Abstract
Generative simulation models and the mathematical analysis of their outputs can help us to quantify, visualise and understand the impact of anthropogenic land cover change on terrestrial ecosystems. In this thesis I describe a socio-ecological simulation model of land-cover change in the Iberian Peninsula called AgroSuccess. This model integrates previous work from the literature describing ecological succession and ecological disturbance in the form of both fire and anthropogenic subsistence activities. AgroSuccess is an agent-based simulation model that enables users to explore the effect of agricultural land management practices and different climatic conditions on the emergent state of simulated landscapes. I demonstrate AgroSuccess by investigating the changes to land cover resulting from the introduction of agriculture during the mid-Holocene at six study sites in the Iberian Peninsula.AgroSuccess requires input data to specify boundary conditions, as well as reference data against which to compare outputs and calibrate parameters. I have developed a collection of reusable software tools to obtain and prepare paleo-ecological pollen abundance data collected by previous researchers, as well as morphological data from remote sensing to characterise study sites. To the best of my knowledge the way in which I have synthesised these data from disparate sources is novel, and the approach I have taken can be easily replicated by others using the open source software tools that I have made available online.
To address my research questions AgroSuccess represents various ecological and anthropogenic processes. Consequently, it is an example of a complicated model that requires the collection and assimilation of multiple forms of data to parameterise and initialise simulation runs. The management, documentation, communication, and reuse of such a model is difficult. An example of how I have mitigated these challenges is the development of a software application, called Cymod, that helps users to visualise the state-and-transition model (STM) that is integral to AgroSuccess’ ecological succession submodel. I argue that the measures taken to ensure the correct implementation of scientific simulation models are arbitrary, and often inadequate to provide scientists with the confidence they need to use each others’ code. In response to this challenge, the software implementation of the AgroSuccess simulation model, and the scripts that process its input data, are modular and well-tested. By distributing these modular components in public software repositories, I aim to make my work transparent and help others understand and reproduce my results.
| Date of Award | 1 Aug 2023 |
|---|---|
| Original language | English |
| Awarding Institution |
|
| Supervisor | James Millington (Supervisor) & Simon Miles (Supervisor) |