Geoscience and Environment

Making DEM's from Maps


The U.S. Geological Survey has used paper topographic maps to produce DEM's. This is not the same as copying the map. Rather it is a process that extracts the contour data from the map and interpolates elevations between the contours. Anyone with a scanner and suitable software can do the same.

The IDRISI program provides several paths from a raster file of contours to a DEM and it is all explained well in the manual. However John Childs makes the whole process more interesting and provides lots of tips and examples. Nobody who has the IDRISI program will use the software he describes, but John does explain how to get contours from paper maps.


The method of scanning and pre-processing depends on the paper map. Aim to scan a bigger area than needed by at least 25% to ensure good results at the edges of the final (cropped) map. First georeference the map, noting the coordinates of the four corners and then export it as a BMP file. Everything except the contour lines has to be regarded as noise, unwanted marks on paper. John Childs recommends Paintshop Pro for this job. In the image shown above left there is a lot of noise that has to be removed. Also there are broken lines to be patched where colored features have been removed. Adjusting the contrast, brightness, and saturation of the image makes it easier to discriminate between contour lines and other features and thus reduces the amount of patching needed.

If the map is scanned at 600 dots per inch (24 dots per millimeter) and the scale of the map is 1:25,000, then one pixel represents one meter on the ground. The smaller contour lines are about 0.2 mm in width, about 5 pixels on the paper and 5 meters on the ground. Depending on the curvature, contour lines in the map vary in width from 4 to 7 pixels. Experiments with scanning a map at 200 dots per inch indicated that pixels would be lost when removing color from the map. So 600 dots per inch seems to be the minimum for some maps.

Depending on the printing technology a contour line will be one color or several. This may present a problem when removing color information from the map. Experimenting with a small piece of a map pays off in time savings. Making good decisions before removing colors will leave fewer broken lines to patch and fewer noise elements to remove.

In places the width of a contour line drops to one pixel or none and that means patching by hand. When patching the contour lines it does not matter if the final line is thicker than the original line, so long as the centerline is the same as the original centerline, because IDRISI uses the center of raster lines to make vector lines.

When the contour line image has been cleaned up and all the lines have been healed import the image into IDRISI using the georeferencing information noted before exporting. (Altenatively replace the RDC file of the imported image with a copy of the RDC file of the georeferenced image.) The rest of the process follows the IDRISI manual. Using the TIN route to the DEM produces a satisfying result.

The experiment I conducted was interesting and useful because my map was for an area outside the USA for which no DEM is available. However, this laborious method is not needed for the USA, because high quality DEMs of varying scale are available free of charge for the whole of the USA, including Mount Shasta. Soon DEMs from the Space Shuttle will be available free of charge for most of the world's land surfaces.


The URL of this site is [www.geoscience2000.info]

Author: Fred Colbourne