set of tests was conducted. Although testing
Statistical correlations between
apparatus differ, this is in agreement with obser-
site characterization parameters
Statistical analyses were conducted on site char-
vations by Houston (1995) who explored relations
acterization and performance data to 1) explore
among cone resistance, water content, and soil
relationships between site characteristics/soil pa-
suction in the context of subgrade variability.
rameters measured using various testing tech-
Surprisingly poor correlations were observed
niques, 2) quantify site variability, and 3) investi-
between moisture contents determined by Vitel
gate the influence of site variability on test section
RF moisture probes and from small soil samples
performance. To investigate relationships between
collected in moisture tins and oven dried. Al-
parameters measured by various testing tech-
though one method measures volumetric water
niques, coefficients of correlations were deter-
content and the other gravimetric water content,
mined between every possible pair of parameters
correlations observed in other studies at CRREL
measured using methods outlined by Harr (1991).
to date have been good. A probable explanation is
Quick simple tests might be used in substitution
attributed to small rocks becoming lodged be-
for time-consuming, complex tests (or tests re-
tween Vitel probe tines. This is known to appre-
quiring missing or inaccessible equipment) in in-
ciably alter apparent water content, and was ob-
stances when coefficients of correlation are high
served during testing at Fort McCoy's wooded
(e.g., close to +1 or 1). Conversely, if a particular
trail on numerous occasions. Although expedient,
test is recommended for testing a soil before pro-
the Vitel moisture probe is not recommended for
ceeding with vehicle passage, information ob-
soils containing small rock fragments. The above
tained by substitute test equipment that had
relationships can be seen in the figures provided
shown low correlations. Included in this particu-
in Appendix A. Analyses of variance (ANOVA)
lar analysis were the following:
were also conducted using site characterization
parameters. Correlations were similar to those dis-
Initial percent coverage of untreated travel-
cussed using only pairs of parameters. Regres-
way with standing water,
sions developed were in general only minimally
Initial rut depths,
improved by inclusion of multiple parameters. It
Gravimetric water content,
is believed that tighter quality control of test meth-
Volumetric water content,
ods would yield improved results for all statisti-
CBR determined by the Clegg impact tester,
cal and variability analyses. Additional statistical
CBR of the uppermost 0.13-m (5-in.) layer
detail on Fort McCoy site characterization is pro-
determined by the DCP,
vided in Kestler (1996).
Depth at which CBR reaches a value of 10,
Static cone index corresponding to 0.15-m-
Influence of site variability
(6-in.-) thick layers,
on test section performance
Thaw depth,
To determine the influence of subgrade strength
Density.
variability on test section performance, a geo-
statistical variability analysis was conducted on
Correlation coefficients between centerline CBR
site characterization parameters. A geostatistical
determined by the Clegg and by the DCP were in
variogram shows variance of measurements made
the range of 0.7, and those between the Clegg and
as a function of separation distance. A brief expla-
cone index indicated by the static cone were
nation of the variogram function is provided in
slightly greater than 0.6. Correlation coefficients
Appendix C. For more detailed variogram de-
between Clegg CBR and water content were also
velopment, the reader is referred to the texts by
moderate (0.6) and in conformance with those
Journel and Huijbregts (1978) or Isaaks and
observed by others (Alkire 1986) and with unpub-
Strivastava (1989).
lished data from other CRREL site characteriza-
Geostatistical variograms were developed for
tion and variability testing (Kestler in prep.). A
several of the preconstruction site characteristics
moderate correlation coefficient of approximately
and post-trafficking rut depths (App. C). The
0.7 was determined between gravimetric water
analysis (Kestler 1996) indicated that a rank can-
content and CBR corresponding to the uppermost
not be assigned (based upon rut depths) to stabi-
0.13-m (5-in.) layer determined by the DCP. The
lizing techniques located at greater than two test
figure includes only points for which a complete
sections apart due to variations in the subgrade.
10
Previous Page
