Source-linked AI summary
Modeling the Multi-layer Nature of the European Air Transport Network: Resilience and Passengers Re-scheduling under random failures
Alessio Cardillo, Massimiliano Zanin, Jesús Gómez-Gardeñes, Miguel Romance, Alejandro J. García del Amo, Stefano Boccaletti
TL;DR
The paper asks how the multilayer structure of the European Air Transport Network affects resilience when flights fail and passengers must be rescheduled. It models airline-specific flight layers and compares passenger rescheduling in multiplex and aggregate representations. The comparison shows that multiplexity strongly affects robustness because passengers cannot generally combine connections from different airlines in one alternative path.
Problem
The study addresses the missing multilayer analysis of ATN resilience, since passengers cannot freely combine connections operated by different airlines.
Method
The authors model the ATN as 15 airline-specific layers and simulate passenger rescheduling after random flight failures, using shortest paths, capacity tolerance, and an aggregate-network comparison.
Results
Multiplexity strongly affects robustness: the aggregate network has considerably fewer no-fly passengers than the multiplex ATN, especially at low failure probabilities.
Takeaways & Limitations
The multiplex representation captures airline-layer constraints that materially change ATN resilience and passenger rescheduling outcomes compared with single-layer projections.
Takeaways & Limitations
The model disregards factors such as flight duration and cost when defining airport distance and analyzes additional load tolerance only for 0 ≤ ftol ≤ 0.3.
Abstract
from arXiv · showhide
We study the dynamics of the European Air Transport Network by using a multiplex network formalism. We will consider the set of flights of each airline as an interdependent network and we analyze the resilience of the system against random flight failures in the passenger's rescheduling problem. A comparison between the single-plex approach and the corresponding multiplex one is presented illustrating that the multiplexity strongly affects the robustness of the European Air Network.
1 Introduction
The European Air Transport Network is a socially and economically important system whose local disruptions can produce broader operational consequences. Because passengers cannot freely combine airlines within itineraries, its resilience requires a multilayer analysis beyond the conventional single-layer representation.
- Air transport evolved from a sparsely connected system into a redundant network carrying 2.7 billion passengers in 2011.
- Multiplex networks represent systems in which nodes participate in multiple interacting layers with layer-dependent neighborhoods and interaction types.
- Multilayer structures can reduce network resilience by increasing vulnerability to cascading failures across interconnected physical and logical networks.
- The ATN is economically and socially significant, carrying 2.4 billion passengers and 43 million tonnes of cargo in 2010 while supporting 32 million jobs.
- Local disruptions such as thunderstorms can cause delays and missed connections that impair overall system performance, with future traffic growth expected to intensify these pressures.
- Previous ATN resilience studies treated all airport connections as equivalent, whereas passengers cannot bypass the costs of switching airlines when constructing itineraries.
- The study therefore examines how multilayer architecture influences ATN robustness and passenger-flow recovery under random flight failures.
2 The European ATN as a multilayer network
The model represents Europe’s air transport system as a 15-layer multiplex of major airlines and compares its structure with single-layer views. Airline layers differ substantially, and aggregating them changes degree distributions and network heterogeneity.
- The model contains 15 airline layers, 308 airport nodes, and commercial IFR flights recorded on 1 June 2011, covering 20% of European airspace operations.
- Each layer represents one major airline, with its airports as nodes and its operated flights as links in an undirected network.
- Major-airline layers are scale-free with headquarters acting as hubs, whereas low-cost layers are more uniform because of point-to-point organization.
- Figure 2 contrasts a traditional major airline network with a low-cost network and marks each network’s hubs using large blue circles.
- Multiplex global degree counts a node’s connections across layers, producing degrees larger than in the corresponding one-layer network.
- The degree enhancement is nonuniform because layer heterogeneity disperses node-level effects, yielding a degree distribution unlike the classic aggregate model.
- Cumulative degree distributions also differ when individual airline layers are considered separately.
3 The model
The model represents the European Air Transport Network as a multiplex of airline-specific layers, assigns passenger routes by shortest within-layer paths, and simulates random flight failures followed by constrained re-scheduling.
- Multiplex structure: Each airline is modeled as an independent network layer, while airports are represented across the multiplex layers.The layer-specific links are flights operated by that airline.
- Passenger initialization: Passenger origins and destinations are randomly selected with probabilities proportional to airports’ global degrees.Trips beginning and ending at the same node are excluded.
- Route assignment: Passengers are assigned to the layer with the minimum hopping distance between their origin and destination, with ties selected uniformly.Flight duration, cost, and other factors are disregarded.
- Failure and re-scheduling: Randomly removing links with probability p forces affected passengers to seek replacement paths with lengths dij(n) = dij + n.The procedure first tries n = 0 and iterates through larger allowances up to n = 2.
- Failure and re-scheduling: Re-scheduling prioritizes the original layer, then other layers, while respecting active paths and link capacities before classifying passengers as fly, re-scheduling, or no-fly.The process repeats while passengers remain unassigned.
4 Results
The experiments vary link-failure probability and load tolerance, comparing multiplex and aggregate representations under increasingly permissive re-scheduling. The aggregate network is more robust because it permits hybrid routes across airlines, whereas multiplex constraints keep passengers within single layers.
- Multiplex results: The experiments measure passenger outcomes across link-failure probability p and load tolerance ftol for re-scheduling allowances n = 0, 1, and 2.Outcomes include no-fly, other-layer, and same-layer passenger fractions.
- Multiplex results: For n = 0, almost all successfully re-scheduled passengers change airline, while the no-fly fraction remains extremely large.At p ∼10^-2 and about 10% tolerance, no-fly passengers exceed 50% of those initially affected.
- Multiplex results: Allowing n = 1 or n = 2 substantially lowers the no-fly fraction and enables some passengers to use alternative routes within their original layer.The n = 1 and n = 2 results are quite similar, indicating little benefit from allowing n > 2.
- Aggregate comparison: The aggregate network merges airline layers into one network, with airport-pair connection multiplicity reflecting the number of operating airlines.The same link-removal and re-scheduling procedure is applied to this single-layer representation.
- Aggregate comparison: In the aggregate network, the no-fly fraction decreases considerably relative to the multiplex network, especially for low p and n > 0.For some parameter regions, almost all affected passengers can be re-scheduled; n = 1 and n = 2 remain nearly identical.
- Interpretation: The aggregate network shows improved robustness because passengers can combine connections from different airlines into hybrid alternative paths.The multiplex architecture prevents such cross-layer path mixing and reduces route-optimization options.
5 Conclusions
The paper models passenger re-scheduling in the European Air Transport Network as a multiplex system and compares it with an equivalent single-layer representation. It concludes that layered structure changes network characteristics and resilience estimates, while motivating extensions for policy use.
- The model represents each airline’s flights as a separate layer in a multiplex European Air Transport Network.Passenger re-scheduling follows available routes and free seats, prioritizing the affected passenger’s former airline before considering another airline.
- Random flight failures are modeled through link-failure probability p and link-load tolerance f_tol.The procedure measures passengers successfully re-scheduled and passengers for whom re-scheduling fails.
- Multiplex and equivalent single-layer representations differ qualitatively and quantitatively in their topological characteristics.
- Single-layer projections overestimate the resilience of the Air Transport Network.
- The framework is positioned for future policy applications if airport distances, airline alliances, and re-routing costs are incorporated.These elements were excluded for simplicity in the presented model.