{
  "id": 719160,
  "title": "Researchers test the GlobalBuildingAtlas, the world’s newest global map of buildings in our cities",
  "url": "https://urgent.news/2026/08/13/researchers-test-the-globalbuildingatlas-the-worlds-newest-global-map",
  "topic": "science",
  "section": "Science",
  "published": "2026-08-13T00:30:00.000Z",
  "source": {
    "name": "Research Matters India",
    "slug": "research-matters-india",
    "url": "https://researchmatters.in/news/researchers-test-globalbuildingatlas-worlds-newest-global-map-buildings-our-cities"
  },
  "original_language": "en",
  "account": "Researchers have conducted an independent assessment of the GlobalBuildingAtlas (GBA), a cutting-edge digital dataset designed to map every building on Earth. The study, which centered on rapidly urbanizing regions of India, reveals that while the GBA excels in pinpointing building locations, it consistently underestimates their heights, frequently depicting skyscrapers shorter than their actual height by more than half. This finding is crucial for urban planners and climate scientists, who depend on 3D models to project energy consumption and flood dynamics within cities. Conducted by ISRO's National Remote Sensing Centre, the research surveyed various urban centers such as Mumbai, Bengaluru, and Hyderabad. India's rapid urbanization underscores the necessity for a digital map of its cities, and the GBA, released as an open-access resource, now contains over 2.75 billion building footprints worldwide. It is the first map of its kind to provide three-dimensional data at this scale, offering not only outlines but also estimated heights and simplified 3D shapes. However, the accuracy of these digital maps hinges on the underlying artificial intelligence and satellite imagery, leading to variable precision depending on regional conditions. To evaluate the GBA, researchers employed NASA's ICESat-2 satellite, which utilizes a laser altimeter to measure height with centimeter-level precision from space. Comparing the satellite's measurements with the GBA's estimates for 20 diverse structures ranging from low-rise warehouses to tall residential towers, the team discovered a consistent underestimation of building heights. For horizontal mapping, the team utilized high-resolution imagery from ISRO's Bhuvan portal, processed with an AI model to ensure ground truth accuracy. The results were striking: horizontally, the GBA achieved an impressive completeness ratio of over 99% in planned residential areas of major Indian cities, confirming its reliability for urban density and sprawl studies. However, the vertical mapping fell short. The researchers identified a systematic negative bias, meaning the GBA regularly underestimated building heights, with errors growing more pronounced as buildings increased in height. For instance, a 266-meter skyscraper in Mumbai was estimated to be only 113 meters tall, a deviation of over 150 meters. Mid-rise buildings experienced errors exceeding 40%. The GBA employs monocular optical imagery, relying on 2D photos to estimate height instead of direct 3D measurements. While AI can infer height from shadows and perspectives, it often misinterprets complex roof shapes, nearby trees, and dense urban layouts. Previous global datasets, like those from Microsoft and Google, concentrated solely on 2D footprints. The GBA ambitiously aims to incorporate the third dimension—height—which is vital for contemporary sustainability goals. Nonetheless, the ISRO study underscores a significant drawback. The AI models that created the GBA were predominantly trained on data from Europe and North America, failing to account for the architectural diversity, construction materials, and urban densities unique to the Global South, particularly in India. Furthermore, the GBA's simplified 3D representation, known as Level of Detail 1, treats every building as a flat-topped block, disregarding intricate spires, sloped roofs, and other distinctive features common in Indian landmarks. The study concludes that while the GlobalBuildingAtlas is invaluable for understanding the horizontal fabric of our cities, it must be applied with caution when vertical accuracy is paramount. As we leverage these maps for smart city design, disaster preparedness, and solar energy calculations, this research serves as a critical reminder. We must ensure that digital maps accurately reflect the real world, and by pinpointing these height inaccuracies, the ISRO team has paved the way for future advancements that could eventually deliver a digital representation as robust and complex as our physical environment itself.",
  "summary": "Research Matters Staff Writer(s) Jodhpur 13 Aug 2026 Researchers have conducted one of the first independent stress tests on the GlobalBuildingAtlas (GBA), a massive new digital dataset that aims to map every building on Earth. The study, focusing on the rapidly growing urban landscapes of India, shows that while the GBA is incredibly accurate at identifying where buildings are located on the…",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}