Processing Methodology

• The one-degree square is not a rectangle in the UTM projection
• 30-meter DEMs are not yet available for all quadrangles
• The DEMs from the USGS are not all in the same projection
• A few of the source DEMs appear to contain errors

The meridians of longitude and parallels of latitude which define the one-degree square become, in general, curved lines when projected into UTM coordinates. Accordingly, the smallest bounding rectangle in UTM coordinates which includes the entire one-degree square will extend into portions of the adjacent one-degree squares. In creating this dataset, therefore, all DEMs which covered any part of the bounding UTM rectangle were identified, and the portions of these DEMs within the bounding rectangle were included in the final concatenated DEM.

The USGS is still in the process of compiling 30-meter DEMs for each 7.5 minute map quadrangle, and our database contains mainly DEMs for the Commonwealth of Pennsylvania and the Susquehanna River Basin. For grid points within quadrangles for which 30-meter DEMs were not available, and for a few grid points on the boundaries between pairs of adjacent 30-meter DEMs which were included in neither DEM (typically, because the DEMs were compiled using different reference datums), the elevation values were estimated by bilinear interpolation from the USGS 3-arcsecond DEMs. Since the 3-arcsecond DEMs have horizontal grid spacings on the order of 90 meters, the amount of terrain detail provided by the interpolated data is significantly less than that provided by the 30-meter DEMs.

For the one-degree square extending from 75° to 74° west longitude and 43° to 42° north latitude, 47 of the 64 quads within the one-degree square had to be filled by interpolation from the 3-arcsecond DEMs.

Most of the 30-meter DEMs are in the NAD27 reference datum, but a few are already in the new NAD83 datum. The 3-arcsecond DEMs are not in UTM coordinates, and were compiled using the WGS72 reference spheroid, which differs slightly from the spheroid used for NAD83. In addition, for one-degree squares on the edge of a UTM zone, some of the DEMs needed to cover the UTM bounding rectangle were located in the adjacent UTM zone, and needed to be converted. Accordingly, most DEMs had to be resampled to NAD83 in the correct UTM zone before they could be concatenated. The 30-meter DEMs were resampled using nearest-neighbor interpolation, which has the effect of assigning an elevation to each grid point which may actually belong to a point up to 15 meters away in both the east-west and north-south directions. The 3-arcsecond DEMs were resampled to the 30-meter grid using bilinear interpolation; this tends to smooth out small-scale changes in topography.

Conversion between spheroids and datums can be somewhat complicated. Because the LAS software used to perform the resampling does not fully support datum and spheroid conversion, the horizontal offsets between points having the same longitude and latitude in NAD27 and NAD83 at the center of each DEM were estimated by using the PROJ.4 software package developed by Gerald I. Evenden of the USGS (available for anonymous ftp download from kai.er.usgs.gov). Since these offsets differ by less than 1 meter over the region covered by any DEM, the offsets at the center were applied to all points of the DEM. Similarly, the offsets between the WGS72 spheroid and the WGS84 spheroid used for NAD83 were computed at the centers of the 3-arcsecond DEMs using the conversion formulae given in Table H1 of the USGS Data Users Guide 5, "Digital Elevation Models", 1990.

The concatenated DEM may contain erroneous values resulting from errors in the individual DEMs from which it was assembled. In some cases, these errors have not yet been corrected by the USGS; in other cases, there may be a corrected version of the USGS DEM but it has not yet been added to our local database. Clearly visible errors in some DEMs include pits and level regions which do not actually exist. There are sometimes small discontinuities in elevation across the boundary between adjacent 7.5-minute quadrangles, and sometimes changes in the amount of fine-scale structure visible in the DEM. Also, it appears that a few of the 30-meter DEMs contain erroneously low values at the edge of the quadrangle.