Deep learning breakthrough creates the world's most detailed national land-cover map, tracking forests, cities, and farms across 9.6 million square kilometers.
For the first time in history, scientists can see every square meter of China in unprecedented detail. The SinoLC-1 dataset represents a quantum leap in satellite mapping technology, using artificial intelligence to analyze millions of satellite images and create a land-cover map so precise it can distinguish individual buildings from surrounding vegetation. This breakthrough required processing over 850,000 satellite images through a revolutionary deep learning system called L2HNet.
The implications extend far beyond academic research. Urban planners can now track city expansion in real-time, environmental scientists can monitor deforestation down to individual trees, and disaster response teams can assess damage with pinpoint accuracy. The dataset has already been downloaded over 539,000 times by researchers, governments, and organizations worldwide, making it one of the most accessed geographic datasets ever released.
What makes this achievement even more remarkable is its open-access nature and continuous evolution. The research team has released not just the data but the entire AI framework, allowing anyone to update and improve the maps. By 2026, they expanded the system to map individual building functions across 109 Chinese cities, identifying whether structures serve as homes, schools, hospitals, or factories—creating an unprecedented digital twin of China's built environment.
Breakdown of major land types identified by the AI mapping system
Scientists can now see every square meter of China in unprecedented detail, distinguishing individual buildings from surrounding vegetation.
The 1-meter resolution enables breakthrough research in ecology, urban planning, and climate science that was impossible with lower-resolution datasets.
Policymakers can track environmental commitments, urban development, and agricultural changes with accountability previously impossible at national scales.
This breakthrough democratizes advanced mapping technology, potentially enabling similar detailed environmental monitoring capabilities worldwide.
Share this story