Optimization of Aircraft Taxiing Strategies to Reduce the Impacts of Landing and Take-Off Cycle at Airports
Abstract
:1. Introduction
2. Methods
2.1. Emissions Assessment
2.2. Parameters for the Estimation of the Emissions
2.3. Taxiing Strategies
- Single-engine taxiing (SET);
- Dispatch towing (APU);
- Taxing with onboard systems (MES);
- Reducing taxiing time (RED).
2.3.1. Single-Engine Taxiing
2.3.2. Dispatch Towing
2.3.3. Onboard Systems
2.3.4. Reducing Taxiing Time
2.4. Costs
2.5. Case Study
3. Results
3.1. Pollutant Emissions in the Basic Scenario
3.2. Pollutant Emissions According to the Optimization Scenarios
3.3. Cost Analysis
4. Discussion
5. Concluding Remarks
Author Contributions
Funding
Institutional Review Board Statement
Informed Consent Statement
Data Availability Statement
Conflicts of Interest
References
- IATA. The IATA Technology Roadmap Report. 2009. Available online: https://www.escholar.manchester.ac.uk/api/datastream?publicationPid=uk-ac-man-scw:106699&datastreamId=FULL-TEXT.PDF (accessed on 28 June 2022).
- European Commission. Reducing Emissions from Aviation. 2017. Available online: https://ec.europa.eu/clima/eu-action/transport-emissions/reducing-emissions-aviation_en (accessed on 5 July 2022).
- International Civil Aviation Organization (ICAO). Trends in Emissions that Affect Climate Change. Available online: https://www.icao.int/environmental-protection/Pages/ClimateChange_Trends.aspx (accessed on 13 July 2022).
- Eyring, V.; Kohler, H.W.; Lauer, A.; Lemper, B. Emissions from international shipping: 2. Impact of future technologies on scenarios until 2050. J. Geophys. Res. Atmos. 2005, 110, D17306. [Google Scholar] [CrossRef]
- Vennam, L.P.; Vizuete, W.; Talgo, K.; Omary, M.; Binkowski, F.S.; Xing, J.; Mathur, R.; Arunachalam, S. Modeled Full-Flight Aircraft Emissions Impacts on Air Quality and Their Sensitivity to Grid Resolution. J. Geophys. Res. Atmos. 2017, 122, 13472–13494. [Google Scholar] [CrossRef] [PubMed]
- Riley, K.; Cook, R.; Carr, E.; Manning, B. A systematic review of the impact of commercial aircraft activity on air quality near airports. City Environ. Interact. 2021, 11, 100066. [Google Scholar] [CrossRef]
- Hudda, N.; Durant, L.W.; Fruin, S.A.; Durant, J.L. Impacts of Aviation Emissions on Near-Airport Residential Air Quality. Environ. Sci. Technol. 2020, 54, 8580–8588. [Google Scholar] [CrossRef] [PubMed]
- Bendtsen, K.M.; Bengtsen, E.; Saber, A.T.; Vogel, U. A review of health effects associated with exposure to jet engine emissions in and around airports. Environ. Health 2021, 20, 10. [Google Scholar] [CrossRef]
- Klöwer, M.; Allen, M.R.; Lee, D.S.; Proud, S.R.; Gallagher, L.; Skowron, A. Quantifying aviation’s contribution to global warming. Environ. Res. Lett. 2021, 16, 104027. [Google Scholar] [CrossRef]
- Wells, C.A.; Williams, P.D.; Nichols, N.K.; Kalise, D.; Poll, I. Reducing transatlantic flight emissions by fuel-optimised routing. Environ. Res. Lett. 2021, 16, 025002. [Google Scholar] [CrossRef]
- International Civil Aviation Organization (ICAO). Document 9889. Airport Air Quality Manual, Second Edition. Available online: https://www.icao.int/publications/Documents/9889_cons_en.pdf (accessed on 2 July 2022).
- Celikel, A.; Duhanian, N.; Peeters, S. Preliminary Local Air-Quality Study; Report EEC/BA/ENV/Note No. 001/2002; Environmental Studies Business Area, EUROCONTROL Experimental Centre: Brussels, Belgium, 2002. [Google Scholar]
- European Union Aviation Safety Agency (EASA). European Aviation Environmental Report, Chapter 7. Available online: https://www.easa.europa.eu/eaer/topics/aviation-environmental-impacts (accessed on 1 July 2022).
- European Union Aviation Safety Agency (EASA). Agency Research Agenda 2020–2022. Available online: https://www.easa.europa.eu/downloads/117222/en (accessed on 1 July 2022).
- Postorino, M.N. Environmental effects of airport nodes: A methodological approach. Int. J. Sustain. Dev. Plan. 2010, 5, 192–204. [Google Scholar] [CrossRef] [Green Version]
- Masiol, M.; Harrison, R.M. Aircraft engine exhaust emissions and other airport-related contributions to ambient air pollution: A review. Atmos. Environ. 2014, 95, 409–455. [Google Scholar] [CrossRef] [Green Version]
- Brennan, E. EUROCONTROL 2050 air traffic forecast and Objective Skygreen provide guidance for a purposeful long-term aviation sustainability. In EUROCONTROL Aviation Sustainability Briefing; EUROCONTROL: Brussels, Belgium, 2022; Volume 6, pp. 2–3. [Google Scholar]
- EUROCONTROL Objective Skygreen 2022–2030. The Economics of Aviation Decarbonisation towards the 2030 Green Deal Milestone. 2022. Available online: https://www.eurocontrol.int/sites/default/files/2022-05/eurocontrol-objective-skygreen-2022-2030-report-20220523.pdf (accessed on 1 July 2022).
- European Union Aviation Safety Agency (EASA). Sustainable Aviation Fuel ‘Monitoring System. Available online: https://www.easa.europa.eu/downloads/115346/en (accessed on 1 July 2022).
- Fan, W.; Sun, Y.; Zhu, T.; Wen, Y. Emissions of HC, CO, NOX, CO2 and SO2 from civil aviation in China in 2010. Atmos. Environ. 2012, 56, 52–57. [Google Scholar] [CrossRef]
- Stettler, M.E.J.; Eastham, S.; Barret, S.R.H. Air Quality and public health impacts of UK airports. Part I: Emissions. Atmos. Environ. 2011, 45, 5415–5424. [Google Scholar] [CrossRef]
- Yim, S.H.L.; Stettler, M.E.J.; Barret, S.R.H. Air Quality and public health impacts of UK airports. Part II: Impacts and policy assessment. Atmos. Environ. 2013, 67, 5415–5424. [Google Scholar] [CrossRef]
- Lawal, A.S.; Russell, A.G.; Kaiser, J. Assessment of Airport-Related Emissions and Their Impact on Air Quality in Atlanta, GA, Using CMAQ and TROPOMI. Environ. Sci. Technol. 2022, 56, 98–108. [Google Scholar] [CrossRef]
- Tait, K.N.; Khan, M.A.H.; Bullock, S.; Lowenberg, M.H.; Shallcross, D.E. Aircraft Emissions, Their Plume-Scale Effects, and the Spatio-Temporal Sensitivity of the Atmospheric Response: A Review. Aerospace 2022, 9, 355. [Google Scholar] [CrossRef]
- Balakrishnan, H.; Deonandan, I.; Simaiakis, I. Opportunities for Reducing Surface Emissions through Airport Surface Movement Optimization; Technical Report ICAT 2008/7; MIT International Center for Air Transportation: Cambridge, MA, USA, 2008; Available online: http://hdl.handle.net/1721.1/66491 (accessed on 1 July 2022).
- Miller, B.; Minoque, K.; Clarke, J. Constraints in Aviation Infrastructure and Surface Aircraft Emissions; Massachusetts Institute of Technology: Cambridge, MA, USA, 2000; pp. 1–15. [Google Scholar]
- Di Mascio, P.; Carrara, R.; Frasacco, L.; Luciano, E.; Ponziani, A.; Moretti, L. Influence of tower air traffic controller workload and airport layout on airport capacity. J. Airpt. Manag. 2021, 15, 408–423. [Google Scholar]
- Guo, R.; Zhang, Y.; Wang, Q. Comparison of emerging ground propulsion systems for electrified aircraft taxi operations. Transp. Res. C 2014, 44, 98–109. [Google Scholar] [CrossRef]
- Li, J.; Yang, H.; Liu, X.; Yu, N.; Tian, Y.; Zhou, X.; Zhang, P.; Wang, K. Aircraft Emission Inventory and Characteristics of the Airport Cluster in the Guangdong–Hong Kong–Macao Greater Bay Area, China. Atmosphere 2020, 11, 323. [Google Scholar] [CrossRef] [Green Version]
- International Civil Aviation Organization (ICAO). Manual of the ICAO Standard Atmosphere—3rd Edition 1993 (Doc. 7488). DOC-07488-003-01-Q-P. Available online: http://www.aviationchief.com/uploads/9/2/0/9/92098238/icao_doc_7488_-_manual_of_icao_standard_atmosphere_-_3rd_edition_-_1994.pdf (accessed on 5 July 2022).
- Ivković, I.; Čokorilo, O.; Kaplanović, S. The estimation of GHG emission costs in road and air transport sector: Case study of Serbia. Transport 2018, 33, 260–267. [Google Scholar] [CrossRef] [Green Version]
- Du Bois, D.; Paynter, G.C. Fuel flow method 2 for estimating aircraft emissions. SAE Trans. 2006, 115, 1–14. [Google Scholar]
- Mazaheri, M.; Johnson, G.R.; Morawska, L. Particle and Gaseous Emissions from Commercial Aircraft at Each Stage of the Landing and Takeoff Cycle. Environ. Sci. Technol. 2009, 43, 441–446. [Google Scholar] [CrossRef]
- Chati, S.Y.; Balakrishnan, H. Analysis of aircraft fuel burn and emissions in the landing and take-off cycle using operational data. In Proceedings of the International Conference on Research in Air Transportation, Istanbul, Turkey, 26–30 May 2014. [Google Scholar]
- International Civil Aviation Organization (ICAO). Aircraft Engine Emission Data Bank (07/2021). Available online: https://www.easa.europa.eu/domains/environment/icao-aircraft-engine-emissions-databank (accessed on 2 June 2022).
- Wilkerson, J.T.; Jacobson, M.Z.; Malwitz, A.; Balasubramanian, S.; Wayson, R.; Fleming, G.; Naiman, A.D.; Lele, S. Analysis of emission data from global commercial aviation: 2004 and 2006. Atmos. Chem. Phys. 2010, 10, 6391–6408. [Google Scholar] [CrossRef] [Green Version]
- Wasiuk, D.K.; Lowenberg, M.H.; Shallcross, D.E. An aircraft performance model implementation for the estimation of global and regional commercial aviation fuel burn and emissions. Transp. Res. D 2015, 35, 142–159. [Google Scholar] [CrossRef] [Green Version]
- Koudis, G.S.; Hu, S.J.; Majumdar, A.; Ochieng, W.Y.; Settler, M.E.J. The impact of single engine taxiing on aircraft fuel consumption and pollutant emissions. Aeronaut. J. 2018, 122, 1967–1984. [Google Scholar] [CrossRef] [Green Version]
- Lukic, M.; Hebala, A.; Giangrande, P.; Klumpner, C.; Nuzzo, S.; Chen, G.; Gerada, C.; Eastwick, C.; Galea, M. State of the art of electric taxiing systems. In Proceedings of the IEEE International Conference on Electrical Systems for Aircraft, Railway, Ship Propulsion and Road Vehicles International Transportation Electrification Conference (ESARS-ITEC), Nottingham, UK, 7–9 November 2018; pp. 1–6. [Google Scholar]
- Environmental Protection Agency (EPA). Technical Support for Development of Airport Ground Support Equipment Emission Reductions; Environmental Protection Agency: Washington, DC, USA, 1999.
- California Air Resources Board (CARB). Air Pollution Mitigation Measures for Airports and Associated Activity; National Technical Information Service: Springfield, VA, USA, 1994.
- Airport Cooperative Research Program (ACRP). Handbook for Evaluating Emissions and Costs of APUs and Alternative Systems; Report 64; Transportation Research Board: Washington, DC, USA, 2012.
- Di Mascio, P.; Moretti, L. Hourly Capacity of a Two Crossing Runway Airport. Infrastructures 2020, 5, 111. [Google Scholar] [CrossRef]
- Di Mascio, P.; Cervelli, D.; Comoda Correra, A.; Frasacco, L.; Luciano, E.; Moretti, L.; Nichele, S. A Critical Comparison of Airport Capacity Studies. J. Airpt. Manag. 2020, 21, 307–321. [Google Scholar]
- Di Mascio, P.; Cervelli, D.; Comoda Correra, A.; Frasacco, L.; Luciano, E.; Moretti, L. Effects of Departure manager and arrival manager systems on airport capacity. J. Airpt. Manag. 2021, 15, 204–218. [Google Scholar]
- Di Mascio, P.; Carrara, R.; Frasacco, L.; Luciano, E.; Ponziani, A.; Moretti, L. How the Tower Air Traffic Controller Workload Influences the Capacity in a Complex Three-Runway Airport. Int. J. Environ. Res. Public Health 2021, 18, 2807. [Google Scholar] [CrossRef]
- Čokorilo, O. Human Factor Modelling for Fast-Time Simulations in Aviation. Aircr. Eng. Aerosp. Technol. 2013, 85, 389–405. [Google Scholar] [CrossRef]
- Simaiakis, I.; Balakrishnan, H. Queuing models of airport departure processes for emissions reduction. In Proceedings of the AIAA Guidance, Navigation, and Control Conference, Chicago, IL, USA, 10–13 August 2009. [Google Scholar]
- IATA. Jet Fuel Price Monitor. 2021. Available online: https://www.iata.org/en/publications/economics/fuel-monitor/ (accessed on 29 June 2022).
- Ministero dello Sviluppo Economico. Prezzi Medi Settimanali dei Carburanti e Combustibili. 2021. Available online: https://dgsaie.mise.gov.it/prezzi-settimanali-carburanti (accessed on 14 July 2022).
- Corazza, M.V.; di Mascio, P.; Esposito, G. Airports as sensitive areas to mitigate air pollution: Evidences from a case study in Rome. Environments, 2022; submitted. [Google Scholar]
Phase | TIM (Minutes) |
---|---|
Taxi-in | 6.3 |
Taxi-out | 12.6 |
Aircraft Class | Fuel Type—Energy | Power | Load Factor | Fuel Consumption | HC Emissions | CO Emissions | NOX Emissions | PM Emissions | CO2 Emissions |
---|---|---|---|---|---|---|---|---|---|
bhp | % | gal/bhp-h | g/bhp-h | g/bhp-h | g/bhp-h | g/bhp-h | g/gal | ||
Narrow Body | Diesel | 175 | 80 | 0.061 | 1.2 | 4 | 11 | 0.55 | 9797.2 |
Petrol | 130 | 80 | 0.089 | 4 | 240 | 4 | 0.03 | 8932.8 | |
Electric | --- | 80 | --- | --- | --- | --- | --- | --- | |
Wide Body | Diesel | 500 | 80 | 0.053 | 1.2 | 4 | 11 | 0.53 | 9797.2 |
Petrol | 500 | 80 | 0.089 | 4 | 240 | 4 | 0.03 | 8932.8 | |
Electric | --- | 80 | --- | --- | --- | --- | --- | --- |
Aircraft Class | FFj,APU | HC Emissions | CO Emissions | NOX Emissions | PM Emissions | CO2 Emissions |
---|---|---|---|---|---|---|
kg/s | g/kg of Fuel | g/kg of Fuel | g/kg of Fuel | g/kg of Fuel | g/kg of Fuel | |
Narrow Body | 0.021 | 6.53 | 31.75 | 5.45 | --- | 3155 |
Wide Body | 0.035 | 0.87 | 10.26 | 7.55 | --- | 3155 |
Aircraft Class | FFj,MES | HC Emissions | CO Emissions | NOX Emissions | PM Emissions | CO2 Emissions |
---|---|---|---|---|---|---|
kg/s | g/kg of Fuel | g/kg of Fuel | g/kg of Fuel | g/kg of Fuel | G/Kg of Fuel | |
Narrow Body | 0.038 | 0.29 | 4.94 | 7.64 | --- | 3155 |
Wide Body | 0.064 | 0.13 | 0.98 | 11.53 | --- | 3155 |
Aircraft Type | Engine Type | Number of Engines | LTO Cycles | Aircraft Type | Engine Type | Number of Engines | LTO Cycles |
---|---|---|---|---|---|---|---|
A30B | PW4158 | 2 | 9 | CRJ7 | GE CF34-8C5 | 2 | 39 |
A310 | GE CF6-80C2A2 | 2 | 3 | CRJ9 | GE CF34-8C5 | 2 | 2676 |
A318 | CFMI CFM56-5B9/3 | 2 | 1069 | D328 | PW306B | 2 | 12 |
A319 | CFMI CFM56-5B5/P | 2 | 1659 | DC8 | CFMI CFM56-2C1 | 2 | 1 |
A320 | CFMI CFM56-5B4/3 | 2 | 4310 | DH8C | PWC PW123 | 2 | 162 |
A321 | IAE V2533-A5 | 2 | 346 | DH8D | PWC PW150A | 2 | 466 |
A330-200 | RR Trent 772B-60 | 2 | 74 | E135 | AN AE3007A1 | 2 | 654 |
A330-300 | RR Trent 772B-60 | 2 | 1 | E145 | AN AE3007A1 | 2 | 37 |
AN26 | Ivchenko AI 24VT | 2 | 135 | E170 | GE CF34-8E5A1 | 2 | 111 |
AT45 | PWC PW127 | 2 | 315 | E190 | GE CF34-10E5A1 | 2 | 1102 |
AT72 | PWC PW120 | 2 | 973 | F100 | RR Tay Mk650-15 | 2 | 260 |
B190 | P&W PT6A-65B | 2 | 1 | F50 | PWC PW125B | 2 | 3 |
B350 | SHP P&W PT6A-60A | 2 | 1 | F70 | RR Tay Mk650-15 | 2 | 1144 |
B461 | ALF502R-5 | 4 | 9 | GL5T | RR BR710A2-20 | 2 | 12 |
B462 | ALF 502R-5 | 4 | 322 | GLEX | R-R BR710-48-C2 | 2 | 16 |
B463 | ALF 502R-5 | 4 | 770 | J328 | P&W 306B | 2 | 11 |
B737-200 | P&W JT8D-9A | 2 | 129 | L188 | Allison T56-A14 | 4 | 1 |
B737-300 | P&W JT8D-9A | 2 | 1166 | MD80 | PW JT8D-217C | 2 | 2 |
B737-400 | P&W JT8D-9A | 2 | 1822 | MD82 | PW JT8D-217C | 2 | 1617 |
B737-500 | P&W JT8D-9A | 2 | 538 | MD83 | PW JT8D-217C | 2 | 74 |
B737-600 | CFMI CFM56-7B20 | 2 | 72 | MD87 | PW JT8D-217C | 2 | 14 |
B737-700 | CFMI CFM56-7B22 | 2 | 93 | MD90 | IAE V2525-D5 | 2 | 2 |
B737-800 | CFMI CFM56-7B26 | 2 | 6859 | RJ1H | LY LF507-1F | 4 | 150 |
B747-200 | PW JT9D-7Q | 4 | 4 | RJ70 | LY LF 507-1F | 4 | 2 |
B747-400 | GE CF6-80C2B1F | 4 | 3 | RJ85 | LY LF507-1F | 4 | 514 |
B757-200 | RR RB211-535 E4 | 2 | 459 | SB20 | AN AE2100A | 2 | 144 |
B767-200 | GE CF6-80A2 | 2 | 267 | SF34 | GE CT7-5A2 | 2 | 4 |
B767-300 | PW4060 | 2 | 140 | SH36 | P&W PT6A-67R | 2 | 1 |
CRJ1 | GE CF34-3A1 | 2 | 2 | SW3 | Garrett TPE331-10U-503G | 2 | 3 |
CRJ2 | GE CF34-3A1 | 2 | 1019 | SW4 | Garrett TPE331-12 | 2 | 9 |
Phase | TIM | Thrust | HC | CO | NOX | SOX | PM | CO2 | Fuel |
---|---|---|---|---|---|---|---|---|---|
Approach | 4.0 | 30 | 0.8 | 9.7 | 39.8 | 4.3 | 0.5 | 13,416 | 4252 |
Taxi-in | 6.3 | 7 | 5.7 | 58.9 | 11.2 | 2.6 | 0.2 | 8205 | 2601 |
Taxi-out | 12.6 | 7 | 11.4 | 118.7 | 22.6 | 5.2 | 0.5 | 16,537 | 5242 |
Take off | 0.7 | 100 | 0.3 | 1.6 | 63.0 | 2.6 | 0.4 | 8228 | 2608 |
Climb out | 2.2 | 85 | 0.8 | 4.1 | 130.5 | 6.7 | 1.0 | 21,281 | 6745 |
LTO cycle | 25.8 | --- | 19.0 | 193.1 | 267.0 | 21.4 | 2.6 | 67,668 | 21,448 |
Taxiing Mode | Phase | HC | CO | NOX | SOX | PM | CO2 | Fuel Consumption |
---|---|---|---|---|---|---|---|---|
Ton | Ton | Ton | Ton | Ton | Ton | Ton | ||
SET | Taxi-in | 5.1 | 53.0 | 10.1 | 2.3 | 0.2 | 7379 | 2339 |
−10.1% | −10.1% | −10.1% | −10.1% | −10.1% | −10.1% | −10.1% | ||
Taxi-out | 8.0 | 82.9 | 15.8 | 3.7 | 0.3 | 11544 | 3659 | |
−30.2% | −30.2% | −30.2% | −30.2% | −30.2% | −30.2% | −30.2% | ||
LTO-cycle | 15.0 | 151.3 | 259.0 | 19.6 | 2.4 | 61,849 | 19,603 | |
−21.2% | −21.6% | −3.0% | −8.6% | −6.5% | −8.6% | −8.6% | ||
RED1 | Taxi-in | 4.8 | 49.5 | 9.4 | 2.2 | 0.2 | 6895 | 2185 |
−16.0% | −16.0% | −16.0% | −16.0% | −16.0% | −16.0% | −16.0% | ||
Taxi-out | 10.5 | 109.3 | 20.8 | 4.8 | 0.4 | 15,227 | 4826 | |
−7.9% | −7.9% | −7.9% | −7.9% | −7.9% | −7.9% | −7.9% | ||
LTO-cycle | 17.2 | 174.3 | 263.4 | 20.6 | 2.5 | 65,047 | 20,617 | |
−9.5% | −9.8% | −1.3% | −3.9% | −2.9% | −3.9% | −3.9% | ||
RED2 | Taxi-in | 3.9 | 40.1 | 7.6 | 1.8 | 0.2 | 5585 | 1770 |
−31.9% | −31.9% | −31.9% | −31.9% | −31.9% | −31.9% | −31.9% | ||
Taxi-out | 9.6 | 99.9 | 19.0 | 4.4 | 0.4 | 13,916 | 4411 | |
−15.8% | −15.8% | −15.8% | −15.8% | −15.8% | −15.8% | −15.8% | ||
LTO-cycle | 15.4 | 155.4 | 259.8 | 19.8 | 2.4 | 62,427 | 19,787 | |
−19.1% | −19.5% | −2.7% | −7.8% | −5.9% | −7.8% | −7.8% | ||
RED3 | Taxi-in | 3.0 | 30.7 | 5.8 | 1.4 | 0.1 | 4274 | 1355 |
−47.9% | −47.9% | −47.9% | −47.9% | −47.9% | −47.9% | −47.9% | ||
Taxi-out | 8.7 | 90.5 | 17.2 | 4.0 | 0.4 | 12,606 | 3996 | |
−23.8% | −23.8% | −23.8% | −23.8% | −23.8% | −23.8% | −23.8% | ||
LTO-cycle | 13.6 | 136.6 | 256.2 | 19.0 | 2.3 | 59,806 | 18,956 | |
−28.6% | −29.2% | −4.0% | −11.6% | −8.8% | −11.6% | −11.6% | ||
APU diesel | Taxi-in | 8.4 | 64.0 | 23.4 | 2.7 | 0.8 | 8481 | 2478 |
48.5% | 8.6% | 109.0% | 3.4% | 240.1% | 3.4% | −4.7% | ||
Taxi-out | 13.6 | 87.2 | 39.2 | 2.9 | 1.4 | 11,043 | 3077 | |
19.0% | −26.6% | 73.4% | −33.2% | 199.8% | −33.2% | −41.3% | ||
LTO-cycle | 23.9 | 166.6 | 295.8 | 19.2 | 4.1 | 62,449 | 19,161 | |
25.9% | −13.7% | 10.8% | −7.7% | 59.3% | −7.7% | −7.7% | ||
APU petrol | Taxi-in | 10.5 | 261.5 | 14.5 | 2.7 | 0.2 | 8488 | 2478 |
84.3% | 343.8% | 29.6% | 3.4% | −9.4% | 3.4% | −4.7% | ||
Taxi-out | 17.7 | 485.2 | 21.2 | 3.5 | 0.2 | 1105 | 3077 | |
54.7% | 308.6% | −5.9% | −33.1% | −49.6% | −33.1% | −41.3% | ||
LTO-cycle | 30.0 | 762.0 | 269.0 | 19.8 | 2.3 | 62,470 | 19,161 | |
58.1% | 294.7% | 0.7% | −7.7% | −10.1% | −7.7% | −7.7% | ||
APU electric | Taxi-in | 7.1 | 59.5 | 11.2 | 2.5 | 0.2 | 7819 | 2478 |
25.0% | 1.1% | −0.4% | −4.7% | −20.2% | −4.7% | −4.7% | ||
Taxi-out | 10.9 | 78.2 | 14.5 | 3.1 | 0.2 | 9709 | 3077 | |
−4.6% | −34.1% | −36.0% | −41.3% | −60.4% | −41.3% | −41.3% | ||
LTO-cycle | 19.9 | 153.2 | 258.8 | 18.6 | 2.2 | 60,453 | 19,161 | |
4.7% | −20.7% | −3.1% | −10.7% | −13.1% | −10.7% | −10.7% | ||
MES | Taxi-in | 4.7 | 49.7 | 13.3 | 2.6 | 0.2 | 8304 | 2632 |
−17.4% | −15.6% | 18.3% | 1.2% | −20.2% | 1.2% | 1.2% | ||
Taxi-out | 4.8 | 52.4 | 17.6 | 3.2 | 0.2 | 10,083 | 3196 | |
−57.6% | −55.8% | −22.0% | −39.0% | −60.4% | −39.0% | −39.0% | ||
LTO-cycle | 11.4 | 117.6 | 264.1 | 19.4 | 2.2 | 61,313 | 19,434 | |
−39.8% | −39.1% | −1.1% | −9.4% | −13.1% | −9.4% | −9.4% |
Mode | Kerosene (€) | Diesel/Petrol (€) | Total (€) | Difference with Respect to FET (€) |
---|---|---|---|---|
FET | 11,825,396 | 0 | 11,825,396 | - |
SET | 10,808,411 | 0 | 10,808,411 | −1,016,985 |
RED1 | 11,367,408 | 0 | 11,367,408 | −457,988 |
RED2 | 10,909,420 | 0 | 10,909,420 | −915,976 |
RED3 | 10,451,432 | 0 | 10,451,432 | −1,373,963 |
APU diesel | 10,564,583 | 1,236,851 | 11,801,434 | −23,961 |
APU petrol | 10,564,583 | 1,460,274 | 12,024,857 | 199,461 |
APU electric | 10,564,583 | 0 | 10,564,583 | −1,260,813 |
MES | 10,714,738 | 0 | 10,714,738 | −1,110,658 |
Phase | HC | CO | NOX | SOX | PM | CO2 | Fuel Consumption |
---|---|---|---|---|---|---|---|
Ton | Ton | Ton | Ton | Ton | Ton | Ton | |
Approach | 0.8 | 9.7 | 39.8 | 4.3 | 0.5 | 13,416 | 4252 |
0% | 0% | 0% | 0% | 0% | 0% | 0% | |
Taxi-in (SET) | 5.1 | 53.0 | 10.1 | 2.3 | 0.2 | 7379 | 2339 |
−10.1% | −10.1% | −10.1% | −10.1% | −10.1% | −10.1% | −10.1% | |
Taxi-out (SET) | 8.0 | 82.9 | 15.8 | 3.7 | 0.3 | 11,544 | 3659 |
−30.2% | −30.2% | −30.2% | −30.2% | −30.2% | −30.2% | −30.2% | |
Take off | 0.3 | 1.6 | 63.0 | 2.6 | 0.4 | 8228 | 2608 |
0% | 0% | 0% | 0% | 0% | 0% | 0% | |
Climb out | 0.8 | 4.1 | 130.5 | 6.7 | 1.0 | 21,281 | 6745 |
0% | 0% | 0% | 0% | 0% | 0% | 0% | |
LTO cycle | 15.0 | 151.3 | 259.0 | 19.6 | 2.4 | 61,849 | 19,603 |
−21.2% | −21.6% | −3.0% | −8.6% | −6.5% | −8.6% | −8.6% |
Phase | HC | CO | NOX | SOX | PM | CO2 | Fuel Consumption |
---|---|---|---|---|---|---|---|
Ton | Ton | Ton | Ton | Ton | Ton | Ton | |
Approach | 0.8 | 9.7 | 39.8 | 4.3 | 0.5 | 13,416 | 4252 |
0% | 0% | 0% | 0% | 0% | 0% | 0% | |
Taxi-in (RED2) | 3.9 | 40.1 | 7.6 | 1.8 | 0.2 | 5585 | 1770 |
−31.9% | −31.9% | −31.9% | −31.9% | −31.9% | −31.9% | −31.9% | |
Taxi-out (MES) | 4.8 | 52.4 | 17.6 | 3.2 | 0.2 | 10,083 | 3196 |
−57.6% | −55.8% | −22.0% | −39.0% | −60.4% | −39.0% | −39.0% | |
Take off | 0.3 | 1.6 | 63.0 | 2.6 | 0.4 | 8228 | 2608 |
0% | 0% | 0% | 0% | 0% | 0% | 0% | |
Climb out | 0.8 | 4.1 | 130.5 | 6.7 | 1.0 | 21,281 | 6745 |
0% | 0% | 0% | 0% | 0% | 0% | 0% | |
LTO cycle | 10.6 | 108.0 | 258.4 | 18.6 | 2.2 | 58,593 | 18,572 |
−44.2% | −44.1% | −3.2% | −13.4% | −14.2% | 13.4% | −13.4% |
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations. |
© 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
Share and Cite
Di Mascio, P.; Corazza, M.V.; Rosa, N.R.; Moretti, L. Optimization of Aircraft Taxiing Strategies to Reduce the Impacts of Landing and Take-Off Cycle at Airports. Sustainability 2022, 14, 9692. https://doi.org/10.3390/su14159692
Di Mascio P, Corazza MV, Rosa NR, Moretti L. Optimization of Aircraft Taxiing Strategies to Reduce the Impacts of Landing and Take-Off Cycle at Airports. Sustainability. 2022; 14(15):9692. https://doi.org/10.3390/su14159692
Chicago/Turabian StyleDi Mascio, Paola, Maria Vittoria Corazza, Nicolò Rocco Rosa, and Laura Moretti. 2022. "Optimization of Aircraft Taxiing Strategies to Reduce the Impacts of Landing and Take-Off Cycle at Airports" Sustainability 14, no. 15: 9692. https://doi.org/10.3390/su14159692