A sensitivity analysis is performed in order to examine the sensitivity of the Airline Response Model results to changes in key input parameters, for which values are particularly
ORD ATL DFW LAX IAH DEN DTW PHL EWR IAD JFK LGA BOS MIA SFO SEA DCA MDW OAK HOU DAL ONT >100% 80% 60% 40% 20% 0% 20% 40% 60% 80% >100%
Model Under-Prediction Model Over-Prediction
Airport Hub Airport 1-5 Flights per day
5-14 Flights per day 15-29 Flights per day >30 Flights per day
Model Validation: Domestic United States, 2005
uncertain. The parameters for which the model sensitivity is analysed are airport capacity and aircraft fuel burn. No other parameters are varied because all are defined in the data with little uncertainty. This includes other performance characteristics that do not typically differ significantly between aircraft type, such as cruise speed, and non-fuel related operating costs, which only vary across airlines, each of which is modelled independently.
Airport capacities, which are reported by air traffic service providers on an hourly basis, represent the air traffic service provider’s best judgement for the rate at which aircraft that can be served by the airport, given weather and wind conditions. These reported airport capacities may, however, under- or over-predict the achievable capacity because of differences in the way individual air traffic controllers operate. In order to examine the sensitivity of the model results to uncertainty in airport capacities, average hourly airport capacities at all airports are increased and decreased by 15% relative to their average reported capacities calculated from FAA ASPM data (FAA, 2008). This uncertainty level was identified by comparing effective and achievable airport capacities identified by Evans and Idris (2005) to reported capacities listed by the FAA ASPM database (FAA, 2008).
The sensitivity of the model to aircraft fuel burn rates is examined because only three aircraft types are modelled, i.e., small, medium and large, represented in each case by the most widely operated aircraft of that size class. In reality, however, airlines operate many more aircraft types, each of which has different performance characteristics. The sensitivity of the model results to aircraft fuel burn rates is examined by increasing and decreasing aircraft fuel burn rates for all types modelled by 15% relative to values extracted from BADA data (EUROCONTROL, 2004) and the ICAO Aircraft Engine Emissions Databank (ICAO, 2008). This uncertainty level was identified by calculating the standard deviation of cruise fuel burn rates for all aircraft of the same size category within the BADA database.
The Airline Response Model was rerun using the modified values of airport capacities and aircraft fuel burn described above. In Table 6-1, the sensitivity of the model is examined by comparing system O-D passenger demand, system flight operations, system CO2
emissions, and average system arrival delay for each of the sensitivity cases described above, and a baseline case with no changes to airport capacities or aircraft fuel burn. The percentage difference between the sensitivity results and the baseline result are presented in brackets
Chapter 6
Table 6-1. Model Sensitivity Results Sys. O-D Pax
Demand (per yr) Sys. Flight Ops. (per yr) Sys. CO2 (Tonnes per yr)
Avg. Sys. Arr. Delay (min) Baseline 104,794,000 1,310,000 24,548,000 11.9 Airport Capacities + 15% 106,252,000 (+1.4%) 1,345,000 (+2.7%) 25,207,000 (+2.7%) 10.6 (-11%) Airport Capacities - 15% 97,111,000 (-7.5%) 1,306,000 (-0.31%) 24,785,000 (+0.97%) 17.2 (+44%)
Aircraft Fuel Burn + 15% 101,273,000 (-3.4%) 1,329,000 (+1.5%) 28,154,000 (+15%) 12.0 (+0.84%)
Aircraft Fuel Burn - 15% 105,576,000 (+0.75%) 1,381,000 (+5.4%) 22,108,000 (-9.9%) 12.6 (+5.9%)
The Airline Response Model behaves as expected in response to changes in airport capacity. An increase in airport capacity leads to a decrease in average flight delay. Lower flight delays result in an increase in passenger demand, and reduce airline operating costs. This leads to a reduction in fares, which also contributes to the increase in passenger demand. The increase in demand results in an increase in flight operations to serve the demand, which, in turn, results in an increase in system CO2 emissions. A decrease in airport capacity
generally has the opposite effect. Flight delays increase, resulting in a decrease in passenger demand and flight operations. However, system-wide CO2 emissions increase under
decreased capacity. This is because of the emissions associated with the large (44%) increase in flight delays, much of which is incurred on the taxi-way and in airborne holding, where the engines are running. These delay-related emissions offset the decrease in emissions associated with the small (0.31%) decrease in operations.
The model result with greatest sensitivity to airport capacity is average system arrival delay. With an increase in airport capacity of 15%, average arrival delays decrease by 11%, while a decrease in airport capacity of 15% results in an increase in average arrival delays of 44%. The greater sensitivity to the decrease in capacity is because of the exponential relationship between flight delays and aircraft operations. System O-D demand is also relatively sensitive to airport capacity in the case where capacity is decreased (system O-D demand decreases by 7.5%). This sensitivity is because of the passenger response to the increase in flight delays. As described in Section 5.6, the passenger value of delay time
Model Validation: Domestic United States, 2005
applied is high – about three times the passenger value of travel time. The other results – system aircraft operations and system CO2 emissions – are not sensitive to changes in airport
capacity. Notably, even though O-D demand decreases by 7.5% in the case where airport capacity is decreased, system operations only decrease by 0.31%. This is because competition effects keep flight frequencies high, even with reduced demand.
The Airline Response Model also behaves as expected in response to changes in aircraft fuel burn. An increase in fuel burn increases airline operating costs, leading to an increase in fares, and a decrease in passenger demand. An increase in fuel burn, and thus CO2
emissions, also results in a small increase in flight operations, despite the decline in passenger demand. The reason for this is a shift in the flight network towards greater use of point-to- point operations in preference to hub-and-spoke operations. Although hub-and-spoke operations allow airlines to take advantage of economies of scale at hub airports, reducing traffic and passenger servicing costs, fuel costs are higher than in point-to-point operations because passengers are flown longer total distances. Thus, with an increase in fuel costs, airlines shift to greater use of point-to-point operations in order to reduce fuel costs. Point-to- point networks require more flights than hub-and-spoke networks to serve the same markets, and hence total flight operations increase. In this case the increase in flight operations from the shift in network offsets any decrease in flight operations required to serve the lower passenger demand. The increase in flight operations also results in an increase in flight delays.
A reduction in fuel burn has generally the opposite effect. Airline operating costs decrease, leading to a decrease in fares, which results in an increase in passenger demand and thus flight operations. Although there is also some shift in the flight network towards greater use of hub-and-spoke operations, any decrease in flight operations resulting from this change in flight network is limited by frequency competition effects, and does not offset the increase in flight operations required to serve the increased passenger demand. The increased flight operations result in an increase in flight delay, but do not offset the decrease in CO2
emissions resulting directly from the decrease in fuel burn.
The model result with greatest sensitivity to aircraft fuel burn is system CO2
Chapter 6
emissions does not match the change in fuel burn exactly, however, because of small changes in the flight network, aircraft types and flight frequencies operated, all of which are induced by the change in fuel burn. The other results – system O-D demand, system aircraft operations and average system arrival delays – are not sensitive to fuel burn.