Open-Source LCA Tool for Estimating Greenhouse ... - ACS Publications

May 1, 2013 - and James E. Duffy. ‡. †. Department of Energy Resources Engineering, Stanford University, Stanford, California 94305, United States...
0 downloads 0 Views 1MB Size
Article pubs.acs.org/est

Open-Source LCA Tool for Estimating Greenhouse Gas Emissions from Crude Oil Production Using Field Characteristics Hassan M. El-Houjeiri,† Adam R. Brandt,*,† and James E. Duffy‡ †

Department of Energy Resources Engineering, Stanford University, Stanford, California 94305, United States California Environmental Protection Agency, Air Resources Board, Sacramento, California 95812, United States



S Supporting Information *

ABSTRACT: Existing transportation fuel cycle emissions models are either general and calculate nonspecific values of greenhouse gas (GHG) emissions from crude oil production, or are not available for public review and auditing. We have developed the Oil Production Greenhouse Gas Emissions Estimator (OPGEE) to provide open-source, transparent, rigorous GHG assessments for use in scientific assessment, regulatory processes, and analysis of GHG mitigation options by producers. OPGEE uses petroleum engineering fundamentals to model emissions from oil and gas production operations. We introduce OPGEE and explain the methods and assumptions used in its construction. We run OPGEE on a small set of fictional oil fields and explore model sensitivity to selected input parameters. Results show that upstream emissions from petroleum production operations can vary from 3 gCO2/MJ to over 30 gCO2/MJ using realistic ranges of input parameters. Significant drivers of emissions variation are steam injection rates, water handling requirements, and rates of flaring of associated gas.



INTRODUCTION Emissions of greenhouse gases (GHGs) from crude oil production vary significantly depending on production practices and crude oil qualities.1−6 The use of energy-intensive secondary and tertiary recovery technologies1,6,7 can have significant impacts on emissions. Other major factors are venting, flaring and fugitive (VFF) emissions, which are difficult to measure and estimate.8−10 Previous studies show that upstream, well-to-refinery gate (WTR) emissions vary by a factor of 10 from low emissions to high emissions fields.2,6 This variability highlights the importance of having the capability to assess the different types of crude oil production operations and under different conditions. Regulatory approaches, such as the California Low Carbon Fuel Standard (LCFS) and European Fuel Quality Directive (EU FQD), seek to regulate the life cycle GHG emissions for transport fuels.11−13 These regulations rely on life cycle GHG emissions models, such as the GREET and JRC WTW studies, respectively.14,15 These existing fuel cycle models include broad coverage of a wide range of transport fuels. For example, one can compare palm oil biodiesel to oil-sands-derived diesel. These models have the advantage of being comprehensive, publicly available, and transparent. But because of their breadth, they lack process-level detail for any particular fuel pathway. For example, GREET calculates GHG emissions for petroleum products using a single average fuel mix and energy efficiency for production of crude oil. This level of modeling detail is not designed to address variations in crude oil fields and operations. Recent life cycle models of GHG emissions from crude oil production provide a more detailed treatment of crude oil emissions.2−5 Work by TIAX LLC3 used field characteristics, such as depth and electricity consumption, but is limited to the © 2013 American Chemical Society

study of specific crude types (Maya, Arab Med, etc.). The model of Jacobs Consultancy2 has the advantage of taking an engineering-based approach in the assessment of GHG emissions from crude oil production. This model is the most detailed treatment of oil production operations performed to date. Unfortunately, this model is not available for public review, and the data sources are not fully transparent. Other crude oil models such as that developed by NETL5 calculate GHG emissions by country and as such are not field-specific. To advance the modeling of crude oil production GHGs in a transparent manner, the Oil Production Greenhouse Gas Emissions Estimator (OPGEE) has been developed. 16 OPGEE is built with the goals of achieving more accuracy and better transparency in the assessment of life cycle GHG emissions from crude oil production. OPGEE calculates the energy use and emissions from crude oil production using engineering fundamentals of petroleum production and processing. This allows the model to flexibly estimate emissions from a variety of oil production emissions sources. OPGEE is constructed using Microsof t Excel for transparency and accessibility by stakeholders, including industry, governments, and members of the public.17 Also, OPGEE includes long-form model documentation, which explains model calculations, assumptions, and data sources. In this paper, OPGEE is introduced by explaining its scope and focus, structure, modeling methods, and data sources. An example is given of the methods and assumptions used in the Received: Revised: Accepted: Published: 5998

November 13, 2012 April 30, 2013 May 1, 2013 May 1, 2013 dx.doi.org/10.1021/es304570m | Environ. Sci. Technol. 2013, 47, 5998−6006

Environmental Science & Technology

Article

modeling process for key subprocesses (e.g., reservoir fluids lifting). Then OPGEE is run in default mode and on a small set of fictional fields (approximate representation of real fields), which serve to anchor sensitivity analysis. The results show the GHG emissions breakdown and the sensitivity of emissions to selected input parameters.



fugitive emissions; and accounting for upstream (fuel cycle) emissions from producing consumed fuels. A full description of OPGEE structure and a user guide is provided in sections S1 and S2 of the Supporting Information (SI). Data Requirements and Default Specifications. OPGEE models oil production emissions in more detail than previous transport LCA models. Increased modeling detail results in an increase in the number of required model input parameters. A full list of OPGEE main input parameters is described in section S2.4 of the SI. Table 1 shows key input parameters mentioned in the paper discussion.

OPGEE MODEL CONSTRUCTIONS

Model Functional Unit. The functional unit of OPGEE is 1 MJ of crude petroleum delivered to the refinery entrance (a well-to-refinery, or WTR system boundary). Emissions are presented as gCO2 equiv GHGs per MJ of crude at the refinery gate. This functional unit is held constant across different production processes included in OPGEE, and the energy content of crude oil at the refinery gate is calculated based on API gravity (no account of effects of other crude oil characteristics such as sulfur content). OPGEE defaults to lower heating value (LHV) basis for all calculations, but model results can also be presented on higher heating value (HHV) basis. Model Scope and Structure. OPGEE is modular in structure, with interlinked sheets representing each process stage. Within each major process stage, a number of activities and processes occur (e.g., fluid production or fluid injection). Seven process stages form the core of OPGEE, and emissions calculations are centered around these process stages. The process stages are described as follows: (i) Exploration, Preproduction emissions that occur during primary exploration for petroleum (generally very small in magnitude when amortized over the production life of an oil field). (ii) Drilling and development, Emissions that occur during field development and construction. Important sources include drilling and land clearing and conversion. Drilling and development emissions tend to be small because they only occur at the outset of production or sporadically during field life. (iii) Production and extraction, Emissions from the work required to lift fluids from the subsurface and to inject fluids into the subsurface. OPGEE includes the two most common artificial lifting technologies, downhole pumps (e.g., sucker-rod pumps) and gas lift. Also included are the injection energy requirements of water flooding, gas flooding, and steam flooding. (iv) Surface processing, Emissions from handling of produced crude, water, and associated gas. This includes water−oil separation, gas−oil separation, crude stabilization, gas processing, and water treatment. (v) Maintenance, Venting and fugitive emissions associated with maintenance (e.g., compressor blowdowns, well workovers and cleanups). (vi) Waste disposal, Emissions associated with waste disposal. (vii) Crude transport, Emissions from the transport of crude oil from the production site to the refinery gate. Output sheets gather the information from the process stage calculations and compile them into summed energy consumption (including energy coproduction credits) and summed GHG emissions (including any offsets from coproduced energy and debits associated with off-site emissions embodied in on site consumption). A control sheet allows users to input key parameters and display summary results. Supplementary calculations are required throughout OPGEE: gas balance accounting to track produced gas through processing, venting, flaring and sales; modeling of steam generation using simple and cogeneration schemes; calculating electricity generation emissions; estimating venting, flaring and

Table 1. Example of OPGEE Input Parameters parameter

unit

symbol

field depth reservoir pressure

ft psig

hel pres

productivity index API gravity gas−oil ratio (GOR) water−oil ratio (WOR) steam−oil ratio (SOR) flaring−oil ratio (FOR)

bbl liquid/ psi °API scf/bbl oil

PI API GOR

normal range 0−20 000 ∼0 to ≥20 000 1 to ≥100 8−45 0 to ≥3000 0 to ≥100

OPGEE default

notes

7240 0.5 ((h[ft])/ (2.31 ft/psi)) 3.0

a

30 var.

b

bbl water/ bbl oil

WOR

bbl steam/ bbl oil scf flared/ bbl oil

SOR

2.5−6.0

2.5 [(bbl)/ (bbl)] e[0.035(t−t0)] 3.0

FOR

0−3000

177

c d

a

The volume of liquid produced per unit of pressure drop between the reservoir and the bottomhole. bThe GOR default is computed from ≥100 California oil fields by binning fields by API gravity and assigning default GOR based on API gravity of crude in question. c WOR is a measure of the volume of water produced. Depleted fields have high WOR. The exponential growth rate in WOR is modeled as 0.035 [1/year] using data from multiple producing locations. t is the year being modeled t0 is the year of field discovery. dIn the case of TEOR, this measures the volume of water boiled per bbl of oil produced.

All required inputs to OPGEE are assigned default values. These defaults can be used if needed, or if data are available, can changed to match the characteristics of a given oil field or crude blend. If only a limited amount of information is available for a given oilfield, most of the input values will remain equal to defaults. In contrast, if detailed field-level data are available, a more accurate emissions estimate can be generated. For some processes and subprocesses, defaults rely on correlations and relationships with other parameters. We call these smart defaults. For example, the amount of water produced with oil (water−oil ratio or WOR) affects the energy consumed in lifting, handling, and separating fluids. If the WOR is known it can be inputted directly. If WOR is unknown, OPGEE includes a statistical relationship for water production as a function of reservoir age.16 OPGEE uses data from a variety of technical reference works. For example, emissions factors of oil production equipment fugitives (e.g., connectors, flanges, pump seals, and valves) are derived from American Petroleum Institute (API) references.18 A large number of technical references, journal articles, and fundamental data sources were consulted during the construction of OPGEE. All data sources are listed in p. S14 of the 5999

dx.doi.org/10.1021/es304570m | Environ. Sci. Technol. 2013, 47, 5998−6006

Environmental Science & Technology

Article

Figure 1. Basic structure of OPGEE. Most user inputs can be placed in the “User Inputs & Results” sheet. Calculations flow up through calculation worksheets to gathering sheets, where they are collected for display in the main sheet.

thermodynamics packages,29 achieving good agreement with published efficiencies, turbine exit temperatures, etc.16 The separation and treatment of reservoir fluids (oil, water, and associated gas) is modeled using fundamentals of process engineering.30−35 Process flow diagrams are included in the surface processing sheet for improved readability. Process flow diagrams are included for associated gas processing (aminebased acid gas removal, gas dehydration) and other fluid processing (water treatment). Flaring is used to dispose of associated gas where there is no economic use for the gas. Flaring emissions are calculated from the chemical composition of the associated gas using stoichiometry and combustion efficiencies CO2.9 Flaring volume is estimated on a country level using data from NOAA and EIA.36,37 Venting is the purposeful release of noncombusted hydrocarbon gases to the atmosphere. Venting emissions generally occur during maintenance operations and other intermittent, infrequent activities. Fugitives are the nonpurposeful or nonplanned emissions of noncombusted hydrocarbon gases to the atmosphere. Fugitive emissions commonly result from leaking equipment and tanks. The heterogeneous nature of venting and fugitives sources makes venting and fugitives emissions difficult to monitor and track. OPGEE emissions factors for venting and fugitives are mainly generated from California industry survey data.38 Venting of CO2 from the acid gas removal (AGR) unit is calculated from OPGEE gas balance assuming that all CO2 remaining in the system after flaring and other venting is removed in the AGR unit. Emissions from equipment (valves, pump seals, flanges., etc) are calculated using emissions factors from API documentation.18 Transport emissions are modeled using the method established in CA-GREET.39 The crude transport sheet allows variation in transport modes (e.g., tanker, pipeline) and in the distance traveled.

SI and described throughout the model documentation. This includes full description of the default values.



GENERAL METHODOLOGY The calculations in OPGEE are constructed using a bottom-up engineering-based approach. OPGEE relies on dozens of calculations across all stages of oil production, processing and transport. Complete details of OPGEE calculations with process flow diagrams where applicable is provided in the technical documentation part of the SI (e.g., see SI Figures S3.12 and 3.13 for flow diagrams for gas processing modeling). Figure 1 shows the overall structure and organization of the model, illustrating the flow of input data and the flow of results. User inputs generally occur on the “User Inputs & Results” sheet but can also be entered in the calculation sheets or the underlying data support sheets if desired. Lifting and production technologies included in OPGEE are the two most common artificial lifting methods: downhole pumps (e.g., sucker-rod pumps) and gas lift.19−21 Also included are the injection energy requirements of water flooding, gas flooding, and steam flooding. The lifting model used for calculating lifting energy is a single phase flow model that neglects gas phase flow for simplicity. Injection horsepower calculations are based on operating pressures and temperatures using fundamental physics.21−25 The energy required for steam flooding requires rigorous modeling of steam generation. Steam generation is modeled with two technologies: once through steam generators (OTSGs) and combined heat and power systems using gas turbines plus heat recovery steam generators (HRSGs).26−28 Energy and mass balances are used to track energy flows in these systems with good accuracy. An additional complexity is caused by the modeling of electricity cogeneration. Gas turbines are modeled at a high level using open-source 6000

dx.doi.org/10.1021/es304570m | Environ. Sci. Technol. 2013, 47, 5998−6006

Environmental Science & Technology

Article

To illustrate the level of modeling specificity included in OPGEE and the meaning of engineering-based assessment, one example is outlined below: lifting of fluids from the subsurface. Methods Example: Reservoir Fluids Lifting. Energy is required to overcome the pressure drop between the bottomhole and the well head and lift the reservoir fluids to the surface at the desired wellhead pressure. The pressure drop arises due to two factors (i) flow against gravity and (ii) frictional losses. A downhole pump is used when the reservoir pressure is not sufficient to transport crude oil and water to the surface. The main parameter in the calculation of the pressure required for lifting is the pressure drop between the bottomhole and the well head. The head is a measure of the pressure drop in feet (ft). The total head is calculated as htot = hel + hf [ft]

plift = (ptrav + pwh ) − pwf [psi]

where plift = pressure for lifting [psi], ptrav = total pressure drop [psi], pwh = wellhead pressure [psi], and pwf = bottomhole pressure [psi]. Having calculated the pressure required for lifting the power of the downhole pump is determined using24 BHPP =

fLv l2 [ft] 2Dgc

(1)

⎡ bbl liquid ⎤ ⎥ ⎢ (pres − pwf ) ⎣ psi‐d ⎦

[hp] (6)

RESULTS AND DISCUSSION For this paper, OPGEE was run on 5 sets of sample input data representing 4 fictional fields, and a generic case run using OPGEE model default. The four fictional fields are based approximately on real California oil fields. The Californiaderived fictional fields were selected to have varying characteristics and are meant to represent a variety of possible operations. We name these fields A, B, C, and D, and the set of key characteristics for each field is shown in Table 2. The Table 2. Modeled Sample Fields Characteristics

(3)

where ptrav = total pressure traverse [psi], 0.43 = fresh water gradient at 60 °F [psi/ft], htot = total head [ft], and γl = the specific gravity of the liquid (oil−water) mixture. The specific gravity of the oil−water mixture is approximated by the average of the specific gravities of water and oil weighted based on the volume fraction of each component. Given the reservoir pressure, the bottomhole pressure is calculated by subtracting the pressure drawdown between the reservoir pressure and the bottomhole pressure, indicated by productivity index (PI). This pressure drawdown causes the flow of reservoir fluids into the well21 PI =

ηP



(2)

where f = Moody friction factor, L = pipeline length [ft], vl = pipeline flow velocity [ft/s], D = pipeline diameter [ft], and gc = gravitational constant, 32.2 [lbm-ft/lbf-s2]. The pipeline length is assumed equal to the well depth. The pipeline flow velocity is calculated from the rate of liquid production (barrels (bbl) of liquid per well-day) and the cross-sectional area of the pipeline. The fluid is a mixture of oil and produced water. A column of fresh water at 60 °F exerts a gradient of ∼0.43 psi/ft.24 For a fluid other than water this gradient is multiplied by the specific gravity of the mixture at a given temperature. Accordingly the total head is converted to pressure drop in pounds per square inch (psi) using22 ptrav = 0.43htotγl [psi]

1.701 × 10−5Q dΔp

where BHPP = brake horsepower [hp], Qd = pump discharge rate [bbl liquid/d], Δp = pumping pressure [psi], and ηP = pump efficiency [%]. The pumping pressure is the difference between pump discharge pressure and suction pressure. The default suction pressure is 0 psig. Therefore the pumping pressure is equal to the pressure for lifting as calculated in eq 5. Direct combustion GHG emissions of downhole pumps are calculated from the fuel consumption using emissions factor (e.g., NG reciprocating engine emissions factor). The fuel consumption of the downhole pump is calculated by choosing the appropriate driver. For example, if the user chooses natural gas fuel, the appropriate natural gas driver is retrieved from a built-in database based on power output per driver.

where htot = total head [ft], hel = well depth [ft], and hf = friction head [ft]. The gravitational head is equal to the well depth [ft]. The frictional head is calculated using Darcy formula22 hf =

(5)

field

API gravity

depth [ft]

oil volume [bbl/d]

GOR [scf/bbl oil]

WOR [bbl water/bbl oil]

SOR [bbl steam/bbl oil]

field A field B field C field D generic

14 25 30 25 30

850 4000 8000 3000 7240

50 000 5500 2000 3000 1500

20 400 300 550 843

10 40 7 5 8.5

3 NA NA NA NA

data for the underlying California fields are derived from the online production and injection database and technical reports from the California Department of Conservation, Division of Oil, Gas, and Geothermal Resources (DOGGR).40−47 Field A is characterized by the use of TEOR which injects steam to decrease crude viscosity. Field B is characterized by very high WOR, which represents an inefficient lifting process and significant energy use to manage large amounts of water at the surface (e.g., treatment and reinjection). Field C is characterized by average depth and moderate WOR. Field D is characterized by low depth, low WOR, and higher gas−oil ratio (GOR). The “generic” case uses only the default parameters used to run OPGEE when no data are available. Variation of Output Results. Results in OPGEE are presented in a graphical format as shown in Figure 2. The variation across fields A−D is ∼3−17 gCO2eq/MJ at the refinery gate or a factor of ∼6 in variation between the lowest and highest cases.

Ql

(4)

where PI = well productivity index [bbl liquid/psi-d], Ql = rate of liquid production [bbl liquid/well-d], pres = average reservoir pressure [psi], and pwf = wellbore pressure [psi]. Increasing production rate requires an increase in pressure drawdown at a constant productivity index. Once the pressure traverse and the bottomhole pressure are calculated the pressure required for lifting is calculated as 6001

dx.doi.org/10.1021/es304570m | Environ. Sci. Technol. 2013, 47, 5998−6006

Environmental Science & Technology

Article

Field C are from production, but these are small because of moderate WOR and average depth. On the other hand, the bulk of the emissions from field D are from venting and fugitives because of higher GOR of 550 scf per bbl of oil produced, which results in higher emissions from acid gas removal and other vents and fugitive emissions. It is instructive to compare the “generic” case in Figure 2 to the four fields modeled. Note that the VFF emissions in the generic case are much higher than the simplified California fields. This is because in the absence of information about field location, OPGEE assumes the global average for the flaring rate, which is much higher than the flaring rate in California (Continental U.S. average flaring rate of 10.7 scf per bbl of oil produced).36,37 The use of defaults in OPGEE is an important factor in the uncertainty of results and is explored in detail in SI sections S3.1.3, S3.2.3−S3.8.3, S4.2.3, and S4.3.5. We explore flaring in more detail below. It is also important to note that model inputs can be in error and should be carefully vetted by model users. To prevent serious errors in analysis, OPGEE contains numerous error checks to ensure that data are consistently entered and not contradictory. For example, if gas treatment data are entered that are inconsistent, a gas balance error can result if the entered data imply a violation of conservation of mass. These error checks are described in SI section S2.4.3. Sensitivity to Input Parameters. In this section, we investigate the sensitivity of GHG emissions results to fieldspecific activity data by varying results from our four fictional oil fields. We explore variation due to WOR, field depth, oil production volume, SOR, application of a heater/treater in surface oil−water separation, and flaring rate. These parameters were chosen to be varied because they have been suggested in previous works, or believed by the authors, to be possibly significant drivers of emissions. The aim is to compare the

Figure 2. WTR GHG intensity of California fields compared to OPGEE default.

The GHG intensity of field A is high because of the use of energy-intensive TEOR. As seen in Figure 2 the bulk of the GHG emissions of field A are from the production stage because of steam generation of 3 bbl of water per bbl oil produced. The sensitivity of the GHG intensity of TEOR to steam−oil ratio (SOR) is discussed below. The GHG intensity of field B is relatively high. Field B is depleted, with WOR = 40 (e.g., it produces 40 bbl of water per bbl oil). Lifting and handling this amount of fluid is inefficient and consumes large amounts of energy. The water produced is assumed to be reinjected into the reservoir to maintain pressure, increasing the energy intensity of production. Fields C and D have relatively low GHG intensity because they do not use energy-intensive secondary and tertiary production technologies and have moderate to low WOR. The bulk of the GHG emissions of

Figure 3. Sensitivity of GHG emissions to selected input parameters. 6002

dx.doi.org/10.1021/es304570m | Environ. Sci. Technol. 2013, 47, 5998−6006

Environmental Science & Technology

Article

sensitivities under varying field characteristics and identify interactions between field characteristics. For fields C and D, WOR was increased by increments of 5 bbl/bbl. The sensitivity of WTR GHG emissions is shown in Figure 3a. Increasing field D WOR from 5 to 25 bbl/bbl causes a GHG emissions increase of ∼80%. WOR sensitivity is larger with deeper fields. Increasing field C WOR from 7 to 27 bbl/ bbl caused a GHG emissions increase of ∼170% because field C is over twice as deep as field D. WOR is generally an important input parameter and is more important in fields of greater depth. Field depth sensitivity depends on the amount of fluid lifted. For fields B and D, the field depth was increased by increments of 2000 ft and the sensitivity of WTR GHG emissions is shown in Figure 3b. Increasing field D depth from 3000 to 11000 ft caused a GHG emissions increase of ∼51%. The field depth sensitivity is significantly higher with higher WOR. Increasing Field B depth from 4000 to 12000 ft caused a GHG emissions increase of ∼120%. Field depth is an important input in fields with high WOR. Field depth otherwise has marginal importance because OPGEE assumes by default that remaining gas is exported and not reinjected. Depending on the nature of the production stream, the separation of crude oil and water at the surface using gravitational/chemical means without heat may not be sufficient to produce crude oil with the desired water content. Additional treatment may be provided by a heater/treater.32 The application of heat in the dehydration of crude oil is a significant source of fuel consumption in surface processing. Before flowing the crude oil mixture into the fired tube of the heater/treater a fraction of free water is removed to save fuel; however, significant amounts of water remain in the flowing stream and are heated. As the WOR increases more water remains in the flowing stream. Figure 3c shows that GHG emissions of Field D (low WOR) are not sensitive to the application of heat in water−oil separation. However, the application of a heater/treater in Field B (high WOR) increases GHG emissions by ∼57%. The application of the heater/treater is an important input parameter at high WOR. Future versions of the model will investigate the use of heater/treaters in high WOR oil fields (e.g., field with high WOR may operate heater treater such that water volume treated scales as a set fraction of oil treated, rather than as a fraction of produced water). Figure 3d shows that the WTR GHG emissions are not sensitive to oil production volume. An increase in energy consumption associated with lifting and processing larger amounts of reservoir fluids is offset by the increase in the energy output with higher production rate (i.e., denominator of the WTR GHG intensity grows as emissions grow). Flaring volume is a key parameter in GHG emissions from crude oil production. Global oil production by country [1000 bbl oil/d] and flaring volume by country [BCM/year] were collected for year 2010.36,37 The fraction of global oil production was computed and binned by flaring rate [scf/bbl oil]. The WTR GHG intensity of flaring for each bin was computed using midpoint values of each bin. Uncertainty is assessed using three cases. The low (90%), average (95%), and high (99.8%) flaring efficiency cases. The flare efficiency is the fraction of flared gas that is oxidized. The remaining gas is assumed to undergo fuel stripping and is emitted as unburned hydrocarbons. Flare efficiency varies with flare exit velocities and diameters, cross wind speed, and gas composition.9,10

Figure 4 shows the distribution of global oil production when plotted with flaring rate. The distribution includes 99.8% of the

Figure 4. Sensitivity of GHG emissions to flaring rate. The uncertainty shown above does not capture the full uncertainty of GHG emissions from flaring (e.g., uncertainty of satellite-based flaring estimates and uncertainty associated with other factors, such as gas composition).

global oil production. Excluded countries include Cameroon, Chile, South Africa, and Uzbekistan which together produce only 0.02% of the global oil production and have outliers flaring rates in the range of 1300−3010 scf/bbl oil. The flaring rates are country average. Field specific flaring rates may be much greater. First, more than 55% of the global oil operations have a flaring rate between 0 and 150 scf per bbl of oil produced with flaring WTR GHG intensity of ∼1.25 to 2.0 gCO2eq/MJ. This group includes U.S. continental, Canada, China, Mexico, Saudi Arabia, and Venezuela. Another 19% of global oil production has a flaring rate between 200 and 300 scf/bbl, with flaring GHG intensity of ∼2.7 to 3.0 gCO2eq/MJ. This group includes Angola, and Iran. In this bin, uncertainty contributes to a ± 0.5 gCO2eq/MJ range between the low and high efficiency cases. The bin with flaring rate of 350−400 scf/bbl represents mainly Russian oil operations, with GHG intensity of flaring between 3.0 and 4.5 gCO2eq/MJ. Lastly, the 550−600 scf/bbl oil bin represents mainly Nigerian oil operations with flaring WTR GHG intensity of 4.0−6.3 gCO2eq/MJ. Russia and Nigeria together produce ∼20% of global oil output and have the highest flaring rates among major oil producers. Another important driver of GHGs from oil operations is the generation of steam for TEOR. Data from California TEOR operations was collected for year 2009 oil production and steam injection.44 Incremental and total SOR by project was computed and binned by SOR for project size (see Figure 5, with total SOR in yellow bars, incremental SOR in blue bars). The GHG intensity for each bin was computed using field A as the basis field. Starting with field A input parameters, SOR was varied to midpoint values for each bin for a variety of cases. Default OPGEE settings were used with 0%, 50%, and 100% cogeneration volumes. Also calculated were a high efficiency case with high inlet water temperature (140 °F instead of 40 °F), and a low efficiency case with high OTSG and HRSG flue gas exit temperature (450 °F instead of 350 °F) and high thermal losses from shell (6% of fuel HHV or inlet exhaust enthalpy instead of 4%). Variation in these cases is minor 6003

dx.doi.org/10.1021/es304570m | Environ. Sci. Technol. 2013, 47, 5998−6006

Environmental Science & Technology

Article

Figure 6. Sensitivity of GHG emissions to secondary input parameters.



Figure 5. Sensitivity of GHG emissions in sample field A to variations in SOR. Yellow and blue bars represent distributions of California SORs (total and incremental SORs, respectively).

OPGEE LIMITATIONS OPGEE includes within its system boundaries over 100 listed emissions sources from oil and gas production.16 Most of these sources are not explicitly modeled because of significance cutoffs: assumed small emissions sources are neglected as (likely) insignificant in magnitude. This cutoff is applied because it would be infeasible to attempt to model the magnitude of every emissions source. OPGEE assumes single-phase fluid flow in the calculation of the pressure drop between the well-reservoir interface and the well head. In reality, there is a simultaneous flow of both fluid (oil and water) and vapor (associated gas). Results show that pressure drop calculated using a two phase flow model can be lower than that calculated using a single phase flow linear model.51 The accuracy of OPGEE results for estimating the carbon intensity of a given field is fundamentally related to data inputs available. All inputs to OPGEE are assigned default values that can be kept as is or changed to match the characteristics of a given oil field or crude blend. If only a limited amount of information is available for a given field, most of the input values will remain equal to defaults. Care must be taken in interpreting results with limited data inputs: if very limited inputs are available, the results will be similar to the generic OPGEE field. The specificity of results will improve as more data are added to the model. Flaring rates (MMscf per bbl of oil) used in OPGEE default to country-level measurements, which do not account for variations at the field level.36,37 Most fugitive and venting emissions in OPGEE are calculated using emissions factors derived from California industry survey data.38 California emissions factors are used except for CO2 venting from the AGR unit, venting from storage tanks, and fugitives from production equipment (valves, connectors, seals, etc). These data are specific to California, where environmental regulations and practices may be different than other regions. WOR is a major driver of GHG emissions. OPGEE includes a statistical relationship for water production as a function of reservoir age. This relationship is based on a set of data from a limited number of regions. This relationship looses accuracy in predicting WOR for giant fields with very high per well productivity. There are some uncertainties that are fundamental to life cycle modeling, which affect OPGEE as well. For example, in a

compared to differences between cases with and without cogeneration of steam and electric power. Figure 5 shows the distribution of California TEOR operations when plotted by SOR. Most California TEOR production (∼66%) comes from fields with SOR between 2 and 4 bbl steam/bbl water. Similar SOR distributions are seen in Alberta TEOR operations. Small amounts of production occur at SORs greater than 8 bbl/bbl, but these operations are marginally economic in general. No public data are available for other major global TEOR regions (e.g., Indonesia). A major driver of emissions results from TEOR with cogeneration is the assumed type of electricity displaced by exported power. In the default OPGEE settings, natural gas power produced in a simple cycle gas turbine is assumed to be displaced. If, instead, the grid average power mix is displaced (from the CA-GREET model, as OPGEE uses in general), then the cogeneration of power roughly offsets the steam generation emissions, resulting in a flat emissions profile with SOR. In contrast, if nuclear power or coal are displaced, the emissions slope with increasing SOR is much steeper than the default (nuclear), or even negative (coal). For an SOR of 3.5, 100% cogeneration case emissions range includes (by electricity displacement type): 10 (default, simple gas turbine), 5 (gridaverage), 35 (nuclear), and −13 g/MJ (coal). Such uncertainties with electricity displacement have been noted before,48−50 and these will be explored in more detail in future work. The sensitivity of results to values of secondary input parameters on equipment efficiencies (e.g., pump efficiency) and modeling choices is shown in Figure 6. These results were calculated by varying inputs to the OPGEE generic case with ranges of values from the literature for a number of parameters. Generic case results are not sensitive to our assumptions about operational efficiency, allocation method, driver type, well diameter, and electricity mix. The maximum sensitivity of these parameters is