Researchers have developed a new deep learning model called the Trimodal Coherent Co-attention Transformer (TCCT) to improve the estimation of Above-Ground Biomass (AGB) in tropical forests. This model uniquely fuses optical reflectance data from Landsat-5 with complex-valued Polarimetric SAR Interferometry (PolInSAR) data from P and L bands. The TCCT uses complex-valued encoders to maintain phase coherence and a co-attention mechanism to dynamically adjust reliance on optical or SAR data, mitigating issues like cloud cover and signal saturation. When fine-tuned, the TCCT achieved an RMSE of 3.78 m for Canopy Height Models and a 4.51% rRMSE for AGB in dense forest areas, meeting the European Space Agency's BIOMASS mission requirements. AI
IMPACT This model offers a novel approach to fusing multimodal data for environmental monitoring, potentially improving carbon stock mapping accuracy.
RANK_REASON This is a research paper detailing a novel AI model for a specific scientific application. [lever_c_demoted from research: ic=1 ai=1.0]
- Above-Ground Biomass
- BIOMASS
- Canopy Height Models
- CNN
- European Space Agency
- Landsat-5
- L band
- Random Forest
- Trimodal Coherent Co-attention Transformer
- Vision Transformer
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
