the basis for comparisons, program RF was developed, also by the 
author. Incidentally, the program was found to be capable of 
computing optimum sampling interval also. Programs LINEAR and RF are 
based on linear interpolation. For details about the four programs, 
see /1/. 
All of these programs were then tested with real data using 1:60 000 
wide-angle photography used for the 1:20 000 provincial topographic 
series. Results were compared relatively and absolutely from which 
guidelines for using the programs and a sampling strategy were 
formulated /1/. 
These guidelines and strategy are summarized as follows: 
a) Data acquisition of one or more sample profiles depending on 
the model's roughness. 
b) Estimation of the error budget for interpolating heights from 
the DEM data to be acquired. 
c) Running of program RF, given the sample profile(s) and error 
budget, to determine the roughness factor of each profile. 
d) Selection of the most suitable program to use, knowing the 
roughness factor, length and sample spacing of each profile. 
e) Running the selected program(s), given the sample profile(s) 
and error budget, to determine the optimum sampling 
interval(s). 
To test the feasibility of using these programs with large-scale 
photography, a similar experiment was undertaken. This paper deals 
with data, procedures and results of such an experiment. 
2. TEST DATA 
2.1 Raymond (1:4 000) 
This data was obtained from the photogrammetric control coverage for 
the town of Raymond located in southern Alberta, 30 miles north of the 
Alberta - Montana, U.S.A border. Sample profiles were measured along 
the north-south boundaries and bases of models: 2152, 2153 and 2154 
(see Figure 1). Sample spacing was 3 m with the use of an analytical 
stereoplotter. 
Profiles 152, 153, 154 and 155 were categorized as long profiles. 
Profiles 1521, 1531 and 1541 were categorized as short profiles, 
supplemented with subdivisions of the long ones: 152-1, 152-2, 153-1, 
153-2, 154-1, 154-2, 155-1 and 155-2. Profiles of varying sample 
spacings were obtained by deriving new profiles of 6 and 9 m sample 
spacings from the long ones. 
2.2 Crowsnest Pass (1:8 000) 
To obtain rough terrain profiles, one stereomodel was chosen from the 
photogrammetric control coverage for the municipality of Crowsnest 
Pass located near the Rocky Mountains, 50 miles north of the 
Alberta-Montana, U.S.A. border. Sample profiles were measured along 
the north-south edges and its base at 4 m sample spacing with an 
analytical stereoplotter (see Figure 2). 
- 41 -