ECA Resilience#
GitHub · Documentation · Development Data Partnership
This project leverages large-scale GPS mobility data to study human activity and urban space usage in response to atypical events. Using the Urban Space Usage Index, a normalized measure of relative human presence derived from anonymized mobile device location pings, the project quantifies deviations from typical mobility patterns and examines how cities respond to a range of shocks, from sudden natural disasters to planned public events and slow-onset climate stressors.
Case studies span two countries and four distinct event types: the 2023 Turkey-Syria earthquakes, Republic Day celebrations in Istanbul, heatwaves in Metro Manila, and monsoon-driven flooding in Manila amplified by Typhoon Doksuri.
Project Overview#
Data Source#
The analysis is based on the Veraset Movement dataset, provided as part of the Mobility Data collection from the Development Data Partnership. The dataset consists of anonymized, high-frequency GPS pings collected through a network of mobile applications and SDKs. Each record includes geographic coordinates, a UTC timestamp, and an anonymized device identifier.
Mobility observations are spatially aggregated using the Uber H3 hierarchical spatial index at resolutions 7 and 8 (average cell areas of ~5 km² and ~0.74 km², respectively).
Methodology#
The analytical framework follows three steps:
1. Define a measure. The Urban Space Usage Index (I) is defined as the daily share of total active users visiting each H3 hexagon, normalizing for day-to-day fluctuations in overall data volume:
I(h, d) = U(h, d) / U(d)
where U(h, d) is the number of unique users in hexagon h on day d, and U(d) is the total number of active users on that day.
2. Quantify deviations. Deviations from typical conditions are measured through Z-scores computed relative to a stable baseline period:
Z(h, d) = (I(h, d) − μ(h)) / σ(h)
3. Interpret deviations. Z-scores are analyzed temporally and spatially, and stratified by land-use category and functional layer (POI-based), enabling characterization of where and how urban activity changes in response to events.
Full methodological details are available in the Methodological Framework and Spatial Characterization of Urban Units notebooks.
Geographies#
Country |
Area of Interest |
Resolution |
|---|---|---|
Philippines |
Metro Manila |
H3 resolution 8 (~0.74 km²) |
Turkey |
Istanbul |
H3 resolution 8 (~0.74 km²) |
Turkey |
11 earthquake-affected provinces |
H3 resolution 8 (~0.74 km²) |
Case Studies#
A key design feature of this project is that the four case studies deliberately span four distinct event typologies, each presenting different challenges for mobility analysis and policy response:
Event |
Location |
Typology |
Period |
|---|---|---|---|
Istanbul, Turkey |
Planned public event |
Oct 2023 |
|
Southern Turkey (11 provinces) |
Sudden, unpredictable natural disaster |
Feb 2023 |
|
Metro Manila, Philippines |
Foreseeable natural disaster (typhoon-driven) |
Jul 2023 |
|
Metro Manila, Philippines |
Slow-onset climate shock |
Apr 2023 |
This typology-driven structure allows the framework to be tested across events with fundamentally different warning horizons, impact profiles, and behavioral responses:
Planned events (Republic Day) produce sharp, predictable spikes in activity that amplify existing spatial patterns city-wide.
Sudden disasters (earthquake) generate delayed but extreme anomalies driven by emergency response, displacement, and humanitarian operations, with no anticipatory behavioral signal.
Foreseeable disasters (typhoon-driven flooding) show a characteristic two-phase pattern: a pre-event increase in activity consistent with anticipatory behaviors (stocking, relocation), followed by a sharp collapse during peak impact.
Slow-onset shocks (heatwaves) produce weaker aggregate signals but reveal systematic spatial and functional redistributions of activity, with people shifting toward climate-controlled or shaded environments rather than reducing mobility altogether.
Data Quality Assessments#
Prior to analysis, comprehensive Exploratory Data Analysis and Quality Assessments (EDA+QA) were conducted for each country dataset, documenting temporal coverage, spatial distribution, regime shifts, and user-level heterogeneity.
Report |
Key findings |
|---|---|
~1.1B GPS points, 18.9M users; 96.7% temporal coverage; three anomalous regimes identified |
|
~4.4B GPS points, 27.2M users; 96.6% temporal coverage; structural break on 10 July 2023 |
Getting Started#
Prerequisites#
Python 3.8 or higher
Jupyter Lab for running notebooks
Installation#
Clone the repository:
git clone https://github.com/worldbank/eca-resilience.git cd eca-resilience
Create and activate the conda environment:
conda env create -f environment.yml conda activate eca-resilience
Usage#
For detailed documentation and analysis notebooks, visit the project documentation.
Contact#
For questions, feedback, or contributions, please contact the Development Data Partnership at datapartnership@worldbank.org.
You can also open an issue in the GitHub repository.
License#
This project is licensed under the MIT License together with the World Bank IGO Rider. The Rider is purely procedural: it reserves all privileges and immunities enjoyed by the World Bank, without adding restrictions to the MIT permissions. Please review both files before using, distributing or contributing.
Code of Conduct#
This project maintains a Code of Conduct to ensure an inclusive and respectful environment for everyone. Please adhere to it in all interactions within our community.