Objectives

The first scientific and technical challenge of the TRACES project concerns the design and operation of a digital object called a semantic environmental trajectory of a territory (SETT), which, to our knowledge, is the first attempt to represent this concept using a semantic web-based approach.

The difficulty lies in defining such a SETT in terms of potentially different types of indicators and measurements, and in coupling it with its intrinsic spatial and temporal dimensions. A SETT appears to be a complex, multidimensional and multigranular object, the modelling, visualisation and analysis of which are difficult and as yet unexplored tasks.

An ontological model of SETT should also include standard vocabularies to facilitate data integration and ensure the reusability of KGs representing SETTs that will be created from this conceptual description and the values of the environmental indicators considered.

The transition from a longitudinal series of environmental indicator measurements to a SETT first requires a synthesis and/or segmentation phase. It is necessary to study the extent to which the specificity of the indicators considered and the multidimensional nature of the expected SETTs complicate the determination of the components of these trajectories.

Subsequently, exploring the possibilities of applying machine learning clustering techniques to SETTs aims to:

1) provide similarity measures adapted to multimodal data representations evolving over time;
2) enable multimodal clustering of SETTs for the definition of trajectory profiles;
3) enable temporal prediction on multimodal data representations.

Similarly, the main challenges posed by modelling SETTs using a multi-agent system lie in:

1) exploiting SETTs to derive, if possible automatically, multi-agent models that reproduce past environmental trajectories;
2) combining and visualising, in the same hybrid models, spatially explicit agents modelling human behaviour and territorial changes;
3) exploiting a series of scenarios highlighting relevant policies in order to provide useful prescriptive analysis.

Environmental indicators and related measurements will be extracted from the Swiss Data Cube platform, which stores Earth observation satellite data spanning more than four decades (1984–2025) from the Landsat and Sentinel satellite mission programmes.
A longitudinal and comparative analysis will be conducted across three territories: Switzerland (Canton of Fribourg), France (Pays d’Évian) and both sides of the border (Grand Genève). The approach adopted is intended to be generic and applicable to various territories around the world.