Modelling DEM data uncertainties for Monte Carlo Simulations of Ice Sheet Models

Abstract

For realistic modelling of digital elevation model (DEM) uncertainty, information on the amount and spatial configuration is needed. However, common DEM products are often distributed with global error figures at best. Where no higher accuracy reference data is available, assumptions have to be made about the spatial distribution of uncertainty, that are often unrealistic. In order to assess the impact of DEM uncertainty on the results of an ice sheet model (ISM) for an area where no higher accuracy reference data was available, we quantified DEM error of comparable regions with available reference data. Deriving good correlation of error magnitude and spatial configuration with DEM characteristics, these dependencies were incorporated into an uncertainty model containing both deterministic and stochastic components. The developed uncertainty model proved to reproduce amount and spatial correlation of DEM error well while producing uncertainty surfaces suitable for Monte Carlo Simulations (MCS). Applying the model to a DEM of Fennoscandia, a MCS was conducted using an ISM during the first 40ka of the Last Glacial Maximum (LGM). Results showed DEM uncertainty to have significant impact on model results during nucleation and retreat of the ice sheet.