Abstract
Population datasets accounting for the full range of routine human activities are needed to address many global human security challenges, including disasters, conflict, and infrastructure demand. LandScan High Definition (HD) supports this need through gridded ambient population estimates that measure average human presence between daytime and nighttime at a high spatial resolution of 3 arcseconds (roughly 90 m). Although LandScan HD has traditionally been produced on a country-specific basis, advances in global foundational data and computational resources now enable scaling its methodology to the world. Combining aspects of top-down and bottom-up gridded population methods, LandScan HD allocates subnational population totals from authoritative statistics to built-up areas based on occupancy estimates for multiple facility types (e.g., residential, commercial) and then reaggregates these estimates to a global population grid. We scale this approach by organizing the LandScan HD data stack into a 1° resolution tileset of vector analytic features, enabling an efficient and repeatable workflow for all countries worldwide. Examining the Philippines as an output of the global LandScan HD baseline dataset, we contrast the ambient population with a gridded population representing residential activities (WorldPop) by (1) exploring a practical application for flood risk assessment and (2) evaluating congruence with outcomes of collective human activities (subnational CO2 emissions). Finally, we discuss confronting current LandScan HD limitations through data/modeling and uncertainty quantification improvements and provide outlook for workflow automation and extending the model to social, demographic and economic population characteristics.
Introduction
Accurately mapping human distribution is critical for disaster response, urban planning, and public health. However, most existing gridded population datasets, such as WorldPop or Gridded Population of the World (GPW), primarily represent residential populations—often referred to as "heads in beds." While essential for certain applications, these datasets fail to capture the dynamic nature of human movement. During the day, populations shift from residential neighborhoods to commercial districts, schools, and industrial hubs.
LandScan High Definition (HD) addresses this gap by providing a global, high-resolution (3 arcseconds, ~90m) dataset of the "ambient" population. Ambient population represents the average human presence over a 24-hour period, integrating both daytime and nighttime activities. This methodology enables a more realistic assessment of human presence for emergency management and infrastructure planning.
Methodology: A Hybrid Modeling Approach
The LandScan HD methodology employs a hybrid top-down and bottom-up approach to achieve high spatial precision while maintaining statistical integrity.
Top-Down Disaggregation
The model begins with authoritative subnational population totals (typically at the Administrative Level 1 or 2) sourced from the U.S. Census Bureau’s International Database and national government agencies. These totals serve as the "control" to ensure the final grid sums to official statistics.
Bottom-Up Occupancy Modeling
To distribute these totals, the researchers utilize Oak Ridge National Laboratory’s (ORNL) Population Density Tables (PDT). The PDT is a Bayesian machine learning framework that estimates occupant density (people per 1,000 square feet) for over 60 different facility types, including residential, commercial, educational, and healthcare buildings. By modeling how many people are likely to be in a specific type of building at different times, the researchers can estimate a 24-hour average presence.
The MOLECULE Framework
To manage the massive data volume required for global modeling, the team developed the "MOLECULE" framework. This vector analytical approach intersects building footprints with land-use data and administrative boundaries to create the smallest possible units of population distribution. By organizing the global data stack into 1° resolution tiles, the researchers moved away from bespoke, country-by-country modeling toward a repeatable, automated global workflow.
Data Infrastructure
The LandScan HD "data stack" integrates diverse and massive geospatial datasets:
- Building Footprints: Primarily ORNL’s own CNN-derived features, supplemented by Microsoft and Google footprints, with GHSL Built-S used as a fallback.
- Land Use/Attribution: Data from OpenStreetMap (OSM), the Urban Tactical Planner (UTP), and the Multinational Geospatial Co-production Program (MGCP).
- Validation Baselines: WorldPop (for residential baseline) and EDGAR (for CO2 emissions correlation).
Key Insights and Validation
The effectiveness of the ambient population model is demonstrated through a case study in the Philippines, highlighting the "ambient advantage."
Ambient vs. Residential Dynamics
In Iloilo City, Philippines, LandScan HD estimated an ambient population of approximately 660,000, significantly higher than the residential-only estimate of 430,000 provided by WorldPop. This delta represents the "daytime surge" of commuters and students, which is vital for emergency managers planning for daytime disaster scenarios.
CO2 Emissions as a Proxy
In a novel validation step, the authors compared population maps to subnational CO2 emissions. They found that LandScan HD correlates significantly better with emissions (ρ = 0.761) than residential maps (ρ = 0.682). Because CO2 is a byproduct of human activity (transport, industry, commercial heating), this correlation validates that LandScan HD more accurately captures where human activity actually occurs.
Flood Risk Assessment
By overlaying population data with flood hazard maps, the study found that more daily activity occurs in "moderate" flood-risk zones than residential data suggests. This provides more actionable insights for urban planners to identify vulnerable populations that might be overlooked by traditional census-based maps.
Limitations and Future Directions
Despite its advancements, LandScan HD faces several challenges. Global data on building heights is still sparse, leading the model to default to a "two-floor" assumption in many areas, which can cause errors in dense urban cores. Additionally, while OSM is extensive, many buildings remain unlabeled. The authors plan to integrate "MapSpace," an ML-driven inventory of land-use labels, to fill these gaps.
Future iterations will focus on moving from deterministic to probabilistic modeling by developing a Confidence Index to quantify uncertainty. The team also aims to automate the pipeline further to support rapid population estimates during active humanitarian crises or conflicts.
Conclusion
LandScan HD represents a significant leap forward in global population modeling. By shifting the focus from where people sleep to where they spend their lives, this methodology provides a more realistic and actionable foundation for addressing global security, infrastructure, and environmental challenges.
Summary updated