Figure 6. Details of rut surface profile measured with the profile meter.
points, out of over 20,000, as statistical outliers
and excluded them from further analysis.
To determine whether significant changes in
rut profiles occurred over time, we calculated the
standard deviation of the 145-pin height readings
of each profile for each date. We used two-way
analysis of variance (ANOVA) with the Decem-
ber 1995 standard deviations to compare initial
differences between the two sites and between
straight and turning ruts. We evaluated changes
in profile smoothness over time using nonpara-
metric tests including Friedman's two-way anal-
ysis of variance, the KruskalWallis test, and the
KolmogorovSmirnov test. We selected nonpara-
metric statistics to relax classical assumptions
a. Frost gauge.
about spatial and temporal independence of the
data and about the shape of the sample distribu-
tions. All statistics were calculated using Systat
7.01 (SPSS Inc. 1997b).
We measured snow depth and used a visual
frost gauge (Schellekens and Williams 1993) to esti-
mate frost depth at each site during the 199596
winter to establish baseline values for sites E and
C (Fig. 7).
On 13 May 1996 we measured saturated
dry-track locations at both sites (Fig. 4). We sam-
pled compacted rut soil and adjacent, uncom-
b. Reading frost guage.
pacted soil lying within 1 m of the center of ruts.
Figure 7. Frost gauge and frost gauge being read. Depth
We chose this distance because our initial mea-
of freezing is indicated by a change in color.
surements showed the zone impacted during tank