Effects of DEM sources on hydrologic applications

https://doi.org/10.1016/j.compenvurbsys.2009.11.002Get rights and content

Abstract

Digital elevation model (DEM) data have been used to derive hydrological features which serve as inputs to various models. Currently, elevation data are available from several major sources and at different spatial resolutions: the National Elevation Dataset (NED), Shuttle Radar Topography Mission (SRTM) data, and the Light Detection and Ranging (LIDAR) data. This article shows the variability of analysis results from these data sources and of different scales. A standard algorithm in GIS was used to extract river networks from these DEM data for part of Kansas. Extracted networks were compared with the one derived from the National Hydrography Dataset (NHD). Flood simulation results obtained from various sets of DEM data were also compared. Using higher resolution DEMs can derive more accurate river networks, but highest resolution data may not necessarily offer the best results, depending on the resolutions selected to compare the extracted networks. DEM spatial resolution may have minor impacts on flood simulation results, but inundation areas from flood simulations vary significantly across different DEM data sources. Results highlight the caveats on using DEM-derived river network data for hydrologic applications and the difficulties in reconciling differences among elevation data from various sources and of different resolutions.

Introduction

Representing surface topography accurately and effectively is critical to support various types of inquires in geography and geosciences. A continuous depiction of surface topography is provided by the US Geological Survey’s (USGS) Digital Elevation Model (DEM) with regularly spaced elevation data. Although equating elevation data with DEM is a misnomer, most scientists seem to accept the general usage that DEM refers to digital data representing topography.

Two directions of development in the use of DEMs are relevant to this article. Because of the general availability of DEMs, scientists have been increasingly using DEMs for various applications, such as flood control and hazard mapping (e.g., Jensen, 1991, Wise, 2000). Many algorithms have been developed to derive basic topographic characteristics or features from DEMs. Such algorithms include extracting river networks (e.g., Mark, 1984) and delineating watersheds (e.g., Band, 1986). Basic geomorphic or topographic attributes extracted from DEMs often serve as inputs to other models. Thus, DEMs and related algorithms provide the foundations of scientific inquiries related to environment and topography.

Parallel to the increasing use of DEMs is the expansion of DEM sources. USGS used to be the only one providing DEM in the US – the National Elevation Dataset (NED). Currently, another source of DEMs is the Shuttle Radar Topography Mission (SRTM) data. Many local jurisdictions also have acquired the Light Detection and Ranging (LIDAR) data, an increasingly important data source for high resolution DEMs. These different sources of DEMs offer different levels of spatial resolution. Most data in the USGS NED are at 30-m resolution. SRTM data have variable resolutions, although the 30-m data are the default standard products. Resolutions of LIDAR data are in general higher than other DEM sources.

A major use of DEM data in Earth Science is in hydrologic analysis. However, many hydrologic studies fail to provide consistent results. Some studies attribute the differences to factors such as data sources, spatial resolutions of DEMs, and the adopted algorithms in analyzing the data (e.g., Baker et al., 2006, Gyasi-Agyei et al., 1995). However, different data sources offer different spatial resolutions and levels of accuracy, both horizontal (or planimetric or positional) and vertical. Therefore, examining the roles of DEM sources and implicitly their associated accuracy characteristics in affecting hydrological modeling are necessary. Such studies may provide some insights and guidelines on how DEMs should be selected for hydrological modeling.

Therefore, the primary objective of this study is to examine whether different sources of DEMs, while controlling for their spatial resolutions, will yield consistent results in hydrological models. As DEMs are available from different sources, some sources offer multiple resolutions. Even if the source offers a single resolution, multiple lower resolution data can be derived from resampling if the original data have sufficiently high resolution. Therefore, another objective is to explore how spatial resolutions of DEMs, from the same data source and across different sources, affect hydrologic studies. To demonstrate the sensitivities of using different DEM sources, we chose two hydrologic applications of delineating drainage networks and simulating flood events, using DEMs for part of the Kansas River area. DEMs of various resolutions are derived from NED, LIDAR and SRTM data. River networks were extracted from these multiple resolution DEM datasets using a standard GIS algorithm. The extracted river networks were compared with the one provided by the relatively high resolution (1:24,000 or better) National Hydrography Dataset (NHD), which may be regarded as a reasonably accurate data source. Second, flood events were simulated using each DEM. Flooding results across different DEMs were compared. Hence, the study examines the effects of horizontal resolution of DEMs, data sources and their associated vertical accuracy levels on hydrologic applications.

Section snippets

Literature review

SRTM, which was completed in 2000, provides the first high resolution DEM data of near-global scale (Farr & Kobrick, 2001). The standard data products are produced by resampling the raw data to approximately 30-m resolution, comparable to the resolution of USGS NED data (Farr et al., 2007). Such global-scale coverage enables terrain studies of extensive areas and provides an alternative source of DEM for the US and other countries (Sanders, 2007). Featured with high spatial resolution and

Methodology

We have identified several major sources of DEM data for our study area (detail below): the 10-m and 30-m DEMs from USGS NED, the 30-m DEM from NASA/JPL SRTM, and 2-m LIDAR data from a local jurisdiction. The 2-m LIDAR are resampled to 10-m and 30-m resolutions for comparisons. Haile and Rientjes (2005) concluded that methods of resampling do not influence the quality of resampled DEMs significantly. We used the nearest neighbor method, one of the more popular methods for resampling. These

River network extraction

River networks were first extracted from each DEM and they are shown in Fig. 4. While the extracted networks may include elevation information, the NHD data do not. Therefore, the focus is on positional accuracy, comparing the central lines of the extracted river networks with those from the NHD data. Visual assessment shows that all DEM-derived networks provide relatively accurate depictions of the main streams, but not too accurate for the lower order streams. These lower order streams are

Conclusions

In this study, we evaluate how different sources of DEM data may affect the outcomes of two hydrologic applications, namely river network extraction and flood simulation. Among the sources we used, LIDAR 2-m data are expected to be the most accurate and SRTM data are suggested to be the least reliable according to the literature. To a large degree, our findings agree with the literature. High resolution DEM derived from LIDAR offers superior results in extracting river networks, but only if the

Acknowledgements

The authors are extremely grateful to the detailed and highly constructive comments provided by the anonymous referees. Their insights and suggested directions help improve the quality of this paper tremendously. The assistance provided by the Editor of CEUS during the review process is also greatly appreciated.

References (51)

  • S. Wu et al.

    A study on DEM-derived primary topographic attributes for hydrologic applications: Sensitivity to elevation data resolution

    Applied Geography

    (2008)
  • K.F. Adkins et al.

    Accuracy assessment of elevation data sets using the Global Positioning System

    Photogrammetric Engineering and Remote Sensing

    (1994)
  • M.E. Baker et al.

    Comparison of automated watershed delineations: Effects on land cover areas, percentages, and relationships to nutrient discharge

    Photogrammetric Engineering and Remote Sensing

    (2006)
  • L.E. Band

    Topographic partition of watersheds with digital elevation models

    Water Resources Research

    (1986)
  • L. Bian et al.

    Scale dependencies of vegetation and topography in a mountainous environment of Montana

    The Professional Geographer

    (1993)
  • Brady, L. L., Grisafe, D. A., McCauley, J. R., Ohlmacher, G. C., Quinodoz, H. A. M., & Nelson, K. A. (1998). The Kansas...
  • B.H. Carlisle

    Modeling the spatial distribution of DEM error

    Transactions in GIS

    (2005)
  • S.-M. Cho et al.

    Sensitivity considerations when modeling hydrologic processes with digital elevation model

    Journal of the American Water Resources Association

    (2001)
  • S.J. Dark et al.

    The modifiable areal unit problem (MAUP) in physical geography

    Progress in Physical Geography

    (2007)
  • M. Di Luzio et al.

    Integration of watershed tools and SWAT model into BASINS

    Journal of American Water Resource Association

    (2002)
  • Environmental Systems Research Institute (ESRI) (2003). ArcGIS Spatial Analyst. Redlands, CA, USA:...
  • M. Fabris et al.

    Automated DEM extraction in digital aerial photogrammetry: Precisions and validation for mass movement monitoring

    Annals of Geophysics

    (2005)
  • T. Farr et al.

    The shuttle radar topography mission

    American Geophysical Union EOS

    (2001)
  • T. Farr et al.

    The shuttle radar topography mission

    Reviews of Geophysics

    (2007)
  • A.S. Fotheringham

    Trends in quantitative methods I: Stressing the local

    Progress in Human Geography

    (1997)
  • Cited by (165)

    • A continuous digital elevation representation model for DEM super-resolution

      2024, ISPRS Journal of Photogrammetry and Remote Sensing
    • Effect of DEM resolution on assessing hydrological connectivity in tropical and semi-arid basins of central Mexico

      2022, Journal of Hydrology
      Citation Excerpt :

      For instance, in our study, the highest DEM resolutions can produce larger or smaller areas and perimeters and vice versa. This inconsistent change relationship between these basin characteristics has also been reported by Li and Wong (2010) with DEM spatial resolutions < 30 m. In Site 1, the largest perimeter and area were obtained with ALOS, the highest resolution DEM in this basin.

    View all citing articles on Scopus
    View full text