BioCube: A Multimodal Dataset for Biodiversity Research

Accurate forest biomass estimation is essential for climate change mitigation, biodiversity monitoring, and sustainable forest management. Recent advances in remote sensing and machine learning have opened new avenues for large-scale, high-resolution biomass mapping. In this work, we introduce Grunblick, an AI-powered service designed for scalable forest biomass estimation, leveraging multi-sensor Earth Observation (EO) data, including Sentinel-1 and Sentinel-2 imagery. The Grunblick pipeline integrates modular deep learning models, notably U-Net architectures with interchangeable feature extraction backbones, to perform pixel-wise above-ground biomass (AGB) regression. We validate our system using the public Biomassters benchmark, demonstrating significant performance gains through multi-modal sensor fusion and self-supervised pretraining strategies. Future extensions will include uncertainty quantification and global deployment capabilities.