The data points were first imported to ARC/INFO and converted to a TIN (triangular irregular network). Triangles with an edge length greater than 500 meters were selected, and this set was edited by hand. These triangles were used to create a mask within which interpolation between survey points would not be used at all.

In places which the DNR hydrology coverage showed as water, the USGS DEM was given a weight equal to the longest triangle edge divided by five, plus fifty. For the remaining area, a weight was assigned to each 10-meter grid cell, based on the length of the longest edge of the principle tringle in that grid cell.

The USGS data was given a weight equal to the longest edge (in meters) divided by five. The final DEM is a weighted average of the USGS 10-meter DEMs and the DEM derived directly from the TIN, with the weight given to the TIN approaching zero as the length of triangle edges approaches 500 meters. The red areas give zero weight to the survey data and the bluest areas give full weight to survey data. You see that survey data is densest near roads.

The resulting DEM is more accurate than the USGS DEM, and contains fewer artifacts. This DEM can be merged with the USGS 10-meter DEM to form a seamless product.

Values range from 18.5 meters higher (bluest) than the USGS DEM to 38.9 meters lower (reddest).

This is an image of the new DEM with the colors repeating in a cycle of 1.6 meters. It shows a smooth transition between elevations in most places.

This is a similar image based on the USGS DEM. Many artifact can be seen here.

And here is the data as an ARC/INFO export file. The Z values are integer decimeters. The projection is UTM zone 10, NAD27. The e00 file is 127155958 bytes, zipped to 6503389 bytes.

Web Curator: Harvey Greenberg