Daniela Fernanda Milón-Flores

Abstract 

The negative impact of human activities on our planet has increased the interest of both expert and non-expert users in monitoring and understanding local environmental changes. Satellite imagery is a valuable resource for observing the Earth, due to its spectral and spatio-temporal resolution. These features enable the construction of Earth Observation Time Series (EO-TS) based on satellite-derived indicators across various themes, such as vegetation and temperature, supporting long-term environmental monitoring. To manage large EO datasets effectively, EO Data Cubes (EODC) offer a structured organization along spatial, temporal, and thematic dimensions.

Despite being openly available, EO data remains complex, sensor-oriented, and lacks semantic context, which makes it difficult for non-expert users to interpret. While EODCs support structured access and facilitate analysis, the temporal sequences they contain still require synthesis to reveal what is actually happening over time. The main challenge lies in semantically characterizing satellite-derived indicators at spatial scales relevant to a wide range of stakeholders, including citizens, spatial planners, and researchers, and in producing clear, interpretable syntheses of key environmental events and trends within territorial units (TUs).

To address these challenges, we combine Semantic Web technologies with semantic trajectory models to identify, synthesize, and enrich meaningful patterns in EO-TS.

Building on this approach, we introduce the SETT ontology, designed to model semantically enriched trajectories of TUs. SETT adopts a multi-layered model that captures the environmental evolution of territories, such as municipalities. It transforms raw EO time series into sequences of states and transitions, semantically annotated to enhance interpretability and support temporal change analysis.

To operationalize SETT, we designed and developed OFf-SETT, a framework that transforms raw EO-TS into linked, semantically enriched environmental trajectories through four modules:

(1) preprocessing and aggregation of EO data at the TU level;

(2) modeling the data as semantic EODCs for improved management;

(3) detecting and annotating patterns (e.g., trends and transitions) in the EO-TS; and

(4) generating and exploring semantic environmental trajectories in the form of knowledge graphs.

We applied the framework in a real-world case study involving municipalities in Switzerland and France. The resulting semantic trajectories synthesize and enrich the environmental evolution of these TUs. OFf-SETT also enables advanced analyses, such as identifying municipalities with similar evolution patterns that may indicate resilience to broader climate change dynamics. Furthermore, we demonstrate the adaptability of OFf-SETT to other types of indicators by applying it to demographic data, enabling correlations between demographic dynamics and EO-based trajectories. Our contribution offers a reproducible, extensible, and interpretable approach for representing and analyzing the evolution of TUs.