Article pubs.acs.org/est
Regional Air Quality Impacts of Increased Natural Gas Production and Use in Texas Adam P. Pacsi,† Nawaf S. Alhajeri,† Daniel Zavala-Araiza,† Mort D. Webster,‡ and David T. Allen*,† †
Center for Energy and Environmental Resources, University of Texas at Austin, 10100 Burnet Road, Austin, Texas 78758, United States ‡ Engineering Systems Division, Massachusetts Institute of Technology, 77 Massachusetts Avenue, Building E40-235, Cambridge, Massachusetts 02139, United States S Supporting Information *
ABSTRACT: Natural gas use in electricity generation in Texas was estimated, for gas prices ranging from $1.89 to $7.74 per MMBTU, using an optimal power flow model. Hourly estimates of electricity generation, for individual electricity generation units, from the model were used to estimate spatially resolved hourly emissions from electricity generation. Emissions from natural gas production activities in the Barnett Shale region were also estimated, with emissions scaled up or down to match demand in electricity generation as natural gas prices changed. As natural gas use increased, emissions decreased from electricity generation and increased from natural gas production. Overall, NOx and SO2 emissions decreased, while VOC emissions increased as natural gas use increased. To assess the effects of these changes in emissions on ozone and particulate matter concentrations, spatially and temporally resolved emissions were used in a month-long photochemical modeling episode. Over the month-long photochemical modeling episode, decreases in natural gas prices typical of those experienced from 2006 to 2012 led to net regional decreases in ozone (0.2−0.7 ppb) and fine particulate matter (PM) (0.1−0.7 μg/m3). Changes in PM were predominantly due to changes in regional PM sulfate formation. Changes in regional PM and ozone formation are primarily due to decreases in emissions from electricity generation. Increases in emissions from increased natural gas production were offset by decreasing emissions from electricity generation for all the scenarios considered.
■
INTRODUCTION Production of natural gas in the United States has increased significantly due to technological advances in horizontal drilling and hydraulic fracturing in shale formations. The annual production of dry natural gas from shale formations (shale gas) is expected to increase nationally from 4.99 trillion cubic feet (tcf) in 2010 (23% of total natural gas production in the United States) to 13.63 tcf in 2035 (49% of total projected natural gas production).1 Shale gas developments are of particular importance in Texas, which accounted for 29% of domestic natural gas production in 2010.2 The Barnett Shale in north central Texas, the Eagle Ford Shale in south central Texas, and the Haynesville Shale in eastern Texas along the Louisiana border are among the areas with the most extensive activity. Currently, the Barnett, Haynesville, and Eagle Ford shales are estimated to contain 6%, 10%, and 3% of the undeveloped dry shale gas reserves in the lower 48 states.3 A variety of studies have been undertaken to understand the environmental impacts of new natural gas developments, including impacts on water quality and availability4,5 and greenhouse gas emissions.6−8 This paper focuses on the regional air quality impacts of natural gas developments. The approach to be used in analyses presented in this work will be to consider spatially and temporally resolved air pollutant © 2013 American Chemical Society
emissions along the supply chain of natural gas, from production to use. Increased natural gas production in Texas will result in emissions of ozone (O3) and particulate matter (PM) precursors, but when that natural gas is used in generating electrical power, increased natural gas use results in decreases in ozone and particulate matter precursor emissions.9 Previous studies8,10 have found net reductions in the overall emissions of the ozone precursor NOx and the fine particulate matter precursor SO2 when natural gas is used as a replacement for coal-fired electricity generation. The analyses presented here will extend those analyses by considering the spatial and temporal patterns of the emissions and by using that information in air quality models to predict the spatially and temporally resolved impacts of the emissions. Emissions from natural gas production occur at a relatively constant rate and occur in natural gas production regions. Emissions from electricity generation have a strong diurnal variability and occur at sites that are relatively remote from the production emissions. This work will examine the net impact of these Received: Revised: Accepted: Published: 3521
November 2, 2012 February 15, 2013 February 26, 2013 February 26, 2013 dx.doi.org/10.1021/es3044714 | Environ. Sci. Technol. 2013, 47, 3521−3527
Environmental Science & Technology
Article
Emissions Inventory Development. The emissions used in this work were 2012 Future Year projections of the 2006 episode, developed by the TCEQ for the Dallas-Fort Worth SIP.12 A brief description of the model performance is available in the Supporting Information. The changes made, in this work, to the publicly available CAMx-ready emissions files14 are described below. Particulate Matter. The SIP emission inventory, used to develop plans for reducing ozone concentrations, did not contain estimates of primary particulate matter (PM) emissions. In this work, a primary PM emission inventory developed by Simon15 was used. The inventory is based on a year 2000 inventory and used size bins associated with the CMU aerosol chemistry mechanism for CAMx. For this work, the CMU species were converted to CF aerosol chemistry species, and it was assumed that primary PM emissions remained constant over time. Since the air quality impacts reported in this work focus on differences in PM concentrations between different natural gas production and use scenarios, and since primary PM emissions from EGUs are small relative to changes in PM formation due to SO2 emissions, the assumptions made in the primary PM inventory have negligible impact on the results reported here. Sulfur Dioxide. As with primary PM emissions, the SIP emission inventory did not contain SO2 emissions. In the modeled domain, SO2 emissions are primarily due to EGUs, and the methods used to estimate EGU SO2 emissions are described separately. For SO2 emissions from other sources (e.g., diesel vehicles), an inventory developed by Simon15 was used. Again, this inventory was for 2000, and because of changes in sulfur concentrations in diesel fuels, the 2000 nonEGU SO2 inventory was multiplied by a factor of 0.094 to estimate 2012 emissions. This factor was determined based on the difference in low level SO2 emissions between a TCEQ non-EGU SO2 inventory for 200616 and the year 2000 inventory that was used in this study15 over an overlapping section of the inventory domains in eastern Texas. EGU Emissions. The response of electric power generation and EGU emissions to changes in natural gas pricing was estimated using PowerWorld Simulator 16 (available from PowerWorld Corporation; www.powerworld.com).17 The model determined hourly generation in each EGU, using a nonlinear optimization algorithm that minimizes operating cost subject to meeting demand, enforcing transmission line constraints, generator unit minimum and maximum power levels, and accounting for line losses. A linear programming (LP) approach was used, which allowed the inclusion of inequality constraints. Electricity demand used in the PowerWorld simulations was the day specific generation in the Electric Reliability Council of Texas (ERCOT) grid from May 31−July 2, 2006,18 grown by 2.1% per year19 to 2012 (Figure S2, available in the Supporting Information). Additional details of the PowerWorld modeling framework used in this work for applying price signals to dispatch electricity generation within ERCOT can be found in Alhajeri9 and Alhajeri et al.20 Emissions for SO2 and NOx were based on emission factors (lb/MWh) developed using the Emissions and Generation Resource Integrated Database (eGRID) 2007.21 In this work, the response of EGU emissions to changes in the relative prices of coal and natural gas was estimated. Four scenarios, with natural gas prices of $1.89, $2.88, $3.87, and $7.74 per Million British Thermal Units (MMBTU), and all with a coal price of $1.89 per MMBTU, were used. The $7.74
spatial and temporal patterns of emissions on ozone and particulate matter formation. In addition, since the extent of use in electricity generation depends on the relative prices of natural gas and coal, this work also will examine the sensitivity of air quality impacts to natural gas prices. The maximum extent of changes to the electricity generation is limited by the natural gas capacity of the grid. In 2011, the Electricity Reliability Council of Texas (ERCOT), which is the grid that services the majority of Texas, utilized a fuel mix that was 39% coal, 40% natural gas, 12% nuclear, and 8.5% wind. Coal is typically used for base load electricity generation, with natural gas used to meet peak loads. Overall generation capacity in ERCOT is 23% coal, 57% natural gas, 7% nuclear, and 13% wind,11 so available natural gas capacity exists in the current operation of the grid. The combination of extensive natural gas production, pipeline capacities that limit the ability to distribute natural gas, and excess natural gas electricity generation capacity makes Texas an interesting test bed for modeling the air quality impacts of natural gas production.
■
METHODS Air Quality Modeling. The overall goal of this work is to estimate impacts, on regional air quality, of increases in natural gas production, coupled with changes in electricity generation driven by changes in natural gas prices. The sections below describe (1) the air quality model used to assess regional air quality impacts and (2) the methods used to estimate changes in emissions driven by increased natural gas use in electricity generation and the changes in emissions due to increases in natural gas production. Air Quality Model Development. The air quality model used in this work was developed by the Texas Commission on Environmental Quality (TCEQ) for evaluating air quality management plans for the Dallas-Fort Worth area. The air quality episode was used in the State Implementation Plan (SIP) modeling and employs meteorology from May 31−July 2, 2006. Performance evaluations of the model, comparing observed air pollutant concentrations in 2006 to predictions of the model using 2006 emissions, are reported by the TCEQ.12 In this work, meteorology and biogenic emissions from 2006 will be used, together with estimates of 2012 anthropogenic emissions, to project estimated 2012 air pollutant concentrations. Most of the estimated 2012 anthropogenic emissions, such as on-road and area source emissions, are extrapolations of 2006 emissions, performed by the TCEQ as part of the SIP development process. As described below, in this work, emissions from natural gas production and electric generation units (EGUs) will be modified to estimate impacts of changes in natural gas production and use. The air quality simulations were performed with the Comprehensive Air Quality Model, with extensions (CAMx, www.camx.com). CAMx is a three-dimensional Eulerian model which calculates the effects of emissions, chemistry, deposition, advection, and dispersion on chemical concentrations in the atmosphere. A detailed description of the model treatment of these processes as well as the computation schemes can be found in the CAMx User’s Guide.13 CAMx version 5.40 with CF aerosol chemistry and plume-in-grid (PiG) treatment for large point source plumes was the version employed. The CAMx domain (Figure S1, available in the Supporting Information) has 36 × 36 km grid cells over the eastern United States and finer grid resolution over eastern Texas (12 km × 12 km) and the Dallas-Fort Worth area (4 km × 4 km). 3522
dx.doi.org/10.1021/es3044714 | Environ. Sci. Technol. 2013, 47, 3521−3527
Environmental Science & Technology
Article
In the Barnett Shale production region, the TCEQ undertook a large data gathering campaign to obtain better spatial resolution of 2009 emissions associated with natural gas production.23 NOx and VOC emission data on approximately 20,000 individual sources were assembled. In this work, this 2009 Barnett Shale Special Inventory was used with emissions grown by 5% per year, based on production data retrieved from the Texas Railroad Commission.22 The differences between the Barnett Shale Special Inventory and the existing inventory were (1) revised emission estimates for VOC and NOx and (2) latitude and longitude spatial locators, rather than county emissions, for 60% of the emissions. The enhanced spatial distribution information was used to create spatial distributions of NOx and VOC emissions, shown in Figure S4 (in the Supporting Information). Since CO emissions were not available in the Special Inventory, CO emissions in this work were based on NOx emissions and CO to NOx ratios, by source type, in the original TCEQ inventory. For sources that did not have latitude and longitude locators, the relative spatial distributions by grid cell were assumed to be the same as for the sources that did have latitude and longitude locators. A performance evaluation of this inventory has been conducted, comparing observed hourly VOC concentrations at a location near the center of the production region and VOC concentrations predicted based on the inventory and lagrangian and eulerian air quality modeling. The performance evaluation indicates that the inventory led to ambient VOC concentration predictions that had little (
