In recent years, a broader application of flood simulation models has been observed due to the steady advancement in numerical modelling and computational resources. Despite these advances, the major flood-prone regions worldwide have not benefitted from state-of-the-art inundation models. This limitation is primarily due to the unavailability of high-quality terrain data or the lack of accessibility of these datasets. Hence, enhancing terrain data quality holds greater significance for reliable flood simulation models. Global terrain datasets, such as the Shuttle Radar Topography Mission (SRTM), have played an essential role in flood inundation modelling in flood-prone regions of the developing world. However, the global DEMs, suitable for large floodplain inundation mapping, have limitations for applications to reach scale problems. This study introduces a novel bias correction procedure to reduce errors in the SRTM data and provide more accurate topographical information for flood inundation mapping, specifically at a reach scale. In addition, this study comprehensively evaluates various spatial sampling methods to determine their effectiveness in reducing the error and maximizing the fitness index for inundation estimates with a limited number of sample survey points. The framework is developed for the Brazos River, Texas, USA, where high-quality terrain datasets are available for experimenting with various spatial survey designs and benchmarking the flood inundation model. Our research has demonstrated a remarkable reduction of the vertical elevation error values and a more pronounced spatial correlation of elevation values in the modified DEM compared to the original DEM. The applicability of the modified DEM is validated for reach-scale flood inundation mapping using the HEC-RAS 2-D model for a major flood event in study reach. The proposed bias correction process is found to have significantly improved the flow connectivity, resulting in a more accurate inundation extent prediction. In addition, the spatial correlation characteristics of the error (semi-variogram analysis) are employed to simulate many statistically plausible topographies to assess the effects of uncertain topography on predicted flood extents and depth values. It is found that using an ensemble of the modified DEM simulated using semi-variograms further improves the inundation estimates and gives inundation depth values closer to a benchmark flood model.