtained through samples had standard deviations in C1 the ERCB, approximately 8500 distinct pools concentration of less than 5.6% of the mean and (whether geological or 75% had standard deviations less than 8.6% of comingled/administrative) were represented, the mean. This compares to a standard deviation indicating that the gas from many pools had of C1 concentration of 9.7% of the mean for all been sampled multiple times. With respect to available gas samples, and 10% among the comingled pools, using available data reported means for each pool. Thus, although the limits to ERCB, there is unfortunately no satisfactory of available data necessitated neglecting spatial method to determine the proportions produced and temporal variations in gas composition from each geological pool within a comingled within a pool, the analysis suggests that this is a pool, or even whether those proportions would reasonable assumption for an aggregate analysis. remain constant over time. Gas compositions

6 from comingled pools were therefore similarly would expect to find the vented gas had lower assigned the arithmetic average of all associated average H2S concentrations relative to flared composition samples. Once the compositions of gas. The resulting maps for C1 (methane) individual geological and concentration are shown in Figure 2 for (a) comingled/administrative pools were identified, exclusively-flaring batteries and for (b) batteries solution gas composition profiles were then reporting any amount of venting. assigned to individual batteries using available The data shown in Figure 2 were then smoothed production stream information, weighted by the and extended into grid elements without fraction of production attributed to each pool. assigned compositions through the application of This method allowed composition profiles to be spatial low-pass filters. Digital low-pass filters directly assigned to roughly 6000 separate provide a method of smoothing that reduces batteries. noise in an array of data while largely For the remaining facilities, in the absence of maintaining its integrity 25. Such tools are additional gas samples, compositions were commonly used in digital image processing to estimated based on geographic location. Gas remove noise, and work by assigning a value to composition maps were created by first each individual element (or pixel) based on the overlaying a spatial grid on the Province with values of the surrounding elements 25. The elements that measured 0.15° latitude by 0.2° number and position of the surrounding elements longitude such that they were approximately to consider are defined by the filter kernel, with square over Alberta’s latitudes. The average larger kernels leading to greater degrees of composition of each element was determined as smoothing, and hence loss of high-frequency the arithmetic mean of the available composition data. Different filter types are primarily profiles from any facilities within that element. identified by the operation used to assign a value The grid size was chosen to be much smaller to a cell. Common examples include mean and than the typical pool dimension, and as small as median filters, which replace the value of the possible while still enabling the available data to cell being filtered with either the mean or be extended to regions without available median value of all elements within the kernel measurements using filtering techniques as (including itself). The use of mean filtering was described below. Based on analysis of the chosen since it offers low-pass smoothing while distances between batteries connected to the preserving the constraint that the component same pool, the median and average pool sizes fractions must sum up to unity. Although mean (i.e. the horizontal dimension of the pool) were filtering is commonly used when differences in ~90 and ~145 km or ~4-7 times larger than the neighboring cells are due to random noise rather grid element size. than from separate sources (as is nominally the case here, where different batteries would be Since regulatory distinction is made between producing from different well(s) that may be flared and vented gas, separate composition drawing from different proportions of pools), for maps were generated using data for facilities that the grid scale at which the filter was applied have reported venting activity and facilities that (which is ~4-7 times smaller than the relevant have not (i.e. those that exclusively flare). This scale of a typical pool as noted above), distinction was made under the assumption that compositions variations can be considered average compositions of gas vented and flared random for the purposes of interpolation. could be expected to be different. For example, Potential effects of this averaging procedure are because of the extreme toxicity of H2S, one further considered in the results section below.

Figure 2: Map showing binned C1 concentrations at (a) exclusively-flaring batteries and (b) batteries that reported any amount of venting.

When implementing the filtering operation for RESULTS interpolation, a kernel size of 3x3 was chosen, Based on the approaches outlined above, the such that only the values held by each element’s solution gas composition of each oil or bitumen immediate neighbours were considered. Two battery in Alberta active in 2008 was determined sequential passes were made over the entire either by direct linkage to gas analyses for pools Province to fully encompass the necessary tied to that battery, or by geographic proximity interpolation area. Each battery with an to other batteries that were themselves linkable unmeasured composition was then assigned the to measured pool composition data and appropriate average composition of the grid segregated by flaring or venting activity. Of all element in which it was located, and using the batteries active in 2008, slightly more than different source data depending on whether that half were directly linkable to pool gas samples. battery predominantly flared or vented. The Considering only the subset of batteries that resulting spatial concentration profiles are reported flaring and venting in 2008, again presented with the results. slightly more than half these (representing slightly less than half of the total volume of gas flared or vented) were directly relatable to

8 measured pool composition data. The proportions of flared and vented volumes directly linkable to pool composition data were less even. Roughly two-thirds of gas flared in 2008 could be linked with pool gas samples, compared to only a quarter of the vented gas. This difference is considered further below. Figure 3 shows mean gas composition profiles at sites with linked samples compared with sites where composition data was interpolated from maps, with error bars to represent the 10th and 90th percentile concentration values for each component gas species. The inset graph shows histograms of H2S concentrations for the two categories. Overall, the solution gas Figure 3: Mean composition profiles for data compositions being flared or vented are heavily linked to samples and interpolated (geographically dominated by methane (note the broken vertical linked) from maps. Note the broken vertical axis to permit plotting of methane concentrations axis on the figure necessary to plot mole alongside concentrations of other species. Error fractions of methane alongside mole fractions of bars represent 10th and 90th percentile values for other species). Although subsequent figures each species. Inset graph shows histograms of H2S reveal noticeable variability among sites and concentration for both groups. th th across regions of the Province, the 10 and 90 percentile limits in Figure 3 suggest the On average, the interpolated profiles shown in variability is confined within a reasonably Figure 3 closely match the directly linked narrow range of the mean compositions (mole profiles, with mean mole fraction deviations of fraction variation of <±0.067 for C1 and less than 0.01 for most species. The interpolated <±0.024 for all other species). Table 1 provides samples do show a slight (+0.027 mole fraction) a detailed statistical summary of the composition shift toward greater C1 concentrations compared data shown in Figure 3, aggregated from all to the directly linked samples, which is active batteries in the Province that reported consistent with the larger proportion of vented flaring and/or venting between 2002 and 2008. gas represented by this category, predominantly These data are further segregated to calculate from heavy oil production. On the other hand, separate composition profiles for batteries that the inset H S histogram in Figure 3 reveals a exclusively flared and batteries that reported any 2 th slight bias towards higher H S contents within amount of venting. For each case the mean, 10 2 the 0-1% mole fraction range at sites with percentile, and 90th percentile component compositions assigned via interpolation, even fractions and gross heating values (GHV) are though the larger full figure shows a negligibly shown. small decrease in the overall mean H2S concentration at these same sites. This is most likely an effect of spatially smoothing the

inherently high-frequency H2S data (i.e. the raw

H2S data in particular show sharp geographic variations). Although this effect appears to be

9 limited to values below the sour threshold of costs (since sour gas processing equipment is 10 mol/kmol20, this does indicate a potential bias typically more expensive than sweet). As is toward falsely labelling some interpolated sites stressed throughout this paper, it is therefore as sour. However, within the context of the crucial that non-aggregate (i.e. site-by-site) overall mean H2S concentration remaining evaluation of mitigation options be informed by constant after interpolation, this effect can be accurate site-specific measurements of solution considered conservative both from a health gas composition. perspective and in terms of potential mitigation

Table 1: Summary of composition profiles as assigned to oil and bitumen batteries in Alberta. Composition values in mole fractions, Gross Heating Value (GHV) and Lower Heating Value (LHV) in MJ/m3 of solution gasa. All Batteries Venting Batteriesb Non-Venting Batteriesc th th th th th th Mean 10 90 Mean 10 90 Mean 10 90

H2 0.0001 0.0000 0.0003 0.0001 0.0000 0.0002 0.0002 0.0000 0.0003 He 0.0006 0.0002 0.0008 0.0006 0.0003 0.0009 0.0005 0.0002 0.0009

N2 0.0335 0.0162 0.0510 0.0354 0.0190 0.0516 0.0311 0.0131 0.0496

CO2 0.0141 0.0055 0.0262 0.0126 0.0055 0.0232 0.0181 0.0068 0.0314

H2S 0.0033 0.0000 0.0088 0.0022 0.0000 0.0060 0.0063 0.0000 0.0166 C1 0.8579 0.7862 0.9201 0.8672 0.7913 0.9215 0.8351 0.7756 0.8910 C2 0.0475 0.0217 0.0706 0.0433 0.0214 0.0687 0.0564 0.0378 0.0754 C3 0.0239 0.0076 0.0399 0.0215 0.0074 0.0380 0.0291 0.0160 0.0433 IC4 0.0042 0.0017 0.0069 0.0038 0.0016 0.0062 0.0051 0.0031 0.0079 NC4 0.0068 0.0019 0.0116 0.0061 0.0019 0.0109 0.0085 0.0043 0.0127 C5 0.0045 0.0017 0.0075 0.0041 0.0017 0.0074 0.0053 0.0029 0.0076 C6 0.0016 0.0006 0.0026 0.0015 0.0006 0.0025 0.0020 0.0011 0.0028 C7P 0.0019 0.0006 0.0032 0.0017 0.0006 0.0029 0.0023 0.0012 0.0035 GHV 38.236 36.890 39.640 37.981 36.863 39.465 38.747 37.648 39.962 LHV 34.567 33.359 35.746 34.337 33.335 35.620 35.036 34.064 36.132 a Heating values calculated at a pressure of 101.325 kPa and temperature of 15°C. b Venting batteries include all batteries that report any amount of venting (i.e. batteries that vent exclusively as well as batteries reporting both flaring and venting) c Non-venting batteries reported flaring exclusively

The maps shown in Figures 4–6 represent the might logically be different from gas that is smoothed geographical distribution of C1 flared exclusively, especially with respect to concentration, H2S concentration, and the gross content of toxic H2S. heating value of solution gas throughout the Several broad trends are apparent from these Province of Alberta, segregated by batteries that distributions. Methane concentrations, seen in flared exclusively versus those that reported any Figure 4, are generally higher near the city of amount of gas venting activity. As noted above, Lloydminster, where predominantly heavier oils this distinction was made since it is reasonable are produced, and lower in the northwest of the to assume that gas that is vented as well as flared

Province. These differences also correspond flaring sites than venting ones, this implies with the relative amounts of flaring and venting proportionally fewer measurement-linked gas in these areas, where much greater proportions compositions in the Lloydminster region, and are venting are correlated with the heavier oil the results for this region therefore rely production in the Lloydminster region 3. Given particularly heavily on geographical linking. the greater proportion of samples linked to

Figure 4: Map showing smoothed C1 concentrations at (a) exclusively-flaring batteries and (b) batteries that reported any amount of venting.

Higher H2S concentrations are noted near the shows that H2S concentrations are generally cities of Edmonton, Calgary, Grande Prairie, and lower in gas that is vented as well as flared, in the northwest region of the Province, while although the in some cases it appears non- low concentrations are seen near Lloydminster, negligible. However, as discussed above, a Brooks, and in the mid-west of the Province. smaller proportion of vented gas is linked to

(Note that the H2S maps seen in Figure 5 use measurements, and the significant impacts that different colour contours than the other figures such small quantities of H2S have on mitigation to highlight the “sour” threshold of options further necessitates measurements on a 10 mol/kmol.) As expected, this figure also case-by-case basis.

Figure 5: Map showing smoothed H2S concentrations at (a) exclusively-flaring batteries and (b) batteries that reported any amount of venting.

Gross heating values seen in Figure 6 are upstream sites are consistently well above the reasonably well-distributed throughout the 20 MJ/m3 minimum lower-heating value Province, with lower values seen near threshold for permitted flaring as specified in Lloydminster and in the area south of Calgary ERCB Directive 60 17, and thus well above the and Brooks. From Table 1, net or lower heating range of heating values shown to lead to poor values (LHV) would be lower in all cases by flare conversion efficiencies 7,8. There appears approximately 3.5 to 4 MJ/m3. It is noted that to be little difference between the heating values the heating values of solution gas flared at of gas that is flared or vented.

Figure 6: Map showing smoothed gross heating values (GHV) at (a) exclusively-flaring batteries and (b) batteries that reported any amount of venting.

produces two moles of CO2, and so on. Calculation of Greenhouse Gas Emission Considering also any CO2 present in the raw Factors flare gas, the total number of moles of CO2 Greenhouse gas emission factors can be emitted ( nCO ) per mole of raw flare gas determined from composition of the gas being 2 either flared or vented. As discussed above, and ( n flare gas) can be calculated as in eq. (2), where assuming ideal stoichiometric combustion, i is the mole fraction of species i in the raw flaring the gas results in oxidation of any carbon flare gas and the C7+ category is conservatively atoms present in the fuel to produce CO2. This assumed to contain only C7 species. is expressed by the generalized hydrocarbon combustion reaction shown as eq (1): n CO2  C1  2C2  3C3  n flare gas  y  y (2) C x H y   x  O2  xCO 2  H 2 O (1)  2  2  4 IC 4   NC4 5C5   6  7   From inspection of eq (1), the number of moles C6 C7P CO2 of CO2 produced by the flare depends only on Invoking the ideal gas model, it is possible to the carbon content of the raw flare gas. compute a GHG emission factor for a flare Therefore, a mole of methane (CH4) produces ( EF f ) according to eq (3), evaluated as one mole of CO2, a mole of ethane (C2H6)

13 kilograms of emitted CO2 per cubic metre of flares is dependent on many factors, such as flare gas (equivalent to tonnes per 103 m3), and cross-wind speed, gas exit velocity, flare exit for the moment still assuming 100% flare diameter, composition of the gas, and steam conversion efficiency (i.e. that all carbon bound assist rate (when relevant) 7,8,26-28. To better up in hydrocarbons within the fuel stream are reflect real-world conditions, emission factors converted to CO2 in the products): were also calculated in Table 2 for cases of non- ideal combustion. For this calculation, it was n PM CO2 CO2 assumed that unburned gases retained their EF f  (3) n flare gas RT initial compositions; that is, the emission factor for a flare with 98% efficiency is a linear where P and T are pressure and temperature (for combination of 98% flaring and 2% venting. all calculations in this paper these are specified This scenario inherently assumes that at industry standard values of 1 atm and 15°C), inefficiencies are dominated by stripping of 6 M the molar mass of CO2, and R is the CO2 unburned fuel . universal gas constant. Finally, the GWP of methane as published by Emission factors for vented gas depend only on the Intergovernmental Panel on Climate Change 10 the concentrations of GHGs present in the fuel, (IPCC) , has seen some revision between namely methane (C1) and CO2. Some C2–C4 publication cycles. In the most recent species can also be considered GHGs as they assessment report, AR4, the 100-year horizon have been implicated as having indirect warming GWP value is 25, whereas in AR2 the value was 10 11 effects, however the magnitude of these effects 21 . Even more recent analysis , in which are small compared to that of methane direct and indirect effects of aerosol responses to (particularly so given their comparatively small oxidant changes associated with methane concentrations in solution gas) and are subject to emissions are also considered, suggests that significant uncertainties 10. The potential actual 100-year horizon GWP values for indirect GHG contributions of these species have methane may be 10-40% higher than the AR4 thus been omitted. In this case, a GHG emission value. Nevertheless, due to legacy issues and legal frameworks, some government bodies factor for venting ( EFv ) can be determined continue to require the use the methane GWP according to eq (4), where GWP is the mass- CH4 value from AR2 (despite the fact that these data based global warming potential factor of are currently more than fifteen years out of methane, and all other variables are determined date). Recognizing this reality, emission factors as above: for venting were separately derived using both AR4 and AR2 GWP values and included in P EF   M GWP   M  Table 2 to extend its potential applicability. For v C1 CH4 CH4 CO2 CO2 RT (4) the emission factors derived for flaring, the variation associated with using different GWP Categorized flaring and venting emission factors values was not significant (there is no difference are reported in Table 2, assuming 100-year time if 100% carbon conversion efficiency is horizons for the cited GWP values 10. In assumed), so only the most recent IPCC GWP practice, however, the combustion efficiency of value for methane was considered.

3 3 a Table 2: Summary of GHG emission factors in tonnes of CO2 equivalent per 10 m of solution gas evaluated on a 100-year horizon derived using data from all batteries, from batteries reporting any amount of venting batteries, or from batteries that flared exclusively. Exclusively Flaring All Batteries Venting Batteriesb Batteriesc Mean 10th 90th Mean 10th 90th Mean 10th 90th d EFf (100%) 2.10 1.90 2.28 2.07 1.90 2.25 2.16 2.03 2.31 d EFf (98%) 2.35 2.13 2.55 2.32 2.13 2.51 2.40 2.25 2.56 d EFf (95%) 2.72 2.48 2.95 2.70 2.48 2.92 2.77 2.59 2.95 e EFv (25) 14.58 13.38 15.62 14.73 13.45 15.64 14.20 13.20 15.13 e EFv (21) 12.25 11.24 13.12 12.38 11.30 13.14 11.93 11.09 12.71 a All volumes assume a pressure of 101.325 kPa and temperature of 15°C. b Venting batteries include all batteries that report any amount of venting (i.e. batteries that vent exclusively as well as batteries reporting both flaring and venting) c Exclusively flaring batteries did not report any amount of venting d Percentages shown refer to flare combustion efficiency; calculations performed assuming incomplete combustion emissions occur via a fuel stripping mechanism5 and a GWP for methane of 25. e Value refers to GWP of methane used in the calculation.

Though GWP values are most commonly quoted within 10-20 years 30,31. Table 3 compares mean assuming a 100-year time horizon, this is not GHG emission factors for flaring and venting universally the best choice. For short-lived calculated using 20- and 100-year time horizons. climate forcers such as CH4, which has a steady These results show that near-term climate state lifetime in the atmosphere of about 9 forcing impacts from venting are much more years29, the 100-year time frame understates the severe than for flaring. Conversely, this opportunity for near-term climate forcing difference illustrates a significant opportunity 11 reductions . Unlike emissions of CO2 which for near-term mitigation of climate forcing, and once released may persist in the atmosphere for shows how substitution of flaring for venting centuries, mitigation of CH4 emissions would leads to a factor of ~20 reduction in CO2 lead to near-term reductions in atmospheric equivalent emissions per unit volume of solution concentrations and consequent climate forcing gas over a 20-year time horizon.

Table 3: Comparison of Mean GHG Emission Factors for Flaring and Venting Evaluated over 20- and 100- year Time Horizonsa derived using data from all batteries, from batteries reporting any amount of venting batteries, or from batteries that flared exclusively. Exclusively Flaring All Batteries Venting Batteriesb Batteriesc

Time Horizon 20-year 100-year 20-year 100-year 20-year 100-year d EFf (100%) 2.10 2.10 2.07 2.07 2.16 2.16 d EFf (98%) 2.92 2.35 2.89 2.32 2.93 2.40 d EFf (95%) 4.14 2.72 4.13 2.70 4.09 2.77

EFv 42.86 14.58 43.27 14.73 40.72 14.20 a 3 3 Emission factors are presented in units of tonnes of CO2 equivalent per 10 m of solutions gas (all volumes assume a pressure of 101.325 kPa and temperature of 15°C). Calculations assume GWP values for CH4 of 25 (100-year time horizon) and 72 (20-year time horizon) as given in IPCC AR410. b Venting batteries include all batteries that report any amount of venting (i.e. batteries that vent exclusively as well as batteries reporting both flaring and venting) c Exclusively flaring batteries did not report any amount of venting d Percentages shown refer to flare combustion efficiency; calculations performed assuming incomplete combustion emissions occur via a fuel stripping mechanism5.

representative of associated gas worldwide, the Estimation of GHG Emissions from Flaring estimated global total of 139 billion m3 of gas and Venting in Alberta flared as determined by satellite imagery 2 In 2008, 5945 batteries in Alberta reportedly translates to an annual emission of 292 Mt of flared or vented solution gas. Flaring activity CO2 equivalent. Given the lack of detailed was reported by 2360 of these batteries, totalling information on global venting volumes in the 305106 m3 of gas flared. Conservatively literature, for the purpose of making a very assuming ideal combustion and using the 100- rough estimate, if venting trends in Alberta year time horizon emission factor derived for could be considered representative of worldwide exclusively flaring (i.e. non-venting) sites, this activity, then the GHG contribution of implies GHG emissions totalling 0.664 Mt of worldwide associated gas venting would be on

CO2 equivalent from solution gas flaring at the order of 2 Gt of CO2 equivalent. While the upstream battery sites. Venting activity, uncertainty inherent in this type of gross reported by 4263 batteries, totalled a similar raw estimate should not be understated, GHG gas volume of 382106 m3. The GHG impact of emissions of this magnitude would represent venting, however, was an order of magnitude significant source globally (~5% of the world 32 greater at 5.74 Mt. The combined total 2008 total of 38.75 Gt in 2005) . Uncertainties in GHG emissions from solution gas flaring and sources of this potential magnitude highlight the venting at battery sites in Alberta was thus found need for closer analyses of industry reported to be at least 6.41 Mt. GHG emissions from all data and ultimately for better direct flow and upstream flaring and venting sources in Alberta composition monitoring of flared and vented gas (i.e. including additional flaring during well- streams globally, and are a further example of tests and at gas plants, etc. which raised total the challenges faced in reconciling bottom-up flare and vent volumes to 1.11 billion m3 in global GHG reporting with top-down estimates 33 200816) were approximately 7.64 Mt in 2008. derived from atmospheric measurements . Finally, assuming that the provincial average flaring emission factor is reasonably

CONCLUSIONS ACKNOWLEDGMENTS A method has been developed to estimate the This project was supported by Natural Resources composition of the solution gas produced at each Canada CanMET Energy (Project manager of the 13,144 oil and bitumen batteries in the Michael Layer), and would not have been Province of Alberta that reported flaring or possible without the invaluable support and venting activity between January 2002 and cooperation of James Vaughan, Jim Spangelo, December 2008. Measured gas samples from Jill Hume, Harvey Halladay, and Jim Dilay of specific sources were tied to geographic location the Alberta Energy Resources Conservation through production facilities, and this Board (ERCB). information was then used to estimate the compositions of nearby sites lacking measured REFERENCES samples. For this most significant production region within the Western Canadian 1. U.S. Energy Information Administration. International Energy Statistics, 2008. Sedimentary Basin, analysis revealed that gas 2. Elvidge, C. D.; Ziskin, D.; Baugh, K. E.; Tuttle, flared and vented has reasonably consistent B. T.; Ghosh, T.; Pack, D. W.; Erwin, E. H.; composition with a site-weighted mean methane Zhizhin, M. A Fifteen Year Record of Global concentration of 85.8% with 10th and 90th Natural Gas Flaring Derived from Satellite Data. percentile concentrations of 78.6% and 92.0% Energies 2009, 2, 595-622; doi: 10.3390/en20300595. respectively. Heating values were also found to 3 3. Johnson, M. R.; Coderre, A. R. An Analysis of be well above the 20 MJ/m minimum limit Flaring and Venting Activity in the Alberta linked in regulation to poor flare combustion Upstream Oil and Gas Industry. Journal of the Air efficiencies. Composition variations are & Waste Management Association 2011, 61, 190- 200; doi: 10.3155/1047-3289.61.2.190. apparent across the region, with higher methane 4. McEwen, J. D. N.; Johnson, M. R. Black Carbon content gas linked with heavy oil producing Particulate Matter Emission Factors for Buoyancy regions in particular. Greenhouse gas emission Driven Associated Gas Flares. Journal of Air & factors were derived under various scenarios, Waste Management Association 2011, submitted assuming different composition ranges and March 25, 2011. different assumed flare combustion efficiencies. 5. Johnson, M. R.; Devillers, R. W.; Thomson, K. A. Quantitative Field Measurement of Soot Emission For a reference case assuming 100% flare from a Large Gas Flare Using Sky-LOSA. conversion efficiencies, GHG emissions from Environmental Science & Technology 2011, 45, flaring and venting at batteries in Alberta were 345-50; doi: 10.1021/es102230y. determined to amount to 6.41 Mt in 2008, and 6. Johnson, M. R.; Wilson, D. J.; Kostiuk, L. W. A 7.64 Mt including all reported upstream flare fuel stripping mechanism for wake-stabilized jet diffusion flames in crossflow. Combustion and vent sources. Applying these emission Science and Technology 2001, 169, 155-174; doi: factors globally, the 139 billion m3 of flaring 10.1080/00102200108907844. estimated from satellite data would equate to 7. Johnson, M. R.; Kostiuk, L. W. Efficiencies of 292 Mt of GHG emissions annually. The low-momentum jet diffusion flames in derived data and methods can be used in efforts crosswinds. Combustion and Flame 2000, 123, 189-200; doi: 10.1016/S0010-2180(00)00151-6. to quantify and manage flaring and venting 8. Johnson, M. R.; Kostiuk, L. W. A parametric emissions and to better assess mitigation model for the efficiency of a flare in crosswind. opportunities. Proceedings of the Combustion Institute 2002, 29, 1943-1950; doi: 10.1016/S1540-7489(02)80236- X.

9. Strosher, M. T. Characterization of emissions Resources Conservation Board, Calgary, Alberta, from diffusion flare systems. Journal of the Air & 2008. Waste Management Association 2000, 50, 1723- 24. ERCB Guide 01: 2011 Catalogue: Publications, 33. Maps, and Services, Alberta Energy Resources 10. IPCC Solomon, S.; Qin, D.; Manning, M.; Chen, Conservation Board, Calgary, Alberta, 2011. Z.; Marquis, M.; Averyt, K. B.; Tignor, M.; 25. Acharya, T.; Ray, A. K. Image processing: Miller, H. L., Eds.; Cambridge University Press: principles and applications; Wiley-Interscience, Cambridge, UK and New York, USA, 2007; p. 2005. 996. 26. McDaniel, M. Flare efficiency study; United 11. Shindell, D.T.; Faluvegi, G.; Koch, D.M.; Stated Environmental Protection Agency: Schmidt, G. A.; Unger, N.; Bauer, S.E. Improved Research Triangle Park, NC, 1983; p. 133. attribution of climate forcing to emissions. 27. Pohl, J. H.; Lee, J.; Payne, R.; Tichenor, B. A. Science, 2009, 326, 716-718; doi: Combustion efficiency of flares. Combustion 10.1126/science.1174760. Science and Technology 1986, 50, 217–231. 12. BP BP Statistical Review of World Energy June 28. Siegel, K. D. Degree of conversion of flare gas in 2011; 2011 (available at: refinery high flares. Ph.D. Thesis, Fridericiana bp.com/statisticalreview) University Karlsruhe, Germany, 1980, p. 192. 13. National Energy Board Canadaʼs Energy Future - 29. Prinn, R. G.; Huang, J.; Weiss, R. F.; Cunnold, D. Reference Case and Scenarios to 2030; NE23- M.; Fraser, P. J.; Simmonds, P. G.; McCulloch, 15/2007E, Calgary, AB, 2007. A.; Harth, C.; Reimann, S.; Salameh, P.; 14. National Energy Board Estimated Production of O’Doherty, S.; Wang, R. H. J.; Porter, L. W.; Canadian Crude Oil and Equivalent 2009 Miller, B. R.; Krummel, P. B. Evidence for (Available at: http://www.neb-one.gc.ca/clf- Variability of Atmospheric Hydroxyl Radicals nsi/rnrgynfmtn/sttstc/crdlndptrlmprdct/stmtdprdct Over the Past Quarter Century. Geophysical n-eng.html). Research Letters, 2005, 32, 2-5; doi: 15. Statistics Canada Energy Statistics Handbook; 10.1029/2004GL022228. Minister of Industry, 57-601-X, 2010. 30. Montzka, S.A.; Dlugokencky, E.J.; Butler, J.H. 16. ERCB ST60B-2010: Upstream Petroleum Non-CO2 Greenhouse Gases and Climate Change. Industry Flaring and Venting Report; Alberta Nature 2011. 476, 43–50; doi: Energy Resources Conservation Board, Calgary, 10.1038/nature10322. Alberta, 2010. 31. Hansen, J.; Sato, M.; Ruedy, R.; Lacis, A.; Oinas, 17. ERCB Directive 060: Upstream Petroleum V. Global Warming in the Twenty-First Century: Industry Flaring, Incinerating, and Venting; An Alternative Scenario. Proc. Natl. Acad. Sci. Alberta Energy Resources Conservation Board, U.S.A. 2000, 97, 9875-9880; doi: Calgary, Alberta, 2006. 10.1073/pnas.170278997. 18. ERCB Directive 007: Volumetric and 32. Climate Analysis Indicators Tool (CAIT) Version Infrastructure Requirements; Alberta Energy 8.0. (Washington, DC: World Resources Institute, Resources Conservation Board, Calgary, Alberta, 2011). 2007. 33. Weiss, R.F.; Prinn, R.G. Quantifying 19. Dandekar, A. Y. Petroleum reservoir rock and Greenhouse-Gas Emissions from Atmospheric fluid properties; CRC Press: Boca Raton, FL, Measurements: A Critical Reality Check for 2006. Climate Legislation. Phil. Trans. R. Soc. A, 2011, 20. Province of Alberta Pipeline Regulation. Alberta 369, 1925-1942. Regulation 91/2005. 21. Province of Alberta Oil and Gas Conservation ABOUT THE AUTHORS Act. Alberta Regulation 151/71. 22. ERCB Directive 017: Measurement Requirements Adam Coderre has an M.A.Sc. degree in for Oil and Gas Operations, Alberta Energy Mechanical Engineering from Carleton Resources Conservation Board, Calgary, Alberta, University and worked as a research engineer in 2011. the Energy & Emissions Lab within the 23. ERCB Directive 056: Energy Development Applications and Schedules, Alberta Energy Mechanical & Aerospace Engineering

18 department of Carleton University. He is currently a Project Engineer at Clearstone Engineering Ltd., Calgary, AB. Matthew Johnson is the Canada Research Chair in Energy & Combustion Generated Pollutant Emissions and an associate professor at Carleton University where he heads the Energy and Emissions Research Lab. Please address correspondence to: Matthew Johnson, Mechanical and Aerospace Engineering, Carleton University, 1125 Colonel By Drive, Ottawa, ON, Canada, K1S 5B6; phone: +1-613-520-2600 ext. 4039; fax: +1-613-520-5715; email: [email protected].