A GIS-based approach for digital elevation models (DEM) correction to derive bathymetric data with improved conveyance

Digital Elevation Model (DEM) is an important input for topography in modeling floods, especially for remote and inaccessible regions. However, DEMs often lack in accuracy to properly depict the ground elevation on river and other water bodies. Uncertainty in DEM may propagate through the simulation and thereby challenge the reliability of the flood modeling results. The objective of this research is to present a GIS-based DEM correction technique by incorporating conveyance and predicting thalweg location (deepest point along a cross section) along rivers for achieving improved accuracy of hydrodynamic simulation. To demonstrate this technique, a 1/3 arc second National Elevation Dataset (NED) DEM for a 60 km section of the Cumberland River near Nashville was adjusted to create a modified DEM. A 1D HEC-RAS model using surveyed topography was calibrated for a high flood event (May 2010). Two 1D HEC-RAS models were set up using the original NED DEM (or base DEM) and modified DEM respectively. These two models were then simulated with calibrated roughness factors to validate for intermediate flood (2003), high flood (May 2010) and low flood (May 2013) events. The corresponding error in flood depths, discharges and flood inundation extents for both models were then estimated with respect to observed data. It was found that the model with base DEM was stable at very high stage but failed during low and intermediate stages. The applicability of base DEM was also limited for any event above 127 m and 3000 m3/s with specific biasness. The model using modified DEM could be used for simulating large array of flood events. The root-mean-square error (RMSE) for simulated stage using modified DEM for 2003, 2010 and 2013 flood events with the observed stage were 0.86 m, 0.23 m and 0.52 m respectively. Comparison of simulated flood map for May 2010 flood event using Modified and base DEM with observed flood extent showed error of 2.66% (overestimation) and 13.38% (overestimation) respectively. This application of DEM correction technique for the study area showed significant improvement in the quality of DEM data with corresponding increase of the HEC-RAS model accuracy. This technique could be helpful for inaccessible areas where no credible terrain data is available, except for remotely produced DEMs.