DETERMINATION OF OPTIMUM SAMPLING INTERVAL IN GRID SAMPLING OF DIGITAL 
ELEVATION MODELS FOR LARGE-SCALE APPLICATION 
A.E. Balce 
Survey Control Branch 
Alberta Bureau of Surveying and Mapping 
Alberta Forestry 
Edmonton, Alberta, Canada 
ABSTRACT 
Four computer programs for determining the optimum sampling interval 
in grid sampling of digital elevation models (DEM) by photogrammetric 
methods, were tested with real data using large-scale photography and 
results compared. These programs were SPECTRA, LOGKV, LINEAR and RF. 
They were tested and compared using sample profiles obtained using an 
analytical stereoplotter from large-scale photography: 1:4 000 with 
flat terrain, and 1:8 000 with rough terrain. For the purpose of 
having absolute comparisons, one stereomodel from each photo scale was 
sampled completely up to three times at various grid sampling 
intervals determined by the programs. Together with the measurements 
of breaklines, hilltops and depressions, contour lines were 
interpolated and plotted from each grid sampling. Contour lines of 
each plot were then tested for accuracy using an analytical 
stereoplotter. From the results, guidelines for using the programs 
were formulated and are recommended. 
1. BACKGROUND 
When Alberta Energy and Natural Resources (now Alberta Forestry) 
embarked on acquisition of digital elevation models required basically 
for the new provincial topographic series at a scale of 1:20 000 with 
10 m contour interval in early 1984, the department and the private 
mapping companies were both uncertain as to the optimum grid sampling 
interval. Details of this project are presented in /3/. 
In an effort to learn from experts regarding DEM data acquisition, 
processing, databanking and applications, ABSM held a DEM workshop in 
Edmonton on October 10-12, 1984 /2/. The workshop was organized and 
chaired by Mr. M.A.G. Toomey, Mapping Branch, ABSM. It was attended 
by DEM experts from Canada, United States of America, Federal Republic 
of Germany, Austria and Switzerland, in addition to local mapping 
contractors and departmental staff engaged in DEM. 
After the workshop, one of the participants Dr. D. Fritsch, Technical 
University of Munich, F.R.G. sent us the source code listing of pro- 
gram SPECTRA which is based on Fourier transformation. About the same 
time, Dr. K. Kubik, The Ohio State University, U.S.A stopped over in 
Edmonton from Denmark while on his way to Australia. In two days he 
introduced and programmed LOGKV which is based on self-similarity 
concept. Both programs SPECTRA and LOGKV were tested and compared 
using real data. However, results appeared to be inconsistent. 
Consequently, a new program was developed called LINEAR, by the 
author. To determine the roughness factors of the profiles, used as