Professional Documents
Culture Documents
a r t i c l e i n f o a b s t r a c t
Keywords: This paper presents a detailed estimation of fuel consumption and emissions during taxi
Aviation operations using aircraft position data from actual operations at Dallas/Fort Worth Interna-
Fuel burn tional Airport. Making assumptions of the thrust level during each state, fuel flow and
Aircraft emissions emission index values from International Civil Aviation Organization’s databank are
Aircraft taxi operations
extrapolated. This provides a relative comparison of all the taxi phases and their contribu-
Emissions inventory
tion to the total effect. Analysis reveals that stop-and-go situations, resulting primarily
from congestion on airport’s taxiway system, account for approximately 18% of fuel con-
sumed. The states of idling and taxiing at constant speed or braking were found to be
the two largest sources of fuel burn and emissions, and the model estimates are sensitive
to the thrust level assumptions for these states.
Ó 2011 Elsevier Ltd. All rights reserved.
1. Introduction
Efficient planning of airport surface operations is one of the goals identified in research towards the Next Generation Air
Transportation System (NextGen) in the US (Joint Planning and Development Office, 2007). As a part of this research, con-
siderable effort has been put into the capacity and delay aspects of planning, but with little quantification of environmental
effects. Taxi operations, however, are often the largest source of emissions in a standard landing take-off (LTO) cycle around
airports, but many studies that focus on aircraft emissions on the airport surface (Levine and Gao, 2007; Simaiakis and
Balakrishnan, 2009) assume an average value for fuel flow during taxi without explicitly accounting for fuel consumption
during idling, accelerating from a stop position, etc. We address this gap by developing a method for detailed estimation,
and applying it to the Dallas/Fort Worth International Airport (DFW) as a case study.
2. Methodology
The estimation of fuel consumption and pollutants emitted on the airport’s taxiway system is done in four steps:
1. Using detailed surface operations data, calculate the amount of time each aircraft spent on the taxiway partitioned in four
states (stopping, turning, accelerating, moving at constant speed or braking).
2. Assuming a certain thrust level during each of the four states, extrapolate fuel flow values from International Civil Avi-
ation Organization (ICAO)’s databank.
3. Multiply fuel flows with time spent in each state to give the estimate of fuel consumed by each aircraft.
⇑ Corresponding author.
E-mail address: nikoleris@gmail.com (T. Nikoleris).
1361-9209/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.trd.2011.01.007
T. Nikoleris et al. / Transportation Research Part D 16 (2011) 302–308 303
4. Multiply the fuel burn estimates with emission indices, corrected for atmospheric conditions at DFW, to compute the
amount of pollutants emitted per aircraft.
The ASDE-X database provides detailed position data information about each flight that arrived or departed from DFW.
Processing this data with the Surface Operations Data Analysis and Adaptation (SODAA) tool1 yields information about air-
craft type, time from spot to runway threshold, number of stops during taxi, duration of each stops, and number of turns on
taxiway. The data processing is very similar to that done in Kistler and Gupta (2009).
The aircraft position data, however, does not provide the time spent in acceleration after a stop and in perpendicular
turns; it just provides a count for such events. To address this, an average duration of 8 s is assumed for acceleration after
a stop, consisting of 4 s to overcome inertia and 4 s to reach taxi speed. Additionally, 6 s are assumed as the average duration
for a perpendicular turn throttle setting. As a result, the time each aircraft spent on the taxiway can be partitioned into four
parts:
1. Time stopped.
2. Time accelerating.
3. Time turning.
4. Time taxiing at constant speed or braking.
While the aircraft type for every arrival and departure is included in the data, the engine type installed on each aircraft is
not. To use the ICAO Databank’s for fuel flow values and emission indices, each aircraft was mapped to a single engine type.2
Although the ICAO’s databank assumes an aggregate 7% of take-off thrust for all operations on the ground, Wood et al.
(2009), DuBois and Paynter (2006) and others argue ‘‘true ground idle’’ is approximately 4%. Further, ‘breakaway thrust’,
accelerating from stop, has been found as high as 9% in a study by British Airways (Morris, 2005). Estimates of engine thrust
levels for ‘‘taxi at constant speed or deceleration’’ and ‘‘turning,’’ however, are not readily available. Assuming the taxiing
thrust level to be slightly higher than idle thrust and turning thrust to be slightly lower than breakaway thrust, 5% and
7% are used as taxiing and turning thrust levels respectively.
In estimating aircraft emissions, the engine performance models for each aircraft would ideally be used. Such data, how-
ever, is usually proprietary. The need for publicly available aircraft performance data has led to development of engine emis-
sions inventories, which provide fuel flows and emission indices as a function of engine thrust. The two widely used
inventories are the Aircraft Engine Emission Databank, developed and maintained by the International Civil Aviation Orga-
nization (1995), and the Base of Aircraft Data (BADA), developed and maintained by the Eurocontrol Experimental Centre
(Eurocontrol, 2004). The latter is used in estimating fuel consumption as a function of thrust and airspeed primarily for
the airborne phase; using it for calculating aircraft fuel consumption on the ground may not be meaningful.
The ICAO Databank is based on engine performance and emissions data obtained from full-scale engine tests at sea level.
For the vast majority of jet and turbofan commercial engines, it provides values of fuel flow (kg/s) and emission indices (g of
pollutant emitted per kg of fuel burnt) taken at 7%, 30%, 85% and 100% rated outputs. The pollutants included in the Databank
are HC, CO, and NOx.
Each aircraft’s surface trajectory is decomposed into four states: stop, accelerating after stop, turning, taxi at constant
speed or braking, thus giving the fuel consumed by aircraft i from the spot until the runway threshold as:
X
4
TF i ¼ tm;i fm;i ð1Þ
m¼1
where TFi is the fuel consumed from spot to runway threshold from aircraft i, tm,i is the time spent from aircraft i on state m,
and fm,i is fuel flow while aircraft i is on state m.
Similarly, the amount of pollutants emitted by aircraft i is estimated as:
X
4
TEpi ¼ t m;i fm;i EIpm;i ð2Þ
m¼1
1
http://www.mosaicatm.com.
2
For turboprops and piston aircraft, a representative jet engine was assigned.
304 T. Nikoleris et al. / Transportation Research Part D 16 (2011) 302–308
Fig. 1. Emission indices as a function of fuel flow for a CFM56-7B26 jet engine. Note: The four points on each trend line correspond to levels of engine thrust
7%, 30%, 85% and 100%, from left to right.
where TEpi is the amount of pollutant p emitted by aircraft i from spot to runway threshold, and EIpi is the emission index (g of
pollutant emitted per kg of fuel burnt) for pollutant p. Values for the time spent in each state, tm,i, are obtained through SOD-
AA. Fuel flows for each state m are obtained by linear interpolation using the ICAO certification fuel flows at 7% and 30% en-
gine thrust levels.
In contrast to fuel flows, extracting emission indices through linear interpolation may yield highly inaccurate values. The
emission indices of HC and CO are greatest at low throttle and decrease dramatically with increasing power settings. An
example for a CFM56-7B26 engine is shown in Fig. 1, which suggests a negative exponential trend for HC and CO emission
indices (EI). In an approach similar to Boeing’s Fuel Flow Method 2 (DuBois and Paynter, 2006), an exponential curve is fitted
to the four points listed in ICAO’s Databank to model the relationship between EI’s and engine thrust level. In particular, the
fit was performed for seven engines, installed on those aircraft that account for 80% of all movements at DFW. For the
remaining aircraft as well as for NOx emission indices, a linear interpolation was made.
The ICAO fuel flow and emission data is provided under standard atmospheric conditions. To correct for ambient condi-
tions at DFW and for installation effects, the procedure described in Boeing’s Fuel Flow Method 2 was employed, which pro-
vides adjustments for temperature, pressure and humidity conditions. For DFW, the values used are summarized in Table 1.
3. Results
Data from two fair-weather days with pre-dominantly south flow configuration was used to estimate the baseline fuel
consumption and emission during various phases of taxi operations. South flow configuration implies arriving aircraft used
runways 13R, 18R, 17C, and 17L, and departures used runways 18L, 17R, and 13L. Clear weather was assumed to be visibility
greater than 10 statute miles, dry runways, and cloud cover representing visual meteorological conditions (VMC). The two
representative days were October 7th, 2008 (Thursday) and November 6th, 2008 (Wednesday); these days represent week-
day operations on a fair-weather day. It was also assumed that all engines were in operation during all phases of taxiing.
Data from SODAA was used for estimation, and the number of flights split into arrivals and departures is given below in Table
1. The assumptions of time and thrust levels for different operations are given in Table 2.
Table 3 shows the estimated fuel consumption and emissions for October 7th and November 6th, 2008 under baseline
assumptions. The estimates are split into stop, acceleration, constant speed or braking, and turns. Both the value (in kilo-
grams) as well as percentage of total is shown. Further, departure queue stops, acceleration and constant speed or braking
were also estimated and reported (the percentage reported is also out of fuel or emissions). The fuel consumption and emis-
sions are further parsed into arrivals and departures, and the same scheme as above is used for representing these, with per-
centages denoting fractions from either arrival or departure operations. Thus, for each section (fuel, HC, CO and NOx) the first
four rows add up to 100%.
For both days, the stops and resulting acceleration events constitute approximately 18% of fuel spent in surface opera-
tions. Using a rough estimate of 120,000 kg of fuel used in surface operations per day, this yields approximately
21,600 kg or 6980 gallons of fuel. Using an approximate value of jet fuel as $2.20/gallon,3 it translates to about $15,000 a
day. Eliminating such stop-and-go situations would probably reduce the daily and annual fuel consumption as well as emis-
3
Source: IATA Jet Fuel Price Monitor, August 19, 2010. (http://www.iata.org/whatwedo/economics/fuel_monitor).
T. Nikoleris et al. / Transportation Research Part D 16 (2011) 302–308 305
Table 1
Flight traffic and ambient conditions for October 7th and November 6th, 2008 at
DFW.
Table 2
Baseline assumptions for time and thrust levels of taxi operations phases.
Baseline assumptions
Time throttle in acceleration mode but aircraft stopped 4s
Time throttle in acceleration mode, aircraft accelerating 4s
Time required to make a perpendicular turn (s) 6s
Idle thrust 4%
Taxi at constant speed or brake thrust 5%
Breakaway thrust 9%
Perpendicular turn thrust 7%
sions. However, it is possible that a potential mechanism for reducing stop-and-go transfers the stopping time to another
part of the airport (Malik et al., 2010).
The majority of the fuel consumption and subsequent emissions are from taxiing at constant speed or braking. However,
arrival aircrafts’ fuel consumption and emissions during stops is distinctly lower than departure aircraft, indicating more
uninterrupted surface trajectories for arrivals. There are two potential reasons for this difference. First, it could be attributed
to the role that runways play as main bottlenecks in the air traffic system. Departures must join the queue and traverse the
principal bottleneck, whereas arrivals have just cleared it. Second, it could be the Federal Aviation Administration on-time
performance metric,4 which considers an arrival flight delayed if it arrives 15 min beyond the scheduled time. The departure
flight metric is based on gate pushback, and thus there is pressure on arrival flights to get to the gate soon, which translates
into controllers stopping arrivals less than departures.
Although fuel consumption is almost equal between arrivals and departures (especially considering the fact that the data-
set has slightly more arrivals than departures), the trend in CO and HC emissions is different. Even though the number of
departure aircraft is slightly less, the CO and HC emissions from departures are considerably more than arrivals. That reflects
the higher emission indices for CO and HC at lower thrust levels, which occur when the aircraft is stopped in the departure
queue area (Fig. 1). The trend in NOx emissions is not that clear from the data.
To quantify the estimation difference between using different thrust levels for each taxi phase and using a uniform thrust
level for all phases (as done in published literature), a thrust level of 7% was assumed for all taxi phases and the results were
compared to the estimates in Table 3. Comparison revealed the assumption of 7% thrust level over-estimated the total fuel
consumption by 16% on both the analysis days. These results, however, are specific to the traffic conditions at DFW.
The assumptions used in Table 2 are based on Morris (2005) and Wood and Herndon (2008) as well as inputs from com-
mercial airline pilots. However, there is a lack of consensus on what values of thrust and time required for each maneuver
should be used, and the values in Table 2 may represent a conservative estimate.
For sensitivity analysis, a larger sample was used compared to the 2 days in Table 3. Fair-weather days between April 6th
and July 11th 2008 were selected, which resulted in 18,336 aircraft trajectories, of which 10,223 were arrival aircraft and
8106 were departures. Some of the departure trajectories were dropped from the data (approximately 8%) due to noise in
the positioning data. Three sets of assumptions were tested (quantities not mentioned are the same as in Table 2):
4
http://www.bts.gov/help/aviation/index.html.
306 T. Nikoleris et al. / Transportation Research Part D 16 (2011) 302–308
Table 3
Estimated fuel consumption and emissions on October 7th and November 6th, 2008 under baseline assumptions.
Total HC stopped 49 16 12 9 37 22 46 16 10 8 35 22
Total HC accelerating 8 3 2 2 6 3 8 3 2 1 6 4
Total HC constant speed/brake 230 76 112 84 118 70 223 76 109 85 114 70
Total HC turn 16 5 8 6 8 5 15 5 7 6 8 5
Queue HC stopped 27 9 – – 27 16 29 10 – 0 29 18
Queue HC accelerating 3 1 – – 3 2 4 1 – 0 4 2
Queue HC constant speed/brake 46 15 – – 46 27 47 16 – 0 47 29
Total HC 302 133 169 291 128 163
Note: Percentage denotes the fraction for that type, example: in 4th row 4th column, fuel consumption during turns for arrival aircraft is 7% of fuel
consumption during arrival operations.
In all three cases the assumed engine thrust levels are higher than in baseline case, with Cases 1 and 3 examining assump-
tions on breakaway throttle settings that are considered at the high end of possible values (Morris, 2005). The results are
presented in Table 4.
The increase in breakaway thrust (Case 1) has little effect on total fuel and emission values; effects are most prominent in
NOx emission, with 3% increase in emissions over baseline assumptions. In Cases 1 and 3, there is almost no change in the
fractional breakup of the emissions and fuel consumption. Changing the idling and constant speed taxi thrust levels (Case 2),
however, has a large effect in fuel consumption and NOx emissions. These results emphasize the importance of assigning an
accurate engine thrust level for those two kinematic states. The HC emissions appear almost insensitive to the assumptions,
whereas the CO emissions are lower at higher thrust level assumptions. This is in accordance with work by DuBois and Payn-
ter (2006) suggesting a more rapid decrease in CO emission index with fuel consumption, as compared to HC. But even with
higher thrust levels, the fractional contribution of each phase to the total does not change. Thus, stop-and-go conditions con-
stitute about 18% of fuel consumption during surface operations, irrespective of assumptions on thrust level. Although not
included in Table 4, there was almost no change in the fractional breakup of arrival and departure aircraft individually; the
percentages were the same as in Table 3.
As shown in Table 3, stops constitute a significant portion of surface operations. However, stops alone do not completely
represent ‘‘inefficient’’ operations on the surface. An aircraft might be following another and thus could be taxing at a slow
T. Nikoleris et al. / Transportation Research Part D 16 (2011) 302–308 307
Table 4
Fraction of fuel consumption and emissions under different assumptions, and sensitivity of total fuel and emissions.
speed without stopping. But situations like these cannot be filtered out completely from aircraft positioning data alone;
some form of pilot and controller intent information is also required, obtaining which is a challenging proposition.
In the absence of such information, we estimate the degree of inefficiency at the airport using unimpeded trajectories, i.e.
trajectories with the same route but taxiing at a constant speed while allowing for one stop situation (at start from spot/gate)
and the turns required to arrive at the runway. For each aircraft for the representative days, using distance traveled on the
airport surface and number of turns from the positioning data, an unimpeded trajectory was constructed assuming a certain
free-flow speed. A stop situation was added for starting from the spot. Since aircraft taxi at varying speeds, it is possible that
some aircraft’s taxi time is less than the calculated unimpeded time. If this is the case, the time of trajectory, and correspond-
ing fuel and emissions, was labeled unimpeded. If the observed time was more than the unimpeded, then the calculated
unimpeded values were used. It is apparent that such estimation would be very sensitive to the choice of unimpeded speed.
Although speed distributions at DFW have been studied, see Kistler and Gupta (2009), there is little consensus on the most
suitable unimpeded speed. Thus, we used three different unimpeded speeds for this analysis: 15, 20 and 25 knots. It should
be noted that a thrust level of 5% was assumed in all three cases for unimpeded speeds. Table 5 shows the results with unim-
peded fuel consumption and emissions both as values and fraction of total fuel and emissions. As in Table 3, the baseline
assumptions for various operation’s time and throttle settings were used.
The results provide with an estimate of the inefficiency of taxi operations compared to a best-case situation, where the
taxiway system is uncongested. For example, assuming that unimpeded speed for all aircraft is greater than or equal to 15
knots reduces overall fuel consumption, as estimated for 10/7/2008, by at least 21%. The unimpeded fuel consumption and
emissions decrease with increasing unimpeded speed and furthermore, unimpeded fuel and emissions are a higher percent-
age of the total for arrivals as compared to departures, signifying more inefficiency in departure operations.
4. Conclusions
This study provides estimates of fuel consumption and emissions from four states of aircraft taxiing on the airport surface
at the Dallas Fort Worth International Airport. It is found that stop-and-go situations account for approximately 18% of fuel
consumed, about 35% higher than in situations where aircraft taxi in an unimpeded manner at 20 knots. Idling and taxiing at
308 T. Nikoleris et al. / Transportation Research Part D 16 (2011) 302–308
Table 5
Estimated fuel consumption and emissions for unimpeded trajectories using different unimpeded speeds.
Note: Percentage denotes the fraction for that type, example: in 2nd row 2nd column, unimpeded fuel consumption is 77% of fuel consumption during
arrival operations.
constant speed or braking are the largest contributors, and are sensitive to the thrust level assumptions for these states. As
compared to differentiating between taxiing phases, the commonly used assumption of constant thrust of 7% for all phases
over-estimated the fuel consumption by 16% at DFW.
Acknowledgment
The authors would like to thank Frank Ketcham, pilot with Delta Airlines, for his enthusiastic support and his valuable
comments on aircraft taxi operations.
References
DuBois, D., Paynter, G., 2006. Fuel Flow Method 2 for Estimating Aircraft Emissions. Society of Automotive Engineers Paper: 01-1987.
Eurocontrol, 2004. User Manual for the Base of Aircraft Data (BADA), Revision 3.6, EEC Note No. 10/04, Eurocontrol Experimental Center, Brussels.
International Civil Aviation Organization, 1995. Exhaust Emissions Data Bank, Doc. 9646-AN-943, ICAO, Montreal.
Joint Planning and Development Office, 2007. Concept of Operations for the Next Generation Air Transportation System. <http://www.jpdo.gov/library/
NextGen_v2.0.pdf> (accessed on 6/29/2009).
Kistler, M., Gupta, G., 2009. Relationship between airport efficiency and surface traffic. In: 9th AIAA Aviation Technology, Integration, and Operations
Conference (ATIO), Hilton Head.
Levine, B.S., Gao, H.O., 2007. Aircraft Taxi-Out Emissions at Congested Hub Airports and the Implications for Aviation Emissions Reduction in the United
States, Transportation Research Board 86th Annual Meeting, Washington, DC.
Malik, W.A., Gupta, G., Jung, Y., 2010. Managing departure aircraft release for efficient airport surface operations. In: AIAA Guidance, Navigation, and Control
Conference, Toronto.
Morris, K., 2005. Results from a Number of Surveys of Power Settings used during Taxi Operations, EJT/KMM/1126/14.8, British Airways, London.
Simaiakis, I., Balakrishnan, H., 2009. Queuing models of airport departure processes for emissions reduction. In: AIAA Guidance, Navigation, and Control
Conference, Chicago.
Wood, E., Herndon, S., et al., 2008. Aircraft and Airport-Related Hazardous Air Pollutants: Research Needs and Analysis, Transportation Research Board 87th
Annual Meeting, Washington, DC.
Wood, Z., Kristler, M., Rathinam, S., Jung, Y., 2009. A simulator for modeling aircraft surface operations at airports. In: AIAA Modeling and Simulation
Technologies Conference, Chicago.