a. Sample test grid.
Variance Generally
Decreases as
Separation Distance
Same for all Points
Increases
with a Separation
Distance of 20 ft
Variance of Difference
between Values at all
Points Separated by 10 ft
0
10
20
30
40
50
Separation Distance (ft)
b. Variogram development.
Figure 4. Test grid and associated variogram development.
ences between test values at pairs of test points
correlated; those spaced farther than 150 ft are
separated by such a small distance will be mini-
independent of each other. While the variogram
mal. This can be repeated for larger separation
should define test point separation distances as
distances. For this particular test grid, the next
outlined here, this analysis should remain invis-
closest spacing is 14.1 ft. The variance continues
ible to the typical user.
to increase up to a certain separation distance at
There are currently many geostatistical soft-
which it levels out as shown in a geostatistical
ware packages both available for purchase and in
semivariogram (Fig. 4b). This is the distance
the public domain. They all analyze data at one
beyond which the values (in this case, modulus
point in time. We plan to modify an existing
or deflection) are no longer auto-correlated. Fig-
shareware package to continually update the
ure 5 shows the variogram corresponding to
optimal distance to the next FWD test point, as
FWD data at a test cell at the Minnesota Road
the data are collected, based on all previous data
Research Program (Mn/ROAD) (Kestler et al.
collected on that pavement during that test ses-
1994). Points closer than approximately 150 ft are
sion. As pavement strength variability increases,
5
Previous Page
