From Satellite Characterisation to Intelligent Earth Observation: A Conceptual AI-Driven Exploitation Framework for ANGEO-1 Data

The increasing availability of Very-High-Resolution (VHR) satellite data requires approaches that extend beyond conventional processing pipelines towards scalable and increasingly automated Earth Observation (EO) exploitation. In this context, ANGEO-1 (Angola Earth Observation-1), currently under development by Airbus Defence and Space on the S250 optical platform, is expected to represent an important advancement in African EO infrastructure. The satellite is designed to acquire sub-metre panchromatic imagery (approximately 0.5 m) and multispectral imagery (approximately 2 m) across the visible and near-infrared spectral domains, with an expected production capacity exceeding 1,000 scenes per day. Such anticipated data volumes highlight the need to prepare scalable processing and analytical strategies before the satellite becomes operational. This study presents the completed conceptual design and formalisation of an Artificial Intelligence (AI)-driven framework that defines how future ANGEO-1 observations could be transformed into actionable geospatial information, with a view to future deployment in operational EO environments. The proposed framework establishes a structured input–processing–output chain in which ANGEO-1 imagery and complementary multi-sensor observations constitute the inputs, AI-supported processing and spatiotemporal fusion form the intermediate analytical layer, and multi-scale environmental information products represent the intended outputs. It specifies the prospective integration of deep learning models, including convolutional neural networks, for feature learning and temporal reconstruction, together with spatiotemporal fusion strategies intended to produce temporally consistent, high-resolution Earth-surface products from complementary datasets. The proposed framework also specifies cloud-based geospatial infrastructure as a prospective deployment environment for scalable processing under potential local data and computational constraints. These components are intended to address challenges associated with high data throughput, limited revisit frequency and persistent cloud cover. However, their technical performance and operational benefits have not yet been experimentally demonstrated and will require post-launch implementation, calibration and validation using actual ANGEO-1 observations. The contribution therefore lies in providing a structured conceptual architecture that defines how ANGEO-1 data could be incorporated into AI-enabled EO systems, rather than in demonstrating performance improvements. The framework establishes a basis for subsequent experimental assessment of data harmonisation, spatiotemporal fusion and automated information extraction, while outlining a pathway for the future application of ANGEO-1 data to environmental monitoring and sustainable development.