Aircraft-Based Measurements of the Carbon ... - ACS Publications

Sep 11, 2009 - West Lafayette, Indiana 47906, Department of Aviation. Technology, Purdue University, 1401 Aviation Drive, West. Lafayette, Indiana 479...
0 downloads 0 Views 3MB Size
Environ. Sci. Technol. 2009, 43, 7816–7823

Aircraft-Based Measurements of the Carbon Footprint of Indianapolis K E L L Y L . M A Y S , * ,† P A U L B . S H E P S O N , †,‡ BRIAN H. STIRM,⊥ ANNA KARION,§ COLM SWEENEY,§ AND K E V I N R . G U R N E Y †,| Department of Earth and Atmospheric Science, Purdue University, 550 Stadium Mall Drive, West Lafayette, Indiana, Department of Chemistry, Purdue University, 560 Oval Drive, West Lafayette, Indiana 47906, Department of Aviation Technology, Purdue University, 1401 Aviation Drive, West Lafayette, Indiana 47907, National Oceanic and Atmospheric Administration, 325 Broadway, Boulder, Colorado 80305, and Department of Agronomy, Purdue University, 15 West State Street, West Lafayette, Indiana 47907

Received May 8, 2009. Revised manuscript received August 11, 2009. Accepted August 22, 2009.

The quantification of greenhouse gas emissions requires high precision measurements made with high spatial resolution. Here we describe measurements of carbon dioxide (CO2) and methane (CH4) conducted using Purdue University’s Airborne Laboratory for Atmospheric Research (ALAR), aimed at the quantification of the “footprints” for these greenhouse gases for Indianapolis, IN. A cavity ring-down spectrometer measured atmospheric concentrations, and flask samples were obtained at various points for comparison. Coupled with pressure, temperature, and model-derived horizontal winds, these measurements allow for flux estimation. Long horizontal transects were flown perpendicular to the wind downwind of the city. Emissions were calculated using the wind speed and the difference between the concentration in the plume and the background concentration. A kriging method is applied to interpolate the measured values to a vertical plane traced out by the flight pattern within the mixed layer. Results show the urban plume is clearly distinguishable in the downwind concentrations for most flights. Additionally, there is large variability in the measured day-to-day emissions fluxes as well as in the relative CH4 and CO2 fluxes. Uncertainties in the method are discussed, and its potential utility in determining sectorbased emission factors is shown.

Introduction In recent years, measurements of carbon dioxide (CO2) and methane (CH4) have become the focus of many scientific studies due to increasing concentrations of CO2 and the potency of CH4 as a greenhouse gas in the atmosphere (1). Though discussion is often centered on anthropogenic emissions, estimations of the carbon footprint of urban areas are still underreported, limiting our ability to understand the underlying societal drivers leading to such emissions and, * Corresponding author phone: (765)796-2404; fax: (765)494-0239; e-mail: [email protected]. † Department of Earth and Atmospheric Science, Purdue University. ‡ Department of Chemistry, Purdue University. ⊥ Department of Aviation Technology, Purdue University. § National Oceanic and Atmospheric Administration. | Department of Agronomy, Purdue University. 7816

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 43, NO. 20, 2009

in turn, to inform scientifically defensible policy approaches (2). By not only offering information about the magnitude of emissions but also providing evidence of source allocation, reliable emission estimates can provide the necessary information for decision-making regarding the reduction of anthropogenic emissions of greenhouse gases. Shao et al. (3) argue that global-scale CO2 emissions should be quantified using bottom-up approaches beginning at local scales, starting with urban areas. The quantification of emissions of the most potent and the most prevalent greenhouse gases using a method that resolves spatial heterogeneity is necessary to understand the contribution of different components of urban emissions to climate change. The most sensible approach is likely through development of high resolution emission model-data systems that incorporate the fuel consumption, activity levels, source emission characteristics, and source spatial and temporal distribution. However, for some components, especially distributed nonpoint sources such as biomass burning or natural gas leakage, the source functions are poorly characterized, and thus the area-wide emission footprints may be highly uncertain. Thus, atmospheric measurements form an integral part of the development of accurate emission data products, such as Vulcan (4). Historically, a variety of methods have been used to quantify concentrations of trace gases and the fluxes of atmospheric constituents to and from the troposphere. Often, these measurements are made from a fixed point on or near the earth’s surface in a heterogeneous environment, limiting the ability to reliably extrapolate to larger areas, e.g. an entire city. Measurements of CO2 and CH4 concentrations and fluxes have been made using flux towers in or near urban areas (5-8), instrumentation on buildings (9, 10), ground-based sampling methods (11-13), mobile measurements within the urban canopy (14-16), and via aircraft measurements (17, 18). The latter type of measurement is most often centered on regional to continental measurements in the free troposphere with less direct focus on individual city and local scale emissions into the planetary boundary layer (PBL). Aircraft studies of emissions fluxes have been completed but often with a focus on air quality issues (19-22). Trainer et al. (23) and Ryerson et al. (24) both implemented aircraft measurements for urban plume studies with the former citing an uncertainty of a factor of 2, primarily due to uncertainty in the wind speed. In the current study, a mass balance approach was used to determine the emissions flux from an urban area. The method relies on the assumptions of constant emissions and constant PBL characteristics, or stationarity, to calculate the emissions. Recent developments in measurement technology have enabled high frequency and high precision measurements of CO2, CH4, H2O, and other gases from aircraft and other platforms (25). Using these with a mass balance approach affords the opportunity for accurate inference of emission fluxes across a range of spatial and temporal scales. In this study aircraft observations of PBL concentrations of CO2 and CH4 in the downwind urban plume are used in conjunction with horizontal wind speed and direction to calculate an emissions flux from Indianapolis, IN. Here, we discuss the sources of uncertainty in the method and the variability in the determined absolute and relative CO2 and CH4 fluxes.

Materials and Methods Instrumentation. Eight experimental flights, five of which took place in March and April 2008, were conducted in the 10.1021/es901326b CCC: $40.75

 2009 American Chemical Society

Published on Web 09/11/2009

PBL surrounding Indianapolis, IN (39.77° N, 86.16° W). The remaining flights took place in the following fall/early winter of 2008/2009. Measurements of CO2 and CH4 concentrations, in addition to atmospheric temperature and pressure, were conducted from an aircraft platform known as the Airborne Laboratory for Atmospheric Research (ALAR, http://www. chem.purdue.edu/shepson/alar.html). The Beechcraft Duchess aircraft was equipped with a Best Air Turbulence (BAT) probe, a global positioning system and inertial navigation system (GPS/INS), a Picarro ESP-1000 CO2/CH4 cavity ringdown spectrometer (CRDS) and in-flight calibration system, and a Programmable Flask Package (PFP). The BAT probe (26-29) measures pressure variations across the probe hemisphere at 50 Hz. GPS/INS data were also acquired at 50 Hz, using a Novatel Black Diamond System, which when combined with the BAT probe data yields the 3D wind velocities (28, 29). Within the central pressure port of the BAT probe is a microbead temperature sensor. For the present study, uncertainties in the computed two-dimensional horizontal wind from the BAT probe measurements were found to be greater than those from the EDAS 40-km model winds from the National Oceanic and Atmospheric Administration (NOAA) Air Resource Laboratory (ARL) READY archive (30), which we thus use in the flux calculation. BAT probe-derived winds were used to characterize the upper bounds to the uncertainty in the winds. The CRDS was connected to a 5 cm-diameter PFA-Teflon sampling manifold which used a blower to pull atmospheric air (at ∼0.1 s residence time) from the nose of the plane to the location of the analyzer. The CRDS sampled this air through a magnesium perchlorate (MgClO4) dryer to remove moisture, at approximately 300 cm3/min. Once the sample is introduced into the optical cavity of the spectrometer, which is equipped with three high reflectivity mirrors, laser light is injected through one of the mirrors and allowed to decay while being monitored by a photodetector behind the second mirror. The analyte concentrations are related to the time constant of decay, considering losses from the cavity mirrors, for selected absorption lines in the CO2 and CH4 IR spectra. Due to the reflectivity of the mirrors in the cavity, the effective path length for the CRDS is long, allowing for greater sensitivity (25). Concentration data were recorded at 0.2 Hz. The experimental setup allowed for a reference sample to be obtained every several minutes during flight, using cylinders filled with compressed air collected at Niwot Ridge, CO and calibrated at NOAA. The concentrations of CO2 and CH4 in the reference gas were 385.76 ( 0.04 and 1.84 ( 0.01 µmol/mol (hereafter referred to as ppm), respectively, for flights completed in the spring of 2008. Flights flown at the end of 2008 and in early 2009 used reference gas concentrations of 389.08 ( 0.04 (1σ throughout) ppm for CO2 and 1.83 ( 0.01 ppm for CH4. During postprocessing, ambient concentration data from the CRDS were corrected to the reference gas values from in-flight calibrations taken every 4 min for spring flights and approximately every 20 min for fall flights and compared to concentrations in the PFP samples taken at various points in flight and analyzed at NOAA. When calibration data are applied, corrections are on the order of (0.25 ppm for CO2 and (1.80 nmol/mol (hereafter referred to as ppb) for CH4. When the corrected concentration values were compared to the PFP measurements, average differences were (0.31 ppm (0.08%) for CO2 and (2.29 (0.13%) ppb for CH4. Flight Design. The experimental periods were chosen to avoid complications due to negative flux components related to primary production in the growing season in this agriculturally intensive regional environment. Flights were conducted on March 28, April 2, April 14, April 15, April 21, November 23, December 20, 2008, and January 7, 2009. To

the extent possible, the data were acquired during the middle of the day (for a, typically, ∼2 h period) when the PBL was fully developed and temporal variability in sources and winds were minimal. Thus, stationarity in the pressure, temperature, winds, and sources of gas were assumed as well as horizontal homogeneity in the wind speed and direction. The Supporting Information provides a discussion of the meteorological conditions and synoptic situation during each flight. Prior to each flight Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) particle dispersion simulations (31) were used to determine wind speed and wind direction to correctly orient the planned flight path perpendicular to the prevailing wind direction. Several flat, level transects were flown throughout and just above the PBL top, in addition to at least one spiral vertical profile per flight. The flight transects were oriented such that they sampled a vertical plane normal to the plume flowing from the urban source region. Transects were vertically stacked with maximum 200-300 m separation and generally ranged in length from 60 km to 100 km. The average diameter of the Indianapolis urban environment is 41 km. The vertical profiles were used to determine the mixed layer height from potential temperature (θ), using on-board measurements of pressure, temperature, and humidity (Vaisala HUMICAP HMT 330). On March 28, December 20, and January 7, vertical profiles could not be completed due to low cloud layers. For these three flights, the base of the cloud layer was assumed to be the upper limit of the mixed layer, as determined by flying to the base of the clouds and noting the altitude. For the remainder of the flights, the vertical profile data were first binned in 20 m vertical bins and a running mean and standard deviation were calculated beginning from the lowest altitude. The height at which the potential temperature increased beyond a threshold (mean + 3σ) for at least 100 m in the vertical was chosen as the mixed layer height. The Supporting Information shows PBL height estimates for each case. For our calculations, we assume that all emissions from the city have not been mixed vertically any further than the PBL top before they are ventilated out through the sample plane. Due to the time resolution of the measurements (0.2 Hz for concentrations and 50 Hz for GPS/INS and BAT probe measurements), a typical true airspeed of approximately 70 m per s corresponds to approximately 350 m of horizontal resolution of trace gas data and 1.4 m of horizontal resolution for BAT probe and GPS/INS data. All data were interpolated on to a regular grid and used with background concentrations to calculate an integrated net mass flow through the vertical plane. This flow was then used to calculate the emissions flux from the city. Indianapolis includes one large coalburning power plant, several small commercial and private airports, and the larger Indianapolis International Airport. In addition, a network of natural gas and refined product pipelines as well as composting facilities, wastewater treatment plants, and landfills are located within the urban area. Kriging. Interpolation was performed by kriging (32) the data to a regular gridded two-dimensional plane, using Matlab-based ‘EasyKrig3.0’ (33). In this method a semivariogram of the data was produced and modeled using a general exponential Bessel relationship (eq I) (31).

[

( )]+γ

γn ) Co 1 - Jo(bh)e-

h L

P

(I)

o

Here, h is the lag, γo is the nugget, or offset from zero of the y-intercept, Co is the sill-nugget, Jo is a Bessel function, L is the length scale, b is the length scale of the hole effect, and p is the power. These parameters can be adjusted for each data set to produce a good model fit of the semivariogram, as evaluated by the Q1 and Q2 validation criteria. The uncertainty in the estimate is assessed using the variance of the kriged result at each point. VOL. 43, NO. 20, 2009 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

7817

For each flight, the sample data was kriged to a grid of 250 m resolution in both the horizontal and the vertical dimensions, with the exception of the December 20, 2008 flight, which was kriged to a grid of 250 m in the horizontal and 225 m in the vertical, due to a low mixed layer height. The vertical dimension was interpolated from the ground to the top of the mixed layer, while the horizontal dimension was kriged from end to end of the flight transect. Kriging was performed on the CO2, CH4, wind speed, temperature, and pressure observations to obtain both the estimate and the variance of the estimate for each variable at each grid point. Only the component of the wind perpendicular to the flight path was kriged and used in the flux calculation. Flux Calculation. After kriging, the regional background concentrations for both trace gases were estimated using the edges of each downwind kriged plane. By extending the flight path perpendicular to the mean wind vector over the city beyond the limits of the urban plume it was possible to obtain background concentrations of both CO2 and of CH4. The background concentration is determined by taking the average of the edges of kriged plane and is subtracted from each grid point on the downwind plane to determine the net concentration. More information on how the background concentrations were determined can be found in the Supporting Information. The net molecular concentration on the downwind side of the city is ascertained using the ideal gas law, the kriged temperature, and the kriged pressure values and is subsequently multiplied by the wind speed at each grid point to determine the net molecular flow through the vertical plane. Finally, the net molecular flow was integrated in the horizontal and vertical directions and divided by the area of city to determine an area-averaged emissions flux for both CO2 and CH4 from Indianapolis. Equations II and III indicate the flux calculation method.

∫ ∫ zi

FC )

o

x

-x

Results and Discussion ([C]ij - [C]b)*U⊥ijdxdz Acity

(C)b )

∑ (C )

ij edge

n

(II)

(III)

Here, FC is the area-averaged emissions flux (µmol/m2 s) of the trace gas, -x and +x are the minimum and maximum horizontal transect distance limits, respectively, the subscript b refers to the background concentration, and the over bar indicates an average. U⊥ij is the gridded wind vector perpendicular to the flight path, and dx and dz are the horizontal and vertical grid spacing. Acity refers to the area extent of the city. The area was estimated using population density data from the 2000 U.S. Census (34). Using a lower threshold of 460 people/km2 to define the urban area results in a city area (shown in Figure 3, below) of 728.35 km2. This then yields an average population density of ∼1800/km2. Vulcan. The Vulcan Project (4) has quantified hourly fossil fuel CO2 emissions for the United States at the subcounty spatial scale, for the year 2002, and is a key component of attributing CO2 fluxes within the North American Carbon Program. Vulcan approached quantification of fossil fuel CO2 by leveraging information contained within a series of regulatory and monitoring agencies including the EPA’s Acid Rain Program, the EPA’s National Emissions Inventory for the assessment of nationally regulated air pollution, the Department of Energy, the U.S. Census and the Department of Transportation. Utilizing the inventory emissions of carbon monoxide and nitrogen oxides combined with emission factors specific to combustion device technology, the Vulcan Project has calculated CO2 emissions for industrial point sources, power plants, mobile sources, and residential and 7818

9

commercial sectors, with information on fuel used and source classification information. Point sources include geocoded electrical generating power stations and industrial stacks. Nonpoint sources refer to nongeocoded stationary sources, generally residential and commercial, that represent relatively scattered sources throughout the area of study and are downscaled using U.S. Census data and GIS road networks. Finally, mobile source emissions come from county-level aggregated data and subsequently are placed on a geocoded road network (4). In addition to the “native” resolution (geocoded points, census tracts, roadways), emissions are placed onto a regular 10 km × 10 km grid. A subset of Vulcan output for those grid cells overlapping Indianapolis was used for the days and hours of flight corresponding to the model output days in 2002. For example, March 28, 2008 was the fourth Friday of March. Therefore, the corresponding fourth Friday in March of 2002, March 29, was chosen for comparison. The carbon emissions for each hour of flight over the corresponding subset of grid cells were summed and divided by the total city area and the time of emission to calculate an emissions flux. The resulting values were compared to the individual experiment day measurements to evaluate the magnitude to which Vulcan represents an average emissions estimate. Vulcan version 1.0, used in this study, does not include sources such as aircraft emissions, off-road vehicles, and any temporal resolution for the residential and commercial emissions, though temporal structure is available for the point source and mobile emissions sources (4). The Vulcan values from 2002 were adjusted to 2008 or 2009, depending on the date, by using statewide trends in energy consumption data from the Energy Information Administration (35). The adjustments range from an increase of 8.4% to 9.6% over the 2002 values.

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 43, NO. 20, 2009

Kriging. Fluxes of CO2 and CH4 from Indianapolis can come from several sources, including mobile source fossil fuel combustion on the road network in and around the city, distributed point sources such as individual houses and office buildings, and the large coal-fired power plant, the Harding Street facility, located on the southwest side of the city. CH4 can also be emitted from open air landfills, wastewater treatment facilities, and natural gas pipeline leaks. In Figure 1, we show a sample set of the data from the experiment transects for April 21 (for the portion flown in the boundary layer) along with the April 21 kriging results, illustrating both the positioning of the transects as well as the degree to which the observed data are represented by the kriging results. For this day the wind direction was ∼140° at ∼2 to 5 m/s. The calculated background concentrations for CO2 and CH4 were 399.0 ( 1.9 and 1.91 ( 0.010 ppm. The downwind concentration values clearly show the urban plume in each case, captured within the sample vertical plane normal to the wind direction (see the Supporting Information for kriging results from each flight), and that the plume concentrations are well above the uncertainty in the background concentrations. The horizontal axis denotes the distance in kilometers from the center point of the ground track traced out by each horizontal transect during flight. The kriging uncertainties are shown for April 21, 2008 in Figure 2. There are no measurements on the ground to constrain interpolated estimates below ∼200 m. The general centers of the large CO2 and CH4 plumes are colocated on the majority of days, but the plots do not necessarily show the same location of smaller more distributed areas of elevated concentrations (e.g., at ∼-27 km for CH4). This reflects the unique source characteristics for these two species. While both are emitted from combustion sources, CH4 has multiple noncombustion sources. For most cases, elevated concentrations of both gases can be seen in

FIGURE 1. April 21st downwind observed CO2 (ppm) (a) and kriged CO2 (b) observed CH4 (ppb) (c) and kriged CH4 (d) concentrations.

FIGURE 2. April 21, 2008 kriging uncertainty of (a) CO2 and (b) CH4. association with the power plant plume (Figure 3). See the associated methane plot in the Supporting Information. Emissions Flux. Net concentrations are used to calculate the flux through the vertical plane and are defined at each grid point by the average background concentration subtracted from the downwind concentration. The uncertainty of the background concentrations was calculated as the standard deviation of the mean of the edge grid points. The average background concentrations for all flights were 396.5 ( 3.2 ppm and 1917 ( 33 ppb, for CO2 and CH4, respectively. These mean background values for these winter season measurements are greater than the free tropospheric averages of 387 and 1.8 ppm (36), reflecting the fact that the boundary layer background contains slowly varying slightly elevated concentrations of these gases from the much more widely distributed upwind rural emissions sources. Table 1 shows the results of the flux calculation for each day of flight. The

uncertainties reported represent the total propagated uncertainty, which is discussed below. Large day-to-day variability in the calculated flux of each trace gas is evident. The day showing the maximum calculated flux for CO2 is November 23, 2008 at 44.7 ( 1.9 µmol/m2 s and for CH4 is January 7, 2009 at 0.32((0.06) µmol/m2 s. The smallest measured flux of CO2 and CH4 occurred on April 2, 2008. The two main sources of CO2 for this study are considered to be the mobile source and power plant combustion emissions, based on the high traffic in the urban area as well as the production capacity of the power plant. It is assumed that because the flights took place at approximately the same time each day, and took place on weekdays for the majority of flights, traffic fluctuations were not the largest cause of variability in the CO2 fluxes. Figure 4a illustrates the traffic count (number of vehicles passing VOL. 43, NO. 20, 2009 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

7819

FIGURE 3. 3-D view of Indianapolis (blue outline) with CO2 concentrations (ppm) along flight tracks, with approximately perpendicular winds, at ∼140°. Red lines are highways, and the black dot represents the Harding Street power plant.

TABLE 1. Results of the Flux Calculation Indicating the Total Flux of Both CO2 and CH4 from Indianapolis for Each Flight, as an Area-Averaged Total Flux, and the Ratio of the Two Fluxesa date

flux of CO2 (µmol/m2 s)

flux of CH4 (µmol/m2 s)

flux CH4/flux CO2 (µmol CH4/µmol CO2)

adjusted Vulcan flux of C (µmol/m2 s)

March 28, 2008 (Friday) April 2, 2008 (Wednesday) April 14, 2008 (Monday) April 15, 2008 (Tuesday) April 21, 2008 (Monday) November 23, 2008 (Sunday) December 20, 2008 (Saturday) January 7, 2009 (Wednesday) mean ((1σ)

11.1 ( 0.6 3.4 ( 0.5 13.5 ( 0.9 18.7 ( 0.7 8.5 ( 0.7 44.7 ( 1.9 41.4 ( 1.3 11.9 ( 1.2 19.2 ( 15.4

4.5 ((0.12) × 10-2 1.6 ((0.19) × 10-2 7.0 ((0.58) × 10-2 1.4 ((0.04) × 10-1 1.1 ((0.05) × 10-1 1.9 ((0.12) × 10-1 2.4 ((0.07) × 10-1 3.2 ((0.06) × 10-1 0.14 ( 0.10

4.1 ((0.25) × 10-3 4.7 ((0.89) × 10-3 5.2 ((0.57) × 10-3 7.5 ((0.34) × 10-3 1.3 ((0.12) × 10-2 4.3 ((0.33) × 10-3 5.8 ((0.26) × 10-3 2.7 ((0.27) × 10-2 8.9 ((7.9) × 10-3

13.4 14.5 12.6 12.2 12.4 9.5 11.4 12.5 12.3 ( 1.4

a

The final column illustrates the results from Vulcan.

FIGURE 4. (a) Traffic flow (number of vehicles passing sensor per hour) for each day of flight at representative sensor on I-65 on the northwest side of the city. (b) Regression of CO2 and CH4 fluxes against traffic flow. sensor per hour; InDoT) for each day of flight, while Figure 4b shows the regressions of the measured fluxes against traffic count. When the CO2 and CH4 fluxes are regressed against traffic count, r2 values of 0.57 and 0.03 are obtained. The slope of the regression with CO2 is significantly different from 0, but the slope of the regression with CH4 is not, at the 95% confidence level. Thus, we can account for 57% of the variability in CO2 flux with the mobile source variability. Figure 5 illustrates the CO2 emissions from the Harding Street power plant on the experiment days (data from the 7820

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 43, NO. 20, 2009

Indiana Department of Environmental Management). November 23, December 20, and January 7 were estimated using a regression of CO2 emitted against MW generated (r2)0.998), as the late 2008 and early 2009 data were not yet available. Using the area of the city and converting the power plant data into units of µmol/m2 s and taking an average ratio of those emissions to total calculated emissions, it was determined that the power plant emissions accounted for 37% of the emissions flux of the city. With the exception of April 21, there is not much daily variation in emissions from the power plant. We expect that this does not necessarily affect the

FIGURE 5. Time series of CO2 emissions (tons CO2/h) from the Harding Street power plant for experiment days (Indiana Department of Energy Management). variability in CH4 emissions, for several reasons, including that the FCH4/FCO2 is quite variable for combustion sources (depending on flame conditions) and because combustion (e.g., the power plant) does not appear to be a dominant source for this urban environment. Vulcan. The flux estimates from the Vulcan inventory exhibit much less temporal variability than for the aircraftbased estimates. Hourly estimates of CO2 emissions from Vulcan are based on average or ‘typical’ emissions dynamics for all sources other than the electric power production emissions, which are based on hourly CO2 stack monitoring. Hence, strict coherence between the measurement results and the “bottom-up” approach is not expected. While there is less day-to-day variability in the Vulcan results versus the aircraft-based estimates, inventory estimates do fall within the range of CO2 fluxes calculated using our aircraft mass balance approach. See the Supporting Information for an example of the Vulcan output. The final calculated fluxes of carbon for Indianapolis using the Vulcan inventory are shown in Table 1, in µmol/m2 s. The urban average value of 12.3 ( 1.4 µmol/m2 s is not statistically significantly different from our mean of 19.2 ( 15.4 µmol/m2 s. For each grid cell, we calculated its contribution to the total by multiplying the fraction of that grid cell’s area that was within the urban area, multiplied by the grid cell’s emission quantity, and then summed the results for all grid cells. The exception is the cell containing the emissions from the Harding Street power plant, for which the entire cell’s emissions were used. However, since the aircraft detects all incremental CO2 upwind of the aircraft, even if it were outside of the defined urban area, we might expect the aircraft results, when expressed as an average urban flux, to be somewhat larger than the calculated Vulcan flux. A sensitivity test shows that the Vulcan results are not very sensitive to enlargement of the urban area considered. It is prudent to note that November 23 and December 20 were weekend days (a Sunday and Saturday, respectively). The Vulcan-estimated carbon fluxes for both of these days were lower than the other, weekday flights. This is most noticeable for Sunday, November 23, with a flux of 9.5 µmolesC/m2 s. The largest Vulcan-estimated carbon flux corresponded to the April 2, 2008 flight with a flux of 14.5 µmolesC/m2 s. A similar comparison for methane emissions can be done using the Emission Database for Global Atmospheric Research (EDGAR) 3.2 Fast Track 2000 results (http://www. mnp.nl/edgar/), a 1 × 1 degree gridded anthropogenic CH4 emissions model. The difference between the grid cell containing Indianapolis and the average of the surrounding grid cells is 0.005 ( 0.01 µmol/m2 s, indicating that EDGAR does not show the Indianapolis grid cell emissions as

significantly different from the surrounding grid cells. Our observations that show clear CH4 plumes from the urban environment are not consistent with that result. The relatively low resolution of the model grid may be a contributing factor to this result and indicates the need for a higher resolution methane emissions model for comparison to calculated urban fluxes. Sources of Uncertainty. One of the largest sources of uncertainty in the calculation of fluxes is the wind speed. Trainer et al. (23) note the high uncertainty (a factor of 2) in their calculation of emissions rates, mainly due to problems in determining the wind speed, a problem which also applies to this study. The relative uncertainty in the wind speed was determined to be 39%, on average, using the difference between the BAT probe-determined winds and the EDAS winds as an upper limit to the uncertainty. It is possible that some of this upper-limit uncertainty reflects a scale difference in the calculated winds as compared to the EDAS winds. CO2 measurement errors both as compared to the reference gas ((0.25 ppm) and as compared to the flask measurements ((0.31 ppm) were smaller than the resulting kriging standard deviations ((1.04 ppm). The measurement errors for CH4 were on the same order as the kriging standard deviations, and both were considerably smaller than the signal of enhanced methane in the urban plume. Therefore, the standard deviations of the kriged results at each grid point were used to represent the uncertainties for the gases and were subsequently propagated through the molecular density and flux calculations. The same is true for the measurements of ambient temperature and pressure, used in the molecular density calculation. The uncertainties reported with the final flux values were the result of this multivariable error propagation through the equations used to calculate the final fluxes. Another large source of uncertainty derives from the uncertainty of background concentrations, which represents, on average, 66% of the total propagated flux uncertainty. Given that the propagated uncertainty for individual days of flux measurements are small compared to the day-to-day variability, it is likely that measurement uncertainties, including the background uncertinty, are not the source of the flux variability. We recognize that a potentially important source of error is associated with the statistics and variability of sampling the point source plumes. For example, if the plume moves vertically between legs of flight, the plume and thus the flux may be overestimated. Similarly, the plume, if well-defined, may be under-sampled, given the limited number of transects possible at different heights in the boundary layer, leading to an underestimate. It is likely that these sampling errors represent a large component of the day-to-day variability in the measured fluxes. Since a sampling requirement is stationarity in the sources and VOL. 43, NO. 20, 2009 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

7821

atmospheric conditions, this situation could be improved through use of a higher flight speed, which would, however, also require a higher measurement frequency. Additionally, sampling in different mixing regimes may add uncertainty that is difficult to quantify. During strong mechanical mixing in the boundary layer, the urban plume will be broader and less well-defined, and therefore the edges of the plume, the background concentrations, are more difficult to define. If the plume exhibits more diffuse edges, then the background concentrations may be overestimated, leading to an underestimation of urban emissions. During weaker turbulent mixing, the urban plume edges are more well-defined, in which case the background values are much easier to define. However, strong buoyant mixing may cause more exchange with the free troposphere, leading to “lost” emissions out of the top of the boundary layer. In this case, the urban emissions will again be underestimated.

Discussion While emissions inventories are in principle straightforward for CO2, this is not the case for CH4, for which source functions are less well-defined (37). The results of spatially resolved regressions of the fluxes of CO2 and CH4 are helpful when discussing the sources of the gases in an urban area. For our overall fluxes, the average value for FCH4/FCO2 ) 0.0089 ( 0.0079 for the entire downwind integrated plane. Nam et al. (38) determined that the average ratio of CH4 emission to CO2 emission, i.e. the CH4 emission factor, for the U.S. vehicle fleet is (1.38 ( 0.11) × 10-4 µmol CH4/µmol CO2, i.e. 64 times smaller. Based on this, we conclude that mobile sources are not significant contributors to the total CH4 flux, consistent with the poor regression for CH4 in Figure 4. If we regress FCH4 against FCO2 through the vertical plane for only those grid cells within the apparent urban plume for all days, by taking the grid cells with a flux value greater than the mean background flux +3σ, the slope was determined to be (5.3 ( 0.10) × 10-3 µmol CH4/µmol CO2 with an r2 value of 0.50. Thus while 50% of the variability in CH4 flux appears to be spatially related to combustion sources, our average plumeonly flux ratio is 2 orders of magnitude larger than the emission factor for coal combustion, i.e. (1.85 ( 0.24) × 10-5 µmol CH4/µmol CO2, calculated from Supplement E of the EPA AP 42 publication (39) and 1 order of magnitude larger than the emission factor for coal combustion calculated from Babbitt and Lindner (40) of 3.04 × 10-4 µmol CH4/µmol CO2. It must be that for this environment, most of the CH4 emissions derive from noncombustion sources, e.g. natural gas distribution network, wastewater treatment facilities, and solid waste landfill sites. The ranges of fluxes calculated in this study can be compared to fluxes from other urban areas. Using eddy covariance, Velasco et al. (7) calculated an average CO2 flux from a neighborhood in southeast Mexico City, with a population density of 12,000 people/km2 (i.e., ∼7 times greater than Indianapolis) of 9.3 µmol/m2 s, with a range from -5 to 36.4 µmol/m2 s. Additional studies indicate average CO2 fluxes of 26 µmol/m2 s (9) during fall in Edinburgh (population density ∼1700/km2) and a range of 10 (nocturnal) to 38 (hourly averages) µmol/m2 s of CO2 (5) during summer in a Chicago neighborhood. Kuc et al. (13) presented an average CH4 flux of 0.039 µmol/m2 s from Krakow over the course of a 2 year study, about a factor of 2.5 lower than our average. Most urban studies of this nature are completed using eddy covariance or stationary continuous monitoring stations and are therefore limited by the immobility of the measurement platform and the impact of urban scale heterogeneity. The use of an aircraft in the manner presented here provides the ability of the measuring platform to fully integrate the entire urban plume. 7822

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 43, NO. 20, 2009

This study presents a method of emissions flux measurement from an urban area that is inherently useful due to its integrative nature. It is possible with this method to both identify and quantify emissions from large point sources and from integrated large area sources. The results of this study indicate significant variability in the fluxes of CO2 and CH4 from Indianapolis, resulting in part from the variability in the emissions from point sources and, to a lesser extent, distributed sources. There is also likely a contribution from inadequate spatial sampling. The actual variability is currently underrepresented in inventory estimates but is an important part of understanding the emissions and trends of emissions in urban centers. Our relative CH4 and CO2 fluxes indicate that CH4 is derived mostly from noncombustion sources. This is likely from gas pipeline and pumping station leakages and from sources such as sewage treatment facilities and landfill sources (41). The combination of sources can create difficulty in determining the source allotment in a highly variable system, such as an urban area. However, the aircraft platform is a highly useful tool for determining total fluxes from urban areas due to its mobility both in the horizontal and vertical and its ability to perform maneuvers such as vertical profiling to gather more information about the state of the atmosphere throughout the PBL and into the free troposphere.

Acknowledgments Funding for this project was provided in part from the Showalter Foundation and the EPA STAR grant R833750. The authors acknowledge Dan Hancock at the Indiana Department of Energy Management and Dr. Darcy Bullock at Purdue University for their assistance in obtaining power plant and traffic data, respectively. This is paper number 0909 of the Purdue Climate Change Research Center (PCCRC).

Supporting Information Available Information on background concentrations, final kriged concentration plots, the use of vertical profiles, and the meteorological conditions during flights. This material is available free of charge via the Internet at http://pubs.acs.org.

Literature Cited (1) Forster, P.; Ramaswamy, V.; Artaxo, P.; Berntsen, T.; Betts, R.; Fahey, D. W.; Haywood J.; Lean, J.; Lowe, D. C.; Myhre, G.; Nganga, J.; Prinn, R.; Raga, G.; Schulz, M.; Van Dorland, R. Changes in Atmospheric Constituents and in Radiative Forcing. In Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change; Solomon, S., Qin, D., Manning, M., Chen, Z., Marquis, M., Averyt, K. B., Tignor, M., Miller, H. L., Eds.; Cambridge University Press: Cambridge: United Kingdom and New York, NY, USA, 2007. (2) Lieberman, J.; Warner, M. America’s climate security act of 2007, S. 2191, 110th Congress, U.S. Senate, 2007. (3) Shao, G.; Qian, T.; Liu, Y.; Martin, B. The role of urbanisation in increasing atmospheric CO2 concentrations: think globally, act locally. Int. J. Sustainable Dev. Ecol. 2008, 15 (4), 302–308. (4) Gurney, K. R.; Mendoza, D. L.; Zhou, Y.; Fischer, M. L.; Miller, C. C.; Geethakumar, S.; de la Rue du Can, S. High resolution fossil fuel combustion CO2 emissions fluxes for the United States. Environ. Sci. Technol. 2009, 43 (14), 5535–5541. (5) Grimmond, C. S. B.; King, T. S.; Cropley, F. D.; Nowak, D. J.; Souch, C. Local-scale fluxes of carbon dioxide in urban environments: methodological challenges and results from Chicago. Environ. Pollut. 2002, 116, S243–S254. (6) Grimmond, C. S. B.; Salmond, J. A.; Oke, T. R.; Offerle, B.; Lemonsu, A. Flux and turbulence measurements at a densely built-up site in Marseille: heat, mass (water and carbon dioxide), and momentum. J. Geophys. Res 2004, 109 (D24), . (7) Velasco, E.; Pressley, S.; Allwine, E.; Westberg, H.; Lamb, B. Measurements of CO2 fluxes from the Mexico City urban landscape. Atmos. Environ. 2005, 39 (38), 7433–7446. (8) Vogt, R.; Christen, A.; Rotach, M. W.; Roth, M.; Satyanarayana, A. N. V. Temporal dynamics of CO2 fluxes and profiles over a

(9)

(10) (11)

(12) (13) (14)

(15) (16) (17)

(18)

(19) (20) (21)

(22)

(23)

(24)

Central European city. Theor. Appl. Climatol. 2006, 84 (1-3), 117–126. Nemitz, E.; Hargreaves, K. J.; McDonald, A. G.; Dorsey, J. R.; Fowler, D. Micrometeorological measurements of the urban heat budget and CO2 emissions on a city scale. Environ. Sci. Technol. 2002, 36 (14), 3139–3146. Pataki, D. E.; Bowling, D. R.; Ehleringer, J. R.; Zobitz, J. M. High resolution atmospheric monitoring of urban carbon dioxide sources. Geophys. Res. Lett. 2006, 33 (3), . Moriizumi, J.; Nagamine, K.; Iida, T.; Ikebe, Y. Carbon isotopic analysis of atmospheric methane in urban and suburban areas: fossil and non-fossil methane from local sources. Atmos. Environ. 1998, 32 (17), 2947–2955. Jime´nez, M. T. F.; Climent-Font, A.; Anto´n, J. L. S. Long term atmospheric pollution study at Madrid City (Spain). Water, Air, Soil Pollut. 2003, 142 (1-4), 243–260. Kuc, T.; Rozanski, K.; Zimnoch, M.; Necki, J. M.; Korus, A. Anthropogenic emissions of CO2 and CH4 in an urban environment. Appl. Energy 2003, 75 (3-4), 193–203. Soegaard, H.; Møller-Jensen, L. Towards a spatial CO2 budget of a metropolitan region based on textural image classification and flux measurements. Remote Sens. Environ. 2003, 87 (2-3), 283–294. Henninger, S.; Kuttler, W. Methodology for mobile measurements of carbon dioxide within the urban canopy layer. Clim. Res. 2007, 34 (2), 161–167. Henninger, S. Analysis of near surface CO2 variability within the urban area of Essen, Germany. Meteorol. Z. 2008, 17 (1), 19–27. Kort, E. A.; Eluszkiewicz, J.; Stephens, B. B.; Miller, J. B.; Gerbig, C.; Nehrkorn, T.; Daube, B. C.; Kaplan, J. O.; Houweling, S.; Wofsy, S. C. Emissions of CH4 and N2O over the United States and Canada based on a receptor-oriented modeling framework and COBRA-NA atmospheric observations. Geophys. Res. Lett. 2008, 35 (18), . Choi, Y.; Vay, S. A.; Vadrevu, K. P.; Soja, A. J.; Woo, J.-H.; Nolf, S. R. N.; Sachse, G. W.; Diskin, G. S.; Blake, D. R.; Blake, N. J.; Singh, H. B.; Avery, M. A.; Fried, A.; Pfister, L.; Fuelberg, H. E. Characteristics of the atmospheric CO2 signal as observed over the conterminous United States during INTEX-NA. J. Geophys. Res. 2008, 113 (D7), . Alkezweeney, A. J.; Drewes, D. R. Airborne measurements of pollutants over urban and rural sites. J. Appl. Meteorol. 1977, 16 (5), 561–563. Klemm, O.; Schaller, E. Aircraft measurement of pollutant fluxes across the borders of Eastern Germany. Atmos. Environ. 1994, 28 (17), 2847–2860. Krautstrunk, M.; Neumann-Hauf, G.; Schlager, H.; Klemm, O.; Beyrich, F.; Corsmeier, U.; Kalthoff, N.; Kotzian, M. An experimental study on the planetary boundary layer transport of air pollutants over East Germany. Atmos. Environ. 2000, 34 (8), 1247–1266. Kalthoff, N.; Corsmeier, U.; Schmidt, K.; Kottmeier, C.; Fiedler, F.; Habram, M.; Slemr, F. Emissions of the city of Augsburg determined using the mass balance method. Atmos. Environ. 2002, 36 (1), 19–31. Trainer, M.; Ridley, B. A.; Buhr, M. P.; Kok, G.; Walega, J.; Hu ¨ bler, G.; Parrish, D. D.; Fehsenfeld, F. C. Regional ozone and urban plumes in the southeastern United States: Birmingham, a case study. J. Geophys. Res. 1995, 100 (D9), 18823–18834. Ryerson, T. B.; Buhr, M. P.; Frost, G. J.; Goldan, P. D.; Holloway, J. S.; Hu ¨ bler, G.; Jobson, B. T.; Kuster, W. C.; McKeen, S. A.; Parrish, D. D.; Roberts, J. M.; Sueper, D. T.; Trainer, M.; Williams, J.; Fehsenfed, F. C. Emissions lifetimes and ozone formation in

(25) (26) (27) (28)

(29)

(30)

(31)

(32) (33)

(34) (35) (36) (37) (38) (39)

(40) (41)

power plant plumes. J. Geophys. Res. 1998, 103 (D17), 22569– 22583. Crosson, E. R. A cavity ring-down analyzer for measuring atmospheric levels of methane, carbon dioxide, and water vapor. Appl. Phys. B: Laser Opt. 2008, 92 (3), 403–408. Crawford, T. L.; Dobosy, R. J. A sensitive fast-response probe to measure turbulence and heat flux from any airplane. Boundary-Layer Meteorol. 1992, 59 (3), 257–278. Crawford, T. L.; Dobosy, R. J.; Dumas, E. J. Aircraft wind measurement considering lift-induced upwash. Boundary-Layer Meteorol. 1996, 80 (1-2), 79–94. Garman, K. E.; Hill, K. A.; Wyss, P.; Carlsen, M.; Zimmerman, J. R.; Stirm, B. H.; Carney, T. Q.; Santini, R.; Shepson, P. B. An airborne and wind tunnel evaluation of a wind turbulence measurement system for aircraft-based flux measurements. J. Atmos. Oceanic Tech. 2006, 23 (12), 1696–1708. Garman, K. E.; Wyss, P.; Carlsen, M.; Zimmerman, J. R.; Stirm, B. H.; Carney, T. Q.; Santini, R.; Shepson, P. B. The contribution of variability of lift-induced upwash to the uncertainty in vertical winds determined from an aircraft platform. Boundary-Layer Meteorol. 2008, 126 (3), 461–476. Rolph, G. D. Real-time environmental applications and display system (READY); NOAA Air Resources Laboratory: Silver Spring, MD, 2003. http://www.arl.noaa.gov/ready/hysplit4.html (accessed October 20, 2008). Draxler, R. R.; Rolph, G. D. HYSPLIT (HYbrid Single-Particle Lagrangian Integrated Trajectory) Model access via NOAA ARL READY; NOAA Air Resources Laboratory: Silver Spring, MD, 2003. http://www.arl.noaa.gov/ready/hysplit4.html (accessed March 28, 2008). Myers, D. E. Interpolation and estimation with spatially located data. Chemom. Intell. Lab. Syst. 1991, 11 (3), 209–228. Chu, D. The GLOBEC kriging software package - EasyKrig3.0; The Woods Hole Oceanographic Institution: 2004. Available from http://globec.whoi.edu/software/kriging/easy_krig/easy_krig. html (accessed June 24, 2008). United States Census Bureau, Systems Support Division. Census 2000. http://www.census.gov (accessed Oct 1, 2008). Energy Information Administration. State Energy Data System (SDES): Indiana. http://www.eia.doe.gov/emeu/states/state. html?q_state_a)in&q_state)INDIANA (accessed Dec 8, 2008). Carbon Cycle Greenhouse Gases Group. Earth System Research Laboratory: Global Monitoring Division. http://www.esrl.noaa. gov/gmd/ccgg (accessed April 30, 2009). Khalil, M.; Butenhoff, C. L.; Rasmussen, R. A. Atmospheric methane: trends and cycles of sources and sinks. Environ. Sci. Technol. 2007, 41 (7), 2131–2137. Nam, E. K.; Jensen, T. E.; Wallington, T. J. Methane emissions from vehicles. Environ. Sci. Technol. 2004, 38 (17), 2005–2010. Emission Factor and Inventory Group. Compilation of air pollutant emission factors. Volume 1: Stationary point and area sources, Supplement E: Bituminous and subbituminous coal combustion. U.S. Environmental Protection Agency: 1998. http:// www.epa.gov/ttn/chief/ap42/ch01/final/c01s01.pdf (accessed Jan 20, 2009). Babbitt, C. W.; Lindner, A. S. A life cycle inventory of coal used for electricity production in Florida. J. Cleaner Prod. 2005, 13, 903–912. Mosher, B. W.; Czepiel, P. M.; Harriss, R. C.; Shorter, J. H.; Kolb, C. E.; McManus, J. B.; Allwine, E.; Lamb, B. K. Methane emissions at nine landfill sites in the northeastern United States. Environ. Sci. Technol. 1999, 33 (12), 2088–2094.

ES901326B

VOL. 43, NO. 20, 2009 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

7823