Building better maps with open data on building heights

State and local governments can now use a new dataset for location-based apps that includes building height data, a feature that will improve their GIS analysis and map visualizations.

Having building height data can be helpful in a variety of use cases, said Mike Tischler, director of the National Geospatial Program at the U.S. Geological Survey (USGS). The dataset was launched in April as a partnership with Overture Maps Foundation (OMF), a collaborative effort to enable current and next-generation interoperable open map products. The project started in December 2022.

Height data, for example, is critical for expanding broadband. “You need [data on] the physicality of the buildings to understand where [signals] are going to be propagated, where you should put your towers, where they can be received over a city,” Tischler said. “For local governments that want to make sure that they’ve got good coverage in their city, having a 3D representation of the infrastructure there is really key.”

Another potential use case is in solar power. Organizations interested in solar energy can use building height in determining solar panel orientation, for example, he said.

Perhaps the biggest use case for building height data is the development of digital twins. “If city managers or urban planners are wanting to understand their entire environment in 3D, this is the kind of information [they] need,” Tischler said. 

USGS and OMF developed the dataset using several sources: building footprints from OpenStreetMap, Microsoft’s open Building Footprints data and community sources. The foundation is the lidar data from the agency’s 3D Elevation Program, whose goal “is to acquire lidar across the nation and ‘provide the first-ever national baseline of elevation data,’” according to OMF’s “Building Heights” report on the project. 

To estimate a structure’s height above the ground, the team uses a building footprint and lidar samples within that footprint to get an estimate of the absolute elevation of the building’s tallest part. Then they subtract the elevation of the ground, the report states. To validate the approximations, analysts use aerial imagery of buildings and an internal tool to measure their highest point from several views to triangulate the absolute elevation.

“One of the other models we are looking at … is getting information that a state or local government might have,” said Marc Prioleau, OMF’s executive director. “For instance, the city of Des Moines, Iowa, could have road centerline data. It might be very precisely surveyed, might be really high-quality data, but it’s the city of Des Moines’,” he said. That data might get added to the city’s open data portal where someone may use it. But if it is added to Overture, then it “becomes part of a bigger worldwide dataset and actually gets used a lot.”

It’s likely that a state or locality already has data in the Overture dataset, Tischler said, because the 3DEP program provides a way for any entity that collects lidar data to pool that with other agencies. 

“No one really envisioned a building height dataset,” he said. “It’s going to be really interesting to see now that this is out there [to see] how the states and local governments use it. Is it management of digital twins and better infrastructure? Is it for more environmental purposes and looking at migratory patterns of birds through their city? I’m curious to see how this gets picked up.”

The first release of the dataset included 6 million building heights across 34,000 square miles in cities such as Boston, Chicago and Santa Clara. Since then, the set has grown to include more than 20 million building heights, and Prioleau said the plan is to put out a bigger release of 40 million to 50 million in the early second half of next year. Currently, building heights are accurate to within about 50 cm, he added.

The report describes five technical challenges that arose. One is misalignment between building footprints and lidar because the data is collected separately. To address that, analysts are looking at more sophisticated estimators. 

Another is point clouds without building classifications. “In about 85% of [USGS] projects, buildings points are not classified separately,” the report states. “We notice this especially when a tree overhangs the footprint of a building.” As a result, the height could be overestimated. 

“In future releases we will continue to refine our approach including a combination of alternative estimators, taking into account the spatial distribution of the points, using last return lidar points, performing our own point classification, and possibly using co-registered imagery to segment trees,” the report says.

Stephanie Kanowitz is a freelance writer based in northern Virginia.