Understanding the effects of Digital Elevation Model resolution in urban fluvial flood modelling

Abstract

Listen

Translate article

With the extensive use of 2D flood models, the resolution and quality of Digital Elevation Models (DEMs) have come under greater focus especially in urban hydrology. One of the major research areas, in this regard, is the effect of DEM resolution on flood modelling. This study first investigates the root causes of the impact of DEM resolution on urban fluvial flood modelling outputs using DEMs with grid resolutions ranging from 1 m to 50 m. The study then investigates how DEM resolution affects the definition and characterisation of the river channel and the consequences of this for the modelled results. For this purpose, a separate set of merged DEMs was generated where the river channel as defined by the 1 m resolution DEM is merged with coarser resolution DEMs. Data obtained during the flood event caused by Storm Desmond (2015) in Cockermouth (Cumbria, UK) was used for this study. The HEC-RAS 2D model was used for all of the simulations. The benchmark model obtained with the 1 m resolution DEM was calibrated using measured water levels at two locations within the rivers. Results show that there is a 30% increase in flood extent from 58.9 ha to 79.0 ha and a 150% increase in mean flood depth from 1.74 m to 4.30 m when the resolution reduces from a 1 m grid to a 50 m grid. The main reason for this is the increasing lack of definition of the river channel with an associated reduction in the estimated depth of the river resulting in reduced river channel conveyance. This then leads to an increase in the flood extent and depth especially in the immediate vicinity of the river. This effect is amplified when the DEM grid size is greater than the river width. When the 1 m resolution DEM for the river channel is used in conjunction with coarser resolution DEMs for the surrounding areas (merged DEMs), there is a significant improvement in the agreement between the modelled and the reference case (obtained from the benchmark model) flood extents and depths. The use of merged DEMs reduces the error in mean flood depth from 90% to 4% and reduces the overall RMSE in flood depths from 2.6 m to 0.9 m at 30 m resolution. The use of merged DEMs, where a higher resolution DEM is used to characterise the river channel in conjunction with, for example, a 30 m resolution DEM (e.g., the freely available NASA Shuttle Radar Topography Mission DEMs) for the wider area could be a cost-effective solution for locations where higher resolution DEMs may not be available.