Policy Analysis

Cite This: Environ. Sci. Technol. 2018, 52, 947−954 pubs.acs.org/est

Statistically Enhanced Model of In Situ Oil Sands Extraction

Operations: An Evaluation of Variability in Greenhouse Gas
Emissions
Andrea Orellana,† Ian J. Laurenzi,‡ Heather L. MacLean,§ and Joule A. Bergerson*,†
†
Department of Chemical and Petroleum Engineering, University of Calgary, 2500 University Drive NW, Calgary,
Alberta T2N 1N4, Canada
‡
ExxonMobil Research and Engineering Company, 1545 Route 22 East, Annandale, New Jersey 08801-3059, United States
§
Departments of Civil Engineering, Chemical Engineering and Applied Chemistry, School of Public Policy and Governance,
See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

University of Toronto, Toronto, Ontario Canada M5S 1A4

*
S Supporting Information
Downloaded via UNIV OF ALBERTA on August 19, 2020 at 20:38:51 (UTC).

ABSTRACT: Greenhouse gas (GHG) emissions associated

with extraction of bitumen from oil sands can vary from
project to project and over time. However, the nature and
magnitude of this variability have yet to be incorporated into
life cycle studies. We present a statistically enhanced life cycle
based model (GHOST-SE) for assessing variability of GHG
emissions associated with the extraction of bitumen using
in situ techniques in Alberta, Canada. It employs publicly
available, company-reported operating data, facilitating assess-
ment of inter- and intraproject variability as well as the time
evolution of GHG emissions from commercial in situ oil sands
projects. We estimate the median GHG emissions associated
with bitumen production via cyclic steam stimulation (CSS) to be 77 kg CO2eq/bbl bitumen (80% CI: 61−109 kg CO2eq/bbl),
and via steam assisted gravity drainage (SAGD) to be 68 kg CO2eq/bbl bitumen (80% CI: 49−102 kg CO2eq/bbl). We also
show that the median emissions intensity of Alberta’s CSS and SAGD projects have been relatively stable from 2000 to 2013,
despite greater than 6-fold growth in production. Variability between projects is the single largest source of variability (driven
in part by reservoir characteristics) but intraproject variability (e.g., startups, interruptions), is also important and must be
considered in order to inform research or policy priorities.

■ INTRODUCTION
The bitumen in Alberta’s oil sands make up 97% of Canada’s oil
reserves, which are the third largest proven reserves in the
world (approximately 168 billion barrels) following only those
of Saudi Arabia and Venezuela.1 Canadian bitumen production
was 2.3 million bbl/day in 2014 and has been predicted to
increase to 3.7 million bbl/day by 2030.2 More than half of
Alberta’s bitumen is produced using in situ recovery techniques,
among which Steam Assisted Gravity Drainage (SAGD) and
Cyclic Steam Stimulation (CSS) are the most common. Steam
is employed in these techniques to heat the bitumen−and
thereby reduce its viscosity, so that it may be produced from
the subsurface reservoir. Once extracted from sand and water,
bitumen must be either diluted or processed (“upgraded”)
before transporting it to refineries. Steam generation, on-surface
activities (e.g., water treatment) as well as upgrading operations
result in greenhouse gas (GHG) emissions. The result is
generally higher GHG emissions than “conventional” crudes
on average,3 although there is significant overlap in carbon
intensity. Moreover, for in situ oil sands production, there
isadded complexity to how emissions are estimated for different
crudesThis has in turn motivated the development of improved
models to make these assessments.
Confounding the issue of GHG emissions from oil sands is
the variability in emissions from project to project as well as
over time. Variability is differentiated from uncertainty in that it
reflects systematic differences between processes, locations
or in time while uncertainty is associated with lack of data or
an incomplete understanding. There are several sources of
variability in GHG emissions of in situ oil sands operations.
Emissions may vary due to differences in production practices
and reservoir properties (interproject variability). They may
also vary due to differences in the maturity of the operating
field and operating conditions (intraproject variability).3,4
For example, SAGD may require different amounts of
energy−resulting in different GHG emissions−at startup, at
steady state and at the end of the project’s life.
Received: August 31, 2017
Revised: November 28, 2017
Accepted: December 12, 2017
Published: December 12, 2017

© 2017 American Chemical Society 947 DOI: 10.1021/acs.est.7b04498

Environ. Sci. Technol. 2018, 52, 947−954
Environmental Science & Technology
Several peer-reviewed life cycle assessments (LCAs) have
attempted to quantify the GHG emissions associated with
transportation fuel products derived from oil sands5−8 and
several government agencies have released public tools that
include oil sands pathways among a suite of fuel pathways.9−11
However, these studies have generally focused on the
development of a single “point estimate” for the life cycle
GHG emissions associated with fuels derived from oil sands.
Ranges of GHG emissions associated with alternative oil sands
production technologies were first estimated with the “Green-
house gas emissions of current oil sands technologies” (GHOST)
model,3,12 which employed confidential operator data. However,
the GHOST model did not quantify the distributions of GHG
emissions associated with these technologies, nor did it assess the
representativeness of the ranges. A few subsequent studies have
attempted to better understand the variability and uncertainty
associated with oil sands pathways.13−15 However, no study has
comprehensively addressed the statistical attributes of the ranges
of GHG emissions estimates that may result from different oil
sands extraction pathways, or the temporal nature of GHG
emissions over a project’s lifetime.
In this paper, we report the findings of a study of the varia-
bility associated with in situ oil sands extraction as estimated by
a statistically enhanced version of the aforementioned GHOST
model (GHOST-SE). This research improves the statistical
representation of operating data and input parameters using
publicly available data sets and improved process calculations,
and generates project-specific distributions of GHG emissions
associated with in situ oil sands extraction (i.e., the extraction of
raw bitumen). We then apply the model to explore inter- and
intraproject variability, including the nature and magnitude of
the variability, historical GHG emissions trends and the effects
of operating parameters. When combined with robust estimates
of GHG emissions associated with pipeline transportation,
refining, and the combustion of fuels refined from oil sands
(60−80% of the life cycle GHG emissions3), our results may be
utilized to assess the life cycle GHG emissions associated with
oil sands, facilitating comparison with other energy sources on a
life cycle basis.
■ MATERIALS AND METHODS
Our analysis of the variability associated with in situ oil sands
extraction was conducted by improving the GHOST model.
The new statistically enhanced model, GHOST-SE, is
distinguished from GHOST by the following features:
1. Use of publicly available operating data from the Alberta
Energy Regulator.16 These data include production, steam
generation, and electricity usage time series through-
out major oil sands project lifetimes. Data are disaggre-
gated on a per project and per month basis over the life of
each project,
2. Improved process calculations, related primarily to
surface facility activities, such as optional cogeneration,
and
3. Monte Carlo (MC) simulation to representatively select
operating data and parameters for the purposes of
generating statistically meaningful ranges of GHG
emissions associated with in situ oil sands extraction.
Similar to the original GHOST model, the improved model
takes a life cycle approach (i.e., includes both onsite as well as
indirect GHG emissions. For example, indirect emissions
such as offsite electricity generation, natural gas production
948
Policy Analysis
are included. This analysis focuses on in situ techniques at the
extraction stage. Emissions are reported in kg CO2 equivalent
per barrel of undiluted bitumen, with CO2 equivalency defined
by the fifth Assessment Report of the IPCC (AR5) for a
100-year time horizon.17 A flowchart describing the boundaries
of the model are provided in Figure S1 of the Supporting
Information (SI).
Operating Data. Alberta’s Responsible Energy Develop-
ment Act,18 requires in situ oil sands project operators to report
monthly data, including production and steam injection
volumes, electricity supply and demand, and volumes of
associated gas flared and vented. This publicly available data19
replaces the confidential data utilized in the original GHOST
model. Inventory data not reported on a monthly basis are
obtained from the Alberta Energy Regulator’s (AER) Statistical
Reports ST53: Alberta Crude Bitumen In Situ Production20 from
January 1992 to April 2014 and the annual In Situ Performance
Presentations.21 The data from in situ annual presentations and
ST53s are collected for all operating projects reported from the
beginning of in situ operations in 1985 to 2014. Operating
parameters such as the boiler feedwater (BFW) temperature,
pressure and solution gas composition are collected from each
project’s Environmental Impact Assessment (EIA), available at
the Alberta Government Library.22 Provincial electricity grid
emission intensities for the province, reported hourly by the
Alberta Electric System Operator (AESO)23 are converted into
monthly intensity values by aggregating hourly data available
from 2011 to 2014.
To represent commercial oil sands industry operations, only
projects with commercial production of more than 10 000 bbl/day
are considered, leaving pilot projects out of the analysis.
Projects in the startup phase are excluded because they have
not reached steady conditions and data are insufficient to
characterize their operations. Details about each project are
presented in Table S4 in SI.
After applying the selection criterion and filtering the data,
15 operating in situ projects were selected for analysis. Three of
these projects are CSS operations, two of which combine steam
and electricity generation (onsite cogeneration). The other
12 are SAGD operations and half of these have cogeneration
capacity. The 15 projects represent approximately 75% of the
total in situ production of the province (303 million barrels out
of the 405 million barrels produced in 2013−according to the
AER’s ST98), offering a good volume-weighted representation
of the industry. The remaining 25% of production are small
facilities, new projects and pilot facilities. While we do not have
complete data associated with these projects, our understanding
is that none of these projects would drastically change the
industry wide emissions profile. However, we exclude these
projects (using the criteria of <3 years of production and/or
under 10,000 bbl/day) as they could bias the estimates slightly
up or down and we would not have the evidence to explain why.
The inventory data employed by GHOST-SE are summar-
ized in Table 1.
Process Calculation Improvements. GHOST-SE also
includes improvements to the process calculations within the
model. These include a more complete treatment of systems
where steam and electricity are cogenerated onsite and more
transparent calculations of the energy requirements of the
surface facilities. Detailed descriptions of the improvements are
reported in Section 1.2 of SI.
Monte Carlo Simulation. GHOST-SE employs MC
simulation to generate distributions of GHG emissions for
DOI: 10.1021/acs.est.7b04498
Environ. Sci. Technol. 2018, 52, 947−954
Environmental Science & Technology Policy Analysis

Table 1. GHOST-SE Input Parameters for in Situ Bitumen Extraction (SAGD and CSS)a
input SAGD range CSS range distribution source
SOR (m3 steam/m3 bitumen) − dry (steam x = 100%) 2.2−5.9 2.3−5.5 direct data sampling AER ST5320
electricity (kWh/m3 bitumen) 55−247 109−264 direct data sampling AER in situ progress reports21
flared Hydrocarbons (kg CO2e/m3 bitumen) 0.1−7.6 0.2−6.3 direct data sampling AER in situ progress reports21
fugitive emissions (kg CO2e/m3 bitumen) 0.3−1.0 0.4−2.6 direct data sampling AER in situ progress reports21
solution gas (m3/m3 bitumen) 4.1−35.1 43.0−89.2 direct data sampling AER in situ progress reports21
produced steam quality 80% 100% uniform GHOST12
boiler output pressure (MPa) 5.0−10.0 5.0−10.0 uniform Project EIAs22
produced steam enthalpy (kJ/kg) 2,460−2,470 2,460−2,470 function of boiler pressure and Project EIAs and superheated steam
temperature tables for water22
boiler feedwater temperature (°C) 105−202 115−150 custom Project EIAs22
Alberta electricity grid emission factor (g CO2e/kWh) 647−761 647−761 custom Alberta Electricity System Operator23
lower heating value of natural gas fuel (pipeline and 37.9 37.9 NA GHOST12
solution gas), MJ/m3
CASE 1. OTSG BOILER
efficiency−boiler 80% - 85% 80% - 85% uniform GHOST12
CASE 2. COGENERATION (GT + HRSG)
efficiency−gas turbine 30% - 35% 30% - 35% uniform GHOST12
efficiency−heat recovery 50% - 60% 50% - 60% uniform GHOST12
efficiency − HRSG direct firing 95% 95% NA GHOST12
total electricity produced (kWh/m3 bitumen) 150−600 97−199 Direct Data Sampling AER In situ progress reports21
a
Two cases are available in the model. Case 1 employs a natural gas fired boiler for steam generation, and electricity is imported from the grid.
Case 2 utilizes onsite cogenerationa gas turbine and heat recovery steam generatorto supply steam and power. Actual in situ extraction projects
feature one or the other. Notes: SOR: steam-to-oil ratio; OTSG = once-through steam generator; GT: gas turbine (electricity generator); HRSG:
heat recovery steam generator; SAGD: steam assisted gravity drainage; AER: Alberta Energy Regulator; EIA: environmental impact assessment;
NA: not applicable as constant values are utilized.

each project on an annual basis (e.g., kg CO2eq/bbl in 2010) as
well as on a project life cycle basis (e.g., total kg CO2eq/total
bbl since project startup). MC simulation capability is incor-
porated into GHOST-SE using the Oracle Crystal Ball add-in
for Excel. Testing revealed that 10 000 MC runs were sufficient
to generate stable distributions of GHG emissions for in situ
extraction.
For a given project, the Alberta Energy Regulator (AER) data
for a particular month are simultaneously selected, preserving
relationships between related parameters such as bitumen
production, injected steam, etc. In cases where a particular data
field (e.g., purchased electricity) is absent from a project’s time
series for a particular month, it is randomly selected from the
rest of the data pool. Data not appearing in the AER data sets
are not correlated with the AER data, and are largely modeled
as uniform random variables based on data from other sources
(see Table 1).
Three types of MC simulations are performed by GHOST-SE:
1. Industry-wide variability. Simulations of all CSS and
SAGD operations over all years the projects are
operating, to obtain representative distributions of the
extraction GHG emissions for these two technologies.
2. Interproject variability. Simulations of each CSS and
SAGD project over all years the projects are operating,
to obtain representative distributions of the extraction
GHG emissions for each project over its lifetime.
3. Intraproject variability. Simulations of each CSS and
SAGD project during each year of their life cycle to
explore the time evolution of the extraction GHG
emissions of each project over its lifetime.
The first type of simulation may be used to assess the
variability of GHG emissions resulting from CSS and SAGD
technologies (Industry-wide variability), permitting comparison
of their emissions using different extraction techniques, etc.
Two sampling methods are employed to explore industry-wide
949
variability including “historic industry emissions intensity”
where sampling occurs across individual projects and the emis-
sion intensity from each project is weighted by the production
level of that project and “prospective industry emissions
intensity” where sampling occurs randomly across all projects.
The former represents a snapshot of industry performance
historically from first commercial operation to 2013 where
projects that have been operating longer have a larger influence
on industry-wide emissions. The latter treats all sample points
as equally likely and therefore represents the distribution of
possible performance assuming that all projects are equally
weighted and could represent future performance rather than
being more heavily influenced by older projects.
The second type of simulation assesses variability of GHG
emissions associated with each project (Interproject variability).
The third type of simulation assesses how variability changes over
the lifetime of each individual project (Intraproject variability).
Projects that began commercial production relatively recently
(2007 to present) are of special interest for analyzing the effect
of well startup operations. At initial reservoir conditions, there
is negligible fluid mobility due to bitumen’s high viscosity.
For SAGD projects, steam is injected into both injection
and production wells to establish interwell communication and
reduce viscosity of the bitumen, allowing it to be produced.
As a consequence of higher-than-normal steam demand and
lower-than-normal crude production, SAGD operations have
higher GHG emissions per barrel of bitumen at the beginning
of project operations. Subsequently, steam requirements
stabilize and the project reaches what could be considered
“steady state”. At the end of a well’s life, the steam requirements
increases to the point where it exceeds an economic threshold
and the well is shut in. Assessment of intraproject variability
facilitates comparison of GHG emissions during startup- and
steady state operations.
DOI: 10.1021/acs.est.7b04498
Environ. Sci. Technol. 2018, 52, 947−954
Environmental Science & Technology
■ RESULTS
Industry-Wide Variability. In Figure 1 we report the
distributions of GHG emissions associated with CSS (in the
Figure 1. Distributions industry-wide cumulative GHG emissions
associated with in situ extraction of undiluted bitumen, in cumulative
kg CO2eq/cumulative bbl bitumen for (a) cyclic steam stimulation
(CSS) and (b) steam assisted gravity drainage (SAGD). Blue vertical
lines represent the 10th and 90th percentiles of the Monte Carlo GHG
emissions ranges obtained from GHOST-SE, whereas the gray vertical
line represents its median value. The light gray area displays the
original GHOST range.12 “Plus signs” represent findings of previously
published studies.7−10,13
upper panel) and SAGD (in the lower panel) operations.
Vertical lines represent the 10th percentile, median and
90th percentile of the range obtained through Monte Carlo
simulation, whereas the red markers report the distribution
means. Estimates of GHG emissions from the literature
(e.g., refs 7−10 and 13]) are reported as points, for com-
parison. While the previous literature when combined shows
variability in emissions, only the current analysis demonstrates
that the distribution is positively skewed. This means that
observing emissions at the lower end of the distribution is more
likely than at the upper end of the distribution. GHOST-SE
explicitly quantifies statistics related to the distribution. The
industry median GHG emissions for CSS-extracted bitumen is
77 kg CO2eq/bbl, and the variability ranges from 61 to 109 kg
CO2eq/bbl (80% confidence interval). By contrast, the industry
median GHG emissions for SAGD-extracted bitumen is
68 kg CO2eq/bbl, and the variability ranges from 49 to 102 kg
CO2eq/bbl (80% confidence interval). These ranges of GHG
emissions are narrower than those previously reported (e.g.,
refs 12 and 24).
Although the mean and median GHG emissions associated
with SAGD are lower than those associated with CSS, the
distribution of GHG emissions associated with SAGD is multi-
modal, suggesting that a subset of SAGD extraction emissions
(in time, by project, or both) yields higher emissions on
average than CSS. Moreover, the variability associated with
emissions from both extraction operations, while similar, is
likewise affected by this multimodality. Higher emissions may
be a consequence of higher energy requirements associated
with extraction from the reservoirs from which these projects
produce. Multimodality, by contrast, may be the result of the
initiation of project expansions, etc.
Ultimately, cumulative industry-wide GHG emissions for
CSS and SAGD operations have remained relatively stable over
the past 14 years, even as there has been a dramatic expansion
in the province’s in situ operations during that time period.
Box plots detailing these trends for SAGD and CSS are in S10
and S11. In theory, there are many factors that could have driven
the emissions to be larger or smaller than historic performance.
950
Policy Analysis
Examples of factors that could increase emissions include
changes in reservoir quality over time, or expansion of projects
to access alternative reservoirs. Conversely, technological
innovation including adoption of cogeneration, better well
placement, and incremental energy efficiency improvements. A
more detailed analysis of these factors is required to draw
conclusions regarding the degree to which each of these factors
have influenced net industry emissions over the past 14 years.
Based on the sampling method employed, these distributions
represent a snapshot of historic industry performance up
to 2013. This is in contrast to the alternative method of
conducting the Monte Carlo simulation described in the
Materials and Methods section, which treats all sample points
as equally likely and therefore represents the distribution of
possible performance if a new project were to come online.
These latter results are presented in the SI and show higher emis-
sions for the mean and median, more pronounced skewness and
broader distributions with heavier tails. The selection of method
for sampling in such Monte Carlo simulations is therefore an
important decision that can lead to different interpretations if the
analysis is focused on understanding historic performance or the
potential emissions of new projects (using the same technology).
Interproject Variability. In Figure 2 and Figure 3,
we report the distributions of the cumulative GHG emissions
Figure 2. Distributions of cumulative GHG emissions associated with
CSS oil sands projects producing more than 10 kbd. Results are
reported in cumulative kg of CO2 equivalent per cumulative barrel of
undiluted bitumen. Vertical lines represent the median, 10th and 90th
percentiles, and means are denoted by red “×” marks. Lighter green
colored projects have production capacities equal to or greater than
60 kbd. Darker purple colored projects have production capacities
between 10 and 60 kbd. Asterisks indicate projects that employ
cogeneration.
associated with individual in situ oil sands projects. The projects
are grouped by extraction technique (CSS, SAGD) and ordered
by the median project emissions (lowest value at the top).
These distributions quantify the variability of operations over
the full life of each project up to 2013. The figure also identifies
the relative size of each project and whether cogeneration is
included (the latter is indicated by an asterisk following the
project number). Variability in GHG emissions across the
industry appears to be derived from project-specific variability,
which in turn depends on the operations associated with each
project over time.
The distributions reported in Figure 2 and Figure 3 explain
the variability illustrated for CSS and SAGD in Figure 1.
For instance, Projects 13 and 9 drive the variability of SAGD
DOI: 10.1021/acs.est.7b04498
Environ. Sci. Technol. 2018, 52, 947−954
Environmental Science & Technology Policy Analysis

Figure 3. Distributions of cumulative GHG emissions (i.e., over the lifetime of each operating project) associated with SAGD oil sands projects.
Results are reported in cumulative kg of CO2 equivalent per cumulative barrel of undiluted bitumen. Vertical lines represent the median, 10th and
90th percentiles, and means are denoted by red “×” marks. Lighter green colored projects have production capacities equal to or greater than 60 kbd.
Darker purple colored projects have production capacities between 10 and 60 kbd. Asterisks indicate projects that employ cogeneration.

emissions in Figure 1, however, these projects have less influ-
ence on the overall SAGD industry’s emissions than Project 2
contributes to the CSS industry’s emissions. Joint bitumen produc-
tion asssociated with Projects 1*, 3*, 5*, and 6 represents
approximately half of the total production of all projects con-
sidered in this study (518 000 of approximately 1 043 000 bbl
bitumen/day).
For SAGD extraction, the variability in GHG emissions
associated with certain projects may exceed the range of median
GHG emissions across all projects. The median GHG emis-
sions associated with SAGD projects range from 52 kg CO2eq/bbl
bitumen to 172 kg CO2eq/bbl bitumen, whereas the full ranges
of GHG emissions for Projects 9 and 13 are 144−249 and
130−171 kg CO2eq/bbl, respectively. GHOST-SE results
951
displayed in Figure 2 and Figure 3 represent interproject and
intraproject variability: both types need to be considered when
discussing ranges of GHG emissions estimates from in situ oil
sands extraction as both are important.
No correlation of variability in emissions with the use of
cogeneration or size of project was identified for SAGD.
Projects 5* and 6 have similar distributions (means and ranges)
of GHG emissions despite the use of cogeneration in one but
not the other, likely a consequence of differences in reservoir
quality for the two projects, necessitating a higher SOR
(and higher steam load) for Project 5*. Likewise, Projects 7*,
10, 11, and 14* have similar distributions of GHG emissions
despite the fact that only two of the projects use cogeneration.
Projects 4 and 12* are likewise indistinguishable in terms
DOI: 10.1021/acs.est.7b04498
Environ. Sci. Technol. 2018, 52, 947−954
Environmental Science & Technology
Figure 4. Time evolution of the variability of extraction GHG emissions for SAGD Projects 8*, 6, and 9. Emissions are reported as annual kg CO2e
per annual bbl of undiluted bitumen. Distribution means are reported in red, and medians are reported in gray. Blue lines denote 10th and 90th
percentiles. Asterisks indicate projects that employ cogeneration.
of their GHG emissions, despite differences in the use of
cogeneration. We know that cogeneration offers efficiency benefits
over the nocogeneration cases, with accompanying fuel, and thus
GHG, reductions. However, these effects are likely dwarfed by
other drivers such as reservoir characteristics. This is supported by
previous work that confirmed SOR is the single most significant
parameter in determining the emissions.12 In addition, other
work has confirmed that reservoir characteristics (the basic
parameters including depth and thickness of the reservoir,
porosity, permeability, gamma ray measurements and oil
saturation) explain roughly 80% of the variability in SOR in
existing projects.25 There are methods of crediting emissions to
surplus electricity generation sold to the grid that could change
the measure of benefits of congeneration. This could, in turn,
increase the impact of cogeneration on emissions estimates for
in situ extraction.26 On average, the two CSS projects featuring
cogeneration demonstrate lower GHG emissions than the CSS
project that does not include congeneration. However, those
two projects (1* and 3*) have higher production capacities
than the third (2). For CSS, cogeneration may enable lower
GHG emissions as part of higher efficiency operations for larger
scale operations. However, the impact of cogeneration on the
variability of GHG emissions could not be discerned robustly
from the three projects considered.
Sensitivity analyses conducted using the original GHOST
model indicated that the GHG emissions for in situ extraction
(both CSS and SAGD) are driven by the amount of steam used
to stimulate the petroleum reservoir (mostly generated by
burning natural gas): over 90% of the GHG emissions asso-
ciated with extraction.9 This is confirmed in the current analysis.
Intraproject Variability. To explore the impact of
temporal variability, MC simulations of each of the individual
projects were conducted for each year of their operation.
The distributions of GHG emissions for each project (CSS and
SAGD) were then evaluated as time series. Time series of distri-
butions of GHG emissions associated with three projects are
952
Policy Analysis
illustrated in Figure 4 as examples; time series of the remaining
12 projects are presented in S12 to S23 of SI.
Project 8* is a SAGD project that has a cogeneration system
that supplies steam and electricity to the project. Variability in
GHG emissions tends to be constant in the latter years of
operation, but project startup affects the variability of GHG
emissions in the early years of operation. When annual GHG
emissions ranges are weighted by annual bitumen production,
the influence of early variability in emissions upon the project
lifetime variability diminishes over time. Therefore, the distri-
bution of cumulative GHG emissions reported in Figure 3
for Project 8* is approximately the same as the distribution of
emissions associated with operations in 2013. This temporal
behavior is also observed for Projects 4, 5*, 10, 11, 12, and 13*.
Project 6 illustrates an alternative time evolution of a project,
with distinct characteristics that affect variability. In this case,
initial variability in emissions (years 2003−2008) affects the
cumulative variability to the extent that the variability in
emissions associated with the most recent year for which data
were available (2013) is different from the cumulative
variability of the project (Figure 2 and Figure 3). In this case,
the distributions of GHG emissions in early years are multi-
modal. Variability in GHG emissions does not seem to decrease
monotonically over time, and does not appear to be associated
with any particular distribution. This behavior is a consequence
of the project’s production prolificacy and the effect of temporal
discontinuities in project development. This is akin to multiple
“startup effects” as new wells impact the variability of steam-to-
oil ratios (SORs) within a project and consequently, the
variability of associated GHG emissions. Most projects have
gone through different phases of expansion in their production
capacity, with consequent startup effects in later years. The time
evolutions of the variability in GHG emissions associated
with Projects 7, 9, 14*, and 15 share characteristics similar to
those of Project 6, although the effect in Project 6 is the most
pronounced.
DOI: 10.1021/acs.est.7b04498
Environ. Sci. Technol. 2018, 52, 947−954
Environmental Science & Technology
That said, bimodal variability in GHG emissions is observed
for individual projects despite the absence of startup effects
(e.g., Projects 2, 3*, and 14*). For these projects, we could
not attribute the variability to any particular input parameter
(e.g., SOR). We hypothesize that this variability is associated
with heterogeneity in the reservoirs from which these projects
produce bitumen.
In Figure 4 we also illustrate the range of GHG emissions
associated with Project 9, which has been commercially
operating since 2007 and has the largest range of emissions
of all in situ projects considered in this study. Our estimates of
GHG emissions for this project are significantly higher than the
previously reported GHOST range of emissions. This project’s
behavior confirms the presence of outliers and the influence of
project specificities (either reservoir conditions and/or produc-
tion strategies) on GHG emissions, that cannot be attributed to
the industry as a whole.
GHG emissions associated with the more recent projects
(Projects 12* and 15) seem to exhibit bimodal probability distri-
butions. This may be a consequence of variable steam generation
and bitumen extraction during the initiation of the projects.
Estimates of GHG emissions associated with recently initiated
projects are calculated with fewer months of data that result in
these data being more heavily weighted toward the more variable
operations. As projects age, data progressively are more heavily
weighted to the steady-state operations, in turn masking the
peaks in steam demand at startup relative to bitumen production.
Sensitivity. Insofar as the AER data sets are not com-
prehensive, certain parameters were retained from the original
GHOST model. Of these, the heating value of natural gas
purchased from pipelines and the efficiency of HRSGs (direct
fired) had the greatest impact on the results. The dependency
of the distribution of GHG emissions associated with Project 1
is illustrated in SI Figure S24. For this and most projects, either
(a) decreasing the LHV from 37.9 to 33.7 MJ/m3 or (b)
decreasing the default efficiency from 95% to 82.5% (the
default boiler efficiency for noncogen systems) results in an
approximately 10% increase in the estimate of the average GHG
emissions associated with an in situ oil sands project. Reduction
of either parameter also increases the width of the distributions.
■
DISCUSSION
This analysis improves the previously developed GHOST
model with more detailed calculations and the incorporation of
publically available input data for all commercial in situ projects.
A Monte Carlo simulation is conducted to derive insights
about industry-wide trends, inter- and intraproject variability.
Arguably, the most significant insight from this analysis is that
the use of a statistically enhanced model, GHOST-SE, allows
for a better representation of the GHG emissions from oil
sands operations. This study confirms that the original GHOST
range of emissions (based on proprietary operating data) is
generally a good representation of the overall industry-wide
range of emissions, but fails to provide information about
the probability distribution of emissions to effectively inform
decision makers. For example, the positive skewness of the
GHG emissions estimates for both CSS and SAGD indicate
that studies that have implied a uniform distribution have likely
overestimated industry wide estimates. The use of project
specific data reported by the AER on a monthly basis through-
out the lifetime of each in situ project increases transparency
and confirms the calculation of project specific emissions
estimates.
953
Policy Analysis
The wider variability in emissions of some projects could
suggest that operating decisions and the learning within each
individual project might be exerting a larger influence on the
range of GHG emissions for in situ oil sands extraction than
other variables, such as reservoir characteristics (reservoir
conditions are different across projects, thus the variability they
generate will be represented by the variability across projects).
Recent work by Akbilgic et al.25 suggests that the reservoir
characteristics can explain a large portion of the variability in
the SOR (>80%) and by extension, the variability in GHG
emissions. Hence, the rest of the variability must be due to
facilities’ technology selection or deployment strategy. This
may explain why the impact of cogeneration or project scale on
the variability in GHG emissions is challenging to discern from
our results−the variability in extraction GHG emissions
associated with the subsurface-driven demand for steam may
dominate other factors such as technology choice. Insofar as the
Alberta grid has been historically coal based, GHG emissions
associated with purchased power have been higher than those
associated with cogeneration, thereby reducing net emissions.
However, in recent years, the Alberta grid intensity has decreased
from 949 kg CO2eq/MWh in 1990 to 689 kg CO2eq/MWh in
2014. If the Alberta grid continues to “decarbonize”, then the
GHG reductions associated with cogeneration will be reduced.
Although we have exclusively considered the upstream
emissions of the in situ oil sands life cycle in this paper, this
analysis could be expanded to include other extraction methods
(e.g., mining) as well as other life cycle stages (e.g., transport,
refining, combustion of transportation fuels). Indeed, the
upstream emissions assessed in this paper are only part of
total life cycle emissions that need to be accounted for before
alternative product pathways can be compared. Once the life
cycle has been accounted for and further detailed statistical
analysis investigated to show the role that each parameter plays
in determining GHG emissions, this tool could be explored for
its potential use in predicting GHG emissions associated with
new projects to inform decisions at the design and deploy-
ment stages. Our results confirm the need to consider the
distributions of emissions given the variability is large, positively
skewed and the inter-, intra-, and time evolution sources of
the variability are all important. Finally, we note that a broad
view of oil sands GHG emissions, including their variability
and uncertainty, must also consider mined oil sands. We will
consider these in future research.
■
ASSOCIATED CONTENT
*
S Supporting Information
The Supporting Information is available free of charge on the
ACS Publications website at DOI: 10.1021/acs.est.7b04498.
more detailed discussion of the modeling, data and sen-
sitivity analysis (PDF)
■
AUTHOR INFORMATION
Corresponding Author
*E-mail: jbergers@ucalgary.ca.
ORCID
Joule A. Bergerson: 0000-0002-4736-3509
Notes
The authors declare no competing financial interest.
DOI: 10.1021/acs.est.7b04498
Environ. Sci. Technol. 2018, 52, 947−954
Environmental Science & Technology Policy Analysis

■ ACKNOWLEDGMENTS
We wish to acknowledge ExxonMobil Research and Engineer-
(19) Responsible Energy Development Act; Alberta Energy Regulator:
http://www.qp.alberta.ca/documents/Acts/r17p3.pdf (accessed Sep-
tember, 2014).
ing Company for financial support. (20) ST53: Alberta Crude Bitumen In Situ Production 1993−2013,

■ REFERENCES
(1) Facts and Statistics. Alberta Energy, Alberta Canada, 2014;;
Alberta Energy Regulator, Calgary, Alberta.
(21) Alberta In Situ Performance Presentations, 2009 - 2014. Alberta
Energy Regulator: http://www.aer.ca/data-and-publications/activity-
and-data/in-situ-performance-presentations (accessed July 2014).
http://www.energy.alberta.ca/oilsands/791.asp (accessed December (22) Alberta Government Digital Library: In Situ Oil Sands Projects’
2017). Environmental Impact Assessments 2001−2013. Alberta Energy
(2) Statistical Reports: ST-39, ST-43, ST-53; Alberta Energy Regulator: https://external.sp.environment.gov.ab.ca/DocArc/EIA/
Regulator: Calgary, Alberta. 2009−2013. Pages/default.aspx (accessed October 2014].
(3) Bergerson, J. A.; Oyeshola, K.; Charpentier, A. D.; Sleep, S.; (23) AESO Market & System Reporting, 2014, Alberta Electric System
MacLean, H. L. Life cycle greenhouse gas emissions of current oil Operator: http://www.aeso.ca/market/8856.html (accessed Decem-
sands technologies: surface mining and in situ applications. Environ. ber, 2017).
Sci. Technol. 2012, 46 (14), 7865−7874. (24) Choquette-Levy, N. Should alberta upgrade oil sands bitumen?
(4) El-Houjeiri, H. M.; Brandt, A. R.; Duffy, J. E. Open-Source LCA An integrated life cycle framework to evaluate energy system
Tool for Estimating Greenhouse Gas Emissions from Crude Oil investment tradeoffs., MSc. Thesis. University of Calgary: Calgary,
Production Using Field Characteristics. Environ. Sci. Technol. 2013, 47, Alberta, 2011.
5998−6006. (25) Akbilgic, O.; Zhu, D.; Gates, I.; Bergerson, J. Prediction of
(5) McCann, P.; Magee, T. Crude oil greenhouse gas life cycle Steam to Oil Ratio of Steam-Assisted Gravity Drainage from Reservoir
analysis helps assign values for CO2 emissions trading. Oil and Gas Characteristics. Energy 2015, 93, 1663−1670.
Journal. 1999, 97 (8), 38−43. (26) Doluweera, G.; Jordaan, S.; Moore, M.; Keith, D.; Bergerson, J.
(6) Development of Baseline Data and Analysis of Life Cycle Greenhouse Evaluating the Role of Cogeneration for Carbon Management in
Gas Emissions of Petroleum-Based Fuels; National Energy Technology Alberta. Energy Policy 2011, 39, 7963−7974.
Laboratory, U.S. Department of Energy: United States of America,
2008.
(7) Comparison of North American and Imported Crude Oil Life Cycle
GHG Emissions; TIAX LLC. and MathPro Inc. for the Alberta Energy
Research Institute: Calgary, Alberta, 2009.
(8) Life Cycle Assessment Comparison of North American and Imported
Crudes; Jacobs Consultancy for Alberta Energy Research Institute:
Calgary, Alberta, 2009.
(9) The Greenhouse Gases, Regulated Emissions, and Energy Use in
Transportation (GREET) Model, Argonne National Lobratories for the
U.S. Department of Energy, 2007.
(10) GHGenius: A Model for Lifecycle Assessment of Transportation
Fuels; (S&T)2 Consultants for Natural Resources Canada, Office of
Energy Efficiency: Ottawa, Ontario, 2013.
(11) OPGEE: the Oil Production Greenhouse gas Emissions Estimator,
Documentation ; https://pangea.stanford.edu/researchgroups/eao/
sites/default/files/OPGEE_documentation_v1.1b.pdf (accessed De-
cember 2014).
(12) Charpentier, A. D.; Oyeshola, K.; Bergerson, J. A.; MacLean, H.
L. Life cycle greenhouse gas emissions of current oil sands
technologies: GHOST model development and illustrative application.
Environ. Sci. Technol. 2011, 45 (21), 9393−9404.
(13) Brandt, A. R. Variability and Uncertainty in Life Cycle
Assessment Models for Greenhouse Gas Emissions from Canadian
Oil Sands Production. Environ. Sci. Technol. 2012, 46, 1253−1261.
(14) Vafi, K.; Brandt, A. R. Uncertainty of Oil Field GHG Emissions
Resulting from Information Gaps: a Monte Carlo Approach. Environ.
Sci. Technol. 2014, 48, 10511−10518.
(15) Ventakesh, A.; Jaramillo, P.; Griffin, M. W.; Matthews, H. S.
Uncertainty Analysis of Life Cycle Greenhouse Gas Emissions from
Petroleum-Based Fuels and Impacts on Low Carbon Fuel Policies.
Environ. Sci. Technol. 2011, 45 (1), 125−131.
(16) Directive 054: Performance Presentations, Auditing, and
Surveillance of In Situ Oil Sands Schemes; Alberta Energy Regulator:
Alberta, Canada, 2008; http://www.aer.ca/documents/directives/
Directive054.pdf (accessed September 2014).
(17) IPCC Climate Change 2013: The Physical Science Basis.
Contribution of Working Group I to the Fifth Assessment Report of the
Intergovernmental Panel on Climate Change; Cambridge University
Press: Cambridge, U.K.\, 2013.
(18) Steinmann, Z. Z. N.; Hauck, M.; Karuppiah, R.; Laurenzi, I.;
Huijbregts, M. A. J. A methodology for separating uncertainty and
variability in the life cylce greenhouse gas emissions of coal-fueled
power generation in the USA. Int. J. Life Cycle Assess. 2014, 19 (5),
1146−1155.

954 DOI: 10.1021/acs.est.7b04498

Environ. Sci. Technol. 2018, 52, 947−954