Â鶹ÒùÔº

Land use and land cover (LULC) information underpins studies in climate science, disaster management, food security, and ecosystem protection. Advances in satellite imaging have improved resolution, but high-resolution land cover mapping still faces major hurdles.

Traditional machine learning methods often fail to capture fine urban structures, while deep learning approaches require enormous amounts of labeled data, which are laborious and expensive to produce. Weakly supervised methods and foundation models show promise but struggle with accuracy and transferability across diverse urban environments. Due to these challenges, there is a pressing need to develop new strategies that reduce annotation costs while ensuring reliable large-scale submeter mapping.

A team from Wuhan University and Zhejiang Mingzhou Surveying and Mapping Institute has developed a new solution. Their study, in the Journal of Remote Sensing, introduces the "initial and expanded labeling" (IEL) engine and presents EcoVision, a submeter-resolution land cover product covering 42 of China's largest cities. By integrating high-resolution imagery, crowdsourced data, and deep learning models, the researchers achieved an unprecedented 0.5-meter land cover dataset with 83.6% accuracy across more than 23 million validation pixels.

The IEL annotation engine operates in two stages. First, it generates trusted "seed labels" through a priority-based weighted voting strategy, reconciling multiple historical land cover products with varying resolutions and classification systems. Second, it iteratively expands these labels using a semantic segmentation network, progressively refining accuracy through repeated cycles until model performance stabilizes. This hybrid approach overcomes mismatched formats, pixel misalignment, and label scarcity.

Applying IEL, the team produced EcoVision, which classifies urban areas into eight categories: buildings, roads, other impervious surfaces, grass/shrubs, trees, soil, agriculture, and water. Validation involved 2,385 image patches covering 23,850,000 pixels across 42 cities, delivering 83.6% overall accuracy. Key categories such as buildings achieved an F1 score of 90%, while roads and agriculture exceeded 83%.

Compared with five state-of-the-art products—including Hi-ULCM and SinoLC-1—EcoVision offered superior resolution and richer details, particularly in distinguishing impervious surfaces and extracting agricultural patches. Visual comparisons highlighted EcoVision's ability to accurately delineate roads obscured by shadows and to capture fine urban ecological mosaics. This makes it the first large-scale submeter LULC product available for China, now publicly released as an open dataset.

"EcoVision represents a milestone in urban remote sensing," said lead author Encheng Zhang. "By eliminating the bottleneck of manual labeling, our IEL engine enables the creation of high-resolution land cover maps at scales previously unattainable. The accuracy and detail of EcoVision allow us to see Chinese cities as dynamic ecological systems, not just built structures. We hope that by making the dataset publicly available, researchers and planners worldwide can use it to advance sustainable urban development and address pressing challenges such as climate resilience and ecosystem protection."

EcoVision's release provides an invaluable tool for multiple domains. Urban planners can use it to assess green space distribution and infrastructure expansion, while environmental scientists can study heat islands, carbon storage, and water dynamics. Policy makers may apply the dataset in territorial spatial planning, climate adaptation strategies, and monitoring of sustainable development goals.

The open availability of 0.5-meter land cover data across 42 major cities also creates opportunities for machine learning applications, such as urban growth modeling and disaster risk assessment. Ultimately, EcoVision demonstrates how innovative AI-driven annotation can unlock the full potential of high-resolution imagery for shaping resilient and sustainable cities.