https://doi.org/10.5061/dryad.kprr4xhb9

Description of the data and file structure

The field data detail the sequence of events as recognized by official sources, alongside populations in Itaewon during the disaster. Simulations within this dataset employs a hydrodynamic model solved through a combination of the mixed-type finite difference method and the fast sweeping method, methodologies that are further elaborated in references [1-2].

Field data

The field data includes:

• Urban Dynamics: LTE data (1-31 October 2022) showing urban and suburban movements.
• Visitor Monitoring: Data from SDOT001 devices tracking visitors at specific stations (24 October - 6 November 2022).
• Population Data: De facto population figures in Itaewon on 29 October 2022, detailed by localities.
• Police Call History: Timeline and details of 11 critical police calls related to the event.

More specifical details of the data file:

• Basic Data (City/Suburb): Basic data_city_suburb (In English).xlsx contains LTE mobile device data spanning from October 1 to October 31, 2022.
  • (Sheet) S-DoT_WALK_2022 10 31-11 06; S-DoT_WALK_2022 10 24-10 30:
    This sheet utilizes the advanced SDOT001 model devices, this sheet captures the number of visitors at designated monitoring stations from October 24 to November 6, 2022. The primary components of the dataset include:
    • City: Seoul, where the data was collected.
    • Model: The monitoring device model, identified as SDOT001.
    • Serial Number: Unique identifier for each monitoring device.
    • Server Type: Denoted as “00Original,” indicating the data’s original or primary source.
    • Site Name: Tied to the serial number, uniquely identifying the data collection location.
    • Number of Visitors: The count of pedestrians detected during the specified dates.
    • Date: Data collection date, formatted as YYYYMMDDHHMM.
    • Registration Date: The date the data was registered, marking the data entry timestamp.
  • (Sheet) Metro_Inner_Itaewon: Focused on October 29, 2022, this sheet integrates de facto population figures, including intercity (labelled as Metro) and intracity (labelled as Inner) travelled population in Itaewon area, identified by unique Suburb and Local IDs. Key components include:
    • Date: Date of population data recording.
    • Time: Specific collection time or aggregated data over a 24-hour period.
    • Suburb ID and Local ID: Unique identifiers ensuring geographical specificity.
    • De Facto Population: Actual population count at the time of data collection.
  • (Sheet) INNER_PEOPLE_20221029: Similar in structure to “Metro_Inner_Itaewon,” this sheet also concentrates on intracity (labelled as Inner) travelled population figures in Itaewon area on October 29, 2022.
  • (Sheet) METRO_PEOPLE_20221029: Collected between 1st to 29th October, 2022, this sheet focuses on the de facto population figures across of intercity (labelled as Metro) travelled population figures in Itaewon area, distinguished by their unique Suburb IDs. Key components of the “METRO_PEOPLE_20221029” dataset include:
    • Date and Time: For recording the population data.
    • Suburb ID: For geographical specificity.
    • Rank in De Facto Population: Indicating the population size ranking.
    • Residential Code: For more detailed geographical analysis.
    • De Facto Population: Actual number of people present at the time of data collection.
  • (Sheet) LOCAL_PEOPLE_DONG_202210: Collected between 1st to 29th October, 2022, this sheet provides de facto population figures registered in each suburb, distinguished by unique Suburb IDs.
  • (Sheet) Daily local de facto population: The dataset titled “Daily Local De Facto Population” provides a detailed snapshot of the demographic dynamics within Yongsan-gu, Seoul, South Korea, from April 5, 2018, to November 9, 2022. Key components of the “Daily Local De Facto Population” dataset include:
    • Date: For temporal context.
    • City Code and City Name: For geographical and administrative context.
    • De Facto Population: Total population present, including residents and visitors.
    • Segments of the population: Korean nationals, international residents, and international travelers.
    • Daily Max and Min Population, Day and Night Time Population, Daily Max Active Population, Daily External Inflow Population, and Active Population details.
• Time Series Data (Three Regions): Time series ppl_three regions (In English).xlsx
  contains hourly populations for three significant areas in Itaewon from October 1 to October 31, 2022.
  • (Sheet) LOCAL_PEOPLE_20221029: Collected on 29th October, 2022, this sheet provides de facto population figures registered in each local suburb, distinguished by unique Local IDs.
    • Date and Time: For recording the population data.
    • Suburb ID: For geographical specificity for a suburb
    • Local IDL For geographical specificity for a local suburb
    • De Facto Population: Actual number of people present at the time of data collection.
    • Segments of the population: age and gender.
• History of Police Call: History of Police Call (In English).docx details the timeline of police calls throughout the Seoul Halloween crowd crush, detailing a total of 11 critical calls.

Simulation codes and results

• Codes
  • MATLAB main script: MAT_TVDRK2.m
  • MATLAB analysis of results: analysis_*.m
  • MATLAB functions: F90_FDM_God.mexw64, functions_*.m,
  • Fortran codes: F90_FDM_God.f90
• Simulation Results
  • Scenario A: Simulation data-senario(a): The original scenario.
  • Scenario B: Simulation data-senario(b): Crowd management strategy assigned. This strategy involved redirecting the pedestrian flow moving towards the south street to the right exit on the north street.
  • Scenario C: Simulation data-senario(c): Increasing pedestrian flow in the geometry based on scenario (b).

Key Sources

Basic Data, Time Series Data and History of Police Call were derived from the following source:
+ Seoul government website (https://english.seoul.go.kr/)
+ Seoul open data center (https://data.seoul.go.kr/)
+ Bill information in Korean national assembly (https://likms.assembly.go.kr/)

Sharing/Access information

The codes and results for the numerical algorithms and simulations is also available on GitHub:

• https://github.com/lianghy14/Seoul_Halloween_simu.git

Software

The simulation is conducted using a hybrid programming approach with MATLAB and FORTRAN. The following software is required to run the simulation:

Matlab R2022a
Visual Studio Community 2022
Intel oneAPI Base Toolkit 2023
Intel oneAPI HPC Toolkit 2023

Run the MAT_TVDRK2.m in MATLAB to start the simulation. Simulation results will be recorded in Results.

Methods

The Seoul Metropolitan Government’s Information and Planning Bureau collects comprehensive data on Seoul’s total and de facto populations. LTE mobile-phone signal data, characterized by extensive datasets with low variance and an even data-generation pattern, were utilized to estimate the de facto population.

On October 29, 2022, at 12:00, a tourist police force consisting of 10 members was deployed in the Itaewon neighborhood to prevent crime. A total of 93 calls to the police were made within the Itaewon police jurisdiction from 18:34 to 22:11, 11 of which were related to overcrowding. These 11 calls were received at 18:34, 20:09, 20:33, 20:53, 21:00, 21:02, 21:07, 21:10, 21:51, 22:00, and 22:11. Shortly after the second overcrowding-related call, a candlelight protest march near the area ended.

The study introduces a mixed-type hydrodynamic model for multidirectional pedestrian flow. Drawing on the empirical data from the Seoul Halloween crowd-crush disaster, the model established a series of partial differential equations (PDEs) with appropriate boundary conditions. This model-based simulation introduces several numerical algorithms to solve nonlinear problems numerically.

Relevent work

[1] Liang, H., Yang, L., Du, J., Shu, C., & Wong, S. C. (2024). Modelling crowd pressure and turbulence through a mixed-type continuum approach. Transportmetrica B: Transport Dynamics, 12(1), 2328774. https://doi.org/10.1080/21680566.2024.2328774

[2] Liang, H., Du, J., & Wong, S. C. (2021). A Continuum model for pedestrian flow with explicit consideration of crowd force and panic effects. Transportation Research Part B, 149, 100-117. https://doi.org/10.1016/j.trb.2021.05.006