Supplementary Information For: Landing on Empty: Estimating the Benefits from Reducing Fuel Uplift in U.S. Civil Aviation Megan S. Ryerson, Ph.D.* Department of City and Regional Planning, University of Pennsylvania, 127 Meyerson Hall, 210 S. 34th Street, Philadelphia, PA 19104, United States Department of Electrical and Systems Engineering, University of Pennsylvania, 200 South 33rd Street, 203 Moore Building, Philadelphia, PA 19104, United States mryerson@design.upenn.edu, Telephone: 215-746-8236 *Corresponding Author Mark Hansen, Ph.D., Lu Hao, Ph.D. Department of Civil and Environmental Engineering, National Center of Excellence for Aviation Operations Research, University of California, Berkeley 109 McLaughlin Hall, Berkeley, CA 94720-1720, USA Michael Seelhorst, Ph.D. Revenue Analytics, 3100 Cumberland Blvd., Suite 1000, Atlanta, GA 30339 *Corresponding author 1
1. Introduction 1.2 Domestic Flight Planning Basics As flight crew members on the ground, flight dispatchers perform a number of duties to ensure the safe operation of a flight from its origin to destination. The process involves strategic decisions, such as checking weather forecasts and operating conditions, selecting routes and flight levels, and determining the quantity of fuel to be loaded (often termed fuel uplift), as well as tactical decisions, such as providing pilots with real-time updates, coordinating between various parties to resolve maintenance issues, and continuously monitoring the flight from takeoff to landing. In the following subsections we explore how a dispatcher chooses a route for a flight and the quantity of fuel uplifted onto the aircraft. The fuel uplifted is in quantities classified as mission fuel, reserve fuel, and discretionary fuel (sometimes termed contingency fuel by U.S. carriers); while in practice the fuel is uplifted as a single quantity, the classification of fuel into these three categories allows us to investigate any additional fuel uplift that could be reduced for fuel savings. Flight planning and dispatch is greatly different between flights within and outside the continental US (CONUS); the following discussion, and forthcoming analysis, will focus flights within the CONUS. Mission and Reserve Fuel U.S. Federal Aviation Regulations (14 C.F.R. 91, E-CFR 2014) (FARs) require a domestic commercial flight to uplift enough fuel to complete the flight to the intended destination airport, miss the landing approach at the intended destination airport, fly from the destination airport to the alternate airport (if required by weather conditions), and hold in the air for 45 minutes at normal cruising speed (Federal Aviation Administration, 2008). This mandated fuel quantity is broken into two categories: mission fuel and reserve fuel. 2
The mission fuel is calculated by the airline s flight planning system (FPS), which may be proprietary or one several off-the-shelf products. Using the FPS the dispatcher chooses a route of flight among several possible routes. The fuel consumption is calculated automatically by the FPS based on factors including the route choice, weather, and aircraft type. While the dispatcher often has a choice between several possible routes, in practice (in our study airline) typically a dispatcher will choose the route with the lowest fuel consumption, known as the economy or econ route. En route weather and turbulence can lead to another route being selected, however, thus affecting fuel load. The reserve fuel is the quantity of fuel an aircraft needs to fly for 45 min at normal cruising speed, presumably to enter a holding pattern above either the destination airport or an alternate airport or to enter a holding pattern enroute in the case of reduced airport or airspace capacity. The reserve fuel is not input by the dispatcher but rather calculated by FPS. It is at the discretion of the airline if the reserve fuel is considered useable fuel for contingencies or if instead it is to be treated as protected and not to be used. Alternate Fuel The alternate fuel is the quantity of fuel that would be needed to fly from the destination airport to the alternate airport. The designation of an alternate airport, however, requires input from the dispatcher. An alternate airport is an airport in the general vicinity of the destination airport that will serve as the designated destination in the event of some flight disruption at the original destination, such as adverse weather, congestion, airport closures, etc. When an alternate airport is listed on a flight release, the FPS calculates the additional fuel needed to miss a landing approach at the original destination and then fly to the alternate airport. 3
If a dispatcher adds an alternate airport to a flight release, it is for one of two reasons. The first is that the designation of the alternate airport (and the fuel loading that it requires) is required by the FARs because of weather conditions. The FARs require a flight to carry enough fuel to travel to an alternate airport if the weather conditions are such that visibility is less than 3 miles and the ceiling at the destination airport (defined as the distance above the earth's surface of the lowest layer of clouds (e-cfr, n.d.)) is less than 2000 feet at the flight s Estimated Time of Arrival (ETA) ± 1 hour. i The second reason a dispatcher might add an alternate to the flight release is that the dispatcher wants to provide extra buffer in the case that capacity is unexpectedly reduced and the flight needs additional fuel to complete its mission (or divert to the alternate). The dispatcher might have access to a different (and possibly internal) weather forecast or they might perceive there will be high levels of congestion when the flight in question enters the destination terminal airspace. In this case, the act of adding an alternate airport to a flight release is similar to adding contingency fuel, except the dispatcher chooses an alternate airport instead of a number of contingency fuel minutes. Contingency Fuel Discretionary fuel, often termed contingency fuel in the U.S. (not to be confused with the portion of European required reserve fuel termed contingency fuel ), may be uplifted onto a flight. The amount of contingency fuel loaded is discretionary and reflects the airline dispatcher s assessment of the downside risks that may lead to additional fuel burn beyond what is projected by the flight plan. Contingency fuel can come in three different forms. First, if an alternate is not required by weather conditions, but is added onto the flight plan regardless, this is classified as contingency fuel. This also includes the addition of a 2 nd alternate airport 4
even if a 1 st alternate airport is required by weather, as weather conditions never require a 2 nd alternate airport. Next, fuel may be added to a generic contingency category, and simply be a quantity of fuel that can be used in any phase of flight. Karisch et al. (2012) explains that many operators base the contingency fuel quantity on a statistical analysis of fuel consumption that is planned vs. historical fuel consumption burned on the same route. Numerous carriers term this Statistical Contingency Fuel (SCF). SCF is calculated in real-time in the following way: about 2 hours prior to departure, a dispatcher is presented with a flight to dispatch. This flight is between an origin and destination airport and the scheduled time of arrival falls into a small time window (for example, a 2-hour time window). When the dispatcher goes to dispatch the flight, the FPS pulls historical data (the number of years prior to be specified by the airline) of all flights between the same Origin- Destination (OD) pair that were scheduled that were scheduled to arrive in the same hour bank or time window specified by the airline. For each historical flight, the difference between the actual fuel consumption and the planned mission fuel consumption is calculated. This value is either negative meaning that the actual fuel burn was greater than the planned fuel burn (overburn), or the reverse, positive (underburn). The FPS converts the fuel over- and underburn to minutes of fuel (using a flight-specific minutes/lbs conversion factor) and estimates a normal approximation of the distribution of over- and underburn minutes. The FPS estimates the 95 th and 99 th percentiles of the distribution, which become the SCF95 and the SCF99. The dispatchers are presented with the SCF95 and the SCF99 as guidelines for contingency fuel loading. As described by Karisch et al. (2012), because the SCF value is based on actual fuel consumption, it implicitly accounts for weather and other events. It is likely that the right tail 5
therefore will contain some of the worst weather days, which will impact the estimation of the distribution and the percentiles. Note that not every flight will have an associated SCF95 or SCF99 value. As it might be statistically inaccurate to approximate a distribution with a small number of observations (flights), very infrequent flights (such as charter flights), may never have an SCF95 or SCF99. Or, a regularly scheduled flight that experiences a change in schedule might find itself without an SCF if it moves to a new hour bank which contains no similar historical flights. Other Fuel Categories An aircraft may also carry fuel such that it can be used on subsequent flight segments throughout the day, a practice termed tankering. There are numerous reasons an airline may tanker fuel: Fuel may be expensive at one airport and the airline is looking to avoid refueling there; fuel may be unavailable because the destination airport is in a remote location; or the airline may not be permitted to refuel at certain airports because of fuel contracts (Guerreiro Fregnani et al., 2013). There is a cost to carry the tankering fuel, as fuel will be burned enroute just to carry this fuel because it is an extra weight. Airlines will calculate this economic trade-off and estimate the amount of fuel to tanker on a flight. In our study airline, the FPS also requires the dispatcher to uplift fuel specifically for the taxi-out and taxi-in phases of flight. The FPS will suggest a value based on historical information (similar to SCF), in unit of minutes. The dispatcher can make adjustments to the taxi-out and taxi-in fuel in the flight planning process, using their own judgment based on information about traffic and weather at the airport (for example, flights departing congested Northeastern airports are normally assigned a larger amount of taxi-out fuel). When making the decision regarding taxi fuel, airline dispatchers can refer to the airport historical daily arrival/departure rate graph to help 6
determine the congestion level at the airport. They can also use the weather forecast information as an additional source to make the decision. 2. Methodology 2.2 Cost to Carry Factor Estimation To validate the Piano fuel consumption model, we use Equation (1) and estimate fuel consumption for every flight in our airline data set. We then calculate a percent difference between the Piano estimated fuel burn and the actual airline fuel burn. A cumulative distribution plot of the percent difference, with negative values meaning Piano is overestimating actual burn and positive values meaning Piano is underestimating actual values, is shown in Figure 1 SI. Figure 1 SI. Cumulative distribution of the percent difference between the Piano fuel consumption model estimates and actual airline fuel burn. 7
As seen in Figure 1, the Piano estimates track well with the actual airline fuel consumption, with 50% of flights being within 7% difference. For the remaining 50% of flights, Piano underestimates fuel consumption between 7% and 27%. It is expected that Piano underestimates fuel consumption, as the Piano model does not capture the endogenous effects of delay, engine use, and other factors. The implication is that the estimates of the cost to carry additional fuel in the latter part of the manuscript are conservative; consider that during an actual flight, additional fuel loaded is being carried during periods of delay. Figure 2 shows the absolute difference between actual and estimated fuel burn. The relationship is rather flat over distance, indicating that the percent difference between actual and estimated fuel consumption is greater for short distance flights. Figure 2 SI. Validation of the Piano fuel consumption model estimates over distance. 8
The differences between the Piano model and the actual airline fuel consumption are of less concern to the ultimate results of the study compared with the coefficients of Mass and Mass Distance, which are used to estimate the any excess cost to carry fuel uplifted. The full Piano model is not used to estimate entire flight fuel consumption, as that value is provided by the study airline. As these coefficients have a high statistical significance, we have confidence in the ability of the model to capture the cost to carry fuel. The coefficients from estimating Equation 1 in SI units are shown in Table 1 SI. Table 1 SI. Cost-to-carry factor estimates in SI units. Aircraft Type 9 A319 A320 A330-200 A330-300 B737-800 B737-800 Winglets B747-400 B757-200 B757-200 Winglets Variable Coefficient t - Estimate Value Mass 0.020 108.44 Aircraft Type Variable Coefficient t - Estimate Value Mass 0.018 104.01 Mass Distance 2.854 10-5 83.35 B757-300 Mass Distance 2.926 10-5 160.8 Distance 0.713 32.18 Distance 0.749 38.22 Mass 0.019 112.73 Mass 0.020 102.71 Mass Distance 2.879 10-5 105.1 B767-300 Mass Distance 2.767 10-5 106.01 Distance 0.887 47.95 Distance 1.105 28.6 Mass 0.025 59.11 Mass 0.024 74.99 B767- Mass Distance 1.784 10-5 86.9 Mass Distance 2.263 10-5 100.13 300ER Distance 1.758 35.73 Distance 1.183 31.23 Mass 0.023 100.6 Mass 0.023 95.62 Mass Distance 2.149 10-5 148.62 B767-400 Mass Distance 2.488 10-5 155.73 Distance 1.280 40.46 Distance 0.861 28.21 Mass 0.021 124.9 Mass 0.028 58.12 Mass Distance 3.395 10-5 153.86 B777 Mass Distance 1.880 10-5 105.13 Distance 0.521 34.04 Distance 1.882 35.03 Mass 0.020 143.2 Mass 0.027 164.63 Mass Distance 3.008 10-5 141.64 DC9 Mass Distance 4.606 10-5 84.74 Distance 0.638 42.94 Distance 9.917 37.77 Mass 0.027 64.05 Mass 0.022 222.96 Mass Distance 2.13 10-5 106.97 MD88 Mass Distance 3.561 10-5 172.12 Distance 2.655 36.86 Distance 1.046 82.88 Mass 0.019 89.81 Mass 0.016 151.36 Mass Distance 2.726 10-5 94.82 MD90 Mass Distance 3.450 10-5 158.95 Distance 0.874 30.81 Distance 0.762 54.09 Mass 0.019 96.66 Mass Distance 2.525 10-5 98.29 Distance 0.917 35.98
3. Cost to Carry Results 3.1 Per-Flight Cost to Carry Results The quantiles of fuel on board at arrival and contingency fuel boarded in minutes, in kg, and the cost to carry this fuel in both kg and in the percent of total flight fuel consumption are shown in Table 2 SI. Table 2 SI. Fuel on arrival and additional contingency fuel uplifted and the cost to carry this fuel. 1st Qu. Median Mean 3rd Qu. FA FATR ACAF ACF Fuel on arrival (minutes) 84.1 105.5 111.1 131.8 Fuel on arrival (kg) 3409.1 4227.3 4531.8 5363.6 Cost to Carry Fuel on arrival (kg) 182.1 254.8 305.2 379.5 Percent of total per-flight fuel consumed 3.65% 4.48% 4.86% 5.73% Fuel on arrival (minutes) 36.3 55.8 63.9 86.3 Fuel on arrival (kg) 1438.6 2000.0 2421.8 3259.5 Cost to Carry Fuel on arrival (kg) 74.7 127.7 169.5 214.7 Percent of total per-flight fuel consumed 1.55% 2.21% 2.56% 3.39% Fuel uplift (minutes) 10.0 16.0 18.0 24.0 Fuel uplift (kg) 419.1 650.5 710.9 920.5 Cost to Carry (kg) 21.5 35.2 44.4 57.8 Percent of total per-flight fuel consumed 0.57% 1.04% 1.74% 2.54% Fuel uplift (minutes) 13.0 24.0 41.0 60.0 Fuel uplift (kg) 522.3 921.4 1626.4 2414.5 Cost to Carry (kg) 26.6 59.8 102.4 143.6 Percent of total per-flight fuel consumed 0.44% 0.70% 0.77% 1.02% 3.3 Comparison of Savings from Reducing Fuel Uplift to Existing Fuel Saving Initiatives To compare the savings from reducing fuel uplift to savings from reducing taxi out fuel consumption, we calculate the average taxi fuel consumed by a flight in our dataset. Consistent with Chester and Horvath (2009) we select three representative aircraft sizes from our data (Large, Midsize, and Small) and estimate the average fuel consumed during taxi out and the median value of per-flight fuel consumption from carrying ACAF and ACF. We collected the 10
median weight at takeoff and the median distance flown for each aircraft type in the airline data. We then used Equation (1) to estimate the fuel consumption for a flight with the median takeoff weight and the median distance for each aircraft type. Table 4 SI. Aircraft details for three categories of aircraft in Table 4. Category Large Aircraft Midsized Aircraft Small Aircraft Base aircraft model Boeing 767-300 Boeing 757-300 Airbus 319 Median takeoff weight (lbs) 314000 226500 132800 Median distance (miles) 2000 1400 700 Fuel consumption per flight (lbs) 40331 22767 8592 References e-cfr: Title 14: Aeronautics and Space PART 1 DEFINITIONS AND ABBREVIATIONS, Whether that commerce moves wholly by aircraft of partly by aircraft and partly by other forms of transportation., n.d., Electronic Code of Federal Regulations. Federal Aviation Administration, 2008. Comparison of Minimum Fuel, Emergency Fuel and Reserve Fuel. Guerreiro Fregnani, J.A.T., Müller, C., Correia, A.R., 2013. A fuel tankering model applied to a domestic airline network. J. Adv. Transp. 47, 386 398. doi:10.1002/atr.162 Karisch, S.E., Altus, S.S., Stojković, G., Stojković, M., 2012. Operations, in: Quantitative Problem Solving Methods in the Airline Industry, A Modeling Methodology Handbook Series: International Series in Operations Research & Management Science. Springer US. i If the destination is equipped for CAT 1, 2, or 3 Instrument Landing System (ILS) operations, then the required minimums for an alternate are reduced slightly. Company policy may strengthen (but not detract) from these rules. Our study airline, for example, requires adding enough fuel for one alternate airport when thunderstorms are forecast for ETA ± 1 hour. Thunderstorm forecasts can vary in terms of probability. What we have seen from our observations is any chance of a thunderstorm typically results in an alternate airport listed on the flight release. 11