Researchers have developed new wall-to-wall maps of canopy height and aboveground carbon density for the Brazilian Atlantic Forest, achieving higher accuracy than existing global products. A deep learning network, trained on extensive LiDAR data and fused with satellite imagery from Sentinel-2, Sentinel-1, and ALOS PALSAR, predicted canopy height with a mean absolute error of 2.26 m. This new methodology significantly outperforms global products from Meta and the World Resources Institute, which showed more than double the error on the same test data. The study also established a conversion from canopy height to aboveground carbon density, estimating the remaining forest holds 1,944 TgC. AI
IMPACT This research demonstrates how AI can significantly improve the accuracy of environmental monitoring, crucial for carbon accounting and conservation efforts.
RANK_REASON Academic paper detailing a new methodology for mapping forest carbon using deep learning and satellite data. [lever_c_demoted from research: ic=1 ai=1.0]
- aboveground carbon density
- arXiv
- Atlantic Forest
- Hugging Face
- LiDAR
- Matheus Ferreira
- Meta
- São Paulo
- Sentinel-1
- Sentinel-2
- World Resources Institute
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →