This work has three main goals. First of all, it develops a framework by which two digital elevation model (DEM) data sets for the same area may be compared. The framework is implemented in the C programming language. Secondly, it tests this method using study DEMs for areas ranging across southern California. Finally, it identifies areas in which the results of comparisons can be used to evaluate or model the quality of digital elevation data. The next paragraphs identify the background for the present research, and the section finishes by providing a more detailed outline of the remainder of this work.

DEMs are important for a variety of applications, including hydrology, soils, land use planning, flood and landslide control (Petrie & Kennie, 1990). The ability of the DEM to accurately model the terrain can be very important for an application (Fisher 1991; Zhang & Montgomery, 1994; Weibel & Brändli, 1995; Ehlschlaeger & Shortridge, 1996). This research falls generally into the area of DEM accuracy. A variety of directions have been taken by researchers interested in assessing the accuracy of either a particular data production method or a specific elevation data file. The most straightforward method has been to compare accurate field elevation measures with DEM estimates; this has been accomplished in several studies (Shearer, 1990; Monckton, 1994; Bolstad & Stowe, 1994). Frequently, however, survey-quality terrain measurements are not available for an area of interest. In such cases, researchers have three general options.

One option is to make assumptions about intrinsic spatial qualities of landscape and identify deviations from these assumptions as indications of error. For example, the potential for fractal properties of terrain surfaces has inspired research into the detection of artifacts in DEMs using a fractional brownian motion model (Polidori et al., 1991; Brown & Bara, 1994). A second option is production-oriented; by studying problems in the methods used to generate digital elevation data, likely artifacts resulting from the methods can be identified. Robinson (1994) reviews the generation of DEMs from contours. Common conversion routines suffer from a variety of problems relating to contour configuration. Earlier research compared several common contour interpolation algorithms to identify differences and problems (Clarke et al., 1982). Carter (1989) explored individual DEMs produced using different methods to identify artifacts. He found that particular error types tended to be symptomatic of particular production methods.

The third option is to compare independently derived terrain data of the same location. Isaacson & Ripple (1990) compared collocated DEMs of the same type that this research uses (USGS 1:250,000 and 7½’ DEMs). They performed a linear regression on a sample of collocated points within the DEM and determined that the best fit was not significantly different from a 1:1 line with an intercept at zero. The standard error for the regression was 31 meters. Greater differences were found in the relationship between slope and aspect. They found no sign of artifacts in the 1:250,000 DEM they were using. Guth (1992) compared SPOT-derived 30 meter data with a USGS 30 meter DEM. Distribution statistics were very comparable, but patterns were identified for locations with large differences between the two data sets. Larger differences were found to be correlated with steep slopes and rugged terrain. Similar analysis was carried out for an area with two independently derived USGS 30 meter DEMs. Again, areas with large differences were concentrated in high slope portions of the study area.

The research reported on here incorporates elements from the second and third options indicated above. Section II provides background on the USGS DEM formats used in this work. Production methods for the data sets are indicated; previous research indicates the impact data production can have on terrain representation. Locating the elevation locations in space is critical to making a comparison. This section covers this topic as well. A significant comparison issue occurs when the data are not coincident in space. Due to the representation schema developed in Section II, this will almost certainly hold true in every case. Section III covers this issue in depth. It also discusses the ability of the code to transform one dataset horizontally or vertically before making the comparison. This allows one to investigate the role of offsets in difference between the two DEMs. A synopsis of the data sets used in the study area is provided in Section IV. Section V reports on the comparison of the study data sets. Univariate analysis isolates systematic artifacts in one of the DEM formats, while statistics for the comparison of the two reveal more complex patterns. Finally, the ability of the code to transform the DEM position for the improvement of the relationship is evaluated. Section VI places this work within the wider context of accuracy research. Additional directions for research are especially highlighted in this section, as well. Finally, Section VII offers some concluding comments on this work.