Can Landscape Evolution Models (LEMs) be used to reconstruct palaeo-climate and sea-level histories?

Reconstruction of palaeo-environmental conditions over long time periods is notoriously difficult, especially where there are limited or no proxy records from which to extract data. Application of landscape evolution models (LEMs) for palaeo-environmental reconstruction involves hindcast modeling, in which simulation scenarios are configured with specific model variables and parameters chosen to reflect a specific hypothesis of environmental change. In this form of modeling, the environmental time series utilized are considered credible when modeled and observed landscape metrics converge. Herein we account for the uncertainties involved in evaluating the degree to which the model simulations and observations converge using Monte Carlo analysis of reduced complexity `metamodels'. The technique is applied to a case study focused on a specific set of gullies found on the southwest coast of the Isle of Wight, UK. A key factor controlling the Holocene evolution of these coastal gullies is the balance between rates of sea-cliff retreat (driven by sea-level rise) and headwards incision caused by knickpoint migration (driven by the rate of runoff). We simulate these processes using a version of the GOLEM model that has been modified to represent sea-cliff retreat. A Central Composite Design (CCD) sampling technique was employed, enabling the trajectories of gully response to different combinations of driving conditions to be modeled explicitly. In some of these simulations, where the range of bedrock erodibility (0.03 to 0.04 m^0.2 a^-1) and rate of sea-level change (0.005 to 0.0059 m a^-1) is tightly constrained, modeled gully forms conform closely to those observed in reality, enabling a suite of climate and sea-level change scenarios which plausibly explain the Holocene evolution of the Isle of Wight gullies to be identified.