Home | Business News | Browse by Publication | C | Cartography and Geographic Information Science

Lidar elevation data for surface hydrologic modeling: resolution and representation issues.

Article Excerpt
Introduction

The past decade has seen a steady increase in quantity, precision, and accuracy of digital spatial data products. High-resolution elevation datasets are among the most demanded products, and they are necessary for a wealth of important applications ranging from shoreline erosion to flood hazard mapping (Moore et al. 1991; Hutchinson and Gallant 1999). Lidar (light detection and ranging) technologies are emerging as a terrain data production technology of choice due to the published high accuracy of the results and a relatively low data collection cost. By early 2005, several high-resolution lidar-derived elevation products were freely available on the Internet, with spatial coverage including portions of the states of Louisiana, North Carolina, and Washington (Atlas: The Louisiana Statewide GIS 2004; North Carolina Floodplain Mapping Program 2004; Puget Sound Lidar Consortium 2004).

Lidar in the simplest description is the use of lasers to determine distance from the instrument to some target. This is not a new technology. The use of lasers for terrain profiling appears in literature as early as 1971 (Penny 1971), but more recent advances in airborne inertial navigation systems, global positioning systems, and digital data processing and storage capabilities have led to the development of lidar scanning systems for terrain mapping. Other ranging technologies such as radar (radio detection and ranging) and sonar (sound navigation ranging) make use of radio and sound waves, respectively, and operate on similar basic principles as lidar. This principle is that the distance to a target is determined by recording the time it takes for an actively emitted wave--laser light in the case of lidar--to travel from the laser to a target and back. If the three-dimensional location of the laser's source is known along with the direction of the projected beam, the location of the target in three-dimensional space can be calculated.

There are limits to the practical use of lidar that vary with the intended application. This paper considers potential issues--e.g., characterizing flow direction and power, identifying sub-basins in a watershed, and calculating variables such as upstream contributing area--and the challenges for their employment in the development of digital elevation models (DEMs) for surface hydrologic modeling applications. We consider in particular three challenges:

1. The inherent variability in sample size and accuracy with land cover. While all elevation surveys are more problematic in heavily vegetated areas, lidar-derived data are especially affected. Terrain mapping requires that laser pulses from an airborne platform reach the ground. Forest cover greater than 80 percent can result in only 10 percent of laser pulses reaching the ground (Cowen et al. 2000), which results in a low density of sample points, or possibly data voids, which must be interpolated from the surrounding area. Shrublands with multiple levels of vegetative cover can also confound the lidar post-processing algorithms (Hodgson et. al. 2003). In many landscapes, riparian areas are of primary hydrological interest in addition to being densely forested.

2. The sensitivity of particular hydrologic applications. A high-resolution, high-accuracy elevation dataset does not necessarily produce a highly reliable model of surface flow. The very high spatial resolution of lidar DEMs capture features that impact any standard surface hydrology model. While a DEM that includes a raised roadbed across a floodplain might be perfectly suitable for many applications (e.g., visibility analysis, draping aerial imagery), all flow-related derivatives will be grossly incorrect if the elevated feature is on pilings or if culverts are present. Workarounds such as intersecting United States Geological Survey (USGS) blue lines have their own limitations. We search for evidence of artifacts related to the presence of such features as railroad grades and elevated highways, and their impact on surface hydrology applications.

3. The role of spatial resolution. Does the resolution of lidar-based DEMs really provide more accurate and substantially improved results over medium-resolution DEMs such as the National Elevation Dataset (NED), which is the seamless elevation data product supplied by the National Map at 1 and, in some places, 1/3 are second resolution (Gesch et al. 2002)? Research on the effects of spatial scale and hydrologic parameters has suggested that, while DEM resolution is critical, little additional information is gained below 10 meters at the catchment level (Zhang and Montgomery 1994). We investigate the difference in common hydrologic measures and models calculated on lidar data and independently derived 30-meter resolution USGS data.

This paper considers each of these problems in turn for two watersheds in North Carolina with mutually distinctive topography. The following section describes the preprocessing required to produce comparable digital elevation models, and then presents the methods by which we assess the three challenges. Then we report on the results of the different lines of investigation. The paper concludes with a discussion about the implications of these results for hydrologic modeling in particular, and for DEM analysis in general.

Methods

Data Sources and Preprocessing

We chose portions of two watersheds in North Carolina for this study: the western portion of the Lower Neuse River basin, henceforth referred to as "Neuse," and the upper Fishing River basin, henceforth referred to as "Fishing" (Figure 1). Both areas had readily available lidar-derived elevation data, in addition to more conventional medium-resolution USGS DEM data. The Neuse landscape is one of gentle relief with low slopes, while the Fishing watershed is an upland area with relatively substantial elevation change. Figures 2 and 3 show the hill-shaded topography of both watersheds; these,...