Fuel Burn Impacts of Taxi-out Delay and their Implications for Gate-hold Benefits Megan S. Ryerson, Ph.D. Assistant Professor Department of City and Regional Planning Department of Electrical and Systems Engineering University of Pennsylvania mryerson@design.upenn.edu meganryerson.com Lu Hao, Ph.D., Lei Kang, Mark Hansen, Ph.D. Department of Civil and Environmental Engineering University of California at Berkeley
Outline Motivation and objectives Fuel consumption model Benefits assessment Airport benefits assessment Conclusions 2
3 Motivation Pictographic
Motivation Fuel-saving initiative: gate holding Quantifying the benefit is challenging Single-engine taxi Aircraft utilize different engine thrust rates Change of taxi fuel consumption rates when encountering delay 4
Motivation Literature Mechanism of reducing taxi delay Finding Assumption for taxi fuel rate Y. Jung et al., 2011 Surface management strategy that provides release time from the gate Taxi fuel decreased up to 38% from reduced taxi movement and time ICAO emission databank, two engines at a constant thrust I. Simaiakis, 2011 Meter pushback to prevent queuing on taxiways Reduce fuel consumption at BOS significantly Two engines at a thrust setting that varies with aircraft movement states 5
Motivation Unclear taxi procedure results in inaccurate fuel saving benefit pool Unclear and varying taxi procedures Different fuel consumption rate for different taxi phase (nominal v.s. delay) Biased estimations leads to suboptimal investment for ground-based initiatives 6
Objectives Estimate taxi fuel consumption rates from actual airline data Utilize airline fuel consumption and flight statistics data Estimate fuel consumption rate separately for unimpeded taxi time and taxi delay for different aircraft types Employ our results to calculate benefit pool for reduced taxi delay from gate holding Per flight Per airport 7
Outline Motivation and objectives Fuel consumption model Benefits assessment Airport benefits assessment Conclusions 8
Fuel Consumption Model: Data (1) Data from two sources: fuel and flight statistics from a major U.S.-based carrier and flight level performance data from the FAA Fuel and flight statistics data from a major USbased airline All U.S. domestic flights April 2012 - May 2013 (810,227) Planned and actual fuel consumption, planned and actual flight time, equipment, and delay 9
Fuel Consumption Model: Data (2) Flight-level performance data for 77 large airports in the FAA ASPM database Taxi times: Actual taxi-out and taxi-in times Nominal taxi-out and taxi-in times for each flight estimated by calendar year, airline, and airport based on analysis of taxi times under low traffic conditions. When actual taxi times exceed the nominal times, the difference is considered to be delay. Airborne times: Airborne holding time (difference between planned and actual airborne time) Actual airborne time Two datasets are merged by flight number, OD, year month and day of flight Selected aircraft types with sufficient observations 10
Fuel Consumption Model: Formulation Dependent variable: F(n,ac), actual fuel burn (lbs) for individual flight n (specified by OD, year, month and day) with equipment ac Independent variables: Actual taxi (out and in) time in components The minimum of the nominal taxi time and the actual taxi time: Unimpeded taxi time in minutes The positive difference between nominal and actual taxi times: Taxi delay in minutes The square of taxi delay minutes 11
Fuel Consumption Model: Formulation Independent variables: Airborne times The 25 th percentile of actual airborne time for flights in the same month, between same OD pair and with the same aircraft type: Unimpeded airborne time in minutes The positive difference between actual airborne time minus airborne delay (holding time) and the 25 th percentile of actual airborne time: Airborne non-holding time in minutes The holding time portion of the actual airborne time: Airborne delay in minutes 12 Fixed effects Individual origin airport Individual destination airport
Fuel Consumption Model: Estimation Results AC Type DC9 757-300 757-200 A320 A319 MD88 MD90 ICAO 2-engine taxi fuel consumption rate (lbs./minute) 38.5 50.3 50.3 33.9 33.9 36.24 32.8 Variable Description Parameter estimates (Robust standard errors) Unimpeded Taxi-Out time (in Minutes) 34.628*** (2.52) 42.388*** (7.19) 49.187*** (3.12) 36.416*** (1.67) 31.309*** (1.58) 40.938*** (1.51) 26.744*** (3.14) Taxi-Out Delay (in Minutes) 30.157*** (0.87) 24.809*** (1.48) 25.737*** (0.69) 22.853*** (0.61) 21.553*** (0.62) 27.376*** (0.42) 26.332*** (0.77) Taxi-Out Delay Squared (in Minutes Squared) -0.162*** (0.02) -0.055*** (0.02) -0.064*** (0.01) -0.083*** (0.01) -0.083*** (0.01) -0.111*** (0.01) -0.161*** (0.01) Unimpeded Taxi-In time (in Minutes) 39.046*** (4.23) 30.256*** (7.30) 65.273*** (4.06) 31.579*** (2.51) 21.289*** (2.56) 51.852*** (2.73) 53.597*** (4.53) Taxi-In Delay (in Minutes) 27.934*** (2.06) 7.354* (4.36) 22.609*** (1.34) 7.371*** (1.38) 9.002*** (1.46) 16.244*** (0.93) 16.471*** (1.78) Taxi-In Delay Squared (in Minutes Squared) -0.107* (0.06) 0.539*** (0.19) -0.075** (0.04) 0.235*** (0.05) 0.319*** (0.05) 0.060** (0.03) -0.061 (0.06) Number of observations 20,167 15,206 101,298 68,700 39,296 105,278 48,585 Adjusted R-squared 0.969 0.988 0.991 0.986 0.983 0.975 0.979 ***,**,* indicate significant at 1%, 5%, 10%, respectively 13
Fuel Consumption Model: Estimation Results AC Type DC9 757-300 757-200 A320 A319 MD88 MD90 ICAO 2-engine taxi fuel consumption rate (lbs./minute) 38.5 50.3 50.3 33.9 33.9 36.24 32.8 Variable Description Parameter estimates (Robust standard errors) Unimpeded Taxi-Out time (in Minutes) 34.628*** (2.52) 42.388*** (7.19) 49.187*** (3.12) 36.416*** (1.67) 31.309*** (1.58) 40.938*** (1.51) 26.744*** (3.14) Taxi-Out Delay (in 30.157*** 24.809*** 25.737*** 22.853*** 21.553*** 27.376*** 26.332*** Minutes) Nominal taxi-out (0.87) (1.48) fuel (0.69) consumption (0.61) (0.62) rates: (0.42) 30- (0.77) Taxi-Out Delay Squared -0.162*** -0.055*** -0.064*** -0.083*** -0.083*** -0.111*** -0.161*** (in Minutes Squared) (0.02) (0.02) (0.01) (0.01) (0.01) (0.01) (0.01) Unimpeded 50 Taxi-In lbs/min 39.046*** 30.256*** 65.273*** 31.579*** 21.289*** 51.852*** 53.597*** time (in Minutes) (4.23) (7.30) (4.06) (2.51) (2.56) (2.73) (4.53) Taxi-In Delay Comparable (in 27.934*** to ICAO 7.354* 22.609*** 2-engine 7.371*** rates 9.002*** 16.244*** 16.471*** Minutes) (2.06) (4.36) (1.34) (1.38) (1.46) (0.93) (1.78) Taxi-In Delay Squared (in Minutes Squared) -0.107* (0.06) 0.539*** (0.19) -0.075** (0.04) 0.235*** (0.05) 0.319*** (0.05) 0.060** (0.03) -0.061 (0.06) Number of observations 20,167 15,206 101,298 68,700 39,296 105,278 48,585 Adjusted R-squared 0.969 0.988 0.991 0.986 0.983 0.975 0.979 14
Fuel Consumption Model: Estimation Results AC Type DC9 757-300 757-200 A320 A319 MD88 MD90 ICAO 2-engine taxi fuel consumption rate (lbs./minute) 38.5 50.3 50.3 33.9 33.9 36.24 32.8 Variable Description Parameter estimates (Robust standard errors) Unimpeded Taxi-Out time (in Minutes) 34.628*** (2.52) 42.388*** (7.19) 49.187*** (3.12) 36.416*** (1.67) 31.309*** (1.58) 40.938*** (1.51) 26.744*** (3.14) Taxi-Out Delay (in Minutes) 30.157*** (0.87) 24.809*** (1.48) 25.737*** (0.69) 22.853*** (0.61) 21.553*** (0.62) 27.376*** (0.42) 26.332*** (0.77) Taxi-Out Delay Squared (in Minutes Squared) -0.162*** (0.02) -0.055*** (0.02) -0.064*** (0.01) -0.083*** (0.01) -0.083*** (0.01) -0.111*** (0.01) -0.161*** (0.01) Unimpeded Taxi-In 39.046*** 30.256*** 65.273*** 31.579*** 21.289*** 51.852*** 53.597*** time (in Minutes) Delayed taxi-out (4.23) (7.30) fuel consumption (4.06) (2.51) (2.56) rates (2.73) are (4.53) Taxi-In Delay (in 27.934*** 7.354* 22.609*** 7.371*** 9.002*** 16.244*** 16.471*** Minutes) (2.06) (4.36) (1.34) (1.38) (1.46) (0.93) (1.78) lower than nominal rates Taxi-In Delay Squared -0.107* 0.539*** -0.075** 0.235*** 0.319*** 0.060** -0.061 (in Minutes Squared) (0.06) (0.19) (0.04) (0.05) (0.05) (0.03) (0.06) Number Compared of 20,167 to ICAO 15,206 reported 101,298 68,700 rates, 39,296 rates 105,278 for 48,585 observations Adjusted R-squared 0.969 0.988 0.991 0.986 0.983 0.975 0.979 15 fuel consumption in taxi delay are 10%-30% less
Fuel Consumption Model: Estimation Results (2) AC Type DC9 757-300 757-200 A320 A319 MD88 MD90 Variable Description Unimpeded Actual Airborne time (in Minutes) Airborne Delay (in Minutes) Non-holding Airborne Time (in Minutes) Travel Distance (in Nautical Miles) Number of observations 92.861*** (1.66) 99.043*** (1.19) 140.939*** (1.90) 3.732*** (0.22) 133.622*** (1.83) 114.924*** (1.96) 164.973*** (1.72) 0.998*** (0.23) Parameter estimates (Robust standard errors) 119.524*** (0.74) 101.499*** (0.85) 134.969*** (0.85) 0.556*** (0.10) 88.116*** (0.59) 80.009*** (0.57) 119.23*** (0.68) 1.665*** (0.08) 82.505*** (0.72) 72.240*** (0.61) 115.610*** (0.90) 1.038*** (0.10) 102.10*** (0.63) 104.01*** (0.50) 142.07*** (0.76) 1.908*** (0.08) 93.022*** (0.98) 89.99*** (0.75) 150.41*** (1.44) 2.233*** (0.13) 20,167 15,206 101,298 68,700 39,296 105,278 48,585 Adjusted R-squared 0.969 0.988 0.991 0.986 0.983 0.975 0.979 16
Fuel Consumption vs. Taxi-out Time Using Estimation Results 17 The fuel consumption increases linearly with taxi-out time when it is below the nominal taxi-out time Beyond 10 minutes, the combined effect of taxi-out delay and the squared taxi-out delay bend the plot downwards. The * on the graph denote the point of equality between X minutes of taxi out time and 10 minutes of airborne holding
Fuel Consumption vs. Taxi-out Time Using Estimation Results 18 The fuel consumption increases linearly with taxi-out time when it is below the nominal taxi-out time Beyond 10 minutes, the combined effect of taxi-out delay and the squared taxi-out delay bend the plot downwards. The * on the graph denote the point of equality between X minutes of taxi out time and 10 minutes of airborne holding
Outline Motivation and objectives Fuel consumption model Benefits assessment Airport benefits assessment Conclusions 19
Benefit Assessment Ideal scenario across sample data set: Taxi out delay is zero Every flight taxis out with unimpeded time Maximum saving that can be achieved from gate holding Calculated and reported at individual flight level Reduction in fuel consumption: Coefficient for taxi out delay time taxi out delay + Coefficient for taxi out delay squared taxi out delay 2 20
Benefit Assessment: Results Aircraft Type No. of Flights Fuel Saving per Flight (lbs) Total Fuel Consumption per Flight (lbs) Percentage of Fuel Saving CO 2 Emissions Saving per Flight (kg) A319 39,296 71.11 9909.60 0.72% 102.11 A320 68,700 73.3 12392.44 0.59% 105.26 757-300 15,206 107.03 20211.98 0.53% 153.69 757-200 101,298 118.94 16501.28 0.72% 170.80 DC9 20,167 135.56 9822.35 1.38% 194.67 MD88 105,278 114.18 11811.27 0.97% 163.97 MD90 48,585 109.84 12446.1 0.88% 157.73 Fuel consumption reduced by around 70-140 lbs/flight (0.5% - 1.4% of total flight fuel consumption) Environmental externality: saving of CO 2 emission ranges from 100-200 kg/flight 21
Outline Motivation and objectives Fuel consumption model Benefits assessment Airport benefits assessment Conclusions 22
Airport Benefits Assessment Consider the potential fuel savings from reducing taxi-out delay on a per airport basis and a time of day basis High policy relevance Help FAA evaluate possible benefits from different NextGen ground procedures and technologies for individual airports Help FAA make better decisions regarding implementation and maximize investments Cluster flights in our dataset By origin airport By time windows: 0001-0800, 0800-1200, 1200-1600, 1600-2000, 2000-2400 based on their planned departure time 23
Top 20 Airports with the Largest Possible Taxi Out Delay Fuel Savings The highest possible savings from reducing taxi-out delay are from the airports known for high levels of surface congestion At these airports, an average flight could save 60-150 lbs of fuel At the top 10 airports, an average flight consumes about 1% of total fuel in taxi delay 24 Average per flight, pounds of fuel saved At LGA, a flight could save up to 300 lbs of fuel
25 Time of Day Benefits Assessment
Time of Day Benefits Assessment Same analysis but for the 5 airports with the largest taxi out times: JFK, EWR, LGA, PHL and DTW 26
Outline Motivation and objectives Fuel consumption model Benefits assessment Airport benefits assessment Conclusions 27
Conclusions 28 This study shows the possibility of reducing fuel consumption through taxi delay reduction from gate holding About 1% of total fuel consumption can be reduced Fuel reduction varies greatly across airports New clarity on surface fuel consumption Fuel consumption from a minute of taxi-out delay is less than the impact of a minute of unimpeded taxi time by up to 30% As taxi delays grow, the rate of fuel consumption for a minute spent in taxi decreases even further, and the likelihood that an aircraft is employing fuel saving measures during taxi such as taxiing on a single engine is greatly increased Even for rough calculations it is not appropriate to apply the unimpeded rates to convert delayed taxi-out time into fuel burn A publicly available dataset with fuel consumption information could greatly enhance fuel-related research
Megan S. Ryerson, Ph.D. Assistant Professor Department of City and Regional Planning Department of Electrical and Systems Engineering University of Pennsylvania mryerson@design.upenn.edu Lu Hao, Ph.D., Lei Kang, Mark Hansen, Ph.D. Department of Civil &Environmental Engineering University of California at Berkeley 29