Global population map by US scientists shows where people are worldwide 24/7, building by building

Researchers in the US have recently issued a new global population map that can track where people are likely to be throughout the day by modeling building use and occupancy.

The map, named LandScan Mosaic, was developed at the Department of Energy’s (DOE) Oak Ridge National Laboratory (ORNL). It reportedly merges building data, land-use information and machine learning to create a high-resolution picture of population distribution worldwide.

Conventional population maps usually rely on satellite images divided into pixels. The novel system, on the other hand, utilizes buildings to estimate where people are located. Andrew Zimmer, PhD, a geospatial scientist at ORNL, elaborated that buildings are used to identify what is essentially a demographic fingerprint.

“For example, there is a certain age group that attends schools,” he highlighted. “So, when we’re building a demographic dataset, we can use the function of buildings at a fine spatial resolution to inform the distribution of those younger populations to certain locations and capture daytime and nighttime dynamics.”

Mapping people through buildings

LandScan Mosaic is an extension of the existing LandScan Global dataset, which is an annually updated mapping system. It reportedly predicts the average location of people over a 24-hour period, accounting for where they live, work and move. The standard global dataset has a resolution of roughly 0.6 miles (one kilometer).

The main difference is that Mosaic focuses on buildings. When information about a building is unavailable, the researchers can use machine learning to estimate its characteristics. These include building height, number of floors, function and use type.

The information shows how residential, commercial, industrial and other buildings are occupied. LandScan Mosaic combines this with land-use data and occupancy patterns to map population distribution throughout the day.

Instead of just showing where people live, it represents an “ambient population,” the average number of people at a location over a 24-hour period. This allows the dataset to account for daily movement between homes, workplaces, schools and other locations.

ORNL said that the approach provides greater temporal and location specificity than LandScan Global. It also measures uncertainty. It shows how confident the researchers are in its population estimates. This, in turn, provides more context for decision-makers.

Half a century of population change

Daniel Adams, an R&D scientist at ORNL and lead author of the study, said that LandScan Mosaic is the first globally available population dataset released with this set of characteristics. “This is a really important project for showcasing how AI and decision analysis can really be blended together in a trustworthy manner,” he added in a press release.

Combining advanced computing and machine learning with uncertainty analysis could make the data more useful when rapid decisions are necessary, including disaster response and infrastructure planning.

The research team’s next release, LandScan Mosaic Timeseries (LSM-TS), will have annual population distributions spanning 50 years, from 1975 through 2024. The system will use LandScan Mosaic as an input and apply a “back-casting” method to reconstruct historical population distributions.

Zimmer explained that the five-decade timeline could help examine the growth of cities and settlements before they were already well established. “The building level framework and the daytime and nighttime data allow opportunities to look at both temporal granularity within the day and how that may have changed over time but also look at those settlements and identify those as well,” he concluded.

The team is also working to add demographic characteristics to LandScan Mosaic. Building-level data can contribute to this effort as well, according to the team.

The study has been published in the journal Scientific Reports.