Towards modelling cost and risks of infrequent events in the cargo screening process

Galina Sherman, David Menachof, Peer‐Olaf Siebers, Uwe Aickelin

TL;DR

This paper explores how to better evaluate port security policies by using Discrete Event Simulation (DES) instead of traditional, static Cost-Benefit Analysis (CBA). By focusing on "infrequent events"—like illegal immigration attempts at the Port of Calais—the researchers demonstrate that simulation provides a more realistic spread of potential outcomes than mean-based decision trees, helping authorities better plan for risks that are rare but not catastrophic.

What Problem Does This Paper Address?

Port authorities face a constant tug-of-war: they must maintain tight security against threats like smuggling and illegal immigration without causing massive traffic bottlenecks that hurt the economy.

Traditionally, security policies are evaluated using Cost-Benefit Analysis (CBA). However, CBA struggles with two extremes. For common events, it’s overkill; for "disastrous" rare events (like a major terrorist attack), the costs are nearly impossible to quantify accurately. This paper targets the "neglected middle": infrequent events. These are events that happen often enough to have measurable data but are rare enough that a simple "average" outcome doesn't tell the whole story. The authors specifically look at the ferry port of Calais to see if simulation can provide a better decision-making tool than standard decision trees.

Methods

The researchers developed two parallel models of the cargo screening process at the Port of Calais:

  1. A Traditional Decision Tree: A standard CBA tool that uses mean outcomes to calculate expected costs and benefits.
  2. A Discrete Event Simulation (DES): Built using AnyLogic software, this model introduces stochasticity (randomness). It accounts for variations in lorry arrival times and the probabilistic nature of detection sensors.

The study models the flow of lorries through both French and British controls, including various detection technologies like Passive Millimetre Wave (PMMW) sensors, heartbeat detectors (HB), $CO_2$ probes, canine teams, and visual inspections. By running multiple replications of the simulation, the authors could observe a distribution of results rather than a single fixed number.

Datasets

The study utilizes real-world operational data from the UK Border Agency (2008):

  • Throughput: Over 900,000 lorries passed through Calais screening between April 2007 and April 2008.
  • Incident Rate: Approximately 0.4% of these lorries were found to have "additional human freight" (clandestines) on board.
  • Economic Impact: The estimated cost of one clandestine entering the UK is cited as £20,000 per year.
  • Technology Benchmarks: While manufacturer benchmarks for sensors exist, the authors note these are often based on experimental rather than "real-world" environments, highlighting a gap in available data.

Key Insights and Ideas

  • The "Sweet Spot" for Simulation: The authors hypothesize that simulation is most useful when the probability of an event falls within a specific range. If an event is too common, a simple decision tree is more efficient. If it is "too rare," the number of simulation replications required to get a statistically significant result becomes computationally unfeasible.
  • Beyond the Mean: Traditional CBA provides a "theoretic worst-case" or a mean outcome. Simulation allows for a "spread" of outcomes, giving port authorities a "truer picture" of the range of potential risks and costs they might face in a given year.
  • Validation through Comparison: By showing that the average of 10 simulation replications closely matched the results of the static decision tree, the authors validated that their simulation was grounded in reality before moving on to more complex "what-if" scenarios.

Key Results

  • Model Validation: The simulation results (averaged over 10 runs) were found to be "very similar" to the decision tree results when testing the probability of checks between 0% and 100%. This confirmed the simulation's accuracy.
  • Flexibility: Unlike the static decision tree, the simulation model successfully handled "randomness," such as fluctuating arrival rates, which is critical for understanding how queues form and how security policies impact port fluidity.
  • Policy Evaluation: The model allows for the evaluation of different screening policies to estimate the annual cost to the British taxpayer, providing a quantitative basis for where to invest security budget.

Limitations and Future Work

As this paper represents an early stage of research, several limitations and next steps are noted:

  • Defining the "Too Rare" Threshold: The authors have yet to establish the exact mathematical bounds where simulation loses its edge over other methods due to the high number of replications required for very rare events.
  • Scenario Expansion: Future work will involve testing more complex scenarios, such as exceeding queue capacities and analyzing the difference between peak and off-peak operational times.
  • Generalizability: The current model is specific to the Calais-Dover link. The researchers intend to apply this methodology to other sites, starting with the Port of Dover, to see if the findings hold across different geographic and operational contexts.

Summary updated

Abstract

We introduce a simulation model of the port of Calais with a focus on the operation of immigration controls. Our aim is to compare the cost and benefits of different screening policies. Methodologically, we are trying to understand the limits of discrete event simulation of rare events. When will they become “too rare” for simulation to give meaningful results?
BibTeX
@inproceedings{sherman2013towards,
  author = {Galina Sherman and David Menachof and Peer‐Olaf Siebers and Uwe Aickelin},
  title = {{Towards modelling cost and risks of infrequent events in the cargo screening process}},
  booktitle = {Repository@Nottingham (University of Nottingham)},
  year = {2013},
}

Loading related content...

Figure