Evaluation of high quality topographic data for geomorphological and flood impact studies in upland area: North York Moors, UK

Abstract

A flash flood on 19th June 2005 caused more than one hundred landslides in the North-western North York Moors uplands, UK. This project aims to 1) assess digital elevation models (DEMs) in terms of statistical terrain analysis and 2) explore the sensitivity of a 2D FLOWMAP model response to DEMs input data. A variety of topographic data were acquired, generated and processed. These included high-resolution aerial photographs, Ordnance Survey (OS) DEMs, topographic maps, InSAR DEMs, LiDAR data and ground survey data. These DEMs of different horizontal and vertical resolutions were analysed through key topographic parameters calculated using three different software packages. Key topographic attributes such as slope, aspect, profile curvature and the Topographic Wetness Index (TWI) were studied. Results demonstrate that DEMs from different sources or at different resolutions provide different representations of topographic parameters especially in areas where large topographic changes take place. Algorithms used in different packages also had an effect. Degradation in the representation of topographic information is larger between 10 m and 50 m DEMs than between 5 m and 10 m DEMs. Finer resolution and smaller filter size have the same type of impact on slope and aspect. In addition, DEMs at finer horizontal resolutions have smaller minimum profile curvatures and larger maximum values and standard deviations in profile curvature. The TWI is more sensitive to the horizontal resolution than DEM data source and finer DEMs calculate smaller minimum and mean TWI and larger maximum TWI and standard deviations. Modelled hydrological responses are sensitive to both DEM resolution and its data source. Model showed different results when using 5 m LİDAR DEM and 5 m InSAR DEM of the same area, which meant DEM source had impacts on modelling These differences reduced with a larger magnitude flooding. Producing a better representative surface model from the LİDAR data has much larger impact on model response than adjusting a constant roughness coefficient.