SISTEL: Artificial Intelligence and Radar for Continuous Forest Monitoring in Conditions of Extreme Cloud Cover

The SISTEL project (System for Cloud Removal in Optical Satellite Images) represents a significant advance in the application of deep learning and Earth observation technologies to precision forest management. Using a generative neural network architecture, the project has overcome the long-standing limitation of cloud cover, enabling near-continuous monitoring of vegetation status and strengthening the operational capacity of the FORA Forest Technologies team. SISTEL has brought significant technological and strategic value to the FORA ecosystem. Overall, the project has enabled the design and implementation of the SAR2NDVI model, a disruptive solution that uses radar data (Sentinel-1) to ‘see’ through clouds and estimate the NDVI vegetation index (Sentinel-2) with high precision. This capability ensures that regions with persistent cloud cover, such as northern Spain and Europe, have access to useful data without relying on favourable weather conditions. DOCXThe key areas in which this project has generated a tangible impact are detailed below:Cutting-edge Research and Development: VICOMTECH has developed an innovative architecture based on temporal and multi-source information. Unlike static models, this solution uses local NDVI history and radar signal differences to capture subtle variations in vegetation.

This has enabled the development of a dynamic monitoring system capable of detecting forest fires even during periods of persistent cloud cover. DOCXCompetitiveness and Business Continuity: The integration of this technology gives FORA a unique competitive advantage: the Business Continuity service. The system drastically reduces alert times and improves the quality of monitoring in critical regions such as Galicia or the Basque Country, where the lack of optical imagery can last for weeks. DOCXTechnology Transfer and Autonomy: An exhaustive process of knowledge transfer has been carried out with the FORA Forest Technologies team. This transfer has included the handover of code repositories, detailed architecture manuals and consolidation training sessions. As a result, the SME has acquired the autonomy required to manage, retrain and evolve the system according to its future needs. Strategic Scalability: Beyond logging detection, the success of this development opens the door to new business lines in sectors such as forest health monitoring (pests and fires), biomass tracking for carbon markets and the generation of precision inventories.

Vicomtech’s role: Vicomtech has acted as a strategic technology partner, leading the design and training of the SAR2OPT generative models and their evolution into SAR2NDVI. Our contribution has ranged from the analysis and processing of large volumes of satellite data (ad hoc datasets for Galicia) to visual and qualitative validation in real-world use cases. Through a methodology of continuous training and asynchronous knowledge transfer, we have ensured the effective integration of a disruptive digitalisation solution that positions FORA at the forefront of the AgriTech and forestry sector.