Aircraft and Airport-Related Hazardous Air Pollutants: Research Needs and Analysis (2008)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

27 Emissions data are used in conjunction with dispersion/ chemistry models in order to predict actual concentrations in particular locations. Emissions from most pollution sources are expressed or calculated as the product of two separate components: emission factors and activity factors. For exam- ple, the emission factors for on-road mobile sources are com- monly tabulated as grams per vehicle mile. By combining this emission factor with an activity factor, in this example, miles traveled, the total emission can be estimated. For aircraft emissions, emission factors are fuel-based, e.g., g of CO/kg of fuel (also known as an emission index). Off-road emission factors are expressed differently, as g/BHP-hr (grams per brake-horsepower hour). Below is a description of the status of knowledge and uncertainties associated with the emissions produced from the following sources: aircraft, airport operations, ground- access vehicles, stationary sources, and de-icing. 5.1 Aircraft This category includes emissions from gas turbine engines, turboprop engines, internal combustion piston engines for fixed-wing aircraft, helicopters, and nonrigid airships such as blimps. As total fuel consumption at most airports is domi- nated by commercial aircraft activity, the emphasis of this report is on high-bypass turbo-fan equipped commercial jets; however, piston-engine aircraft emissions will also be discussed. 5.1.1 Commercial Aircraft Emission Factors The ICAO has an established testing protocol for new jet engines. The results of this certification process are tabu- lated in the ICAO databank (www.caa.co.uk). The mass emissions of NOx, CO, and unburned hydrocarbons (UHC) are compiled in units of mass per kilogram of fuel for each jet engine at four defined operational points. ICAO defines these operational points as the fraction of rated thrust from the engine for the following named conditions: idle (7%), approach (30%), climb-out (85%), and take-off (100%). The ICAO databank also tabulates the result of a measure- ment of exhaust opacity known as the smoke-number, how- ever, reporting only the maximum smoke number is man- dated. Smoke-number is frequently omitted for other operational conditions. All tabulated values have been ei- ther conducted at standard atmospheric pressure, tempera- ture, and relative humidity or corrected to this reference state. The fuel flow at each power setting is also tabulated in this databank. For the ICAO species, this database serves as the primary source of aviation emissions factor informa- tion. Volatile organic compounds (VOCs) (UHCs) exhibit a different emission profile than does NOx. Figure 6 depicts the average NOx and CO emission indices versus throttle setting. The emission indices of CO are greatest at low throttle and decrease dramatically with increasing power settings. NOx EIs start at an average value of ∼5 g/kg at low thrust and unlike for CO, the EI increases with increasing throttle. This depic- tion is not quantitatively representative of any single specific engine, but is qualitatively true for all jet engines in the sense that CO is relatively high at low engine power and NOx is relatively high at high engine power. The UHC trend is qual- itatively similar to the trend observed in CO emissions. Un- burned hydrocarbons and VOCs are mostly synonymous in this context. Emissions of VOCs exhibit a similar trend with engine power as CO emissions do. This is depicted in Figure 7, in which an average VOC emission factor and fuel flow is depicted versus the throttle setting. Consideration of the emission factors, fuel flow rates, and times spent in mode indicate that the activity leading to the greatest gas- phase HAP emissions (excluding particulate bound PAH species) is idling. S E C T I O N 5 Emission Factors and Activity Factors

28 Figure 6. Average NOx and CO emission indices versus throttle setting. Figure 7. Generalized depiction of UHC emission index and fuel flow rate versus engine power. One of the more definitive speciated measurements of hydrocarbons in jet exhaust is the work of Spicer, Holdren et al. (1994). This work reported on a wide variety of individ- ual hydrocarbons in the exhaust of a TF-39 and a CFM56 engine (similar to those found on Boeing 737 airframes) fueled by JP-5 fuel. Whole air canister samples were collected for off-line analysis by various gas-chromatographic (GC) techniques while the engine was running at ground idle and at 30% and 80% of rated thrust. The work of Spicer was the most authoritative manuscript on this topic for over a decade. Additional work was performed in the late 1990s by Gerstle and co-workers on behalf of the U.S. Air Force, though these measurements focused on military engines (Gerstle, Virag et al. 1999). Slemr et al. reported hydrocarbon emission factors from a Rolls Royce M45H-Mk501 engine measured during flight (Slemr, Giehl et al. 2001). In the last few years, several emission measurement cam- paigns have significantly contributed to the total knowledge base of VOC emissions from commercial aircraft. Table 9 summarizes the aircraft engines characterized during these campaigns. For all of the engines presented in Table 9, the particulate and gas-phase components of the engine exhaust were ana- lyzed by a team of research groups. VOCs were measured by either in-situ instruments (e.g., proton transfer mass spec- trometry [PTRMS], tunable infrared laser diode absorption spectrometry [TILDAS]) or whole air canister samples for subsequent analysis by GC-MS. Additional studies at the Delta/Atlanta Hartsfield and JETS-APEX2 studies focused on the measurement of advected exhaust plumes from in-use aircraft. Such measurements are analogous to on-road re- mote sensing measurements in that emission factor distribu- tions can be measured. VOC measurements during these campaigns have both confirmed past results and revealed new findings. The main findings are summarized below. Engine Airframe Campaign RB211-535E4 B757 EXCAVATE CFM56-2C2 DC-8 APEX CFM56-3B1 B737 JETS-APEX2, APEX3 CFM56-7B22 B737 JETS-APEX2 RB211-535E4-B B757 APEX3 RB211-535E4-B phase 5 B757 APEX3 PW4158 A300 APEX3 AE3007-A Embraer APEX3 AE3007-A1E Embraer APEX3 CJ6108A NASA Lear Jet APEX3 Notes: EXCAVATE: EXperiment to Characterize Aircraft Volatile Aerosol and Trace-Species Emissions, NASA- Langley, 2002. APEX: Aircraft Particle Emissions eXperiment, Dryden Air Force Base, 2004. JETS – APEX2: Oakland International Airport, 2005. APEX3: Cleveland International Airport/NASA-Glenn, 2005. Table 9. Summary of recent aircraft engine exhaust measurement campaigns.

29 Figure 8. Formaldehyde (HCHO) emission ratio as a function of engine power. Credit: Knighton, Herndon et al. 2006. Used with permission. Figure 9. Ratios of several VOCs to HCHO. 5.1.1.1 VOC Emission Factors versus Thrust As observed by Spicer and colleagues (1994), VOC meas- urements are a strong function of engine power and are high- est at low thrust; decreasing dramatically at higher powers. Figure 8 shows the formaldehyde (HCHO) emission ratio as a function of fuel flow rate (power) for the CFM56-2C1 en- gine studied during APEX (Yelvington, Herndon et al. 2007). The HCHO emission ratio (parts per billion [ppb] of HCHO per ppm of carbon dioxide [CO2]) is proportional to a fuel- based emission index (g HCHO per kg of fuel). The emission ratio decreases by two orders of magnitude between low thrust (idle) and high thrust (take-off) settings. Furthermore, the decrease in HCHO emission index between 4% rated thrust and 7% is more than a factor of three. The ICAO cer- tification power setting for idle is 7% rated thrust, though lower thrust levels (e.g., 4%) appear to more accurately reflect true operational ground idle. 5.1.1.2 VOC Emissions Follow a Universal Scaling Law The ratio of any two VOCs in engine exhaust is approxi- mately constant and independent of engine, power setting, fuel content, and ambient temperature. Figure 9 displays this feature. Data from one CFM56-3B1, two RB211-535E4-Bs, and one PW4158 as measured at APEX3 were used in this plot. The ratio of the emission index of any chosen VOC to that of formaldehyde is a constant, as indicated by the linear fits through each group of points. Thus speciation profiles that have relied on the work of Spicer, Holdren et al. (1994) are likely fairly accurate. Formaldehyde is a somewhat ar- bitrary choice for the x-axis, though was measured with separate instrumentation (TILDAS) than the other VOCs (PTRMS). Thus the variation in formaldehyde emission index with power and temperature described elsewhere in this section hold true for all VOC species measured. It is un- known whether this universal scaling is affected by extremely low temperatures or engine age/maintenance. 5.1.1.3 VOC Emission Factors Are Highly Dependent on Ambient Temperature Volatile organic compound emissions increase greatly with decreasing temperature. The variation of HCHO emission index with temperature is evident in Figure 8. Figure 10 depicts the measured temperature dependence of HCHO emissions from the APEX-1 campaign. For comparison, the Boeing Fuel Flow Method 2 (BFFM2) temperature depend- ence has been added to the data (Baughcum, Tritz et al. 1996), depicted as the solid lines. These lines have been com- puted by scaling the measured emission rate at 25°C by the ratio of the temperature dependent function to that at 25°C using the following equation: Emission Rate = BFFM2(T)/BFFM2(25 C) * Measured HCHO ER.

30 Figure 10. APEX-1 measured temperature dependence of formaldehyde (HCHO) emissions. In this way, only the temperature dependence of the BFFM2 modeling function is compared to the data. An absolute comparison is not directly possible, as the procedure was developed to scale reference ICAO UHC. Though the absolute magnitude of the reference value may be arbitrary, the temperature dependence is evidently slightly steeper than predicted. It should be noted that the developers of the BFFM2 protocol have more certainty in its ability to predict NOx emission at take-off and during cruise aloft. In the description of the method, it acknowledges challenges associated with using 30% and 7% as the only model inputs for CO and HC emissions. There are no measurements below 15°C as such knowledge of HAP emissions in cold temperatures is extremely limited. 5.1.1.4 Negligible Variation of VOC Emission Factors with Fuel Composition Variations in aromatic content and sulfur content were found to have a minimal effect on VOC emissions during APEX and are minor compared to variation with engine power and ambient temperature (Anderson, Chen et al. 2006; Knighton, Rogers et al. 2007; Yelvington, Herndon et al. 2007). Note that while the “aromatic” content of the fuels tested varied from 18% to 22%, the C/H ratio was not affected. Clearly the full variation in hydrocarbon matrix of the fuel was not fully explored. 5.1.1.5 Agreement Between GC-FID and Individually Measured VOCs At APEX, the sum of individual VOC measurements and a total HC measurement by GC-FID (flame ionization detec- tor) agreed to within 10% at low powers. At low power (<10% thrust), the sum of compounds measured by TILDAS and PTR-MS was ~10% higher than the GC-FID measure- ment, quite possibly due to the latter technique’s lack of sen- sitivity to formaldehyde. At higher powers, the total VOC concentrations were lower than the detection limit for the TILDAS and PTR-MS as deployed during those campaigns (Yelvington, Herndon et al. 2007). 5.1.1.6 Absolute VOC EIs Vary by More Than a Factor of 10 Between Different Engines This is reflected in both the individually-measured VOC emission indices as well as the ICAO certification data for total hydrocarbon (HC) emissions. Table 10 displays the HCHO emission index (EIHCHO) of the engines observed at the APEX campaigns and the corresponding ICAO certifica- tion HC emission index, both at 7% power (ICAO idle). The CF6 engine is included in the table for comparison. The variation in the HCICAO/HCHOAPEXx ratio illustrates the difficulty in comparing these quantities due to the large variation in VOC emissions with power and temperature. The variations should not be interpreted as actual variations in the HCHO contribution to total HC. Comparison of indi- vidual VOC measurements to the total HC measurement by GC-FID has not been done for JETS-APEX2 and APEX3 yet. 5.1.1.7 Emission Factors: True Idle versus 7% (ICAO) The category labeled “idle” in the ICAO databank of emis- sions is a standardized thrust setting. The certification setting (7% of rated maximum engine thrust) is somewhat greater than the setting at which many modern high-bypass ratio engines actually idle in today’s fleet. Although it is not for- mally correct, numerous data plots and representations attribute “true ground idle” to be approximately 4% of rated thrust. Figure 10 portrays the temperature dependence of actual HCHO emissions measured during APEX1 along with the predicted temperature dependence of Boeing Fuel Flow Method-2. This figure also shows the increase in HCHO emission rate between 7% and 4% (ground-idle). A current problem in emission inventory modeling (such as EDMS) involves the treatment of how the taxi phase of the LTO is treated. This discussion is tightly coupled to the dif- ference in VOC emission rates between 7% and ground-idle. In each of the studies where the engine has been measured at both 7% and “ground idle,” the CO and VOC species are con- sistently greater at the lower effective thrust setting. Though the number of aircraft engines sampled through the EXCAVATE and APEX missions has grown, a systematic characterization of emissions resulting from the range of idle levels used (3% to 10%) is not available.

31 Figure 11. Depiction of the emissions of 1,3-butadiene from aircraft, GSE, GAV, and stationary sources at PHL. The shaded area outside of the circle represents the extra emissions from aircraft when calculated assuming that the power settings used during idle/taxi are equal amounts of time spent at 7% thrust and 4% thrust (versus the standard 7% assumption). Engine EIHCHO (7%, g/kg) ICAO EIHC (g/kg) HC ICAO/HCHO(APEXx) CFM56-2C1 0.43 NA NA CFM56-3B1 0.33 2.3 7.0 CFM56-7B22 0.31 2.5 8.1 RB211-535E4B 0.05 0.14 2.8 RB211-535E4-B phase 5 0.22 0.28 1.3 PW4158 1.00 1.8 1.8 AE3007-A 0.41 2.5 6.1 AE3007-A1E 0.54 3.5 6.5 CF6-6D NA 21 NA Notes: EIHCHO formaldehyde emission index g/kg grams per kilogram ICAO International Civil Aviation Organization EIHC hydrocarbon emission index APEX Aircraft Particle Emissions eXperiment NA not applicable Table 10. Comparison of formaldehyde and ICAO HC emission indices for several engines. Measurements of in-use aircraft at Oakland Airport dur- ing JETs-APEX2 have helped to characterize the real-world emissions at true operational idle. Although use of 7% thrust was known to be too high by aircraft operators, the use of 7% idle in engine calculations has remained common in numer- ous studies (e.g., Pison and Menut 2004; Unal, Hu et al. 2005). Figure 11 depicts the underestimate of 1,3-butadiene emis- sions from PHL resulting from the assumption of 7% for the thrust value used during the idle phase. The shaded area out- side of the circle represents the conservative underestimated amount, assuming a temperature of 18°C. 5.1.1.8 Reactive Aldehyde Emission Factors Due to the importance that this report and others assign to acrolein as an air toxic, a few details regarding its emission in- dices are included here. The emissions of acrolein from air- craft engines are greater than anticipated by simple extrapo- lations of other combustion sources. The ratio of acrolein to formaldehyde (HCHO) in gasoline “engine out” (pre- catalytic converter) is 0.4% by mass (Schauer, Kleeman et al. 2002). This same ratio in diesel truck exhaust is slightly greater, 15% by mass (Schauer, Kleeman et al. 1999). In the Spicer, Holdren et al. (1994) determination of VOC content in the exhaust of the CFM-56, the measured ratio of acrolein to HCHO is 29%, by mass. Preliminary analysis of wind-advected plumes at the taxi- way of Oakland International Airport using the fast re- sponse online instrumentation during the JETS/APEX2 campaign supports the observations of elevated acrolein emissions from aircraft engine. One such event is depicted in Figure 12, where the exhaust from an in-use CF-6 engine was sampled. The time series show high correlation between the sum of the butene isomers and acrolein, HCHO, and CO2. The PTR-MS instrument used to measure acrolein + butene was not using a recently developed scrubber system which can distinguish the two compounds independently. As a result, this analysis must rely on the Spicer et al. frac- tion of acrolein to total acrolein + butene isomers (0.64). The Spicer, Holdren et al. fraction is very similar to that ob- served in diesel truck exhaust (Schauer, Kleeman et al. 1999). The advected plume data in Figure 12 suggest the ratio of acrolein to HCHO is 56%. This further underscores

32 Figure 12. Acrolein sampled from CF-6 engine exhaust. the need to characterize the acrolein emission index with robust measurements. 5.1.1.9 Aircraft Emissions at Engine Start At many airports, aircraft can plug-in to an electrical ser- vice at the gate, which provides basic power to the aircraft during preparation for flight and deplaning, and allows the APU to be turned off. At other airports the APU is used to generate power for the aircraft while at the gate and, in either case, the APU is used to start the main engines. In starting the engine, the turbo-machinery is first accelerated to a nominal starting rotational speed, at which point, fuel is supplied to the combustor and a fuel spray is generated. With the initia- tion of the fuel flow, the spark ignition system is activated, and a flame is quickly established in the combustor. Once the flame is established, the rotation speed and the fuel flow are increased together to the nominal ground idle operating point. Over the course of 30 to 90 sec, the temperature of the engine reaches a stable condition and the emission indices of hydrocarbons and CO stabilize. As it pertains to total emissions accounting, any fuel that is pushed through the combustor before ignition is “un- burned.” The nature of these emissions will have the same speciation as evaporative emissions from the fuel stock. This pre-ignition contribution will have a very different speciation profile than the engine-on profile of HAP emissions, and they are much less toxic, since they consist mainly of alkanes. There are a few different approaches to assessing the mag- nitude of start-up emissions relative to the total emissions during an LTO. Direct determinations, based on fuel flow measurements and analytical UHC instruments, are chal- lenging because the unburned fuel can foul sampling lines, and standard methods using combustion reference gases can- not be used prior to the initiation of the flame in the com- bustor. Alternately, the combustion efficiency can be esti- mated from engine exit plane temperature measurements. If assumptions about the HC profile are associated with com- bustion efficiency, the start-up emissions can be estimated. Such calculations using unpublished test data yield esti- mates that pre-ignition emissions of unburned fuel are be- tween 30 and 90 g of fuel (hydrocarbons). The range is largely a result of the uncertainty in the ignition timing. The compo- sition of this emission source should resemble evaporated fuel and be less similar to the exhaust HC profile. Pre-ignition emissions are estimated in Table 11. This estimate uses a fuel flow rate (kg/s) that is half of the ICAO 7% value for a CFM56- 7B22 engine during the initial act of starting the engine. The next line involves estimating the emissions at the crit- ical point of ignition. The contribution of the near-ignition phase is small compared to the total LTO; however, there is Condition Emission Inde x (g kg -1 ) Time Fuel Flow (kg s -1 ) Emission UHC (g) Emission ~HAP (g) Pre-ignition 1000 0.3-0.5 s 0.05 15-25 - Ignition 1000-10 0.01-10 ms <1 Warm-Up 3 – 6 30 – 90 s 0.105 10 – 50 10 – 50 Taxiway Activity 2.5 24 min 0.105 410 410 Pre-ignition —The time and fuel flow values have been esti mated using an unpublished draft of a wo rking paper on start-up emissions (Will Dodds, GE, personal communication). Ignition —These large ranges are meant to indicate the vast uncertainty in this transition state in the combustor. Warm-up —The approximate “doubling” of the emission index during wa rm-up as we ll as the estimate of how long the machinery of the engine requires to come to a stable temperature is based on an unpublished analy sis of an RB211-535-E4-B engine during APEX3. Taxi wa y acti vi ty— As the template for these estimates, the emission index and fuel flow rate for a CFM56- 7B22 have been drawn from the engine certification value. Table 11. Single engine emissions of hazardous air pollutants at start-up.

33 no certainty as to the compositional make up of these emis- sions. There is no known measurement of in-use engine emissions at the point of ignition, however these estimates in- dicate it will be small relative to the emissions during an LTO. An unpublished analysis of an engine test from APEX3 indicates that the emissions stabilize about 40 sec after start. Other anecdotal material suggests this is a reasonable esti- mate of a typical warm-up time. Preliminary analysis of APEX3 data indicates that the initial magnitude of the HC emission indices was approximately twice as high as idle emission indices. For example, the initial CO emission index was measured to be 70 g kg−1 and this settled to just less than 35 g kg−1 after warm up. The direct measurements of formaldehyde and benzene saw very similar trends. The total impact of the warm-up period relative to the taxi- way portion is fairly straightforward to estimate. If the emis- sion rate is approximately double for the first minute, then the warm-up phase “adds 1 min” to the idle phase. If the real time spent in the taxiway mode is 20 min, then the start-up emissions account for 5% of HC emissions that stem from the combined start-up + idle emissions. Due to the uncertainty in the composition of the emissions during the ignition phase, and the possibility that the aircraft analyzed in APEX3 is not representative of the fleet emission rate behavior, a more conservative estimate of the HAP emissions associated with “start-up” is < 10%. Further, it is important to note that the preceding analysis is based on comparing start-up emis- sion indices to those at “ground idle.” If the ICAO 7% idle is used as a reference point, correspondingly different ratios would result. 5.1.2 General Aviation The general aviation category comprises aircraft that use piston engines, turbojet engines, and low bypass turbofan en- gines such as business jets. The VOC emissions of all of these are unregulated, and as a result are mostly unknown. Piston engine aircraft activity accounts for a small fraction of total fuel consumption at U.S. airports, however piston engines are the biggest source of airborne lead (Pb) from airports due to the continued use of leaded aviation gasoline (“AvGas”). For example, total throughput of AvGas at PHL in 2003 was 144,000 gallons, compared to 401,000,000 gallons of Jet Fuel A. Thus at most airports the contribution to total VOC emis- sions is minor. There are many general aviation airports in the U.S. at which piston engine aircraft are the most common type of aircraft, however, and so they are considered here. Although technically a criteria pollutant, lead is often included as a hazardous air pollutant as well. The most common grade of leaded AvGas is “100 LL,” which has a maximum lead con- tent 0.56 g Pb/L (Chevron 2006). Between 75% and 95% of the lead in the fuel is emitted in the exhaust; therefore lead emission factors are quantitatively related to the lead fuel con- tent and total lead emissions can be calculated to a high degree of certainty. For example, annual AvGas throughput at FLL in 2005 was 6x106 gallons (higher than most other commercial airports). The FLL environmental impact statement (Lan- drum and Brown 2007) used a fuel-based emission factor of 0.15 g Pb/L AvGas, which resulted in total Pb emissions (at the airport and aloft) of 3.7 tons, of which 14% (0.53 tons/year) appears on FLL’s speciated HAP emissions inventory. Very little is known regarding the emission of hydrocar- bons from general aviation, though a recent study has been released by the Swiss Federal Office of Civil Aviation (Rindlisbacher 2007). This lack of knowledge regarding HAP emissions from piston engines represents a large informa- tion gap. 5.1.3 Activity Factors for Jet Engines The ICAO databank also defines a standard landing/take- off cycle (LTO) by specifying the time spent in each mode. As illustrated in Figure 13, when the emissions vary by large fac- tors with the operational state, knowledge of the time spent in each mode is crucial for accurately calculating the total emissions. A factor that often complicates the time-in-mode estimate involves the approach and climb-out phases. In the context of assessing airport-related HAP emissions, the climb-out and landing phases of an LTO are mostly minor since VOC emissions are dominated by the idle phase. Fur- thermore, approach emissions are spread out over a very large distance. In contrast, assessment of NOx emissions re- quires knowledge of the height of the mixing (boundary) layer and time in all LTO modes since NOx emissions increase with engine thrust. The ICAO certification data sheets use 3000 feet as a nominal mixing height, and defines the climb- out period as aircraft movement from a height of 500 ft up to 3000 ft and lasting 2.2 min. The true mixing height will vary with season, time of day, and daily meteorology, and there- fore the true amount of time spent in climb-out phase will vary accordingly. Such accounting for time spent in climb- out and approach is not very relevant for assessing VOC/HAP emissions since the idle phase dominates the emissions of these species. The fraction of total HAP emissions for an LTO is domi- nated by consideration of the “idle” mode (>90%). This is pre- sented in Figure 13, in which the unburned hydrocarbon (UHC) emission rate (g/s) is plotted versus time during an LTO cycle. The total emission (g/LTO) from each LTO phase is equal to the area of the relevant “block.” Evident is the pre- dominance of the idle phase over the other phases as well as APU operation. Small deviations from the 7% ICAO thrust level at idle have large effects on the total HAP emissions dur- ing the idle mode. Evidence to date indicates that true ground

34 Figure 13. UHC (VOC) emissions during a landing/take-off cycle. The area of each block represents the total VOC emis- sions from that phase of a landing take-of cycle. VOC emis- sions are dominated by the idle phase. The error bars indi- cate the minimum uncertainty due to lack of knowledge regarding the true thrust values used and the temperature- dependence of VOC emissions. idle for most aircraft is significantly less than 7%. Additionally, the emission indices of HAPs are greatly affected by ambient conditions (temperature, relative humidity) as discussed ear- lier. The error bars indicate the minimum uncertainty associ- ated with those parameters. In order to save fuel and reduce engine noise, some air- ports and airlines encourage the practice of single-engine taxiing, whereby one engine is shutdown while idling. According to the Chicago Department of Aviation, most planes at ORD engage in the practice (Johnsson and Wash- burn 2006), and American Airlines has publicized its use of single-engine taxiing (Arpey 2007). Furthermore, Boston Logan Airport recommended use of the practice on a volun- tary basis, although its prevalence is unknown (Vanasse Hangen Brustlin 2006). Anecdotally the practice is consid- ered to be uncommon. Idle times are recorded by airports. Emission inventories, however, are based on annual averages and only express annual emissions. Such an annual average conveys little information that is relevant to acute effects, for example, the frequency of severe delays. Outside the realm of HAPs is the criteria pollutant ozone (O3), which is produced by photochemical reactions involv- ing VOCs and NOx. The NAAQS for ozone is an 8-hr average of 80 ppb. Nonattainment is based on the number of days in which the NAAQS is exceeded. Severe delays could also im- pact O3 concentrations, yet to our knowledge there has been no investigation of the link between delays and number of instances when ozone NAAQS has been exceeded. 5.2 Airport Operations 5.2.1 Auxiliary Power Units Auxiliary power units (APUs) are used to provide electri- cal power to an aircraft while on the ground so that the main engines can be turned off, and for starting the turbine engines. Total use of APUs is decreasing as jet bridges are commonly outfitted with 400 Hz plug-in electrical power and conditioned air. EDMS-recommended activity factors are 7 min when plug-in power is available and 26 min when it is not. The total emissions, even when used for 20 min, are neg- ligible compared to other sources. APU emissions might be relevant to the exposure of passengers inside cabin air expo- sure or of ground service employees. Measurements of APU emissions are available in the lit- erature. Emission factors can be accessed through EPA’s NONROAD inventory (which is used for EDMS when creat- ing emission inventories). 5.2.2 Ground Support Equipment Ground support equipment (GSE) consists of numerous vehicles such as belt loaders, baggage tugs, pushback vehicles, tractors, cabin service trucks (e.g., water, lavatory, catering, fuel trucks), de-icing vehicles, and airfield rescue and fire fighting equipment (ARFF). EPA’s NONROAD inventory and CARB’s OFFROAD are the models most commonly used for creating emission in- ventories, similar to the case for APUs. GSE emission factors

35 are expressed in g/BHP-hr. Total emissions are calculated by the product of the time-in-mode for each type of GSE, the emission factor, the horsepower, and fractional load. The extent to which such off-road inventories accurately depict the emissions from GSE is unknown, as are the effects of fleet age and maintenance. Variation between real-world “fleet” GSE emission factors and database values, however, is likely small compared to that observed in the on-road fleet (dis- cussed in Section 5.2.3). Whereas the fleet emissions of on- road vehicles are greatly influenced by vehicles with faulty emission control technology devices, the absence of such emissions control technology in most off-road engines im- plies that the variability is expected to be minor. Nevertheless, it is unclear if the off-road inventories used are appropriate for airport GSE fleets. There are few peer-reviewed publications in the litera- ture regarding GSE emission factors. Among them are a study of PAH emission factors from gasoline, diesel, and JP-8–fueled military GSE using chassis dynamometers and characterization by GC-MS. Much early work on charac- terization of GSE emission was done by the Air Force (Wade 2002). Most airports employ a mix of gasoline and diesel-fueled GSE. The Inherently Low Emission Airport Vehicle (ILEAV) Pilot Program provided funds for replacement and/or upgrades of conventionally-fueled (gasoline/diesel) GSE to alternatively-fueled vehicles (FAA 2006). Upgrades of GSE fleets are becoming increasingly common, especially among airports located in polluted areas in nonattainment of the EPA’s air quality standards (e.g., Dallas-Ft. Worth and LAX) (GAO 2003). Although no longer active, the ILEAV program has been succeeded by the FAA’s Voluntary Airport Low Emission (VALE) program, which has been successful in encouraging the modernization of many airports’ GSE fleets and applicable infrastructure. 5.2.3 Ground Access Vehicles The GAV category consists of on-road vehicles driven on airport roadways, mostly for delivering airport passengers and employees to the airport terminals and buildings. It in- cludes but is not limited to private passenger vehicles, taxis, shuttle vans (hotel, rental car, etc.), public transportation (buses), and security vehicles. It consists of a mix of gasoline, diesel, CNG, and electric-powered vehicles. Emission factors for emission inventories typically rely on EPA’s MOBILE database, which relies on chassis dynamometer measure- ments. “Real-world” measurement techniques such as on- road remote sensing (Zhang, Stedman et al. 1995), tunnel measurements (Miguel, Kirchstetter et al. 1998), “chase” ex- periments (Herndon, Shorter et al. 2004; Herndon, Shorter et al. 2005), and near-roadway ambient measurements (Miller, Hidy et al. 2006) provide an important supplement to the knowledge base. These real-world measurements have focused on criteria pollutants (CO, NOx, etc.) and only few studies have made HAP measurements; however, their find- ings are relevant to HAP emissions as well. Although few if any on-road remote sensing studies have been able to measure individual HAPs, they have been in- valuable for providing information on the distribution of a fleet’s emission factors. The most noticeable result revealed by remote-sensing measurements is that for modern gasoline- fueled cars, a small fraction of vehicles is responsible for most of the total emissions (Zhang, Stedman, et al. 1995). For example, in Denver (and many other locations) ap- proximately 5% of the cars are responsible for 50% of the fleet CO emissions. These vehicles are known as “super- emitters” or “gross-emitters,” and are usually the result of poor maintenance or in some cases emission-controls tam- pering. The emission rate of CO from a super-emitter can be 100 times higher than that from properly functioning vehicles. Similarly skewed distributions are observed for NOx and hydrocarbon emissions. The relative importance of super-emitters is likely to increase in the future as vehi- cles become cleaner due to increasingly stringent emissions standards. The accuracy of the GAV portion of EDMS-derived inven- tories depends on how well MOBILE can describe the actual fleet of vehicles at an airport. Since MOBILE’s emission fac- tors are based on dynamometer tests, assumptions regarding the frequency of super-emitters are required. The extent to which MOBILE accurately reflects any given airport fleet is unknown. A remote-sensing study at LAX from the mid- 1990s (Klein and Saraceni 1994) indicated that the fleet of taxis operated by the Bell Cab Company contained a dispro- portionately large number of super-emitters; subsequent investigation revealed that significant emission-control device tampering had occurred. A similar study in the late 1990s found that taxis at ATL were among the dirtiest fleets recorded. The researchers’ permission to operate on airport ground was even revoked after complaints from the taxi driv- ers (M. Rodgers 2007, personal communication). A different part of the same project found that New York taxis were rel- atively clean. This story highlights the challenges presented to MOBILE in accurately characterizing on-road emissions. These uncertainties are important for determining the risk of exposure groups such as airport employees that work close to the terminal ground traffic (as opposed to the airfield), but have little effect on overall airport emissions of gas-phase HAPs since aircraft at idle are the dominant HAP emitters. Further findings regarding the accuracy of MOBILE were presented at the recent 16th Coordinating Research Council

36 (CRC) On-Road Vehicle Emissions Workshop (Cadle, Ayala et al. 2007): 1. MOBILE6 predicts hydrocarbon and NO emissions rea- sonably well for new vehicles, but underestimates hydro- carbon emissions from older vehicles; 2. MOBILE6 seriously underestimates hydrocarbon emis- sions at low temperatures; and 3. The relative importance of nontailpipe (evaporative) emis- sions is highly uncertain. On-road and source apportion- ment studies suggest that EPA and CARB models greatly overpredict the importance of evaporative emissions. Another technique used to assess the accuracy of MOBILE relies on ambient measurements near roadways. Comparison of such ambient measurements with MOBILE6 has shown that there are significant flaws in the hydrocarbon speciation. For example, the benzene/acetylene ratio of MOBILE6 is likely high by a factor of three based on comparisons with ambient measurements (Parrish 2006). Further evaluations of the EPA’s National Emission Inventory (including on-road, nonroad, and area VOC emis- sions) have been reviewed recently (Warneke, McKeen et al. 2007). Miller et al. states that our estimated confidence levels in on-road mobile HAP emissions are only “medium” (Miller, Hidy et al. 2006). 5.2.3.1 Activity Factors The relevant activity factors for GAV depend on the desired level of inclusivity in an emission inventory. Some in- ventories such as those at IAD (DOT, FAA et al. 2005) include only on-airport complex miles traveled, whereas others such as PDX and ORD (FAA 2005; S. Hartsfield, personal com- munication) include much greater lengths of roads, depend- ing on the purpose of the emissions inventory. MOBILE’s emission factors are mileage-based, expressed in grams of pollutant per mile traveled. Since mileage-based emission factors tend to decrease with vehicle speed, some emission in- ventories meticulously divide the on-road sections into smaller segments and measure vehicle speed and total vehicle counts at each roadway segment (Vanasse Hangen Brustlin 2006). In such studies the uncertainty in the distribution of speeds is likely smaller than the uncertainty in the fraction of super-emitters at airports. When considering only on-airport miles traveled, total HAP emissions from GAV are minor compared to total airport emissions, which are dominated by aircraft emissions at idle power. 5.2.4 Stationary Sources Stationary sources at airports include airport power plants (HVAC systems), back-up electrical generators, parked motor vehicles (evaporative emissions), fuel storage and han- dling, airport restaurants, training fires, painting, solvent use, and engine run-up activity. Although there are large variations in the size of stationary source VOC emissions in airport emission inventories, these inventories all indicate that HAP emissions are negligible and would have to be several orders of magnitude higher than currently believed for these sources to constitute a significant (>10%) source of HAPs. The speciated HAP emission inven- tories presented in the environmental impact statements from PHL, FLL, and ORD indicate stationary sources are of minimal importance. The one exception is toluene emissions at FLL, in which 12% of total toluene emissions are ascribed to painting activities, though we note that toxicity of toluene is low compared to other HAPs emitted at airports. The rele- vant national emission inventories are used for creating emis- sion inventories (e.g., EPA’s TANKS for emissions from fuel storage tanks). Of the potentially large VOC (but not HAP) sources, the two largest contributions are from evaporative emissions from motor vehicles and fuel handling and storage. The VOC speciation of evaporative emissions closely resembles the fuel composition and is much different than that from exhaust emissions. VOC measurements at Zurich Airport (Schur- mann, Schafer et al. 2007) revealed that evaporative emission plumes contained enhanced concentrations of aromatics and C2-C9 alkanes but no enhancement of alkenes compared to background concentrations. Oxygenated VOCs were not measured, but are likely negligible as Jet A contains a very small oxygenated content. 5.2.5 De-Icing Activities The use of de-icing compounds is more a concern for water contamination than it is for air contamination. One study concluded that de-icing activities had a negligible impact on air quality (Celikel, Fleuti et al. 2003). Most envi- ronmental impact studies do not even address de-icing com- pounds in air quality chapters.