Context-Aware Displacement Estimation from Mobile Phone Data: A Methodological Framework

A team of researchers led by Rajius Idzalika has published a methodological framework on arXiv that leverages mobile phone data to deliver more accurate population displacement estimates during disasters. The framework addresses a critical flaw in existing approaches: they often misclassify regular commuters as displaced persons. The new method introduces three key innovations: mobility profile classification to distinguish residents from commuters, context-aware detection that accounts for expected locations based on user type and day of week, and operational uncertainty bounds derived from baseline variability with a disaster adjustment factor.

In a case study following Super Typhoon Nando (2025) in Aparri, Philippines, the team applied the framework to daily location data from Globe Telecom. The context-aware approach reduced estimated between-municipality displacement by 1.6 to 2.7 percentage points on weekdays compared to naive methods, primarily due to the commuter exception. The framework produces three complementary metrics scaled to population with uncertainty bounds: displacement rates, origin-destination flows, and return dynamics. While the single-case demonstration establishes proof of concept, the authors note that external validity requires application across multiple events and locations. The framework is designed for humanitarian decision support and preserves individual privacy through data aggregation.