From discrete to continuous references: QSM distribution for deep learning

Accurate estimation of aboveground biomass (AGB) is a fundamental pillar for sustainable forest management, carbon modeling, and climate change studies. For decades, conventional methods have relied on forest inventories based on discrete plots, where field measurements are aggregated and correlated with LiDAR data. However, this approach introduces a bias known as the edge effect: in small plots, uncertainty increases because trees on the boundaries partially contribute to the signal, degrading the performance of machine learning models. A recent study (arXiv:2607.05260) proposes a radical alternative: replacing aggregated references with a continuous horizontal biomass distribution (HBD) derived from quantitative structure models (QSM). By training a 3D U-Net on simulated broadleaf forest structures, the researchers demonstrated that, for 100 m² plots, the HBD method reduces the relative mean squared error by 16.84% and increases R² by 0.22 compared to the classic inventory. This advance not only corrects edge artifacts but also opens the door to more robust models even with limited training data. At Q2BSTUDIO, we understand that innovation in remote sensing and deep learning requires custom software platforms capable of processing massive point clouds and training complex architectures. Our experience in artificial intelligence for businesses allows us to design solutions that integrate everything from LiDAR data acquisition to production inference, whether on cloud infrastructure like AWS and Azure cloud services or through custom applications that manage geospatial data pipelines. Additionally, the combination of AI agents with business intelligence tools like Power BI enables real-time visualization of biomass maps, facilitating decision-making. The transition from discrete to continuous references is not just an academic achievement: it represents a paradigm shift that, with the right technological support, can be scaled to global forest monitoring projects. At Q2BSTUDIO, we help organizations implement these models through consulting in AI for businesses, ensuring that every innovation translates into tangible value. Cybersecurity also plays a key role in protecting critical field data, while process automation ensures workflows are efficient and reproducible. Ultimately, the fusion of QSM techniques with deep learning and cloud platforms marks the beginning of a new era in biomass estimation, and at Q2BSTUDIO we offer the know-how to make it a reality.