Environ. Sci. Technol. 2009, 43, 1730–1736
Aircraft Hydrocarbon Emissions at Oakland International Airport S C O T T C . H E R N D O N , * ,† E Z R A C . W O O D , † MEGAN J. NORTHWAY,† RICHARD MIAKE-LYE,† LEE THORNHILL,‡ ANDREAS BEYERSDORF,‡ BRUCE E. ANDERSON,‡ RENEE DOWLIN,§ WILLARD DODDS,| AND W. BERK KNIGHTON⊥ Aerodyne Research, Inc., Billerica, Massachusetts, Atmospheric Science Division, NASA Langley Research Center, Virginia, Port of Portland, Portland, Oregon, General Electric Aviation, Cincinnati, Ohio, and Montana State University, Bozeman, Montana
Received May 12, 2008. Revised manuscript received December 4, 2008. Accepted January 7, 2009.
To help airports improve emission inventory data, speciated hydrocarbon emission indices have been measured from inuse commercial, airfreight, and general aviation aircraft at Oakland International Airport. The compounds reported here include formaldehyde, acetaldehyde, ethene, propene, and benzene. At idle, the magnitude of hydrocarbon emission indices was variable and reflected differences in engine technology, actual throttle setting, and ambient temperature. Scaling the measured emission indices to the simultaneously measured formaldehyde (HCHO) emission index eliminated most of the observed variability. This result supports a uniform hydrocarbon emissions profile across engine types when the engine is operating near idle, which can greatly simplify how speciated hydrocarbons are handled in emission inventories. The magnitude of the measured hydrocarbon emission index observed in these measurements (ambient temperature range 12-22 °C) is a factor of 1.5-2.2 times larger than the certification benchmarks. Using estimates of operational fuel flow rates at idle, this analysis suggests that current emission inventories at the temperatures encountered at this airport underestimate hydrocarbon emissions from the idle phase of operation by 16-45%.
Introduction The International Civil Aviation Organization (ICAO) has adopted recommended practices (1) for the measurement of CO, NOx, smoke number and unburned hydrocarbons (UHC) and tabulates the results for commercial aviation engines (2). The ICAO engine certification databank sheets are benchmarks for engine emissions performance and are used in regulatory evaluations. They are also commonly used in inventory assessments, global modeling, and air quality research (3, 4). Emissions of CO, NOx, and UHC are compiled in units of grams per kilogram of fuel for the following named engine conditions: idle, approach, climb-out, and takeoff. NOx is expressed as NO2 equivalent while UHC is expressed * Corresponding author e-mail:
[email protected]. † Aerodyne Research, Inc. ‡ NASA Langley Research Center. § Port of Portland. | General Electric Aviation. ⊥ Montana State University. 1730
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 43, NO. 6, 2009
as CH4 equivalent. Fuel flow rates for each engine condition are also tabulated. The emissions dispersion modeling system (EDMS) uses the ICAO certification data in a wide variety of airport related inventory and modeling applications (5, 6). Potential bias in an emissions inventory assessments could arise from uncertainties in the following sources: time-in-mode, ambient conditions, and actual operational thrust levels. The influence of operational thrust levels on NOx emissions due to the practice of reduced thrust take off has been predicted in research reports and has been observed on operational runways (7, 8). Analysis of the tabulated “out-off-on-in” times for each airport can refine the time spent in each engine mode for a specific airport. The influence of ambient conditions on engine emissions can be significant (9). The Boeing fuel flow methodology (BFFM2) describes a procedure that can be used to correct emissions for temperature, pressure and relative humidity (10). BFFM2 can also be used to interpolate between the tabulated emissions in ICAO (at the named conditions) based on generalized fuel flow. Generally, BFFM2 is believed to be more accurate for NOx emissions, particularly during the takeoff and climb out phases of a landing take-off cycle. Anecdotally, it is well understood that most modern engines idle on the ground at thrust setting lower than the engine certification value (defined as 7% of rated thrust). For pollutants whose emission indices increase with decreasing thrust, such as CO and UHC, the ICAO certification value named “idle” usually underestimates total emissions because the emission index increase outweighs the fuel flow decrease. The certification standard measurement of UHC uses the “flame ionization detection” (FID) analytical technique. It is a very sensitive method for detecting many types of hydrocarbons species; however, it is neither selective nor universal. The FID response is diminished for several oxygenated compounds known to be emitted by aircraft. The UHC idle emission indices in the ICAO databank measured using FID spans a wide range of values for different engine types. In the last 20 years, research programs have conducted more direct speciated hydrocarbon measurements of turbofan and turbojet aircraft engine technologies. The comprehensive measurements reported in Spicer et al. (11) have been a valuable benchmark for the CFM56- class of engine with over 60 specific compounds identified. In a report to the U.S. Air Force, Gerstle and co-workers (12) reported UHC emission rates for several engines not included in the ICAO databank, as well as some emissions from auxiliary power units. During the EXCAVATE campaign Anderson et al. (13) measured the speciated hydrocarbon emissions from an RB211-535-E4 engine for two different fuel sulfur levels (13). The APEX-1 campaign (14) conducted online hydrocarbon speciation using infrared fingerprint absorption spectroscopy and chemical ionization mass spectrometry for a CFM562C1 (9, 15). Using analysis of wind-advected plumes sampled at Boston Logan International Airport, a selected set of hydrocarbon emissions were characterized from in-use aircraft (16). Schu ¨ rmann et al. (17) measured volatile organic compounds using canister sampling of diluted exhaust in an operational taxiway area. They found that nearby refueling activity altered the profile of hydrocarbons considerably. This work focuses on reporting the results from the JETS/ APEX-2 campaign in August 2005 at Oakland International Airport. One primary component of this campaign involved staged aircraft measurements. The results from the staged aircraft testing in JETS/APEX-2 and APEX-3 will be reported elsewhere (18). An additional approach employed during 10.1021/es801307m CCC: $40.75
2009 American Chemical Society
Published on Web 02/12/2009
JETS/APEX-2 involved measurements of wind-advected plumes. This work focuses on the taxiway/runway measurement phases and the data collected opportunistically during the mission’s “down time”. This paper uses the data from the JETS/APEX-2 campaign as well as the previously published results to address the following questions. How much variability is observed in the hydrocarbon emission profile? What levels of hydrocarbon emissions are observed during actual operation, as compared to the tabulated certification values? Does the ICAO UHC benchmark quantitatively reflect differences in speciated emission indices among different in-use engine models?
Experimental Description The results described in this work were collected as part of the JETS/APEX-2 campaign, staged at the Oakland International Airport. Several different sampling approaches were employed between August 20 and 29 of 2005. The most formal approach involved sampling from four different 737’s provided by Southwest Airlines and operated in a “staged” mode in the airport’s ground runup enclosure (GRE). The detailed description of that work is described elsewhere (18, 19) and follows closely on the former APEX campaign (20). The work described here is based on the analysis of data collected on an ambient sampling manifold. This manifold consisted of a 3.8 cm diameter tube, ∼7 m long drawing ∼150 slpm. The various instrument sampled from this manifold at ∼3-4 s after being sampled at the common tip. While configured at the GRE, when the staged aircraft were not being tested, the instruments sampled from this manifold, and any transient plumes of aircraft taxiing upwind from the GRE. On August 26, the vehicles housing the various instrumentation packages were positioned at the end of an active taxiway next to the main runway. The plumes described here were typically present in the atmosphere for 15-180 s before they were sampled (see Meteorological Conditions below). All emission index calculations are performed following methods described in previous work (21, 22) with a CO2 emission index of 3160 g CO2 kg-1 fuel. In this work, the preferred reporting unit for speciated hydrocarbon (HC) measurements is grams of HC per kg of fuel. For the purposes of comparison to previous measurements, ppbv HC per ppm CO2 has been converted to g HC kg-1 using the cited CO2 EI and the molecular weight of the species in question. Each plume encounter is analyzed individually where the excess hydrocarbon and CO2 concentrations are determined by integration of the corresponding peaks. Emission ratios expressed as ppbv per ppmv of CO2 from these analyzes are then converted to g of HC per kg of fuel. One of the fundamental analysis methods of this data set involves normalizing the observed measurements relative to that simultaneously determined for HCHO. These HCHO relative emissions are reported in g of HC per g of HCHO. Analytical Methods. This campaign was a collaboration of the University of Missouri-Rolla’s Center of Excellence for Aerospace Particulate Emission Reduction Center, NASALangley Propulsion Research Group, University of CaliforniaRiverside, Montana State University, and Aerodyne Research, Inc. The most detailed descriptions of the instrumentation and analysis procedures can be found elsewhere (7). The results in this work are derived largely from the tunable infrared differential absorption spectrometers and the proton transfer reaction mass spectrometer measurements, which have been discussed previously (9, 15, 16, 23). Meteorological Conditions. The weather at Oakland International Airport during the measurement period was fairly consistent with early morning lows of 12 C (04:00) rising
to ∼22 C (14:30). Relative humidity was ∼90% at night and dipped to 60% during the day. The wind speed profile was similar to that of temperatures (lagged by ∼1 h) rising from 1 to 3 m s-1 between 06:00 and 15:30. The use of wind advected plumes essentially requires a well characterized wind speed and direction together with careful monitoring of what is occurring along the upwind vector. The geography of the Oakland International Airport is such that the predominant wind pattern (August 20-27, 2005) carried air from the bay, to the runway, taxiway, and then to the sampling location. This work did not have to contend with any appreciable bulk transport of air from any notable local sources, such as the maintenance hangars, terminal area, or ground access roads. Though the regional wind pattern carried air influenced by the emissions in downtown San Francisco to the sample location, these were observed as an increase in the general background. All emission indices are calculated by considering only the high frequency plume encounter. A consequence of this analysis is that the background is always “subtracted”. Boeing Fuel Flow Methodology. Ambient temperature, pressure and relative humidity can have a large effect on the emissions characteristics of a turbofan engine. In order to account for the influence on emissions performance, the Boeing Fuel Flow Method-2 has been developed (10). In this work, for any comparison of measured data to ICAO tabulated values, an adaptation of the procedure described in Dubois et al. (10) has been used to correct reference ICAO values to the ambient conditions. Airframe and Specific Engine Type Determination. Aircraft tail numbers were recorded and entered into time coded notes along with the activity state along the upwind vector whenever operators were present. A substantial number of taxiway plumes were measured while the site was not staffed. This occurred during the down time of the staged aircraft testing in the GRE. Using a time coded web cam and the airline service quality performance (ASQP) record, most of the taxiing airframes could be associated with a specific engine. Tail number to engine/airframe combinations were performed by the VOLPE transportation center (24). For the air freight carriers, personal communication was used to determine the engines in use for aircraft operating at their facilities.
Results and Discussion Emission Index Magnitudes. A time series of the measurement of benzene, propene, acetaldehyde. and formaldehyde is depicted in Figure 1a. This time period is characterized by numerous rapid increases in each of these compounds. This is due to several taxiing aircraft passing ∼30-50 m along the upwind vector from the sample line. The categories of aircraft engines encountered in this time segment include CF6-50C2, -80C2A5F; CFM56-7B22, -7B26, -3C1; JT8D-15; V2527-A5; and two smaller executive jet engines not present in the ICAO databank. The ensemble correlation plots relative to HCHO are shown in Figure 1 b.1 (benzene), b.2 (propene) and b.3 (acetaldehyde) for the time series in section (a) of the figure. The correlations are good for several different plumes. This result supports the suggestion that all of the various hydrocarbon species scale together despite variations in overall emission magnitude due to near-idle throttle variations, ambient conditions, fuel compositions and engine technologies. This point will be further discussed later. To help illustrate the sensitivity the variability in observed HCHO emissions, a probability distribution of the absolute magnitude of the formaldehyde emissions is depicted in Figure 2. This analysis includes 150 plumes from numerous engine types encountered. The median value of this distribution is 0.815 g HCHO kg-1 fuel. The nature of this distribution is driven by the sensitivity of HCHO emissions VOL. 43, NO. 6, 2009 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
1731
FIGURE 1. HCHO and PTR-MS time series data. Taxiway plumes sampled 40-90 s after emission. Engine taxiway plumes from CF6-50C2, -80C2A5F; CFM56-7B22, -7B26, -3C1; JT8D-15; V2527-A5. The spikes in this time series are due to aircraft passing along the upwind vector (30-50 m upwind).
FIGURE 2. Distribution of measured formaldehyde emission indices. The normalized probability distribution for the GRE wind-advected plume sample (bin size ) 0.25 g kg-1). This population encompasses numerous engine types, including those installed on commercial passenger, airfreight and general aviation aircraft. Previous measurements are depicted with statistical representations of this distribution on the upper portion of the figure for comparison. to engine technology, throttle position and ambient temperature. Despite these challenges, the observed profile compares favorably to previous HCHO emission index determinations. The value drawn from the Spicer et al. (11) report is for the CFM56-3 engine. In the Yelvington et al. (9) study, a range of values were observed at ground idle (reflected by the error bar) for the CFM56-2C1 engine. In the Herndon et al. (16) study, 13 wind-advected plumes from 1732
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 43, NO. 6, 2009
in-use aircraft at Logan airport were analyzed (median ( σ). The engine types in the Logan study likely included CF34-, AE3007, CFM56-, JT8D-, and RB211- series engines. In the present work, an active airframe/engine identification procedure was performed and will be discussed later. In the Supporting Information (SI), a subset of the measured hydrocarbon species (drawn from the references cited in the introduction) have been tabulated. A cursory
FIGURE 3. Distribution of the ratios of measured volatile organic carbon (VOC) to formaldehyde. The measured ratio of emission index to EI(HCHO) for benzene, acetaldehye, propene, and ethene. The solid lines are the actual distribution, whereas the corresponding dotted lines are the Gaussian fit results to the distribution.
TABLE 1. Speciated HC Index Ratio (HCHO Relative)a
HCHO acetaldehyde C2H4 propene butenes+acrolein pentenes benzene toluene 1-ring aromatics styrene naphthalene
Spicer et al. staged
APEX-1 staged
Logan advected
1 0.35 1.26 0.36 0.36 0.11 0.14 0.05 0.28 0.03 0.04
1 0.24 0.78 0.31 0.45 0.31 0.14 0.06 0.48 0.03 0.01
1 0.26 0.25 0.11 0.06 0.3
EXC staged
1b 0.32 0.45 0.08 0.01
Zurich advected
1b 0.32 0.26c 0.11 0.11 0.13 0.39 0.04
APEX-2 staged 1a 0.37 0.76 0.45 0.49 0.18 0.09 0.73 0.04 0.04
APEX-3 staged
1b 0.41 0.16 0.86 0.24 0.13 0.57 0.3
this work advected 1 0.32 ( 0.09 1.1 ( 0.3 0.43 ( 0.2 0.15 ( 0.08
a
A table containing the absolute emission indices from the different studies is provided in the supplemental information. All values are in units of grams of HC per gram of HCHO, except for the Excavate, Zurich (17) and APEX-3 column which are grams of HC per gram of C2H4. The APEX-2, Staged aircraft column represents the average result for “ground-idle” including the following engines; three CFM56-7B22, one CFM56-3B1, two CFM56-3B2. Tabulated values in the Advect GRE column represent Gaussian fits to the distribution of measured compound to HCHO ratios. The error bar is one Gaussian width. APEX-3 averaged GC results for three staged aircraft at idle (Beyersdorf, in preparation). Acrolein is assumed to be present in the same ratio as Spicer et al. b The Excavate, Zurich and APEX-3 data sets have been normalized by the emission index for ethene in lieu of formaldehyde. c The Zurich tabulation for butenes + acrolein assumes the ratio of acrolein to the sum of the butene isomers is that found in Spicer et al.
examination of these measurements reveals large differences in the magnitudes of the reported emission indices. Although several of the measurements were performed on different engines, an important finding throughout these measurements is that the magnitude of the hydrocarbon emission index at the low idle settings is very sensitive to small changes in engine power setting, bleed air extraction (to operate the environmental control and anti ice systems), and power extraction (to operate the aircraft flight control and electrical power systems). During these tests, identical engine conditions are often difficult to reproduce. The magnitude of the uncertainty in the reported values (in the SI and cited references) is due to real variability in the emission index. Despite this, these studies also find that the speciation profile (ratios between species) was much less variable, and is discussed further below. Emission Index Ratios. In the correlation diagrams (part b) of Figure 1, plumes from several different engines appear to have some level of correspondence between various HCs. One potential problem with this analysis, though, is that the
plumes with the greatest magnitude may dominate the correlation. In order to determine a more balanced indication of the correspondence between the measured HCs in this work, an analysis of the plume by plume distribution of the HC to HCHO ratios has been performed, although the choice of hydrocarbon as the normalizing quantity is somewhat arbitrary. Here, formaldehyde has been used because of its abundance in the exhaust matrix (only ethene is of a similar magnitude) and the sensitive direct measurement as part of the instrumentation suite. The result for the GRE sampled plumes is depicted in Figure 3. Each normalized distribution with is shown with units of g HC g-1 HCHO. The distributions have been fit to a Gaussian line shape in order to extract a center value and gausswidth. The fit results and widths are noted in the APEX-2/Advect GRE column of Table 1. The widths of these fits are a driven by two factors; the real observed variability (which is desired) and the instrumental noise (which can be estimated). The median peak values of formaldehyde and carbon dioxide for the plumes encountered here were 13.8 ppbv VOL. 43, NO. 6, 2009 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
1733
FIGURE 4. Measured HCHO Emission Index vs ICAO “idle” UHC Emission Index. The measured HCHO emission indices are depicted vs the ICAO tabulated UHC for idle. The general aviation jets use engines not included in the ICAO databank and have been “boxed” and placed arbitrarily on the bottom axis at the right. The measured points have been colored by the simultaneously measured NOx emission index. and 27.2 ppmv, respectively. The strength of the plume “hit” are highly dependent on distance between the point of emission and sampling, atmospheric wind strengths, aircraft wake turbulence and other uncontrollable factors. Assuming this is a representative plume strength, conservative estimates of the noise in the QCL and PTR instruments can be used to estimate how much of the observed width is due only to instrument noise and how much is due to variability in the ratio of the emissions. This also assumes that each of these contributions is Gaussian in nature. For example, the projected noise level in the ratio of benzene to formaldehyde is 0.06 g g-1, which accounts for most of the observed variability in the ratio ((0.08 g g-1). Although the widths in these distributions are driven by the precision noise in the measurement, the widths can at least be used as an upper limit for the real observed variability among different engines operating at various near-idle thrust levels. In this data set, there may be an insufficient number of samples to discern any temperature dependence in the ratio. A cursory analysis of the daytime and nighttime distributions, however, results in nearly the same median ratio for acetaldehyde (
