5.1 Procedures
5.1.1 Raymond stereomodel
To validate the observations made previously, model 2153 (see figure
1) was sampled three times at grid intervals of 20, 51 and 62 m.
Breaklines, hilltops and depressions were also measured. From each
grid sampling combined with breaklines, hilltops and depressions,
contour lines at 0.5 m interval were interpolated and plotted at 1:1
000 scale. Each contour plot was tested for accuracy by spot
heighting intersections of a drawn profile with the contour lines
using an analytical stereoplotter.
5.1.2 Crowsnest Pass stereomodel
This model was sampled twice at grid intervals of 7. and 12 m.
Breaklines, hilltops and depressions were also measured. From each
grid sampling combined with breaklines, hilltops and depressions,
contour lines at 1.0 m interval were interpolated and plotted at
1:3 000 scale. Each contour plot was tested for accuracy as with the
Raymond's.
5.2 Results and Discussion
The accuracy of the discrepancies of contour lines may be estimated as
follows /1/:
2 2 re of Soin 2
9 disc e e spec S AT g setup * a samp
For the Raymond stereomodel,
2 1/2
2 2 2
Saige ^ (0.15 + 0.07 + 0.03 + 0.03 )
s0.17 m
For the Crowsnest Pass stereomodel,
2 1/2
2 2 2
Cis ^ (0.31 - + 0.12 t 0.05 + 0.05)
a 0.34 m
From the RMS's of discrepancies of contour elevations, it is evident
that they all meet the required accuracies (see Table 5). From these,
it was concluded that the true optimum grid sampling interval for the
Raymond stereomodel was 62m, and for the Crowsnest Pass stereomodel
was 12m. Based upon these conclusions, sample profiles that were
representative of each model's roughness were identified. The range of
computed sampling intervals for each profile should include 62 or 12m
as the case may be. Differences of computed sampling intervals from
the true intervals were computed and summarized in tables 6 and 7.
Differences from 12m are within + 3m for long and rough profile 7 (see
Table 7). For long profiles, RF is the most conservative, while
-. 45%
