Assessing resolution and source effects of digital elevation models on automated floodplain delineation: A case study from the Camp Creek Watershed, Missouri

Abstract

Digital elevation models (DEMs) have been widely used in automated floodplain modeling to determine floodplain boundaries. However, the effects of DEM resolution and data source on floodplain delineation are not well quantified. This paper presents a case study to assess these effects from the Camp Creek Watershed, Missouri, using two sets of DEMs. One is the Light Detection and Ranging (LiDAR) DEMs re-sampled from 1-m to 3, 5, 10, 15, and 30-m resolutions. The other is 5, 10, and 30-m DEMs obtained from the U.S. Geological Survey (USGS). Floodplain boundaries are delineated using a combination of hydrological, hydraulic and floodplain delineation models under the Federal Emergency Management Agency’s (FEMA) guideline. Model outputs including stream network, watershed and floodplain boundaries are compared to 1-m LiDAR DEM outputs (as the reference) to assess the uncertainty. Results indicate that re-sampled 3 or 5-m LiDAR DEMs produce similar streams and floodplain boundaries within 10% difference of the reference. In contrast, coarser LiDAR DEMs (such as the 10-m resolution) are more appropriate for watershed boundary delineation because higher DEM resolutions are likely more sensitive to minor topographic changes and may introduce erroneous boundaries. For different data sources, uncertainties introduced by USGS DEMs are much higher than LiDAR DEMs with a distinct relationship between uncertainties and DEM resolutions. Uncertainties of LiDAR DEMs consistently increase with decreasing resolutions, whereas similar levels of uncertainty are observed for different USGS DEM resolutions. This difference is probably due to the inherited difference in their original data source resolutions to make these two types of DEMs.