Geomorphological evidence has frequently been used to infer past environmental conditions, but in recent years the emergence of landscape evolution models (LEMs) has opened the possibility of us- ing numerical modelling as a tool in palaeo-environmental reconstruction. The application of LEMs for this purpose involves retrodictive modelling, each simulation scenario being congured with model variables (e.g. reflecting climate change) and parameters to reflect a specic hypothesis of environmental change. Plausible scenarios are then identied by matching contemporary observed and modelled landscapes. However, although considerable uncertainty is known to surround the specication of model driving conditions and parameters, previous studies have not considered this issue. This research applies a technique of accounting for the uncertainty surrounding the speci- cation of driving conditions and model parameters by using reduced complexity 'metamodels' to analyse the full model parameter space and thus constrain sources of uncertainty and plausible retrodicted scenarios more eectively. This study applies the developed techniques to a case study focused on a specic set of coastal gullies found on the Isle of Wight, UK. A key factor in the evolution of these gullies are the relative balance between rates of cliff retreat (which reduces gully extent) and headwards incision caused by knickpoint migration (which increases gully extent). To inform the choice and parameterisation of the numerical model used in this research an empirical- conceptual model of gully evolution was initially developed. To provide a long-term context for the evolution of the gullies and to identify the relative importance of the various driving factors, the Holocene erosional history of the Isle of Wight gullies was then simulated using a LEM. In a preliminary set of simulations a 'traditional' (i.e. with no consideration of parameter uncertainty) retrodictive modelling approach was applied, in which driving variables were arbitrarily altered and observed and simulated landscape topographies compared, under various scenarios of imposed environmental change. These initial results revealed that the coastal gullies have been ephemeral in nature for much of the Holocene, only becoming semi-permanent once cli retreat rates fall below a critical threshold at 2500 cal. years BP. Next, in an attempt to constrain more detailed erosional histories and to explore the extent to which retrodicted interpretations of landscape change were confounded by uncertainty, a Central Composite Design (CCD) sampling technique was employed to sample variations in the model driving variables, enabling the trajectories of gully response to dierent combinations of the driving conditions to be modelled explicitly. In some of these simulations, where the ranges of bedrock erodibility (0:03-0:04m0:2a).
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:505745 |
| Date | January 2009 |
| Creators | Leyland, Julian |
| Contributors | Darby, Stephen |
| Publisher | University of Southampton |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
| Source | https://eprints.soton.ac.uk/69706/ |
Page generated in 0.0018 seconds