EUROCONTROL Experimental Centre

Transcription

1 EUROCONTROL Experimental Centre Ian Fuller Jean-Claude Hustache Tarja Kettunen Flight Efficiency and its Impact on the Environment 2003 Study EEC/SEE/2004/005 EUROCONTROL

2 Flight Efficiency and its Impact on the Environment 2003 Study Ian Fuller EUROCONTROL Experimental Centre, Centre de Bois des Bordes, BRETIGNY SUR ORGE CEDEX Jean-Claude Hustache ENVISA, 38 rue des Gravilliers, PARIS Tarja Kettunen ISA Software, 38 rue des Gravilliers, PARIS Ref : EEC/SEE/2004/005 European Organisation for the Safety of Air Navigation EUROCONTROL May 2004 This document is published by EUROCONTROL in the interest of the exchange of information. It may be copied in whole or in part providing that the copyright notice and disclaimer are included. The information contained in this document may not be modified without prior written permission from EUROCONTROL. EUROCONTROL makes no warranty, either implied or express, for the information contained in this document, neither does it assume any legal liability or responsibility for the accuracy, completeness or usefulness of this information. ii

3 REPORT DOCUMENTATION PAGE Reference: EEC/SEE/2004/005 Originator: Society, Economics and Environmental Studies Business Area Sponsor: Security Classification: Unclassified Originator (Corporate Author) Name/Location: EUROCONTROL Experimental Centre Centre de Bois des Bordes B.P BRETIGNY SUR ORGE CEDEX France Telephone: Sponsor (Contract Authority) Name/Location: EUROCONTROL ENVIRONMENT DOMAIN A. Watt. EUOROCONTROL environment domain manager EUROCONTROL Rue de la Fusée, 96 B 1130 BRUXELLES Telephone: TITLE: Flight Efficiency and its Impact on the Environment Authors : Date Pages Figures Tables Appendix References Ian Fuller, Jean-Claude Hustache, Tarja Kettunen EATMP Task Specification - 30/04 Project ENV-KPI Task No. Sponsor Period Nov Apr Distribution Statement: (a) Controlled by: EUROCONTROL Project Manager (b) Special Limitations: None (c) Copy to NTIS: YES / NO Descriptors (keywords): Env-KPI, KPI, Emissions, Performance Indicator, Emissions Study 2001, Route Efficiency, Total Fuel burn, Flight Efficiency. Abstract: This report introduces a number of transitory steps necessary to progress towards the production of enhanced flight efficiency indicators. Three issues were addressed. The first was the replication of existing indicators to test their sensitivity to different assumptions and to the evolution of measurement tools. The second was the compilation of results for airport pairs, and a classification that will ease an efficient selection of airport pair candidates to be studied in detail in future steps of the study. The third was the investigation into the unit costs to use for pricing different kinds of flight inefficiencies, while taking into account differences in operating cost between aircraft types.

4 Table of Contents REPORT DOCUMENTATION PAGE...III TABLE OF CONTENTS... IV LIST OF FIGURES... VI LIST OF TABLES... VII LIST OF ABBREVIATIONS... VIII EXECUTIVE SUMMARY... IX 1. INTRODUCTION Context Objectives of the study Scope FLIGHT EFFICIENCY INDICATORS Introduction Study data Data source Data processing Data inventory Data characteristics Indicator calculations Global results Sensitivity analysis Change in AEM version Change in flight level filter Yearly change in flight efficiency parameters Change in the flight sample distribution and the geographical area Conclusion...31 iv

5 3. AIRPORT PAIR SELECTION AND RESULTS Introduction Selection criteria Statistical reliability of indicators Identification of high inefficiency areas Identification of cases with high potential cost Selection of groups of typical cases Results and conclusion on airport pairs ECONOMIC EVALUATION Context Flight efficiency and delay costs Tactical costs airborne delays Strategic costs network design Global impacts Intermediate calculations Final cost estimation Conclusions...57 ACKNOWLEDGEMENTS...58 REFERENCES...59 ANNEX 1. METHODOLOGY...60 ANNEX 2. TOOLS USED IN THE STUDY...62 AMOC...62 AEM...62 ArcView...63 v

6 List of Figures Figure 1: The FIR s and UIR s where the study s CPR data originate from Figure 2: Geographical coverage of CPR data used for the study Figure 3: Example of CPR profile with its corresponding 4D Direct Profile Figure 4: AEM behaviour to complete flight trajectories below FL75 (2002 method) and FL30 (2003 method) Figure 5: Dispersion of departure trajectories at FL30 and FL Figure 6: Flight altitudes for arrivals and departures at Charles De Gaulle airport (Paris) under east configuration (FL30=914m, FL75=2286m) Figure 7: Standard departure profiles for certain aircraft Figure 8: Example arrival profiles Figure 9: Flowchart of data processing and the utilisation rates at different stages Figure 10: Change in the AEM calculations of fuel burn Figure 11: Comparison of flight range distributions between the traffic samples of 2002 and 2003 studies Figure 12: Most important city pairs in terms of amount of traffic in the studied sample Figure 13: Efficiency indicator and distance between airports Figure 14: Identification of groups of airport pairs Figure 15: Group of high inefficiency / long distance / high traffic Figure 16: Group of high inefficiency / long distance / low traffic Figure 17: Group of high inefficiency / short distance / high traffic Figure 18: Group of high inefficiency / short distance / low traffic Figure 19: Group of low inefficiency / long distance / high traffic Figure 20: Group of low inefficiency / long distance / low traffic Figure 21: Group of low inefficiency / short distance / high traffic Figure 22: Group of low inefficiency / short distance / low traffic Figure 23: Distance efficiency versus route length Figure 24: Flight Efficiency cost framework Figure 25: Tactical cost models Figure 26: Strategic cost models Figure 27: Aircraft types and flight range Figure 28: Flowchart of data processing with different tools Figure 29: AEM3 flight profile for emission calculation Figure 30: AEM3 method to create the Landing Take-Off phases vi

7 List of Tables Table 1: Flight data inventory Table 2: All flight plan traffic and study traffic sample categorized based on MTOW Table 3: Top 10 aircraft types in the study traffic sample Table 4: Comparison of direct and CPR profiles in terms of total distance, duration and fuel burn by phase of flight (calculations based on 13 days traffic sample) Table 5: Daily, monthly and total indicators Table 6: Indicators for different direct distance ranges Table 7: Summary of the data sets and AEM versions ran to analyse the changes in the results Table 8: Change in outputs between AEM3 v1.3 and v Table 9: Impact of AEM3 v1.5 on the inefficiency indicators Table 10: Impact of the change in flight level filter on the inefficiency indicators Table 11: Trends after correction Table 12: Division of traffic by flight range in 2002 and Table 13: Top 10 airport pairs in terms of distance inefficiency Table 14: Top 10 airport pairs in terms of cumulative inefficiency Table 15: Separators of airport pair groups Table 16: Shortlist of 35 airport pairs Table 17: Tactical cost items Table 18: Tactical airborne delay costs: en-route and holding (with network effect) based on 15 minutes delay Table 19: Strategic cost items Table 20: Cost of strategic airborne buffer minute (20% used en-route, 80% used holding) Table 21: Flight efficiency costs per flight range.* Table 22: Estimation of the average inefficiency for flight duration (% difference between radar and direct trajectories) Table 23: Global estimation for tactical and strategic cost of flight efficiency [Million 57 vii

8 List of abbreviations AEM AMOC ANSP ATFM ATM BADA CFMU CPR ECAC EEC ETFMS FL ICAO KPI LTO MTOW NM TMA Advanced Emission Model ATFM Modelling Capability Air Navigation Service Provider Air Traffic Flow Management Air Traffic Management Base of Aircraft Data (Aircraft Performance Database) Central Flow Management Unit Correlated Position Report European Civil Aviation Conference EUROCONTROL Experimental Centre Enhanced Tactical Flow Monitoring System Flight Level International Civil Aviation Organisation Key Performance Indicators Landing and Take-off (Cycle) Maximum Take-off Weight Performance Review Unit Terminal Manoeuvring Area viii

9 Executive summary The 2003 report on flight efficiency and its impact on the environment is the fourth annual report of an ongoing activity led by the EUROCONTROL Experimental Centre. The evolutions brought to the 2003 study were conducted with the intention of continuously progressing in this complex research field, and with the purpose of enlarging the scope of future users of this work. These users include the EUROCONTROL Environmental Domain, the Performance Review Unit, and potentially the Airspace Design Unit and the Central Flow Management Unit. The primary objective of the project is to develop indicators that can be used to measure the impact of the ATM system on global environment effects. In practice, the flight efficiency indicators must be used in conjunction with the priorities of a safe and orderly ATM system. Existing indicators measure the efficiency of the actual routes and profiles flown in terms of distance, flight duration, fuel burn and costs to airlines and environment. The indicators are calculated by comparing actual trajectories to fictive direct profiles (great circle distance flown between two airport and optimum vertical profile). The same indicators were computed as in the preceding annual reports to allow the testing of the robustness of the indicators with different traffic samples, and to investigate historical trends. However, the flight efficiency and its impact on the environment 2003 report is more than a simple repetition of past exercises with new air traffic samples. As identified already in previous studies, research in the field of flight efficiency involves complex relations between airspace users strategies, air traffic management operational constraints, as well as a conjunction of circumstances, meteorology, etc. Indicators simply comparing actual to fictive direct trajectories are useful but insufficient to progress in the analysis of root causes of flight inefficiency. Therefore, to better respond to the different stakeholder needs, the EUROCONTROL Experimental Centre work plan for 2004 includes some additional actions in order to progressively be able to specify enhanced indicators. In this context, the 2003 study report introduces a number of transitory steps necessary to progress towards the production of a second report focused on the specification of enhanced indicators, which will be delivered later in Three issues were addressed in the present study. The first was the replication of existing indicators to test their sensitivity to different assumptions and to the evolution of measurement tools. The second was the compilation of results for airport pairs, and a classification that will ease an efficient selection of airport pair candidates to be studied in detail in future steps of the study. The third was the investigation into the unit costs to use for pricing different kinds of flight inefficiencies, while taking into account differences in operating cost between aircraft types. Flight efficiency global results The flight efficiency indicators are intended to measure how closely the 4D path flown by an aircraft approaches the optimum 4D trajectory for the route flown. To achieve this comparison, a process of data collection, conversion, filtering and validation was carried out. This exercise allowed us to identify that, in appearance, flight efficiency indicators in 2003 were almost at the same level as in Actually, the apparent stability hides two opposite effects that partially compensate each other. On one side, the study showed that enhancements to AEM tool and new assumptions in the definition of vertical start and end ix

10 points lead to a diminishment of the observed inefficiency. On the other side, it appeared that the net historical trend from 2002 to 2003 showed a deterioration of flight efficiency. Obviously, as a comparison all else being equal was impossible, one could argue that slight changes in the geographical area and in the traffic flow repartition could explain the observed evolution. On average, the main message coming out from the 2003 flight sample analysis is that distance, duration, and fuel inefficiencies are respectively around 10%, 14%, and 9%. These results are sensitive to flight length, and argue in favour of a slight degradation of performance since 2002, but the degree of confidence is relatively poor. Airport pair results The need to adopt a more microscopic view was a necessary intermediate step to progress towards more precise indicators. Thus, this report started to look at flight efficiency results per airport pair. Almost 1,500 pairs had flight activities in the studied sample, so it was necessary to reduce the sample while keeping the most interesting and typical cases available for future steps of the study. The report shows that a group of 35 airport pairs will allow us to simultaneously capture airport links where inefficiencies are the highest (in relative and absolute values), and also airport links which experience different combinations of operational attributes (distance between airports, number of flights, relative inefficiency). Among the 35 airport pairs selected, distance inefficiency indicators varied widely. Best cases showed only 4% deviation between departure and arrival airports, while worst cases had higher than 30% distance deviation. One third of the sample had an inefficiency indicator below 8.2% and one third above 13.5%. Although the report introduces a first set of proposals for classifying airport pair distance efficiency performance, it only constitutes an initial basis among which enhanced indicators could be investigated. Flight efficiency cost results The report also contains a section presenting the results of applying to flight inefficiency indicators more recent and more precise cost figures than in the previous flight efficiency study report. This was possible thanks to the work of the University of Westminster, who provided us with their working report on the cost of airborne delays. The most important step in the new cost evaluation was the distinction between route network inefficiencies, leading to strategic costs, and airborne delay inefficiencies, leading to tactical costs. As flight efficiency indicators do not allow, so far, the separation of these two effects, the cost estimation was done along two scenarios. The first one assumed 2002 flight efficiency distribution to be 100% airborne delays, and generated an average annual cost of 716 million euros, which is slightly lower that previous year s low bound estimation. The second scenario assumed 100% of inefficiencies were related to route network design, which resulted in a strategic cost in the order of 2,658 million euros, in line with previous estimations. x

11 1. Introduction 1.1 Context This report presents the 2003 study results for the project Flight Efficiency and its Impact on the Environment. The study was conducted by ISA Software, in support of the EUROCONTROL Experimental Centre. Flight efficiency terminology is used in this project to refer to air traffic management (ATM) contributions to efficient flight. Although flight efficiency has a pure flight management dimension lying in the hands of the pilot during the flight, this part of flight efficiency was not addressed in the study. The flight efficiency project was initially set up to develop indicators that can be used to measure the impact of the ATM system on the environment. Existing indicators measure the efficiency of the actual profiles flown in terms of distance, flight duration, fuel burn and costs to airlines and environment. They were previously used by the EUROCONTROL Performance Review Commission for its annual performance review report of air traffic management in Europe. Flight efficiency concerns all parties involved with ATM. z On one hand, flight efficiency investigations are not only relevant to a few EUROCONTROL units, but to many teams, each having its own area of responsibility. These include the Airspace Design Unit, the Central Flow Management Unit, the Performance Review Unit, the EUROCONTROL Environmental Domain, and the Experimental Centre. z On the other hand, it was recognised by all parties involved that there is a need to enhance existing methodology and indicators. This is necessary to progress in the production of simple reproducible indicators, to be able to steer performance, and to generate meaningful environmental impact assessments. In this context, the report was written at a cornerstone of future flight efficiency studies, and therefore had to cover a wide range of issues. 1.2 Objectives of the study With the purpose of fitting as well as possible into the evolving context of flight efficiency studies, the aim of the study was to contribute to the following domains: z Firstly, to apply the existing methodology to the 2003 traffic sample (already successively applied to 2000, 2001, and 2002 traffic samples). The methodology relies basically on radar data and great circle distance, and the purpose is mainly to provide some historical trends for existing indicators. [Chapter 2] z Secondly, to clarify the methodology, as it is actually perceived as a black box. Restating the basic references and key assumptions is not only useful to better understand past indicators, but also to introduce new indicators. Therefore, the methods used to obtain the efficiency indicators are described in detail and an annex [Annex 2] about the specific technical steps is attached to the report. z Thirdly, to progressively investigate enhanced indicators. The first aim for these is to have separate metrics for network design inefficiencies and network utilisation 11

12 inefficiencies. This requires a computation of the flight efficiency indicators at a more microscopic level (airport pairs), after which the investigations of more detailed situations will be easier. [Chapter 3] z Lastly, the aim of the report is also to refine the economic evaluation, as flight efficiency is at the heart of many tradeoffs for which monetary valuation constitutes the most practical common denominator. More recent and more precise cost inputs will be used to estimate more realistic flight efficiency costs, and to validate the annual cost range estimated in previous year s study. [Chapter 4] Overall, the report may seem quite fragmented and probably even disappointing for the reader who wishes to find in-depth analysis of issues that may be covered too rapidly in the following sections. However, it is a necessary transition phase that will enable more stimulating investigations during the next steps of the project. 1.3 Scope Theoretically, the scope of the flight efficiency study should cover all European air traffic. However, results presented in this report do not use the same scope. z The flight efficiency indicator section uses a scope constrained by radar data and by the quality of available flight tracks. Therefore, global flight efficiency results were computed from a limited set of internal European traffic (see Figure 2). It is emphasized that the scope of the study is traffic with flight plans. z The airport section starts from the same scope as the flight efficiency section, but since its aim is to provide a more microscopic view into flight efficiency, the scope becomes limited to the few countries involved in the flight management of the particular studied airport pairs. z The economic section aims at evaluating a cost valid at the European level, but can achieve it only by adding a number of assumptions to the ones already existing in the flight efficiency section. 12

13 2. Flight efficiency indicators 2.1 Introduction The flight efficiency indicators are intended to measure how closely the 4D path flown by an aircraft approaches the optimum 4D trajectory for the route flown. To achieve this comparison, a process of data collection, conversion, filtering and validation has been carried out. This chapter describes the general methods used to obtain the efficiency indicators for the 2003 study as well as the global results. The aim of this chapter is to estimate 2003 flight inefficiency indicators [Ref. 1], while testing new assumptions and tools. This exercise should allow us to identify if there is a positive or negative trend in the evolution of the flight efficiency indicators compared to previous years. As changes applied to the methodology for 2003 obviously impact the observed trend, this report will present several tests isolating the effect of each change made since the study of Study data Data source The data sources for the 2003 study are the same as 2002 Correlated Position Reports (CPR) data and Flight Plan data from CFMU. The data used for the analysis contains records for flights starting and ending within the European Civil Aviation Conference (ECAC) area; intercontinental flights were thus not included in the analysis. Nevertheless, it is recognised that intercontinental flights are also affected by flight efficiency between entry and exit points in the European airspace. The reason for working only with intra-european flights is that the start and end points of the trajectories are defined by real geographical entities (airports). Working with segments of intercontinental flights would add an extra difficulty. The entry and exit points to the ECAC area are determined by factors outside ANSPs (Air Navigation Service Provider) scope of responsibility. The data covers one week in March and one week in June. One day in the March data set (30/03) contained errors (after filtering no records of this day were left for further analysis) so data for this day has not been included in the calculations. z CFMU Last Filed Flight Plan The recordings of CFMU Flight Plans were in blocks of 24 hours. The recordings are comprised of the Last Filed flight plans (that gave the planned itinerary of each flight validated by the CFMU), and the scheduled time of departure. The ground track and vertical profile followed by the flight does not always adhere to the last filed flight plan for tactical and operational reasons. Although the modifications to the flight plan implemented during the flight were not available, the Last Filed Flight Plan gives a representative picture of ECAC-wide daily traffic. CFMU Flight Plan Data were necessary to check the consistency of the CPR data, and also to generate the Direct 4D Profiles (see Chapter ). 13

14 z Correlated Position Report CPR The European Traffic Flow Management System (ETFMS) provides Correlated Position Reports (CPR). CPR data are derived from processed radar track data that are provided by the Air Navigation Service Providers (ANSP) in the states participating in ETFMS. (ANSPs providing the data for this study are shown in Figure 1). The CPR data were obtained by a special agreement with the ETFMS team at CFMU. 1. Scottish UIR 2. Shannon UIR 3. London UIR 4. Fr ance UIR 5. Brussels UIR 6. Amsterdam FIR 7. Hannover UIR 8. Copenhagen FIR 9. Switzerland UIR 10. Wien FIR Figure 1: The FIR s and UIR s where the study s CPR data originate from. Figure 2 shows an example for one day of radar tracks that were studied under this limited scope. 14

15 Figure 2: Geographical coverage of CPR data used for the study Data processing Apart from the necessary data processing to transform the recorded data into the formats required by the analysis tools (AMOC, AEM, ArcGIS etc.) the data preparation process was important to ensure consistency and coherency between the three sets of flight information used in the study (CPR, Flight Plans and Direct trajectories) Direct route profile creation For the purpose of calculating the actual efficiency indicators (presented in detail in Chapter 2.4), a direct trajectory corresponding to each CPR trajectory needed to be created. z The Direct Route (horizontal definition) was defined as the great circle distance between the first and last surveillance data point of each filtered CPR trajectory (see Chapter ). The great circle distance was considered as the best approximation of optimum route since it is the shortest distance between two points on the earth s surface. The direct route was unique to each CPR route even for a same origindestination airport pair due to the fact that each CPR trajectory was unique and thus the corresponding direct route was created based on the first and last radar record points of the CPR trajectory. Creating a single average direct route for each airport pair was considered but it would not have changed anything for the resulting distance efficiency calculations which use the sum of all CPR distances and the sum of all direct distances. 15

16 z The Optimum Profile (vertical definition) is the vertical path which best suits the type of aircraft for the route to be flown. The vertical profile depends on the aircraft type and distance flown. The direct routes/profiles were calculated for each track of the Flight Plans using ATFM MOdelling Capability tool (AMOC) together with the BADA aircraft performance database. It must be noted that the aircraft performance file of AMOC has not been updated since An update should be performed for example to cover more aircraft types and thus to improve the analysis. The arrival times for the direct trajectories were recalculated by AMOC. These correspond to the arrival time after using the direct route and thus are always prior to the actual arrival time. The following diagram shows an example of a calculated direct profile corresponding to the CPR profile CPR DIRECT Flight level event time (s) Figure 3: Example of CPR profile with its corresponding 4D Direct Profile CFMU Flight Plan Processing The CFMU Flight Plan files were processed and completed by AMOC. It was possible for some of the tracks to be incomplete (usually missing waypoint information) which made it necessary to extrapolate trajectories. Care was taken to ensure consistency of information before such trajectories were used in the analysis. 16

17 CPR Processing The CPR flights plot data was first correlated and filtered by AMOC, since a large part of the recorded radar data were of no interest for the study. The original CPR files from ETFMS contain geographical plots with a callsign as identifier, which is not unique. The correlation revealed some flights with an insufficient number of plots as well as some plots without any Flight Plan. These flights were removed from the data set as well as flights with the same departure and destination airport. Also, some flights had multiple radar records from different air traffic control centres which had to be removed. At the end of this first processing step, a large set of tracks remained. In addition to the correlation and initial filtering performed by AMOC, additional filters were applied using Microsoft Access to obtain profiles appropriate for the flight efficiency analysis. These three filters are described below in detail. Filter 1. The first and last CPR data record of each trajectory lower than FL30 1 In the 2002 study the flight level cut-off point was set to FL75 but for 2003 the limit was lowered to FL30 to better capture the ideal trajectories for the analysis from several viewpoints (although it was recognised that a FL30 cut-off would impact the sample size). First of all, having the limit at FL30 will cover the flights tracks almost to the ground and thus a large part of the Terminal Manoeuvring Area (TMA) activity will be included in the flights histories. Setting the limit lower than FL30 would be problematic since the radar coverage is often incomplete or missing from low altitudes and thus the lower the limit is set, the fewer flights can be analysed. Secondly, choosing FL30 as the cut-off point for trajectories made the calculations more convenient because we used the standard ICAO LTO cycle in AEM between the airport and FL30 to all flights according to the type of aircraft. If the first or last plots of the CPR data were from higher than FL30, the vertical profile would be interpolated between the LTO cut-off point (FL30) and the first/last plot. The interpolation was done in a way where the departing aircraft is modelled to move from the airport to the flight level of the first record point with only a vertical change in position (using the airport coordinates). The aircraft is then modelled to fly on cruise phase to the real coordinates of its first data record (and vice versa for arriving aircraft). When using FL30 as the cut-off point for trajectory selection this artificial trajectory completion is less significant and the results are more realistic. Figure 4 illustrates how AEM functions in the case of a very simplistic trajectory with CPR records to both FL 75 and to FL30. 1 When describing the vertical position of the aircraft, a notion of Flight Level (FL) is used instead of altitude since the analysis is based on CPR data which is derived from tracked radar data using Flight Level. FL30 corresponds to 3000 feet (914m), and FL75 corresponds to 7500 feet (2286m). 17

18 Airport Original trajectory AEM trajectory Standard LTO cycle FL75 FL30 Figure 4: AEM behaviour to complete flight trajectories below FL75 (2002 method) and FL30 (2003 method). Lastly, the choice of the cut-off point affects the consistency of the trajectories. Below 3,000 feet the aircraft are normally inline with the runway centreline, and the distance from the departure airport is around 3-4 NM (nautical miles) after take-off. Further away from the airport the trajectories begin to disperse more and at FL75 the flights are around 7-15 NM from the departure airport but can be situated anywhere in a wide range of geographical locations. The choice of FL30 is justified since below this level a flight is still mainly influenced by the constraints of final approach or initial climb and so the start/end points for all flights over all airports are more uniform, resulting in less geographical distribution. On the contrary, at FL75 the error zone is much larger, as illustrated in Figure 5, due to factors such as the aircraft performance, aircraft load, destination of the flight, weather, flow organisation around the airport as well as transfers between sectors. Runway FL75 FL30 Figure 5: Dispersion of departure trajectories at FL30 and FL75. 18

19 As an example, Figure 6 shows the dispersion of trajectories to and from Charles De Gaulle airport. Arrivals Source: Aéroports de Paris Departures Figure 6: Flight altitudes for arrivals and departures at Charles De Gaulle airport (Paris) under east configuration (FL30=914m, FL75=2286m). 10,500 9,000 7,500 B Height [Feet] 6,000 4,500 A320 DHC830 3,000 DHC8 1,500 EMB Distance to the start of take-off run [KM] Figure 7: Standard departure profiles for certain aircraft. 19

20 10,500 9,000 7,500 Height [Feet] 6,000 4,500 3,000 1, Distance from touchdown [KM] Example of standard ILS approach Example of continuous descent approach Figure 8: Example arrival profiles. Filter 2. Flight duration at least 15 minutes The duration requirement was set to 15 minutes quite arbitrarily to exclude short local flights and short distance flights within a TMA for which flight efficiency indicators are irrelevant. The duration was calculated between the first and last CPR record point, so the duration is not the duration of the complete flight from departure airport to destination airport nor is it the duration between the plots starting and ending at FL30 as trajectories are filtered at a later stage. Filter 3. Maximum flight level (cruise) greater than FL75 An assumption was made that flights cruising below FL75 would mostly be general aviation flights which were outside the scope of this study. All of the filtered trajectories were validated visually using ArcGIS tool (described in Annex 2) to check for any remaining inconsistencies in the data after the correlation and filtering process Data inventory The utilisation of the traffic data at different stages of the data processing can be seen from Table 1. The total utilisation rate of the 2003 data is 5.0 % (2002: 16 %). Table 1 shows also that if the filtering criterion would have been FL75, the utilisation rate would have been close to that of

21 Since the selection criteria are more stringent than for 2002 (only flights having radar records to as low as 3,000 feet are included as opposed to the limit of 7,500 feet in 2002) it was anticipated that the total number of flight records after filtering would be lower. Table 1: Flight data inventory. Number of flights Mon 24/03 Tue 25/03 Wed 26/03 Thu 27/03 Fri 28/03 Sat 29/03 Sun 30/03 Mon 16/06 Tue 17/06 Wed 18/06 Thu 19/06 Fri 20/06 Sat 21/06 Sun 22/06 Total A in ALL_FT file B Flight plans processed C CPR tracks after 1st filtering (AMOC) D CPR tracks after 2nd filtering (Access, complete profile) % of CPR flights used out of total flight plans 6.2% 6.3% 6.0% 5.8% 4.7% 4.0% 5.1% 4.4% 4.8% 5.2% 5.0% 3.8% 3.5% 5.0% % of CPR flights used out of total flight plans if start and end FL criteria > % 17.0% 14.4% Definitions of fields A-D in Table 1: A: The number of Last Filed Flight Plans without filtering (in the ALL_FT File) received from the CFMU. B: The number of Last Filed Flight Plans (A) after processing by AMOC to remove flights that do not have complete flight data, e.g. unknown destination airport. C: The number of radar flights after processing by AMOC to remove flights without corresponding Last Filed Flight Plans (B) and after removing outlier radar messages in position and time. D: The number of radar flights after the final selection and filtering with the abovementioned criteria. 21

22 [A] = Last Filed Flight Plans (CFMU) CPR data (ETFMS) AMOC removes incomplete flights (e.g. no destination airport) [B] = Flight plans processed: 93% of [A] AMOC removes outlier radar messages and correlates flight plans to CPR data [C] = Pairs of CPR track & Flight Plan available: 55% of [A] Additional filters: o Complete CPR profiles from FL30 to FL30 o Flights longer than 15 minutes o Maximum altitude > FL75 [D] = Studied traffic sample: 5% of [A] Figure 9: Flowchart of data processing and the utilisation rates at different stages. The geographical coverage remains almost constant compared to year Only the eastern part of Germany that was included in the 2002 traffic sample is absent in the 2003 sample Data characteristics The traffic sample used is different from the global European traffic flying during a complete year. These differences might impact the unit costs to be used in the economic evaluation chapter, as a flight extension for a wide-body aircraft will obviously cost more than for a small jet. Table 2 and Table 3 are shown to characterize the traffic sample used and to justify the adaptation needed for the economics chapter. Table 2 presents both the total ECAC traffic and the used filtered traffic sample categorized in terms of maximum take-off weight (MTOW). It can be seen that the used traffic sample is quite representative of the whole ECAC traffic even though the study traffic sample has a somewhat larger share of light aircraft (under 25 tons) and aircraft between 50 and 100 tons. All other take-off weight categories are under-represented in the final traffic sample used. 22

23 Table 2: All flight plan traffic and study traffic sample categorized based on MTOW. All traffic (flight plans) Flights in study sample Number of Number of Maximum take-off flights in 13 % flights in 13 % weight MTOW (kg) days days - 25, ,2% ,2% 25,000-50, ,9% ,0% 50, , ,0% ,1% 100, , ,8% 274 1,7% 150, , ,0% 242 1,5% 200, , ,8% 21 0,1% 250, , ,1% 12 0,1% 300, , ,1% 0 0,0% 350, ,0% 52 0,3% Total Note: Table 2 includes data only where a MTOW value was available. A MTOW value was found for 97.1% of the total ECAC traffic and for 99.6% of the used traffic sample. Table 3 presents the 10 most common aircraft types in the final traffic sample. Altogether, these aircraft account for 52 % of the whole traffic sample, where a total of 111 different aircraft types were present. Boeing 737 models alone account for 28% of the total. The 10 most common aircraft in the traffic sample constitute 31% of all the traffic in the whole ECAC area. Aircraft Table 3: Top 10 aircraft types in the study traffic sample. Engine type Flights in study sample (13 days) % All traffic (Flight plans) Boeing Jet 2 engines % % Airbus A320 Jet 2 engines % % Fokker 50 Turboprop 2 engines 857 5% % Boeing Jet 2 engines 845 5% % Aerospatiale/Alenia ATR Turboprop 2 engines 831 5% % Airbus A319 Jet 2 engines 772 5% % Embraer 145 Jet 2 engines 748 5% % Boeing Jet 2 engines 644 4% % Fokker 100 Jet 2 engines 641 4% % Boeing Jet 2 engines 569 3% % % 23

24 2.3 Indicator calculations Three flight inefficiency indicators were calculated based on the following data: 1. CPR trajectories 2. Direct trajectories corresponding to the CPR flights and computed by AMOC. The distance inefficiency indicator illustrates the extra ground track distance flown in comparison with the direct ground track route for each flight. (See chapter ) The duration and fuel burn inefficiency indicators measure the extra duration and fuel consumed compared to the direct flight profile (including the optimum vertical profile unlike in the case of distance inefficiency). The inefficiency indicators were calculated using the following formulas: Distance indicator = Duration indicator = Fuel burn indicator = DistanceCPR Distance DurationCPR Duration FuelburnCPR Fuelburn Distance direct Duration direct Fuelburn direct direct direct direct Note that these inefficiency indicators measure inefficiency, thus a lower value represents a better performance. In addition to the general flight inefficiency indicators, a more detailed analysis based on flight phases was attempted. The goal was to identify which phases of the flight experienced the greatest inefficiency. However, the comparison of flight phases between the CPR and direct trajectories found to be unfeasible since the flight phase analysis was not sufficiently robust. As can be seen from Table 4, the shares of different flight phases differ considerably between the CPR and direct trajectories. CPR trajectories indicated shorter cruise phases and longer descent phases than direct trajectories. This is logical since direct trajectories are created with optimum vertical profile which means that the aircraft climb to their cruise altitude as fast as possible and thus the cruise phase is longer than in reality (CPR). Table 4: Comparison of direct and CPR profiles in terms of total distance, duration and fuel burn by phase of flight (calculations based on 13 days traffic sample) Dis tance Duration Fuel burn Phase of Flight Direct CPR Direct CPR Direct CPR Climb 24% 26% 27% 28% 49% 52% Cruise 52% 41% 52% 40% 45% 39% Descent 24% 32% 20% 32% 6% 9% 24

25 2.4 Global results The results have been calculated individually for each day, for each week as well as for the whole data sample (Table 5). In the 2001 and 2002 studies the indicators were calculated based on flights whose direct distance was km, because of data availability constraints and because of underlying assumptions. For 2003 some progress was made in the coverage of flights longer than 1100 kilometres, but results are still displayed for the range of km. This reduced range is justified by the fact that for long flights, winds are expected to play an important role in the choice of routes flown, which would make the observation of extra distances hardly interpretable as an inefficiency. When comparing the indicators for 2002 and 2003 (Table 5), it can be noticed that the distance and duration efficiency have decreased from the previous year, whereas fuel burn efficiency has increased slightly. These results may be affected by the changes in methodology between 2002 and A sensitivity analysis in the following section evaluates the impact of these changes in methodology. Table 5: Daily, monthly and total indicators. Distance inefficiency Duration inefficiency Fuelburn inefficiency Mon 24/03/ % 13.8% 8.7% Tue 25/03/ % 13.1% 7.4% Wed 26/03/ % 14.9% 8.9% Thu 27/03/ % 14.1% 8.6% Fri 28/03/ % 15.6% 10.3% Sat 29/03/ % 14.2% 8.4% Mon 16/06/ % 13.5% 8.6% Tue 17/06/ % 15.2% 9.4% Wed 18/06/ % 13.5% 8.8% Thu 19/06/ % 17.6% 12.2% Fri 20/06/ % 17.7% 12.4% Sat 21/06/ % 15.1% 10.0% Sun 22/06/ % 16.3% 11.1% March 10.3% 14.2% 8.7% June 9.7% 15.6% 10.4% All 10.0% 14.9% 9.6% All km 10.2% 14.8% 9.4% 2002 results km 8.9% 13.5% 9.6% 25

26 Direct distance (km) Table 6: Indicators for different direct distance ranges. Number of flights Distance inefficiency Duration inefficiency Fuelburn inefficiency % 14.0% 8.2% % 16.5% 10.7% % 15.6% 11.3% % 15.9% 9.8% % 12.5% 7.0% % 14.4% 8.6% % 12.1% 6.9% % 15.4% 9.3% % 17.5% 12.0% % 23.5% 17.2% % 17.1% 12.7% % 16.9% 10.1% % 15.3% 12.2% % 13.5% 7.2% 2.5 Sensitivity analysis This section presents a sensitivity analysis to evaluate the impact of the changes in methodology between 2002 and The difference in the results of the 2003 study compared to the 2002 study can be tracked down to three potential sources: 1. Change in the version of AEM used: The 2002 study was carried out with AEM3 v1.3; the 2003 study used the latest AEM3 v1.5. The major change between the two versions concerns the fuel burn calculations. In the earlier version used for the 2002 study the fuel burn rate was constant from the very beginning of a flight leg until the flight reached the end of its current leg, i.e. the same value (value based on the first point of the leg) was applied to the whole leg. In AEM3 v1.5, the fuel burn takes into account the evolution of the flight level between each record point making the calculations more precise than before. The fuel burn rates being applied come from the BADA aircraft performance data file. Figure 10 illustrates the difference in fuel burn calculations between the two versions. AEM3 v1.3 used the fuel burn rate of trajectory point 1 for the whole climb phase; whereas AEM3 v1.5 uses individual fuel burn rates at each point (1-5) calculated based on their flight level. Altitude 5 4 Flight data record Flight trajectory Time Figure 10: Change in the AEM calculations of fuel burn. 26

27 2. Change in the cut-off flight level for trajectories accepted for analysis: The 2002 study used FL75 as the cut-off point for trajectories, whereas the 2003 study uses FL30 as the limit. This change causes the trajectories to be of longer length but on the other hand there are fewer flights in the studied sample. 3. Yearly change in the flight efficiency parameters: This change in results reflects the most interesting factor, namely the development of the flight efficiency between the years 2002 and Three days were chosen to represent the 2002 data for the sensitivity analysis: 20/09/2002, 23/09/2002 and 26/09/2002. Table 7 summarises the different data options needed to carry out the analysis to identify the three above-mentioned factors. Table 7: Summary of the data sets and AEM versions ran to analyse the changes in the results. Reference Data source Flight level AEM version Days of data filter applied A 2002 FL 75 AEM3 v (3 used) B 2002 FL 75 AEM3 v1.5 3 C 2002 FL 30 AEM3 v Change in AEM version The impact of the new version of AEM3 (v1.5) was quantified by comparing the results after running the same data set through both versions of AEM (options A and B in Table 7). change output Table 8: Change in outputs between AEM3 v1.3 and v1.5. CPR Direct distance duration fuelburn distance duration fuelburn in 0% 0% +6% 0% 0% +8% Based on the results shown in Table 8 and Table 9, AEM3 v1.5 gives somewhat different results compared to version 1.3 used in last year s study. The change from old version to new version is noticeable in the case of fuel burn outputs, where the fuel burn values have increased 6-8% depending on the data set (CPR/direct data). Changes in distance and duration are logically negligible, and if any, they are attributable to rounding errors. The reason why the fuel burn outputs have increased more for direct trajectories is that there are much more data (radar) records for the CPR flight trajectories than for the optimally constructed direct trajectories (see Figure 3). A different fuel burn rate value is applied at 27

28 each record point depending on the flight level. Thus, the fuel burn calculations for direct trajectories are not as precise as for CPR trajectories. Table 9: Impact of AEM3 v1.5 on the inefficiency indicators. Distance efficiency Duration efficiency Fuelburn efficiency AEM3 v % 14.7% 10.8% AEM3 v % 14.6% 9.1% As the increase in fuel burn is more important for direct profiles than for CPR profiles with the AEM3 v1.5, it results in an improvement of the measured efficiency (indicator decreasing from 10.8 % to 9.1 %). As a conclusion, changes in the AEM version used for the study have no impact on the distance and duration indicators, but could explain why the fuel indicator shows an improvement trend. Factors explaining the changes in fuel consumption are presented in Figure 10. Table 5 showed that both the distance and duration inefficiency indicators increased by 1.3 percentage points between 2002 and 2003, but decreased by 0.2 percentage points for the fuel indicator. Actually, Table 9 shows that AEM3 v1.5 could explain by itself a reduction of 1.7 percentage points for the fuel indicator. Thus one could conclude that if AEM3 v1.5 had been used for the 2002 study, the evolution observed for fuel inefficiency indicator would have been +1.5 percentage points, which is consistent with the evolution observed for distance and duration Change in flight level filter The impact of changing the cut-off point of the trajectories from FL75 to FL30 was quantified by comparing the results of data sets B and C in Table 7. The two data sets are of different size (data set with filter at FL30 has less than 4,000 trajectories whereas the data set with filter at FL75 has almost 14,000 trajectories), so changes in the raw AEM outputs cannot be presented. However, the resulting efficiency indicators are shown in Table 10. Table 10: Impact of the change in flight level filter on the inefficiency indicators. Distance inefficiency Duration inefficiency Fuelburn inefficiency FL % 14.6% 9.1% FL % 13.5% 9.0% With a lower filter (FL30) the resulting efficiency indicators are lower for distance and duration, but the impact is negligible for fuel burn. Thus, using filter FL30 instead FL75, would have decreased indicators for In conclusion, the change in flight level does not explain the increased inefficiency observed in On the contrary, it shows that with constant assumptions, the observed evolution would have been even worse. (+1.3 percentage points would become +1.8 percentage points for distance and +2.4 percentage points for duration). 28

29 2.5.3 Yearly change in flight efficiency parameters The most interesting factor to identify for the purpose of monitoring the flight efficiency performance is the net yearly evolution of indicators, after corrections for potential bias coming from changes in assumptions and tools. This net yearly evolution was quantified by assuming that impacts from AEM and flight level changes observed on 3 days are applicable to the whole dataset. The lines of the table below should be read as follow: if new assumptions were applied to 2002 traffic, the indicator would be x percentage points higher (+) or lower (-). Table 11: Trends after correction. Distance Duration Fuel Impact of AEM version change Flight level change FL75 ->FL Cumulative impact published results (200-1,100 km) 8.9% 13.5% 9.6% 2002 recalculated results (with cumulative impact) 8.4% 12.3% 7.8% 2003 results (200-1,100 km) 10.2% 14.8% 9.6% Trend after correction In conclusion, we see that flight inefficiency indicators for 2003 are higher. This trend could already be identified in Table 5, except for the fuel consumption indicator. After correcting for the impacts of AEM and flight level changes, the evolution is even worse than indicated in Table 5, and the trend is consistent between all indicators Change in the flight sample distribution and the geographical area So far, no precise validation tests have been performed in the area of flight sample distribution and geographical coverage of the studied sample. However, the evolution trend could be further explained by the two following factors: z 2003 data did not include CPR data for low flight levels for a large part of East Germany (see Figure 1 and Figure 2) compared to 2002 data. It could be that efficiency in this area was better than the average. Then, omitting the traffic would tend to increase new average results. This kind of element will be difficult to quantify, and the only thing that could cancel this source of bias in the trend computation would be to work with well-defined and yearly constant sets of city pairs. z The distribution of traffic (see Table 12), compared to last year is much more concentrated on very short range flights ( km) where observed distance inefficiencies are the highest. It could partially explain why indicators are higher in 2003 than in These differences in the indicators should be studied further. However, a rapid estimation indicates that the impact of this factor on the results is almost negligible. It would actually reduce observed inefficiency, but only by 0.1 percentage points (in the case of distance indicator, as shown in Table 12). 29

30 Range Table 12: Division of traffic by flight range in 2002 and Distance inefficiency 2003 Number of flights % 2002 Number of flights % 2,933 20% 4,805 14% % 2,671 19% 7,371 21% % 2,543 18% 6,747 19% % 1,843 13% 6,032 17% % 1,441 10% 3,989 11% % 1,090 8% 2,376 7% % 459 3% 1,218 3% % 968 7% 1,724 5% % 379 3% 784 2% Total traffic 14,327 35,046 Weighted average inefficiency based on 2003 traffic distribution Weighted average inefficiency based on 2002 traffic distribution 10.9% 10.8% 2003 traffic repartition impact % 25% 20% Share of total traffic within the range 15% 10% 5% 0% flight range (km) Figure 11: Comparison of flight range distributions between the traffic samples of 2002 and 2003 studies. 30

31 2.6 Conclusion This chapter allowed us to identify that, in appearance, flight inefficiency indicators in 2003 were almost at the same level as in Actually, the apparent stability hides two opposite effects that partially compensate each other. On one side, the study showed that new assumptions in the definition of vertical start and end points, as well as enhancements brought to the AEM tool, lead to a reduction of the observed inefficiency. On the other side, it appeared that the net historical trend from 2002 to 2003 showed a deterioration of flight efficiency. Obviously, as a comparison of all else being equal was impossible, one could argue that slight changes in the geographical area and in the traffic flow repartition could explain the observed evolution. On average, the results show that distance, duration, and fuel inefficiencies are respectively around 10%, 14%, and 9%. These results are sensitive to flight length, and argue in favour of a slight degradation of performance since 2002, but the degree of confidence is relatively poor. 31

32 3. Airport pair selection and results 3.1 Introduction As shown in previous sections, assessing flight efficiency on a global scale has some advantages, but rapidly encounters limitations as far as the interpretation of results is concerned. When starting to investigate the reasons for inefficiencies, and how to define new indicators capturing all aspects of flight efficiency, the need to adopt a more microscopic view became evident. Performance levels will vary between different airport pairs, and measuring flight efficiency at this level should present two advantages. First, it should provide the opportunity to observe particular segments of the route network and to identify causes for route design extension as well as strategies for network utilisation. Second, it should allow comparison of flight efficiency across airport pairs, and possibly progress towards some benchmarking indicators. The purpose of this section is not to undertake a precise review on airport pair s relative performance. This activity will be carried out in future steps of the project, and will be presented in a report entitled Enhanced Flight Efficiency Indicators. The work presented here on airport pairs is an initial presentation of how efficiency indicators for airport pairs compare to the global results. Although a partial coverage of the European traffic was available, almost 1,500 airport pairs had flight activity in the studied sample. This figure is obviously too high to realistically present flight efficiency results for all airport pairs. Therefore, this section aims to present some selection criteria that can be used to identify a few relevant cases, and to help categorize different situations. The shortlist presented as a conclusion of this section will then serve as a basis among which airport pair candidates will be chosen for future steps of the project. 3.2 Selection criteria Studying flight efficiency on routes where the potential cost to airlines and the environment are the greatest make sense. These are the places where corrective actions, if feasible, should be targeted in priority. However, understanding the drivers of flight efficiency performance is not possible by looking only at the places where inefficiencies are the highest. Actually, investigating cases where efficiency is good provides a complementary insight into the conditions supporting efficient flights. The selection of airport pairs started from the complete list of pairs available in the traffic sample (1,491) after which a series of criteria listed hereafter was applied Statistical reliability of indicators Number of flights The number of flights per airport pair varied greatly due to CPR data availability and the different filters that needed to be applied. The final number of flights to be analysed for each airport pair was not proportional to the actual number of flights between these airports. 32

33 Therefore, to minimize the risk of attributing to some airport pairs a flight efficiency metric that would mainly rely on exceptional cases, a decision was made to apply a threshold on the number of flights necessary to select an airport pair into our shortlist. This threshold was set arbitrarily to 40 flights in the 13 days studied (3 flights per day on average), assuming that 40 observations was, statistically, a reasonable minimum. This threshold of 40 available flight records also separates the top 100 airport pairs in terms of available flights in our traffic sample. Moreover, this threshold allowed a good trade-off between a high number of flights kept and a reduced number of airport pairs. Actually, Figure 12 shows that keeping only the top 100 airports pairs allowed the elimination of 1391 airport pairs, while keeping 55% of the traffic. 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Number of airport pairs Figure 12: Most important city pairs in terms of amount of traffic in the studied sample Distance between airports Even in the hypothetical case where aircraft could fly along the great circle line between their departure and arrival airport, the total distance flown would exceed the great circle distance because of constraints linked to airport departure and arrival procedures (hereafter referred as incompressible flight phase ). The most obvious constraint is the fact that aircraft take off and land facing the wind. Therefore, using current indicators, even a perfect route network could not lead to zero percent inefficiency. However, as flights get longer, the share of these incompressible flight phases should become decreasingly significant. Logically, in absence of any other sources of inefficiency, plotting distance inefficiency relative to direct distance between arrival and departure airports should reveal a clear decreasing trend. Even if Figure 13 does not reveal such an obvious correlation between the inefficiency and the direct distance between the airport pairs, the distribution of results does not refute our initial expectations. 33

34 Direct distance [KM] Figure 13: Efficiency indicator and distance between airports. Actually, for very short distances the dispersion is wide, but, as distance increases the correlation becomes clearer, and the occurrence of outlying points is less and less frequent. For flight ranges shorter than 200 kilometres, results are spread over such a wide range that these flights were not considered suitable to be included in further steps of the analysis. As a conclusion, all city pairs separated by less than 200 kilometres were excluded from the sample, which, when combined with the first selection criteria, leaves a sample of 90 observations Identification of high inefficiency areas Among the 90 airport pairs remaining after the selection stages mentioned in previous chapters, it seemed logical to select the ones where high route inefficiency was measured. Of course, these high inefficiencies must be explained in the next phases of the project, and some of them will probably be justified by some obvious factors, such as very short distances between the airports or geographical obstacles. However, it was considered that the Top 10 airport pairs in terms of distance inefficiency, shown in Table 13, should definitively be included in the final shortlist of candidates available for the stages of the project. It should be noticed that city pairs which are displayed in this report were selected from a reduced sample. The ranking of these pairs do not constitute a proper performance benchmarking, but simply a selection criteria corresponding to the study purpose. 34

35 Table 13: Top 10 airport pairs in terms of distance inefficiency. Distance Inefficiency [%] Average Direct distance [KM] Number of Flight Plans Cumulative Inefficiency [KM] Geneva - Zurich ,364 Nice - Geneva ,316 London Heathrow - Leeds ,350 Amsterdam - Luxembourg ,033 Manchester - London Heathrow ,451 Manchester - London Gatwick ,674 Brussels - Geneva ,570 Brussels - Lyon ,041 East Midlands - Amsterdam ,798 Düsseldorf - London Heathrow , Identification of cases with high potential cost In addition to airport pairs with high distance inefficiency, it was considered that further selection should be based on the notion of potential cost generated by these airport pairs. To reflect the potential cost generated by an airport pair, we considered the cumulative inefficiency. This notion is supposed to identify city pairs having different characteristics, but producing similar cost. Three components were considered to form the cumulative inefficiency: 1. The distance inefficiency indicator (as % of direct distance) 2. The direct distance between airports 3. The total number of flights connecting the airport pair 2 The cumulative inefficiency was obtained by multiplying these three factors. It was decided arbitrarily that any airport pair ranking in the Top 10 of cumulative inefficiency should be included in the final shortlist. These 10 airport pairs are shown in Table At this stage the total number of flights from CFMU flight plans has been used instead of the number of flights for which a valid CPR was available. It is justified by the fact that, statistically, flights for which we couldn t use a valid CPR might experience the same level of inefficiency as other flights. If the ratio between flights with usable CPR trajectory and total number of flight were constant across airport pairs, this distinction would have been useless. However, as it was not the case, we considered that the multiplication by the total number of flights would provide a more realistic potential cost. 35

36 Table 14: Top 10 airport pairs in terms of cumulative inefficiency. Average Distance Number of Cumulative Direct Inefficiency Flight Inefficiency distance [%] Plans [KM] [KM] London Heathrow Zurich ,190 London Heathrow Paris CDG ,842 London Heathrow Amsterdam ,765 London Heathrow Dublin ,849 London Heathrow Geneva ,750 London Heathrow Edinburgh ,832 London Heathrow Oslo Gardermoen ,646 Amsterdam Zurich ,223 Nice Paris CDG ,665 Brussels London Heathrow ,461 Considering the very high potential costs linked to these 10 airport pairs (on only 13 days their cumulated extra distances equal around 8 times the earth s circumference), they should be given the highest importance when investigating flight efficiency dimensions and causes. However, several arguments encouraged us to pursue the constitution of the rest of the shortlist using complementary criteria. These arguments are listed below: z The Top 10 airport pairs in terms of cumulative inefficiency give a partial vision of European diversity. Actually, it allows studying routes from London to the rest of Europe, and there is a risk that focusing only on these 10 pairs restricts our future investigations to Heathrow s runway capacity issues, leading to stack utilization. z The identification of some kind of best performers list is needed to be used as a benchmark against which other airport pairs could be compared. z Since the aim is at understanding the drivers of flight efficiency, observing a variety of different performance levels relative to different basic conditions would allow identifying a wider range of flight efficiency aspects. As a conclusion, a set of complementary selection criteria is proposed. In the following paragraphs, the 90 airport pairs will be classified as a function of their relative position in terms of 1) the flight efficiency, 2) direct distance, 3) the total number of flights. This will allow isolation of 8 cases presented in Figure 14, and selection of the airport pairs belonging to each category. A shortlist representative of route diversity can then be constructed Selection of groups of typical cases The 90 airport pairs selected in the previous steps were divided into three groups of equal size for each factor (efficiency, distance, and number of flights) using 33 rd and 66 th percentile values. These are shown in Table 15. Then, airport pairs belonging to any of the 8 categories shown in Figure 14 were identified and selected for composing the final short list. High 36

37 categories refer to values higher than the 66 th lower than the 33 rd percentile. percentile, and Low categories to values Table 15: Separators of airport pair groups. Distance inefficiency [%] Average direct distance [KM] Total number of flights (13 days) 33 rd percentile th percentile Distance inefficiency High Low Direct Distance Direct Distance High Low High Low Number of flights Number of flights Number of flights Number of flights High Low High Low High Low High Low Group 333 Group 331 Group 313 Group 311 Group 133 Group 131 Group 113 Group 111 Figure 14: Identification of groups of airport pairs. In the following presentation of each group, airport pairs belonging to each group are shown together in a chart allowing visualization of each one s relative position. To ease the reading and comparison across different groups, chart scales were normalized (for each series value, the average value was subtracted and the result was divided by the standard deviation). Thus, in the following charts, values of 0 mean that the value of the variable equals the average, values of 1 or -1 mean that the variables are respectively one standard deviation above or below the average. Each chart also displays as reference a fictive average airport pair. 37

38 Group 333 Inefficiency Traffic Distance Amsterdam - Zurich London Heathrow - Geneva London Heathrow - Zurich Average Figure 15: Group of high inefficiency / long distance / high traffic This group is composed of airport pairs cumulating high inefficiencies, long direct distance, and large number of flights. Logically, all the components for a potential high cost are met, and airport pairs of this group were already identified in the Top 10 cumulative inefficiency pairs. Group 331 Inefficiency Traffic Distance London City - Geneva London Luton - Geneva Birmingham - Copenhagen Average Figure 16: Group of high inefficiency / long distance / low traffic. 38

39 The group shown in Figure 16 is cumulating high inefficiency indicators, and long distances, which could potentially result in high costs to airlines. However, as the number of flights on these routes is relatively low, actual impacts might be moderate in absolute values (but still very annoying for the few airlines operating these routes). The three airports pairs appearing in this group were not identified with previous criteria, but the focus still remains around London and Geneva. Group 313 The group shown in Figure 17 contains six airports pairs cumulating high inefficiency and experiencing large traffic. Three of these airport pairs are also member of the Top 10 cumulative inefficiency group. The very short distance separating the airports of each pair is probably an important factor explaining the outstanding level of inefficiency. Moreover, the high traffic frequency on these routes may also create favourable conditions for using non-direct routes, as dense traffic may add complexity in traffic flow separation and increase the probability for congestion. On a more general basis, these airport pairs cumulate favourable factors for generating high external costs. Therefore, they might be considered in priority for further investigation of tradeoffs around ATM efficiency. In a context of environmental policy, they are probably good case studies where better use of multimodal options could be considered. Inefficiency Traffic Distance Geneva - Zurich Manchester - London Heathrow Manchester - London Gatwick London Heathrow - Paris CDG Brussels - London Heathrow London Heathrow - Amsterdam Average Figure 17: Group of high inefficiency / short distance / high traffic 39

40 Group 311 Inefficiency Traffic Distance Manchester - London Stansted Manchester - London City Brussels - London City London Heathrow - Rotterdam London Heathrow - Leeds Amsterdam - Luxembourg Average Figure 18: Group of high inefficiency / short distance / low traffic This group, partially already captured by the Top 10 inefficiency list, presents high percentages of extra distance flown that can probably be justified by the very short distances separating departure and arrival airports. As traffic is relatively low on these routes, costs might be limited in absolute values. Group 133 Inefficiency Traffic Distance Nice - Paris Orly London Heathrow - Copenhagen Average Figure 19: Group of low inefficiency / long distance / high traffic 40

41 This group, showing good results for flight efficiency but having long direct distances and many flights, includes two interesting airport pairs. Actually, although these pairs achieve a good efficiency, they might generate high costs because any single percent deviation from the shortest distance will apply to many kilometres and to many flights. Group 131 This group contains two airport pairs with relatively good distance efficiency, long direct distances and low traffic. Inefficiency Traffic Distance Amsterdam - Torp Liverpool - Geneva Average Figure 20: Group of low inefficiency / long distance / low traffic Group 113 Two airport pairs appear in this group. Even though the airport-to-airport distances are relatively short and the frequency of traffic is high, these routes perform efficiently. These two city pairs may be interesting for further investigations as examples of good performers. 41

42 Inefficiency Traffic Distance Copemhagen - Karup Copenhagen - Aalborg Average Figure 21: Group of low inefficiency / short distance / high traffic Group 111 Airport pairs in this group have logically the least cost impact, since they have low inefficiency, short distances and few flights. They are shown for completeness and also as examples of good flight efficiency performance although direct distance between airports is very short. 42

43 Inefficiency Traffic Distance Dublin - Kerry East Midlands - Glasgow Average Figure 22: Group of low inefficiency / short distance / low traffic 43

44 3.3 Results and conclusion on airport pairs The combination of selection criteria described in previous paragraphs provides a shortlist of 35 airport pairs. This shortlist contains only the cases where we considered the number of observations (radar trajectories) was enough to provide reliable results Distance Inefficiency [%] Worse Performance all airport pairs Group 333 Group 331 Group 313 Group 311 Group 133 Group 131 Group 113 Group 111 TOP 10 Cumulative Better Performance 0 Direct Distance [KM] Figure 23: Distance efficiency versus route length Figure 23 shows how the groups defined previously cover the full set of possible situations. Logically, each quadrant could be interpreted as follow: z North-East: relatively bad performance in relatively easy conditions worst performance. z South-East: relatively good performance in relatively easy conditions normal performance z South-West: relatively good performance in relatively difficult conditions best performance z North West: relatively bad performance in relatively difficult conditions normal performance For each of these situations, the potential cost should then be proportional to the number of flights between airport pairs. This report does not intend to go further into detail with the presentation of airport pairs. We propose this shortlist and the identified groups to be used as a starting point for the next step of the study that will focus on the drivers of flight efficiency and propose enhanced indicators. 44

45 Airport pairs (both ways) Table 16: Shortlist of 35 airport pairs. Number of Radar Trajectories Average Distance Efficiency [%] 45 Average Direct Distance [KM] Total Number of Flights (13 days) Cumulative Inefficiency [KM] Rank Cumulative Inefficiency Rank Distance Inefficiency London Heathrow Zurich , London Heathrow Paris CDG , London Heathrow Amsterdam , London Heathrow Dublin ,849 4 London Heathrow Geneva , London Heathrow Edinburgh ,832 6 London Heathrow Oslo Gardermoen ,646 7 Amsterdam Zurich , Nice Paris CDG ,665 9 Brussels London Heathrow , Manchester London Heathrow , Geneva Zurich , Düsseldorf London Heathrow , Brussels Lyon ,041 8 Southampton Amsterdam , Brussels Geneva ,570 7 East Midlands Edinburgh , London City Rotterdam , Manchester London Gatwick , Nice Geneva ,316 2 Aarhus Oslo Gardermoen , London City Isle of Man , London Heathrow Leeds , Copenhagen Aalborg , Amsterdam Luxembourg , East Midlands Amsterdam ,798 9 Liverpool Geneva , Brussels London City , Manchester London Stansted , London Heathrow Rotterdam , Manchester London City , Copenhagen Karup , Amsterdam Torp , East Midlands Glasgow , Dublin Kerry , Below 33th percentile Between 33th and 66th percentile Above 66th percentile Group Number

46 4. Economic evaluation 4.1 Context This part of the report attempts to investigate economic aspects linked to flight efficiency indicators presented in previous sections. Flight efficiency impacts airline costs at several levels and puts an extra burden on the environment. These two components of flight efficiency cost can be respectively referred to as internal costs (costs that are actually paid by the actors suffering from inefficiencies) and as external costs (costs that are not captured by markets, i.e. for which there is no formal supply and demand, case of aviation emissions). In ATM Flight Efficiency and its Impact on the Environment 2002 study [Ref. 1] an evaluation for both internal and external costs of flight efficiency was presented. It highlighted the issues and proposed an approach to, first, derive global flight efficiency ranges applicable to the whole European traffic, and second, select unit costs relevant to both internal and external costs. In this report, it is neither intended to reapply the methodology developed in the 2002 study for costing flight efficiency impacts on airlines, nor to develop further the discussion on assumptions and values to use for climate change impacts. Several reasons explain this decision: z The range of costs estimated previously was quite large (1,000 to 2,500 million Euros for airlines costs and 100 to 600 million Euros for environmental costs) and the extent of this interval resulted from uncertainties lying both in the flight efficiency quantification and in the unit cost valuation. z The latest results for flight inefficiency indicators computed from a 2003 traffic sample are more or less the same as those based on the 2002 samples, and they do not allow refinement of the range of flight efficiency. z Sources for environmental costs and knowledge about the impact of aviation emissions on climate changes have not significantly progressed in one year. Therefore, it was considered that reprocessing new flight efficiency indicators into the economic methodology used last year would not bring significant changes to the previous estimation of the cost range. Instead, it was chosen to refine the airline costs estimation using a study recently done by the University of Westminster [Ref. 2], which investigated costs of ground and airborne delays. The particularity of this study is to look into detail at all the categories of airline costs that are affected by airborne delays, depending on delay duration, and aircraft types. Precise figures were thus provided by anonymous airlines, which is an important step towards more realistic cost estimations. In future steps of the flight efficiency project, it is planned to study some enhanced indicators at the airport pair level. Working with a lower level of granularity will allow looking at specific issues, and costing efficiency at the local level will have to take into account aircraft specific costs. This was one more argument for directing the economic evaluation work toward a deeper investigation of airline costs. 46

47 The purpose of this section is to: z Investigate to what extent the costs computed by the University of Westminster [Ref. 2] are applicable to flight efficiency. z Propose models allowing to link aircraft types with flight efficiency costs. z Assess the impact of using models computing flight efficiency costs per aircraft on the cost range estimated in Flight efficiency and delay costs Airborne delay and flight inefficiency are two expressions that, in some cases, refer to the same situation. While airborne delay is a restrictive term, especially referring to extra duration in the air compared to a given time reference, flight inefficiency is a more generic term that can cover several situations, not restricted to a duration reference. Actually, indicators presented in Chapter 2.4 may, for some flights, capture holding stacks. In such cases, flight efficiency is directly comparable to airborne delays, and cost references used for delays could be applied to flight efficiency indicators. However, the flight efficiency indicators may also capture situations where flights depart and arrive according to their normal schedule, but having flown extra distances and duration compared to an optimum situation. In the latter case, airborne delay costs are not an adequate cost metric to use, as the efficiency was known in advance. These flight extensions generated by the network are known in advance and correspond to opportunity costs. Opportunity costs mean that more efficient trajectories would not only allow a reduction in fuel consumption, and any other direct operating costs, but would also allow a better utilisation of the airlines assets. Later on, we will refer to this situation using the terminology of strategic costs, as these costs are actually comparable to one of the options evaluated in the University of Westminster [Ref. 2] report, called strategic cost of one minute buffer included in the schedule and used in operation. Actually, a limit of the global flight efficiency indicators is that they do not allow a separation of tactical delay costs from strategic costs. Therefore, costing flight efficiency will have to be done using two scenarios: z First, it will be assumed that 100% of inefficiencies are airborne delays and the unit cost will correspond to tactical delay costs (with reactionary delays, also known as network effects). [See page 49] z Second, it will be assumed that there is no airborne delay, and that 100% of inefficiencies result from route design constraints that are known in advance by airspace users. In this case the unit cost to be used is a strategic cost of adding a buffer in the schedule and using it operationally. [See page 51]. Obviously, reality is more complex, and each route between two airports is a mixture of these two extreme scenarios, with in addition, some incompressible flight phases that are known to airlines but that could not be removed even in an optimum situation (except in a very hypothetical case of replacing all aircrafts by helicopters!). Figure 24 gives an overall presentation of indicators and the paths towards costing them. Our source for unit costs (both tactical and strategic costs) was proposing costs to airlines for short and long delays (15 and 65 minutes were respectively chosen as reference). In the case of flight efficiency, only the 15 minutes reference was used. The observed inefficiencies within the studied traffic sample confirm that 65 minutes airborne delay would only occur in 47

48 very particular situations. Actually, only 21 flights out of 16,328 (0.1%) experienced extra durations longer than 65 minutes. To account for the fact that flights experience different levels of inefficiencies and that they neither have the same cost structure nor operating practices, the low, base and high scenarios provided by the University of Westminster were kept. Scenario 1: 0% Scenario 2: 100% Global Inefficiency Cost Model Scenario 1: 100% Scenario 2: 0% Route Design Route Utilisation Incompressible flight phases Route inefficiencies (tradeoffs) Airspace user choice (tradeoffs) Airborne delays 0 cost assigned No savings possible Strategic costs Savings possible via tradeoffs optimisation Tactical costs Saving possible via tradeoffs optimisation Primary delays Reactionary delays Similar situation than buffer added to the schedule and used operationally Tactical delays En-Route Arrival Management En-Route Arrival Management Weighted average cost (20% en-route, 80% arrival) Weighted average cost (20% en-route, 80% arrival) Low cost Base cost High cost Low cost Base cost High cost Strategic flight efficiency costs Tactical flight efficiency costs y = 0.37x R 2 = y = x R 2 = y = 0.26x R 2 = y = 0.10x R 2 = y = 0.16x R 2 = Maximum Take Off Weight Low Base High 20 y = 0.07x R 2 = Maximum Take Off Weight Low Base High Figure 24: Flight Efficiency cost framework 48

49 4.3 Tactical costs airborne delays Scenario 1 refers to a hypothetical situation where 100% of inefficiencies captured by our indicators would correspond to airborne delays. Only some of the variable operating costs are accounted for in this scenario. Table 17 indicates the variable direct operating cost items that were included in the different tactical cost estimations. Given the level of complexity, the number of costs items, and their unit costs variation according to low, base, and high hypothesis, the reader should refer to the University of Westminster s report [Ref. 2] for more information, as presenting all details here would deviate from the initial purpose of this report. Table 17: Tactical cost items Direct operating costs - variable Low cost scenario Base cost scenario High cost scenario Fuel x x x Maintenance costs related to utilisation x x x Crew costs related to utilisation x Ground handling (aircraft) (3rd-party) pax handling Airport aeronautical charges En-route ATC Pax delay compensation and costs x Airborne delays may, as ground delays, have repercussions on subsequent aircraft rotations, and even on the punctuality of other branches of the network (especially in a hub and spoke operating mode). Therefore, it was considered more realistic to adopt the tactical airborne costs including network effects (reactionary delays). Both en-route and arrival management costs were available, and a weighted average of 20% en-route and 80% arrival was selected. 49

50 Table 18: Tactical airborne delay costs: en-route and holding (with network effect) based on 15 minutes delay Aircraft Average Low cost Base cost High cost MTOW [Tonnes] Boeing Boeing Boeing Boeing Boeing Boeing ER Boeing Airbus A Airbus A Airbus A ATR ATR Source: University of Westminster report [Ref. 2] Based on these costs per aircraft, three regression models ( low - base - high ) were estimated to link maximum take-off weight to costs. The models shown in Figure 25 will allow a computation of flight efficiency costs adapted to the fleet type in operation within each flight range. Thus, the proposed costs will vary according to ranges, which was not the case in the previous year s study. Tactical flight efficiency costs Cost [Euro per minute] y = 0.25x R 2 = 0.96 y = 0.10x R 2 = y = 0.07x R 2 = Maximum Take-Off Weight [Tonnes] Low Base High Figure 25: Tactical cost models 50

51 4.4 Strategic costs network design Scenario 2 refers to a hypothetic situation where 100% of inefficiencies captured by our indicators would correspond to route network constraints (i.e. observed deviations from great circle distance would correspond to the path of existing routes). As introduced in the beginning of this section, such a situation is different from airborne delays because airspace users know the network in advance and can organise their operations according to the (known) imperfections. However, operating in a non-optimal network 3 would generate some opportunity costs. Reducing these costs would allow airlines to more intensively use aircraft and crews, and to achieve the same operating revenues. As a consequence, these strategic costs are logically higher than tactical costs. Referring once again to the University of Westminster study, the unit cost to apply for route network inefficiency is comparable to the cost of a schedule buffer that would just match the tactical requirement. This cost reference was named strategic airborne buffer used en-route or at arrival. It means that schedules are longer than in an optimum situation, and, that actual flights are longer than they need to be. Table 19 mentions the items included for the computation of strategic costs according to low, base, and high hypothesis, but does not list their percentage allocation. The reader should refer to [Ref. 2] for more information. Presenting all details here would make us deviate from the initial purpose of this report. As for tactical costs, both en-route and arrival management costs were available. A weighted average of 20% en-route and 80% arrival was selected, by consideration that route design around airports may encounter more constraints than en-route (with, for example, preferred noise routes). However, it is important to note that the assumption of repartition between enroute and arrival costs is not crucial, as its impact on final strategic cost is marginal (values for en-route and arrival are very close). Table 19: Strategic cost items Direct operating costs - variable Low cost scenario Base cost scenario High cost scenario Fuel X X X Maintenance costs related to utilisation X X X Crew costs related to utilisation Ground handling (aircraft) (3 rd -party) pax handling Airport aeronautical charges En-route ATC X 3 The current route network organisation is not challenged, and what is meant here is not that ANSPs or EUROCONTROL were unable to optimize the network. We are conscious that the airspace design exercise is a very complex one, subject to many tradeoffs. It is obvious that the strategic costs estimated in this report are an upper bound, corresponding to a fictive scenario, and that only a fraction of this cost could be saved when all operating constraints and other costs sources are considered. 51

52 Pax delay compensation and costs Direct operating costs - fixed Aircraft depreciation, rentals & leases Maintenance cost unrelated to utilisation Fixed crew cost unrelated to utilisation Flight equipment insurance Indirect operating costs Passenger accident / liability insurance Passenger service staff (terminal) Ground equipment, property & staff X X X X X X X X X X Table 20: Cost of strategic airborne buffer minute (20% used en-route, 80% used holding). Aircraft Average Low cost Base cost High cost MTOW [¼PLQ@ [¼PLQ@ [¼PLQ@ [Tonnes] Boeing Boeing Boeing Boeing Boeing Boeing ER Boeing Airbus A Airbus A Airbus A ATR ATR Source: University of Westminster report [Ref. 2] Based on these costs per aircraft, it is proposed to adopt three regression models ( low - base - high ) to link maximum take-off weight to costs per minute. The models shown in Figure 26 will allow a computation of flight efficiency costs adapted to the fleet type in operation within each flight range. Thus the proposed costs will vary according to ranges, which was not the case in the previous year s study. 52

53 Strategic flight efficiency costs Cost [Euro per minute] y = 0.37x R 2 = 0.89 y = 0.26x R 2 = 0.93 y = 0.16x R 2 = Maximum Take-Off Weight [Tonnes] Low Base High Figure 26: Strategic cost models 4.5 Global impacts Intermediate calculations With the 6 cost models presented in the previous sections, we are now able to compute a flight efficiency cost for any aircraft (provided its MTOW is known) covering both a scenario of airborne delays and a scenario of network design. In order to re-aggregate these individual aircraft costs and to generate a global figure for annual flight efficiency costs, it was necessary to compute, from our traffic sample, the average MTOW per flight range. As the flight efficiency metrics are computed per flight range, this allowed a more precise application of unit cost relevant to the aircraft type flying in each range. Table 21 gives average observed MTOW values together with associated costs per flight range, and Figure 27 shows the model used to estimate MTOW values for the ranges where no information was available. 53

54 Aircraft utilisation according to flight range 90 Maximum Take Off Weight [Tonnes] y = 3.05x 0.45 R 2 = Distance [KM] Figure 27: Aircraft types and flight range. Distance [KM] Average MTOW Table 21: Flight efficiency costs per flight range.* Tactical cost Low Tactical cost Base Tactical cost High Strategic cost Low Strategic cost Base Strategic cost High

55 * Cells in light green: MTOW generated by the model presented in Figure 27. As a conclusion, it can be observed that the cost of one minute flight inefficiency varies greatly in function of the considered scenarios and in function of flight ranges because bigger aircraft are usually operated on longer routes. However, the aircraft size effect is somewhat smaller than the effect of the scenario considered. Actually, the difference of unit costs between low tactical cost and high strategic cost is typically between a factor of 8 to 10, whereas the cost differs by a factor of 2 between very short range (200 km) and medium ranges (3,500 km) flights Final cost estimation As explained earlier in the introduction, the aim of the economic section was not to reapply the previous year s methodology to 2003 flight efficiency indicators, but rather to test if applying better models for airline costs would allow a better precision in the total cost estimation. At least, an indication of the most realistic bound (lowest or highest) would be considered useful with regard to its wide scale. Therefore, this paragraph presents total cost estimations of global flight efficiency results. It applies the six cost models presented previously to the duration efficiency indicator that was computed from the 2002 traffic sample. Doing so will allow the isolation of the cost model s impact in the global cost estimation. For the sake of completeness, the main steps and assumptions used for deriving global flight efficiency metrics from the 2002 partial sample are restated hereafter. As the flight efficiency of long-haul flights were not assessed in the study, different assumptions had to be considered, according to the flight length, distributed onto three intervals. z 200-1,100 km range: use of flight efficiency results. z 1,100-1,700 km range: use of the linear trend observed for the 200-1,100 km range. z 1,700-3,600 km range: two hypothesis were considered: o In a low bound hypothesis, 0% inefficiency was assumed. This is justified by the fact that the longer the range, the higher the probability that route extension is a deliberate choice of the airline (either to avoid a congested 55

56 area, or to benefit from favourable winds that decrease the flight cost). This limitation also exists for shorter ranges, but we believed that on long ranges it could be predominant. o In a high bound hypothesis, the linear trend computed from the validated results was used. This was simply derived from statistical observation, and led to a progressive decline toward 0% inefficiency as the flight range increases. Table 22: Estimation of the average inefficiency for flight duration (% difference between radar and direct trajectories) Flight range [KM] All flight plans duration seconds (sum of 10 days) Duration inefficiency (radar - direct) / direct Extra duration % Low bound Extra duration % High bound ,276,890 11% ,200,384 14% ,470,453 14% ,570,566 15% ,196,258 15% ,030,856 14% ,116,820 12% ,731,271 11% ,765,973 13% ,070,928 9% ,678,257 10% ,679,412 11% ,837,852 10% ,145,305 8% ,314,962 9% ,240,781 Na ,815,965 Na ,739,954 Na ,506,811 Na ,557,744 Na ,829,059 Na ,318,082 Na ,444,929 Na ,643,819 Na ,474,802 Na ,410,645 Na ,024,587 Na ,277,159 Na ,775,515 Na ,022,954 Na ,188,220 Na ,002,841 Na ,703,082 Na

57 To compute total annual costs, the six cost models were applied using a different cost per range, multiplied by the relevant inefficiency coefficient shown in Table 22. As Table 22 itself contains two scenarios, this resulted in twelve costs. Table 23: Global estimation for tactical and strategic cost of flight efficiency [Million ¼@ Low inefficiency on long flight ranges High inefficiency on long flight ranges Tactical cost Low Tactical cost Base Tactical cost High Strategic cost Low Strategic cost Base Strategic cost High ,639 1,103 2,277 3, ,219 1,490 3,038 4,825 Average ,929 1,297 2,658 4,227 Among, these twelve results, it is probable that extreme values should not be considered realistic, as the combination of high inefficiencies on long flight ranges with high costs becomes a very hypothetical case. The same is true for combinations of low hypothesis. If it is accepted that such cases do exist for particular flights, it is obviously wrong to generalise these cases to the whole European air traffic. Therefore, the base case for unit costs should be preferred when attempting to give a realistic global value to annual cost of flight efficiencies. These new results do not contradict the range estimated in the flight efficiency 2002 study, but tends to argue in favour of lower costs. The lowest bound estimated in the 2002 study was 981 million, and is now, with more precise cost inputs, 716 millions (with confidence interval between 607 and 826). The highest bound estimations, corresponding to strategic costs, are in the same order of magnitude as in Conclusions The section presented the results of applying more recent and more precise cost figures than in the previous flight efficiency study report. This was possible thanks to the work of the University of Westminster, who provided us with their working report on the cost of airborne delays. The most important step in this new cost evaluation was the distinction between route network inefficiencies, leading to strategic costs, and airborne delay inefficiencies, leading to tactical costs. As flight efficiency indicators do not allow, so far, the separation of these two effects, the cost estimation was done along two scenarios. The first one assumed 2002 flight efficiency distribution to be 100% airborne delays, and generated an average annual cost of 716 millions euros, which is slightly lower that previous year s low bound estimation. The second scenario assumed 100% inefficiencies were related to route network design, which resulted in a strategic cost in the order of 2,658 million euros, in line with previous estimations. 57

58 Acknowledgements The authors wish to thank the following persons and teams for their assistance for this study: EUROCONTROL EATM Environment Domain EUROCONTROL Performance Review Unit EUROCONTROL Experimental Centre (EEC) Network, Capacity and Demand (NCD) research area The CFMU Engineering department 58

59 References [Ref 1] Chesneau, S., Hustache, J-C., Fuller, I. (2003). ATM flight efficiency and its impact on the environment, 2002 study. EUROCONTROL Experimental Centre note EEC/ENV/2003/001. [Ref 2] Cook, A., Tanner, G., Anderson, S. (2004). Evaluating the true cost to airlines of one minute of airborne or ground delay, Final report, edition 4, 17 February Transport Studies Group, University of Westminster, London. 59

60 Annex 1. Methodology Both CPR trajectories and Flight plans were run separately through AMOC. AMOC needs the appropriate environmental data of the corresponding time period of the data. AMOC created direct route trajectories based on the flight plans (see Chapter ). Traffic and flight files were created as output files for both CPR and direct trajectories. CPR tracks Flight plans AMOC CPR Traffic file CPR Flight file Direct Traffic file Direct Flight file MS Access CPR Traffic file CPR Flight file Direct Traffic file Direct Flight file AEM ArcGIS CPR Duration Fuelburn Direct Duration Fuelburn CPR Distance Direct Distance Figure 28: Flowchart of data processing with different tools Next, the CPR traffic and flight files were imported to Microsoft Access (CPR and Direct data were processed in separate databases following a specific methodology and order). Most of 60

61 the processing done by using Access was automated with a pre-created form combining several tasks into a simple click of a button. Firstly the flight table was processed. This included the calculation of the time of day and flight speed, and removing multiple records. After this several filters were applied to the traffic table using queries that searched the correct data from the processed flight table using the following criteria: z Flight s start and end points <FL30 z Flight s maximum flight level >FL75 z Flight s duration >15 min z Origin and destination airports in Europe (based on airport code) A table was created of all the CPR flights meeting the above criteria. Before the processed CPR flights could be exported, the Direct trajectories needed to be processed. Direct traffic and flight files were imported into Access and the flight file was processed in the same way as the CPR flight file. The same filters were also applied to the traffic table including a requirement that the Direct flight had a corresponding CPR flight in the processed CPR database. After this the resulting Direct traffic and flight tables could be exported. In the CPR database the requirement of having a corresponding Direct flight was applied by linking to the table of final Direct flights, and the resulting flights (traffic and flight tables) were exported. The two CPR files and two Direct files were next run by AEM3 v1.5. During the AEM run the option of completing all the flights with a LTO cycle was used. AEM calculated the needed flight durations and fuel burns for both CPR and Direct flights (with a lot of other unused data), which were included in the flight summaries. In parallel with the AEM runs the output flight files from Access were imported into ArcGIS using a specially developed functionality for the specific data format. The same functionality calculated the distances for the flights. The Great Circle distance between the first and last CPR data record was used as the Direct distance. The total distance of the CPR trajectories was calculated based on all the data points. The resulting output files with distances were then exported. Visual validation of all the final trajectories was done in ArcGIS. It was possible to see simultaneously the CPR trajectory and the corresponding Direct trajectory on a map, which enabled the verification of matching flight records and their correctness. At the same time, the visual inspection provided insight into the inefficiencies experienced by the flights. 61

62 Annex 2. Tools used in the study AMOC AMOC is an integrated ATFM simulator developed by the NCD (Network Capacity and Demand management) business area of Eurocontrol Experimental Centre. As the ATFM Simulator for the Future ATM Profile (FAP) project, the AMOC tool allows the identification and evaluation of future capacity problems. AMOC can also translate these problems into delays, and can test various potential solutions. A new version of AMOC has been developed in order to integrate ATFM models and data from the CFMU, such as Flight Plans, CPR and Environment files. AMOC facilitates the simulation process, the traffic preparation, and the statistical calculations. AMOC has been used in this study to transform and filter the CPR files received from the CFMU to obtain clean tracks in our own format. For this purpose, AMOC uses ALL_FT files, which contain the Flight Plan information, and also the ENV files containing information about airports and waypoints. It also uses those files to generate all the trajectories with flight plans that are studied in the economical part. AMOC has another useful option: it creates direct 4D profiles, associated to each track of the ALL_FT file (Flight Plan file). It takes the departure and destination airports, and the aircraft type, with its associated performance table (CFMU or BADA, depending on the option chosen), and generates a direct profile in a predefined format, recalculating the arrival time. AEM The Advanced Emission Model (AEM3 v1.5) is a stand-alone modelling system that uses flight-profile information to calculate fuel burn and emissions. The input information is the type of aircraft, the departure time, departure and arrival airports and characteristics of plots representing the flight trajectory. AEM3 v1.5 calculates the fuel consumption for each phase of flight. Figure 29: AEM3 flight profile for emission calculation. 62