In work published Feb. 17 in Nature Communications, researchers estimated Human Development Index, or HDI, scores for 61,530 municipalities and counties worldwide using a model trained on satellite images and official survey based data. They found that more than half of the global population lives in municipalities where the development tier differs from the tier assigned to their country as a whole, showing that local conditions often diverge from national averages.
The team reports that in states and provinces placed in the bottom two of five development tiers in conventional assessments, about 8.5 percent of residents moved into the top two tiers when development was evaluated at the municipal level. When the model generated estimates for 10 by 10 kilometer grid tiles, roughly the area of a major city such as Paris, the share of people whose development tier shifted rose to about 13 percent, underscoring how spatial detail can alter policy relevant classifications.
While the new estimates provide much more geographic detail than earlier HDI maps, the authors emphasize that they do not identify conditions in individual households or at the neighborhood scale. Instead, they argue that municipal and grid level information can help governments and aid organizations tailor policies and programs more effectively, directing support to localities whose development status has been misrepresented by national level data.
"We frequently target policies and programs based on these aggregate statistics, but we want to support the people who need it, not only the countries that need it," said study co author Solomon Hsiang of the Stanford Doerr School of Sustainability. Co author Heriberto Tapia of the United Nations Development Program's Human Development Report Office noted that the index, introduced in 1990, was designed to shift attention beyond income and economic growth so that policy makers could focus more directly on what is happening to people.
Tapia said that relying on national averages has always risked obscuring inequality and missing opportunities to improve policy within countries, a problem made worse by data gaps in the poorest nations, where only about half have conducted a census in the past decade. According to the authors, those gaps are one reason they turned to satellites, which now generate more data each day than all social media platforms combined, yet remain underused in official development statistics.
Beginning in 2020, Hsiang, Tapia, and collaborators set out to test whether satellite imagery could sharpen the spatial resolution of United Nations HDI data and related indicators. "Our ambition is that, thanks to these estimates that are very granular, people in different localities around the world will be able to assess what is happening with their human development following the same standard," Tapia said.
The need for that insight has grown as global human development has stalled over the past two years after decades of progress, according to United Nations data. Tapia pointed to the lingering impacts of the pandemic, rising water scarcity, extreme weather, and other climate related hazards as "multiple shocks" that are hitting different parts of the world and contributing to the slowdown in human development.
To build their model, the researchers trained a machine learning system on satellite images of states and provinces paired with survey based statistics, including official HDI data. Because provinces are large and irregularly shaped, rather than neatly gridded like typical computer vision inputs, co lead author Jonathan Proctor of the University of British Columbia said the team was initially surprised that their approach performed as well as it did.
Over time, the model learned associations between HDI and visual features in the satellite data, and the team then applied it to predict HDI for municipalities and counties across the globe. Lead author Luke Sherman of the Stanford Global Policy Laboratory said the work shows that with satellite imagery it is possible to obtain approximate estimates of variables such as school enrollment, educational attainment, or HDI itself at much finer spatial resolution than is available from standard household surveys alone.
The researchers also probed which visible features in the images were driving their model's predictions. They found that municipalities with higher road density and building density tended to have higher predicted HDI values. Population was only weakly associated with HDI at the global scale, but within individual countries, more densely populated areas generally showed higher HDI, suggesting that built infrastructure and local population density are important markers of development.
Overall, built infrastructure and population density together explained about one third of the variation in municipal HDI estimates, leaving most of the variation attributable to other factors that are not directly visible from space. The authors note that this unexplained share points to the continued importance of traditional survey data and on the ground knowledge for understanding human development, even as remote sensing expands what can be mapped consistently at large scales.
In a recent pre print, the team extended their approach to more than 100 social, economic, and environmental variables, including crop yields, asset ownership such as cars or livestock, and electricity access. They found that the same technique can predict many of these indicators with useful accuracy, suggesting that satellite based models could be used to increase the spatial resolution of a wide range of administrative data at relatively low cost.
The group designed its system for ease of use at large scale so that researchers, governments, and development practitioners can apply satellite imagery to assess development gaps without building their own complex pipelines. "In some ways, our approach is the Toyota Camry of the remote sensing world," Proctor said. "It does not take the curves as well as a Porsche, but it gets the job done and is accessible to a broad range of people."
By lowering technical barriers, the authors hope to make it easier for agencies and organizations in low and middle income countries to generate their own high resolution development maps. They argue that widespread access to such tools could help local decision makers identify pockets of deprivation inside otherwise prosperous regions and track how shocks such as climate extremes or economic crises affect human development over time.
Research Report:Global high-resolution estimates of the UN Human Development Index using satellite imagery and machine learning
Related Links
Stanford Doerr School of Sustainability
Space Technology News - Applications and Research
| Subscribe Free To Our Daily Newsletters |
| Subscribe Free To Our Daily Newsletters |
