A key challenge to the preservation of Earth's biodiversity is to develop understanding of how terrestrial ecosystems can be used sustainably. This entails the study of how anthropogenic changes to the landscape interact with natural ecological processes and climate over decadal timescales. Focusing on the notably biodiverse Mediterranean Basin, I will use Agent Based Modelling (ABM) to explore the dynamics of forest structure change in the Iberian Peninsula since the arrival of agriculture at the beginning of the Holocene.
The data sources are predominantly paleological, and include pollen sequences and sedimentary charcoal analyses from sites across Iberia. The relative performance of a given model will be determined by how well it can explain data from a variety of study sites simultaneously. Such performance indicators will be calculated within a Bayesian framework. Having inferred a number of optimal model structures and parameterisations based on empirical data, I will use those models to generate synthetic data describing system trajectories through an abstract state space of various emergent social and ecological variables.
This approach is inspired by statistical mechanics and, when combined with clustering techniques from the machine learning literature, will assist in the identification of system states which are expected to be qualitatively meaningful to landscape management professionals. I will then use statistical techniques such as Markov State Modelling to identify the transition rates between these states. My ambition for this work is to provide practitioners with novel insights into the dynamics of coupled socio-ecological systems over decadal time periods.