Task T5 |
Agent-Based Modelling and Prescriptive Analysis |
Task leader: Giovanna Di Marzo – CUI |
| Participants: J. Gensel (LIG); C. Cruz (LIB); G. Falquet, N. Hamel (CUI); Flann Chambers (CUI/LIB) | ||
Objectives
This task develops spatially explicit computational agent-based models encompassing human actor behaviour, human actor-environment dynamics and its impact on land cover, based on land cover policy changes. We will develop forecasting agent-based models by integrating the past evolution (enriched KGs (from T3), trajectory profiling (from T4), and satellite images data (from T1)), with environmental and socio-economic data and investigate future scenarios for a prescriptive analysis on land cover trajectories.
Activities
A5.1 Multi-agent modelling: aims to develop a series of spatially explicit agent-based models, capturing diverse human actors/stakeholder behaviour, decisions, adaptations to policies or environmental conditions, including socio-economic factors, and providing different actors’ strategies. To this end, it is essential to identify all the human actors/stakeholders involved in the observed system, the interactions among them and with the environment (land). We will also consider multi-scale aspects as actors may act individually, collectively or as an institution at diverse levels. This activity leverages information provided by enriched KGs (T3), trajectory profiling (T4), and land images (T1), and provides KGs-driven agent models.
A5.2 – Impact on land cover and trajectories: aims to model the impact of human actors’ behaviour on the environment, as well as various dynamic aspects and behaviour such as choice of crops, urban development and their effect on land degradation. Similarly, we will consider the effect of land change on the human actors’ behaviour (e.g., changing strategy). We will attempt at reproducing land change, land use, or urban sprawl patterns, identified from enriched KGs (T3), trajectory profiling (T4) and land images (T1). These hybrid models, combining human actors’ behaviour and land changes, will display spatiotemporal evolution and multi-scale aspects (micro, meso and macro-scale, local vs regional). For land cover trajectories, we will use 2D multi-scale lattice agent-based models (e.g., cellular automata). This activity provides a combined model of human actors/stakeholder behaviour and their impact on land trajectories. We will validate our models by identifying to which extent they: 1) exhibit similar behaviour, profile, or effect on the land (T4); 2) follow history of land trajectory as provided by Enriched KGs (T3); 3) share spatiotemporal similarities with the observed satellite data (T1).
A5.3 – Forecasting and prescriptive analysis: aims to automate the development of various policies scenarios, by integrating those policies in the models above (A5.1), in order to derive spatiotemporal forecasting of the land cover in various scenarios. Policies are important to consider because they have an impact on human actor’s behaviour, and, through them, on land use and land change. We will, on the one hand, identify the mechanisms at work behind the dynamics of land-use change (e.g., explaining the changes), and, on the other hand, anticipate the effect of policies on a land trajectory (i.e. providing a prescriptive analysis).
Deliverables |
| Activity | N° | Caption | Type | Deadline |
| A5.1 | D5.1 | Agent-based models | Executable models | M18 |
| A5.2 | D5.2 | Impact on land cover and trajectories | Executable models | M30 |
| A5.3 | D5.3 | Forecasting and prescriptive analysis | Executable models, report, Phd thesis | M39 |