Digital terrain data covering most regions in the United States is available in the USGS DEM format from the U.S. Geological Survey. For more information, see the Digital Elevation Model Standards documents available on the Nationalmap.gov website.
It is important to note the differences between raster datasets created from USGS 7.5-minute and 1-degree DEMs. The following table compares raster datasets created from these two sources.
|Raster dataset from a 7.5-minute DEM
|Raster dataset from a 1-degree DEM
7.5' x 7.5'
1° x 1°
12 km x 14 km
90 km x 110 km
1 to 60
Distance between elevation points in x and y
30 meters x 30 meters
3 arc seconds 75 meters x 75 meters (approx. in the United States)
Number of elevation points in x and y
Varies. The typical column counts range from 345 to 385, and row counts from 460 to 475.
Constant of 1201 x 1201.
Variation in ground distance between elevation point spacing across the raster dataset
No variation; the spacing is constant.
Varies with the latitude.
Raster datasets constructed from USGS 1-degree DEMs are not immediately suitable for volume, slope, or accurate visibility analysis because the x,y locations are measured in latitude and longitude, but the z-values are measured in meters. Consequently, the actual distance on the ground, represented by one Ground Unit, is not constant, and the Ground Units and the Surface ZUnits are not in the same units of measure.
Unlike USGS 7.5-minute DEMs that contain surface elevations at a constant spacing of 30 meters, 1-degree DEMs contain ground elevations at an interval of 3 arc seconds. The actual distance in ground units represented by 3 arc seconds varies with latitude even within the raster dataset. For example, in the vicinity of Southern California, at the top of the raster dataset, 3 arc seconds measure 76.86 meters in x and 92.36 meters in y. At the bottom of the raster dataset, 3 arc seconds measure 77.11 meters in x. In the region of the North Pole, the distance in x approaches zero.
The 7.5-minute USGS DEMs are a by-product of the orthophotomapping program. Some orthophotos were produced using semiautomatic profiling devices that yield elevations along profiles. "When these heights are resampled to a 30-meter square grid, the interpolation technique produces estimated elevations which have a much stronger spatial autocorrelation along the direction parallel to the profiles than in the orthogonal direction. Many images produced from these models have clearly visible 'stripes' parallel with the scanning direction of the profiler." For a reference on this topic, see Mark, D., "Automated Detection of Drainage Networks from Digital Elevation Models," in Proceedings of the Sixth International Symposium on Computer Assisted Cartography (Auto-Carto VI), Vol. 2, 1983, pp 288–298.
You can check the data for this effect by creating a hillshade of the raster dataset with the Hillshade tool and displaying it. If any striping is visible, before using the elevation raster any further, you may wish to first smooth out the values with the Filter tool and specifying the Low filter type.
To make a surface model suitable for volume, slope, and accurate visibility analysis, the ground units in the 1-degree USGS DEM must be projected to a nonangular unit of measure, such as UTM or Lambert. In addition, the surface ZUnits should be expressed in the same units of measure as the Ground Units.
Use the Define Projection tool or Project Raster tool (with a Bilinear or Cubic resampling technique) to convert a raster dataset to the desired projection or to change datums.
The WGS72 datum does not share a common transformation method with NAD27 or NAD83, so an equivalent datum must be used.
The Project tool does not perform ZUnits conversion. Use the Times tool to convert the ZUnits by multiplying by a factor (for example, raster * 0.3048).