Geosynthetic Subgrade Stabilization - Field 
Testing and Design Method Calibration
Authors: Eli V. Cuelho and Steve W. Perkins
NOTICE: this is the author’s version of a work that was accepted for publication in Transportation 
Geotechnics. Changes resulting from the publishing process, such as peer review, editing, 
corrections, structural formatting, and other quality control mechanisms may not be reflected in 
this document. Changes may have been made to this work since it was submitted for publication. 
A definitive version was subsequently published in Transportation Geotechnics, [VOL# 10 (March 
2017)] DOI: 10.1016/j.trgeo.2016.10.002.
Cuelho, Eli V., and Steve W. Perkins. "Geosynthetic Subgrade Stabilization - Field Testing and 
Design Method Calibration." Transportation Geotechnics 10 (March 2017): 22–34. doi:10.1016/
j.trgeo.2016.10.002.
Made available through Montana State University’s ScholarWorks 
scholarworks.montana.edu 
Geosynthetic subgrade stabilization – Field testing and design 
method calibration
E.V. Cuelho a,⇑, S.W. Perkins b
aWestern Transportation Institute, Montana State University, PO Box 1742
USA
tabil
plica
e pav
the relative operational performance of geosynthetics used as subgrade stabilization, as
well as determine which material properties were most related to performance.
Unpaved test sections were c
of geogrids and geotextiles.
effect that subgrade strength
had on performance. Even th
good performance as subgra
formance was difficult to es
relevant tests to properly ch
longitudinal rut as the prima
ear regression analysis that
direction correlated best wit
the d
grid
thetic,
n of t
geosynthetics may provide a working platform so that
the base course aggregate layer can be properly con-
structed and overall rutting reduced. Geosynthetics are
stabilization) to
ly deposite
paration fu
ile the rei
ment function may be derived from both geotextil
geogrids; however, in certain circumstances, ge
may also offer separation (Maxwell et al., 2005). Su
stabilization is typically applicable for unpaved temporary
roads such as haul roads, or construction platforms to sup-
port permanent roads. These roads are generally character-
ized by low volumes of heavy vehicles that can tolerateFor low-volume roadways and temporary construction
platforms where excavation and replacement of inferior
subsoils may not be cost effective, soil stabilization using
used in these situations (i.e., subgrade
reinforce and/or separate poorer natural
from the crushed aggregate layer. The se
is primarily attributed to geotextiles, whd soils
nction
nforce-
es and
ogrids
bgradeBackground and introduction planar polymeric materials that have been extensivelyUsing this knowledge,
calibrated to make geo
property of the geosyn
ibration and verificatioonstructed using twelve geosynthetics consisting of a variety
Multiple control test sections were also built to evaluate the
, base course thickness, and/or presence of the geosynthetic
ough the geotextile materials used during this study showed
de stabilization, material properties associated with their per-
tablish due to the limited number of test sections and lack of
aracterize these types of materials for this application. Using
ry indicator of performance, it was determined through a lin-
the stiffness of the geogrid junctions in the cross-machine
h performance in this application and under these conditions.
esign equation associated with the Giroud–Han method was
junction stiffness in the cross-machine direction the primary
thereby replacing geogrid aperture stability modulus. The cal-
his method is described herein.materials. Full-scale test sections were constructed, trafficked and monitored to comparebMontana State University, 205 Cobleigh Hall, Bozeman, MT 59717,
Geogrids and geotextiles are used routinely to s
construction. Typical subgrade stabilization ap
unpaved low-volume roads, but can also includ
Abstract50, Bozeman, MT 59717, USA
ize weak subgrade soils during road
tions are temporary haul roads or
ed roads built on poorer foundation
deeper ruts. According to the National Highway Institute,
geosynthetic stabilization techniques used for these types
of roads are ‘‘one of the more important uses of geosyn-
thetics” (Holtz et al., 2008). Historically, geotextiles were
first used in these applications; however, geogrids are
more commonly used in recent years. The first design for
geotextile stabilization of unpaved roads was created in
the late 1970s by Steward et al. (1977) based on soil
mechanics theory and experimental data generated in the
laboratory and field. Since then, several alternative designs
for geogrid stabilization have also been created (Tingle and
Webster, 2003; Giroud and Han, 2004a; USCOE, 2003).
Limitations within each of these methods, lack of calibra-
tion for a wide variety of products, and a growing variety
of the types, strengths, and composition of geosynthetic
reinforcement products has introduced uncertainty in the
design of geosynthetic-reinforced unpaved roads.
Geosynthetics can improve the performance of weak
subgrades under temporary unpaved roads by the follow-
ing mechanisms: (1) reduction of plastic shear stresses
that cause bearing capacity failure in the subgrade, (2)
reduction of maximum normal stresses on the subgrade
surface by improved load distribution, (3) increase in the
bearing capacity of the subgrade by confining lateral
movement at the subgrade-base interface and a reorienta-
tion of the induced shear stresses, (4) increase in the bear-
ing capacity and stress reduction attributable to the
‘‘tensioned membrane effect” in rutted areas, (5) lateral
restraint and reinforcement of base course aggregates,
and (6) reduction of mixing between subgrade and base
soils (Hufenus et al., 2006; Maxwell et al., 2005; Giroud
and Han, 2004a; Leng, 2002; Perkins et al., 2005; Watn
et al., 2005). Three of these mechanisms are illustrated in
Fig. 1. These improvements in subgrade performance can
facilitate compaction, reduce the gravel surface thickness,
delay rut formation, and extend the service life of unpaved
roads, particularly in cases of very soft subgrades with a
California Bearing Ratio (CBR) less than three (Benson
et al., 2005; Hufenus et al., 2006).
The current practice of using geosynthetics for subgrade
stabilization is primarily based on empirical evidence from
constructed test sections. Field tests constructed strictly
for research purposes, instead of during scheduled rehabil-
itation or reconstruction activities, offer better control over
study variables such as careful preparation of soil and
reduced incidental trafficking. Despite this, it is still diffi-
cult to achieve uniform conditions throughout a project
site utilizing the natural subgrade (e.g., Fannin and
s in suFig. 1. Possible reinforcement functions provided by geosynthetic bgrade stabilization applications (from Haliburton et al. (1981)).
strength. The cyclic tensile modulus, J, is a material prop-Sigurdsson, 1996; Edil et al., 2002; and Hufenus et al.,
2006). Conversely, research studies in which a subgrade
soil was artificially placed demonstrate better consistency
(e.g., Santoni et al., 2001; Perkins, 2002; Tingle and
Webster, 2003; Cuelho and Perkins, 2009; Cuelho et al.,
2014).
While laboratory studies can be conducted more
quickly and usually include more alternatives, they are
only able to simulate field conditions and, as Hufenus
et al. (2006) point out, there ‘‘are no incontrovertible indi-
cations from laboratory tests of the influence that the
geosynthetic will have on the performance of the pave-
ment under trafficking” (p. 23). Thus, the need still exists
for field tests that provide uniform conditions and incorpo-
rate a variety of geosynthetics in order to develop a suffi-
cient database of performance results. The need for such
a database of information is resoundingly clear in light of
the fact that there is still not a universally accepted and
calibrated design method for unpaved roads (or construc-
tion platforms) that incorporate both soil and geosynthetic
material properties.
A simple existing design method published by the Fed-
eral Highway Administration (Holtz et al., 2008) is based
on the U.S. Forest Service method developed by Steward
et al. (1977). A more recent design method for unpaved
roads that attempts to incorporate geosynthetic properties
was theoretically derived based on the stresses that
develop at the base-subgrade interface. The impact of
these stresses and the subgrade bearing capacity were
related to rut depth based on empirical data (Giroud and
Han, 2004a). However, only limited data were used to cal-
ibrate the model: (1) field data from Hammitt et al. (1970)
for unreinforced unpaved sections, and (2) lab data from
Gabr (2001) that involved two versions of one type of
geosynthetic (integrally formed geogrid). One parameter
(the bearing capacity, Nc) in the model can take on three
different values depending on whether the roadway is (1)
unreinforced, (2) geotextile reinforced, or (3) geogrid rein-
forced design. If a geogrid is under consideration, the aper-
ture stability modulus is used, but only if the material
property is within the approximate range of the types of
geogrids tested by Gabr (2001) for which the model was
calibrated. Even though this design method was intended
to be used to design reinforced and unreinforced unpaved
roads, there are inherent limitations in how it models the
contribution of various geosynthetics that should be con-
sidered. While this model is an improvement over less
sophisticated designs from the 1980s (Giroud and Noiray,
1981 and Giroud et al., 1985), there is still a need to inves-
tigate the performance of geosynthetics in controlled field
tests. Calibrations with additional data sets may be suffi-
cient, although geosynthetic material properties other than
aperture stability modulus should be considered.
Based on these limitations, two ambitious projects were
sponsored by the Montana Department of Transportation
(MDT) to evaluate the performance and behavior of a wide
variety of geogrids and geotextiles when used as subgrade
stabilization (Cuelho and Perkins, 2009 – Phase I; Cuelho
et al., 2014 – Phase II). The primary objective of these
research efforts was developed based on deficiencies in
the standard design techniques and lack of agreement aserty that is derived from a recently developed test proce-
dure that describes the tensile properties of the
geosynthetic under cyclic loading representative of pave-
ment reinforcement applications. Similarly, the resilient
interface shear modulus, GI, describes the stiffness of the
interface between the geosynthetic and surrounding
aggregates under small cyclic loads and at various levels
of load and confinement. The junction strength, Xj, is the
average shear strength of the geogrid junctions (or nodes)to which geosynthetic properties are most relevant for this
application. The results of this research were used to
understand which properties are most relevant to this
application, and consequently to update the design
methodology to incorporate these material properties.
The result of which should be a more accurate design
method that more broadly encompasses materials with
which good experience exists.
Experimental program
This research project was specifically planned to quan-
tify differences in performance of various geosynthetic
products under the same conditions (i.e., same subgrade
strength and base course thickness). In addition, supple-
mental test sections were constructed to study the effect
that variations in subgrade strength and base course thick-
ness had on the performance. Specifically, three control
sections (i.e., no geosynthetic) were constructed, each hav-
ing different thickness of base course aggregate, and three
test sections were built using the same integrally-formed
geogrid (test sections IFG-1, IFG-2, and IFG-3), each having
different subgrade strengths. The final arrangement of the
test sections is shown in Fig. 2, which includes the average
subgrade strengths and base thicknesses. Each test section
was 4.9 mwide and 15 m long. The Transcend research test
facility managed by the Western Transportation Institute
at Montana State University was used for this study.
Twelve geosynthetic products (ten geogrids and two
geotextiles) were used in this research project to evaluate
their relative performance under the conditions presented
herein. A summary of the basic material characteristics and
strengths of these products is listed in Table 1, and corre-
sponding photos are provided in Fig. 3. Five laboratory
tests were used to characterize the geosynthetics used in
this research, and include wide-width tensile strength
(ASTM D4595 and ASTM D6637), cyclic tensile modulus
(ASTM D7556), resilient interface shear stiffness (ASTM
D7499), junction strength (ASTM D7737), and aperture sta-
bility modulus (Kinney, 2000). Results from tests con-
ducted in the cross-machine direction are summarized in
Table 2.
Geosynthetic tensile strength is commonly used to
evaluate the ability of the geosynthetic to transmit load,
specifically in its principal strength directions, which gen-
erally lines up with the machine and cross-machine direc-
tions of the material. This material property also allows
designers to ensure it has adequate strength to sustain
construction stresses. Working stresses in roadway appli-
cations generally correspond to less than 5 percent
tion Direc
15 m 15 m 15 m15 mper unit width. Strength values are averaged from multiple
tests on a single tensile member that is pulled from its
junction with a cross-member. Junction stiffness Xja was
determined by taking the secant stiffness of the junction
strength response at 1.3 mm of displacement. Finally, the
4.9 m
W
oG
-6
IF
G
-1
IF
G
-2
IF
G
-3
Kn
G
-9
W
oG
-8
W
oG
-7
W
eG
-4
W
eG
-5
15 m 15 m 15 m15 m 15 m
Regular base thickne
(avg = 27.7 cm)
Regular 
(avg
Weaker subgrade
(avg = 1.64 CBR)
Stronger subgrade
(avg = 2.17 CBR)
Fig. 2. General layout
Table 1
Summary of geosynthetic characteristics.
Geosynthetic
Test Sectiona
Polymer and structureb Mass per un
area (g/m2)
IFG-1, IFG-2
and IFG-3
PP – integrally-formed, biaxial geogrid 302
WeG-4 PP – vibratory-welded, biaxial geogrid 200
WeG-5 PP – biaxial, welded geogrid 203
WoG-6 PMY – PVC–coated, woven, biaxial geogrid 322
WoG-7 PMY – PVC–coated, woven, biaxial geogrid 417
WoG-8 PMY – PVC–coated, woven, biaxial geogrid 309
KnG-9 PP – polymer-coated, knitted, biaxial geogrid 220
ExG-10g PP – extruded, triple-layer, biaxial geogrid 329
IFG-11 PP – integrally-formed, triaxial geogrid 180
IFG-12 PP – integrally-formed, triaxial geogrid 217
WoT-13 PPF – woven geotextile 417
NWoT-14 PP – non-woven, needle-punched geotextile 271
MD = machine direction; XMD = cross-machine direction.
a Acronym meanings (related to manufacturing process): IFG = integrally-fo
ExG = extruded grid, WoT = woven textile, NWoT = non-woven textile; numbers
b PP = polypropylene, PMY = polyester multifilament yarn, PPF = polypropylen
c for a single layer; apparent opening size is reduced when three layers are st
d reported as ‘‘rib pitch” in manufacturer’s specification sheet.
e Apparent Opening Size (AOS) in U.S. Standard sieve size, ASTM D4751.
f WoG-6 and WoG-7 materials experienced some grip slippage at their ultim
g Tested as a composite, i.e., not separately (triple layer material).
h When the IFG-11 and IFG-12 geogrids are tested in the machine direction, te
load, resulting in large distortions of the material and lower and/or inaccurate s
i Grab tensile strength (ASTM D4632) in kN.of Traffic
15 m
Not to scale
15 m 15 m 15 maperture stability modulus, ASM, describes the dimensional
stiffness or torsional rigidity of geogrids under a rotational
load. The torque it takes to rotate the material with respect
to the clamp is called the aperture stability modulus,
reported in units of N-m/deg.
15 m
Co
nt
ro
l 3
Co
nt
ro
l 2
W
oT
-1
3
IF
G
-1
2
IF
G
-1
1
NW
oT
-1
4
Co
nt
ro
l 1
Ex
G
-1
0
15 m 15 m 15 m
Thickest base 
(avg = 63.2 cm)
Thicker base
(avg = 41.4 cm)
ss
subgrade strength
= 1.79 CBR)
of test sections.
it Aperture
Size (mm)
Strength @
2% (kN/m)
Strength@
5% (kN/m)
Ultimate
Strength (kN/m)
MD x XMD MD XMD MD XMD MD XMD
25 
 33 8.5 12.0 15.7 21.8 21.6 28.4
33 
 33 14.1 13.8 26.4 26.7 30.4 39.6
43 
 41 14.6 12.5 29.6 25.9 38.6 34.7
25 
 25 5.8 9.0 10.0 13.5 29.8f 55.2f
25 
 25 5.8 14.4 10.4 21.1 31.3f 84.9f
25 
 25 9.4 10.8 20.1 18.7 38.4 47.0
15 
 15 9.7 13.8 20.8 28.3 27.2 38.2
43 
 51c 8.3 10.1 15.3 19.6 20.6 32.8
41 
 41d 0.5h 4.7 2.6h 9.7 9.1h 12.3
41 
 41d 1.0h 5.7 3.8h 10.9 11.0h 12.9
40e 7.3 21.9 18.8 50.2 82.0 89.2
80e — — — — 1.03i 1.13i
rmed grid, WeG = welded grid, WoG = woven grid, KnG = knitted grid,
represent position of test section.
e fiber.
acked on top of one another.
ate strength values.
nsile members are offset by 30 degrees from the direction of the applied
trength values.
The soil used to construct the subgrade consisted of nat-
ural overburden material that classified as CL (sandy lean
clay) according to the USCS classification system (ASTM
D2487). The base course material for this project consisted
of crushed aggregates and classified as GP-GC (poorly
graded gravel with clay with sand) according to the USCS
classification system (ASTM D2487). It contained 10 per-
cent fines and 55 percent fractured faces. Laboratory
strength tests run on the base course aggregate (ASTM
D1883) resulted in a California Bearing Ratio (CBR) value
greater than 100; however, in-field CBR tests indicated that
the average in-place CBR strength of the base was approx-
imately 20. This difference is due primarily to the
Fig. 3. Photos of geosynthetics: (a) IFG-1, IFG-2 and IFG-3, (b) WeG-4, (c) WeG-5,
12, (k) WoT-13, and (l) NWoT-14.conditions under which the base course was tested in the
lab compared to how these values were obtained in the
field. The laboratory CBR test uses a rigid cylinder to con-
fine the sample, and due to the particle size of this partic-
ular gradation, a replacement was necessary to reduce the
size of the larger particles. This replacement had a large
effect on the strength. In addition, the bearing capacity at
higher penetration (5 mm) were greater than at 2.5 mm
penetration, so the higher values are what are recom-
mended to be reported by the standard. Finally, the shape
of the bearing capacity curve was concave upward making
it necessary to apply a correction, which further increased
the value. The CBR of the base course in the field was
(d) WoG-6, (e) WoG-7, (f) WoG-8, (g) KnG-9, (h) ExG-10, (i) IFG-11, (j) IFG-
Table 2
Geosynthetic material properties in the cross-machine direction used in the analy
Geosynthetic test sectiona Jcyclic (kN/m)
0.5% 1.0% 1.5% 2.0% 3.0%
IFG-1, IFG-2 and IFG-3 933 918 915 911 913
WeG-4 1150 1141 1148 1157 1213
WeG-5 1019 983 971 1005 1034
WoG-6 765 794 823 839 900
WoG-7 1231 1252 1290 1325 1421
WoG-8 919 913 941 971 1070
KnG-9 1064 1061 1063 1103 1160
ExG-10 855 806 810 794 790
IFG-11 320 374 398 412 412
IFG-12 424 443 448 454 470
WoT-13 1647 2120 2258 2285 2344
NWoT-14 NTb NTb NTb NTb NTb
NT = not tested.
NA = not applicable.
a Acronym meanings (related to manufacturing process): IFG = integrally-fo
ExG = extruded grid, WoT = woven textile, NWoT = non-woven textile; numbers
b Material too delicate to test using unconfined cyclic tension.
c Impossible to test with triple layer construction.determined using a dynamic cone penetrometer, the
mechanics of which differs significantly from the labora-
tory method.
Construction of the test sections began with preparing
and placing the subgrade (depicted in Fig. 4), followed by
installing the geosynthetics and instrumentation, and
finally preparing and placing the base course aggregate. A
cross-sectional view of a typical test section is shown in
Fig. 5 with the test vehicle. Preparation and construction
of the subgrade and base course was monitored exten-
sively to ensure that these materials were placed in a con-
sistent and uniform manner.
The subgrade was built in six lifts that were approxi-
mately 15 cm deep for a total depth of about 0.9 m. The
subgrade was processed to reach the target strength by
adding water from a water truck and fire hose. Water was
added until it reached the target moisture content (target
of approximately 23 percent to achieve CBR = 1.70).
Fig. 4. Filling trench with prepared subgrade.Processing was accomplished using a large excavator to
move and mix the material as water was being added. Suf-
ficient material was processed to construct a single 15-cm
deep layer over two test sections at a time (about 30 m3
of material). The subgrade was then placed in the trench
using the excavator and a track-mounted skid-steer tractor
was used to level and initially compact the subgrade. A
smooth, single-drum, vibratory roller was used to compact
the subgrade by making two passes of the roller in three
longitudinal paths of the freshly placed subgrade. The
moisture in the top surface of the subgrade wasmaintained
during construction by periodically wetting the surface and
keeping it covered with plastic until the next layer of sub-
grade or the base course could be placed. Prior to placement
of the geosynthetics and base course, the top surface of the
sis.
GI (MPa) Xj (kN/m) Xja (MN/m/m) ASM (N-m/deg)
4.0%
965 2106 30.1 4.36 0.78
1297 1284 10.1 4.43 1.15
1091 631 8.7 3.43 1.57
983 1227 6.5 2.17 0.25
1552 2013 5.0 1.74 0.27
1174 1657 6.3 3.57 0.35
1136 890 1.8 0.68 1.09
808 1308 NAc NAc NAc
426 609 12.7 2.18 0.28
464 1671 13.1 1.37 0.55
2407 2269 NA NA NA
NTb 1013 NA NA NA
rmed grid, WeG = welded grid, WoG = woven grid, KnG = knitted grid,
represent position of test section.subgrade was smoothed and screeded to the height of the
adjacent pavement surface. The final average strength of
the subgrade in all the test sections was CBR = 1.79.
Preparation of the base course aggregate began by add-
ing water and mixing with an end loader until it reached
optimum water content. A large screed that rested on the
paved surface on both sides of the subgrade trench was
used to level the surface of the gravel layer. The base
course was placed in two layers. The final thickness of
the first layer of base course was about 20 cm when com-
pacted and the second was about 7.6 cm deep for a total of
about 28 cm of gravel, on average. Two of the three control
test sections contained thicker base material. The Control 2
test section was constructed of two layers of about 20 cm
thick, for a total of about 40 cm of gravel when compacted,
and the Control 3 test section was constructed of three lay-
ers of about 20 cm thick, and had a final average thickness
of about 60 cm of gravel when compacted. Compaction
was achieved using a smooth, single-drum, vibratory
roller. In total, eight passes of the roller were made per lift.
Assessment of the base course was evaluated using LWD,
DCP, in-field CBR and nuclear densometer tests. All of
the test sections met the minimum 95 percent density
4.9 m
n of trequirements based on Modified Proctor test results. A
detailed summary and analysis of the physical attributes
of the base course can be found in Cuelho et al. (2014).
Trafficking was accomplished using a three-axle dump
truck that weighed 20.6 metric tons and had 620 kPa tire
pressure. Trafficking was always in one direction, and the
speed was approximately 8 kph to ensure that dynamic
loads were not induced in the test sections from any
unevenness in the gravel surface. Trafficking was applied
until rut levels reached 75 mm. Allowable ruts generally
range from 50 to 100 mm for this application. Ruts greater
Original taxiway
277 mm
Geosynthetic
Fig. 5. Cross-sectional representatiothan 75 mm may cause the undercarriage of the truck to
drag during trafficking, especially if the subgrade experi-
ences bearing capacity failure that results in heaving of
the road surface. Photos of a typical test section during
trafficking are presented in Fig. 6 for rut levels of about
0, 25, 50 and 75 mm, respectively. Once the allowable rut
level was reached, repairs were made by placing additional
gravel in the rutted areas. Repairs within test sections were
made incrementally, so that unfailed portions of test sec-
tions could continue to be trafficked until they reached
failure. No further measures of rut were made in areas that
were repaired. Rut measurements were made at 1-meter
intervals along two longitudinal lines (in the direction of
traffic), corresponding to the outside rear wheels of the test
vehicle. A robotic total station was used to make these
measurements, and the data were used to determine rut
as a function of the difference in the elevation of the mea-
surement points over time.
Data analysis
Longitudinal rut measurements (in the direction of traf-
fic) were the primary means used to determine the relative
performance of each test section. Rut behavior was mainlyaffected by four factors: (1) the strength of the subgrade,
(2) the depth of the base course, (3) the strength of the
base course, and (4) the presence of the geosynthetic.
The field test sections were constructed to have the same
subgrade strength and base thickness (with the intentional
exception of multiple control test sections) to minimize
differences between test sections and facilitate a more
direct comparison of their performance. Despite efforts
during construction to eliminate differences in subgrade
strength and base course thickness, small variations were
inevitable. An empirical correction procedure was imple-
Artificial
Subgrade 0.9 m
Base course
ypical test section (drawn to scale).mented to adjust the rut response for these two properties
so that direct performance comparisons between test sec-
tions were more accurate. Rut data was not adjusted based
on base course strength and stiffness because (1) strength
and stiffness properties were not measured at every rut
measurement point, and (2) there were no controls where
these properties were purposefully varied to determine
their effect on performance. After adjustments for sub-
grade strength and base course thickness were applied to
the rut data, the remaining behavioral differences between
the reinforced test sections could more confidently be
attributed to the geosynthetic reinforcement.
Individual rut measurements were adjusted and aver-
aged together within a particular test section. Individual
values of rut greater than one standard deviation away
from the mean were not used in the analysis. An analysis
of the longitudinal rut responses was conducted to deter-
mine which geosynthetic material properties were most
related to the performance of a particular test section. This
analysis was conducted at various rut depths to determine
whether different material properties affected perfor-
mance at various levels of rut. Predicted values were used
in the regression analysis for test sections that did not
reach failure. The following material properties were con-
sidered in this analysis:
 Wide-width tensile strength in the cross-machine
direction at 2% (WWT-2%).
 Wide-width tensile strength in the cross-machine
direction at 5% (WWT-5%).
 Ultimate wide-width tensile strength in the cross-
machine direction (WWT-Ult.).
 Cyclic tensile stiffness at 0.5, 1.0, 1.5, 2.0, 3.0 and 4.0
percent in the cross-machine direction (CTS-0.5%, CTS-
1.0%, CTS-1.5%, CTS-2.0%, CTS-3.0%, CTS-4.0%).
 Resilient interface shear stiffness in the cross-machine
direction (RISM).
 Junction strength in the cross-machine direction (Junc.
Str.).
 Junction stiffness in the cross-machine direction, deter-
mined by taking the secant stiffness of the junction
strength response at 1.3 mm of displacement (Junc.
Stiff.)
 Aperture stability modulus (ASM).
Linear regression was used in this analysis because
there were too few points to clearly indicate a more
sophisticated regression equation, and it provided suffi-
cient information to be able to compare data fit between
individual factors or to observe changes or trends in data
fit for multiple variables. In this analysis, the number of
truck passes for a particular test section was adjusted by
Fig. 6. Typical Phase II test section at (a) 0 mm, (bsubtracting the number of truck passes in comparable con-
trol test sections to determine Nadd, the number of addi-
tional truck passes a particular test section experienced
in comparison to the unreinforced case. That allowed the
y-intercept to be set to zero because the absence of
geosynthetic reinforcement would result in no benefit to
the test section. R-squared (the coefficient of determina-
tion) is commonly used as the indicator of how well the
data points fit the regression line, and was used in this
analysis to determine how well a particular material prop-
erty related to field performance. R-squared values
approaching 1.0 indicate a better fit, while values less than
that (including negative values) indicate poorer correla-
tions. R-squared values greater than 0.5 were considered
significant for the purposes of this analysis. The results
from these analyses are shown in Fig. 7a (note: the analysis
of some of the properties resulted in R-squared values less
than zero, which are not shown).
Referring to Fig. 7a, the geosynthetic material property
that best related to the performance of the test sections
was the strength and stiffness of the junctions in the
cross-machine direction, and this property correlated bet-
ter with performance as rut increased. R-squared values
in the machine direction (not shown) were all negative
with the exception of the ultimate wide-width strength,
which showed better correlation at lower levels of rut. A
) 25 mm, (c) 50 mm, and (d) 75 mm of rut.
s-macsecond linear regression analysis was conducted excluding
data from geosynthetics that performed poorly (WoG-1
and KnG) due to their low junction strengths. Knowing that
the primary property linked to performance in these test
sections was junction stiffness, these products were unable
to transmit stresses into the cross-machine structural ele-
ments because the junctions were too weak. By eliminat-
ing these products from the analysis other potential links
between the geosynthetic properties and performance
became more apparent. The results of this analysis are
shown in Fig. 7b. These results indicate that by excluding
materials that did not perform well based on their weaker
junctions, tensile strength in the material is also a good
indicator of performance. This is most apparent in the
wide-width tensile strengths at 5 percent and the cyclic
stiffness values. R-squared values are reduced for junction
strength and stiffness in this analysis because of the miss-
ing data.
A linear regression analysis was also conducted using
rut data from an earlier phase of this project (Cuelho and
Perkins, 2009 – Phase I). Six of the test sections from Phase
I used the same geosynthetics as this project (IFG-3, WeG-
4, WeG-5, WoG-6, WoG-8, and NWoT-14). These test sec-
tions had very similar subgrade strengths but 75 mm less
base aggregate thickness, creating a more severe condition
than Phase II. Performance data was analyzed with respect
to the material properties listed above at 25.4, 50.8, 76.2
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
R-
Sq
ua
re
d
25.4 mm rut
50.08 mm rut
63.5 mm rut
a)
Fig. 7. Phase II regression analysis results in the crosand 101.6 mm of rut. The results of this analysis are shown
in Fig. 8. Considering a similar approach as before, the
regression analysis using performance data from Phase I
indicates, overall, that tensile strength in both material
directions relate to performance at higher levels of rut,
while junction strength relates to performance at lower
levels of rut. The relationship with junction stiffness peaks
at 76.2 mm of rut. Aperture stability modulus is also
related to early performance of the Phase I test sections.
Considering both phases of this effort, cross-machine
junction stiffness and cross-machine tensile strength are
the two most important properties associated with good
performance of geogrids used in this application and under
these conditions. These two properties work together to
ensure proper reinforcement of the gravel layer and
increased longevity of the road (evident as an increased
number of load passes). Geogrids with weak junctionswere unable to fully utilize the strength in the cross-
machine direction because the load transfer through the
individual members and into the junctions was weak. All
of the materials in Phase I had adequate junction strength.
Not surprisingly, a greater reliance on tensile strength was
more evident in this case (refer to Fig. 8a). This information
concurs with information from Phase II where tensile
strength in the cross-machine direction was also linked
to good performance (once materials with weak junctions
were removed from the analysis, as shown in Fig. 7b).
Overall, this analytical process helped establish geogrid
material properties associated with good performance.
The geotextile materials used during these studies
showed good performance as subgrade stabilization, but
material properties associated with their performance
were difficult to establish due to the limited number of test
sections and lack of relevant tests to properly characterize
these types of materials for this application. Understanding
that junction stiffness in the cross-machine direction is the
property of the geogrids that most related to their perfor-
mance it can be inferred that the mechanism by which
geotextiles provide reinforcement is related to how well
they transmit stresses from one principal strength direc-
tion to the other. Despite that fact that fibers oriented
orthogonally to one another are not firmly bound, the
interaction between them is significantly enhanced by
the overburden pressure of the road materials above them.
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
R-
Sq
ua
re
d
25.4 mm rut
50.8 mm rut
63.5 mm rut
b)
hine direction using (a) all data and (b) select data.The strong frictional bond between these fibers is able to
transmit stresses from traffic loads into fibers oriented in
the cross-machine direction, similar to a typical geogrid
junction. Multiple fibers and intersections between these
fibers means that there are significantly more paths for
stresses to be transmitted. This helps explain why the geo-
textiles worked well in this application under these condi-
tions. A similar case can be made for the non-woven
geotextile. Its tensile properties are significantly enhanced
by confinement from overburden, and the continuous
sheetlike structure of the material creates an infinite num-
ber of stress paths.
Calibration of the Giroud–Han Design equation
Information from the testing program described above
was used to calibrate the design equation associated with
m thethe methodology developed by Giroud and Han (2004a,b).
The generic design approach developed by Giroud and Han
(also referred to as the G–H method) takes into considera-
tion the geometry of the unpaved structure, the level of
truck traffic, the truck axle configuration and loading, rut
depth and serviceability, the properties of the base course
and subgrade materials, the ratio of base course and sub-
grade strength, and the properties of the geogrid. The gen-
eric equation associated with the G–H method is provided
in Eq. (1).
h¼ 1þk logN
tana0½1þ0:204ðRE1Þ
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
P
pr2
s
f s
 
½1nexp xðrhÞn
 Nccu
vuut 1
2
64
3
75r
ð1Þ
where h = compacted base course thickness {m}.
N = number of axle passes. k = constant dependent on
base thickness and reinforcement. a0 = initial stress distri-
bution angle = 38.5. RE¼min EbcEsg ;5:0
 
¼min 3:48CBR0:3bcCBRsg ;5:0
 
.
P = tire load {kN}. r = radius of equivalent tire contact area
{m}. s = allowable rut depth {m}. fs = reference rut
depth {m}. cu = subgrade undrained shear strength {kPa}.
Nc = bearing capacity factor (5.71 for geogrid-reinforced
roads). n, x, and n are constants calibrated by Giroud and
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
R-
Sq
ua
re
d
25.4 mm rut
50.8 mm rut
76.2 mm rut
101.6 mm rut
a)
Fig. 8. Phase I regression analysis results using data froHan (2004b) using data from unpaved, unreinforced roads
(n = 0.9, x = 1.0, and n = 2.0).
The variable k in Eq. (1) is responsible for describing the
contribution of the geosynthetic, as well as contribution of
the base thickness and equivalent radius of the applied
load [(r/h)1.5]. The general form of the equation to describe
k (Eq. (2)), as published by Giroud and Han (2004a,b), was
based on the calibration of two biaxial geogrids in terms of
the aperture stability modulus (J).
k ¼ ð0:96 1:46J2Þ r
h
 1:5
ð2Þ
The first step in this process was to isolate the constant
k in Eq. (1). The result of this algebraic manipulation
(and subsequent substitution of the constant values for n,
x, and n) is presented in Eq. (3). The variable k was
redefined as k’ to indicate a back-calculated value, as
described below.k0 ¼ h½1þ0:204ðRE1Þ
1:26
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
P
pr2
s
f s
 
10:9exp  rhð Þ2
  
Nccu
vuut 1
2
64
3
75r
1
2
666666664
3
777777775
1
logNðrhÞ1:5
!
ð3Þ
The following material properties of the test sections
were used in Eq. (3) to determine k’ for each test section.
h = 0.276 m; average thickness of base course layer.
RE = 4.8; average CBRbc. field = 20, average CBRsg = 1.79.
P = 37.63 kN. r = 0.139 m. Nc = 5.71. cu = 62.7 kPa.
fs = 75 mm.
The single property that varied within the test sections
was the number of axle passes of the truck (N) for a partic-
ular rut level. The value of k0 was determined at various
levels of rut (s = 38.1 mm, 50.8 mm, 63.5 mm and
76.2 mm) and axle passes. Because k0 represents the mate-
rial property of the geogrid associated with design, it was
plotted with respect to the various properties of the
geosynthetic, as in the regression analysis previously
described above. Similar to that analysis, linear regression
was used to fit the data from these comparisons, the
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
R-
Sq
ua
re
d
25.4 mm rut
50.8 mm rut
76.2 mm rut
101.6 mm rut
b)
(a) cross-machine direction and (b) machine direction.results of which are summarized in Fig. 9.
Not surprisingly, a similar pattern of interdependence
between k0 and junction stiffness emerged from these anal-
yses. Therefore, junction stiffness in the cross-machine
direction emerged from this analysis as the primary mate-
rial property related to rutting performance. This does not
mean that tensile strength in the cross-machine direction
is unimportant, as indicated by the previous regression
analyses. Development of the tensile capacity of the geo-
grid was critically dependent on the ability of the junction
to transmit stresses into members oriented in the cross-
machine direction. Therefore, materials that have sufficient
tensile strength but weak junctions will not perform well.
The converse is also true – strong junctions with weak
members will also not perform well. Reasons why the
aperture stability modulus did not correlate to perfor-
mance are unknown. The individual slopes of k0 versus
junction stiffness varied as a function of rut, as shown in
Fig. 10. The information contained in Fig. 10 can be used to
determine the value of k0 based on junction stiffness (in
units of MN/m/m).
To validate the accuracy of this method, Eq. (3) was
rearranged in terms of N, the number of traffic passes,
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
R-
Sq
ua
re
d
38.1 mm rut
50.8 mm rut
63.5 mm rut
76.2 mm rut
Fig. 9. Regression analysis results for k0 with respect to geogrid material
properties in the cross-machine direction.
m rut
R² = 0.855
0
200
400
600
800
1000
1200
1400
0 200 400 600 800 1000 1200 1400
N
ac
tu
al
Npredicted
Fig. 11. Comparison of predicted axle passes to actual axle passes.and the predicted number of passes was determined based
on junction stiffnesses of the geogrids used in this project.
Values for k’ were determined using the relationships pre-
sented in Fig. 10. The predicted values were compared to
the actual number of passes to reach various levels of rut,
the results of which are shown in Fig. 11. These results
indicate that there is good correlation between the pre-
dicted and actual values.
The final form of the design equation for geosynthetic
reinforced unpaved roadways, based on the development
work by Giroud and Han (2004a) and the calibration based
on field test sections constructed by Cuelho et al. (2014), is
shown in Eq. (4). It should be noted that the base course
layer thickness (h) is present on both sides of the equation;
1.0
1.1
1.2
s = 76.2 m0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
0.0 1.0 2.0 3.0 4.
k'
Juncon Sffne
s = 38.1 mm rut
s = 50.8 mm rut
s = 63.5 mm rut
Fig. 10. Junction stiffness versustherefore, an iterative process is necessary to determine a
single value of h for a given set of conditions. The variable
k’ is based solely on junction stiffness (ASTM D7737), and
can be determined using the relationships presented in
Fig. 10. The information contained in Fig. 10 is shown for
various levels of expected rut, as indicated by the multiple
lines. The regression line is extended to include products
with greater junction stiffnesses than the materials used
in this study.
h¼1:26½1þk
0 r
h
 1:5 logN
1þ0:204ðRE1Þ
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
P
pr2
s
f s
 
10:9exp  rh
 2 h iNccu
vuut 1
2
64
3
75r
ð4Þ0 5.0 6.0 7.0 8.0
ss (MN/m/m)
k0 for various levels of rut.
sity, and sponsored by the Montana Department of Trans-
portation, to build and traffic multiple test sections
containing various types of geosynthetic reinforcement
products. The results of this effort were used to better
understand which properties are most relevant to sub-
grade stabilization of unpaved roads, and consequently
to update the design methodology to incorporate these
material properties.
Test sections were built and trafficked in a large-scale
controlled laboratory environment to study the perfor-
mance of various geosynthetics to stabilize weak sub-
grades. A comprehensive set of material tests were
conducted to more thoroughly evaluate the potential rela-
tionship between geosynthetic material properties and
the relative performance of the test sections. The results
of a regression analysis showed that junction stiffness
was the best indicator of performance under these
conditions.
The Giroud–Han method is currently the most sophisti-
cated method to design geosynthetic stabilized unpaved
roads; however, calibration of this method to date is based
on very limited data. Performance data from the full-scale
field tests built as part of this research effort were used to
calibrate the Giroud–Han design equation. Geogrid junc-
tion stiffness in the cross-machine direction was the pri-
mary indicator of performance in these test sections and
was therefore used as the material property to calibrate
the design equation. Nevertheless, tensile strength in the
cross-machine direction is also important, as indicated by
the regression analyses. Development of the tensile capac-
ity of the geogrid is critically dependent on the ability of
the junction to transmit stresses into members oriented
in the cross-machine direction. Correlations between pre-
dicted axle passes and actual axle passes indicated good
predictions using the calibrated equation. The newly cali-
brated design equation presented herein replaces aperture
stability modulus with junction stiffness in the cross-
machine direction to describe the contribution of the
geogrid.
Acknowledgements
Financial support for this pooled-fund study lead by
Montana Department of Transportation was generously
provided by the following United States departments of
transportation (listed in alphabetical order) – Idaho, Mon-
tana, New York, Ohio, Oklahoma, Oregon, South Dakota,
Texas and Wyoming. Geosynthetic materials were charita-
bly donated by Colbond, Huesker, NAUE, Propex, Synteen
and TenCate.Summary and conclusions
Geosynthetics are routinely used to stabilize weak soils
in transportation applications; however, deficiencies in
the standard design techniques have made widespread
adoption of the existing methods slow. In addition, agree-
ment as to which geosynthetic properties are most rele-
vant to subgrade stabilization applications is lacking. A
rigorous program was undertaken by researchers at the
Western Transportation Institute at Montana State Univer-References
ASTM Standard D1883. Standard test method for california bearing ratio
(CBR) of laboratory-compacted soils. West Conshohocken, PA: ASTM
International; 2007. www.astm.org.
ASTM Standard D2487. Standard practice for classification of soils for
engineering purposes (unified soil classification system). West
Conshohocken, PA: ASTM International; 2011. www.astm.org.
ASTM Standard D4595. Standard test method for tensile properties of
geotextiles by the wide-width strip method. West Conshohocken,
PA: ASTM International; 2011. www.astm.org.
ASTM Standard D4632. Standard test method for grab breaking load and
elongation of geotextiles. West Conshohocken, PA: ASTM
International; 2008. www.astm.org.
ASTM Standard D4751. Standard test method for determining apparent
opening size of a geotextile. West Conshohocken, PA: ASTM
International; 2004. www.astm.org.
ASTM Standard D6637. Standard test method for determining tensile
properties of geogrids by the single or multi-rib tensile method. West
Conshohocken, PA: ASTM International; 2011. www.astm.org.
ASTM Standard D7499. Standard test method for measuring
geosynthetic-soil resilient interface shear stiffness. West
Conshohocken, PA: ASTM International; 2009. www.astm.org.
Standard test methods for determining small-strain tensile properties of
geogrids and geotextiles by in-air cyclic tension tests. West
Conshohocken, PA: ASTM International; 2010. www.astm.org.
ASTM Standard D7737. Standard test method for individual geogrid
junction strength. West Conshohocken, PA: ASTM International;
2011. www.astm.org.
Benson CH, Edil TB, Tanyu BF, KimW-H. Equivalency of crushed rock with
industrial by-products and geosynthetic-reinforced aggregates used
for working platforms during pavement construction. Wisconsin
Highway Research Program, Final Report No. 0092-00-12; 2005. p.
100.
Cuelho E, Perkins S, Field investigation of geosynthetics used for subgrade
stabilization. Final Report to the Montana Department of
Transportation, FHWA/MT-09-003/8193; 2009. p. 140.
Cuelho E, Perkins S, Morris Z. Relative operational performance of
geosynthetics used as subgrade stabilization. Final Report to the
Montana Department of Transportation, FHWA/MT-14-002/7712-
251; 2014. p. 328.
Edil TB, Benson CH, Senol A, Bin-Shafique MS, Tanyu BF, Kim W-H, Field
evaluation of construction alternatives for roadway over soft
subgrade. Wisconsin Highway Research Program, Interim Report;
2002. p. 65.
Fannin RJ, Sigurdsson O. Field observations on stabilization of unpaved
roads with geosynthetics. J Geotechn Eng 1996;122(7):544–53.
Gabr M. Cyclic plate loading tests on geogrid reinforced roads. Research
Report to Tensar Earth Technologies Inc., NC State Univ.; 2001. p. 43.
Giroud JP, Han J. Design method for geogrid-reinforced unpaved roads.
Part I – development of design method. J Geotechn Geoenviron Eng
2004a;130(8):775–86.
Giroud JP, Han J. Design method for geogrid-reinforced unpaved roads.
Part II – calibration and applications. J Geotechn Geoenviron Eng
2004b;130(8):787–97.
Giroud JP, Noiray L. Geotextile-reinforced unpaved road design. J
Geotechn Eng 1981;107(9):1233–54.
Giroud JP, Ah-Line C, Bonaparte R. Design of unpaved roads and trafficked
areas with geogrids. In: Proceedings of a conference sponsored by the
science and engineering research council, polymer grid
reinforcement. London: Thomas Telford Ltd; 1985. p. 116–27.
Haliburton TA, Lawmaster JD, McGuffey VC. Use of engineering fabrics in
transportation related applications. Federal Highway Administration;
1981. FHWA DTFH61-80-C-00094.
Hammitt GM. Thickness requirement for unsurfaced roads and airfields,
bare base support, project 3782-65. CE, Vicksburg, Miss: U.S. Army
Engineer Waterways Experiment Station; 1970. p. 139. Technical Rep.
S-70-5.
Holtz RD, Christopher BR, Berg RR. Geosynthetic design and construction
guidelines. Washington DC: U.S. Department of Transportation,
Federal Highway Administration; 2008. p. 592. Report No. FHWA-
NHI-07-092.
Hufenus R, Rueegger R, Banjac R, Mayor P, Springman SM, Bronnimann R.
Full-scale field tests on geosynthetic reinforced unpaved roads on soft
subgrade. Geotext Geomembr 2006;24(1):21–37.
Kinney T. Standard test method for determining the ‘‘aperture stability
modulus” of a geogrid. Seattle: Shannon & Wilson Inc.; 2000.
Leng J. Characteristics and behavior of geogrid-reinforced aggregate
under cyclic load (Ph.D. thesis). North Carolina State University; 2002.
Maxwell S, KimW-H, Edil TB, Benson CH. Geosynthetics in stabilizing soft
subgrades with breaker run. Wisconsin Highway Research Program.
Final Report No. 0092-45-15; 2005. p. 88.
Perkins SW. Evaluation of geosynthetic reinforced flexible pavement
systems using two pavement test facilities. Montana Department of
Transportation. Final Report No. FHWA/MT-02-008/20040; 2002. p.
136.
Perkins SW, Bowders JJ, Christopher BR, Berg RR. Geosynthetic-
reinforcement for pavement systems: US perspectives. In:
Proceedings: Geo-Frontiers 2005, ASCE Geotechnical Special
Publications 130-142. Reston, VA: American Society of Civil
Engineers; 2005. p. 3039–63.
Santoni RL, Smith CJ, Tingle JS, Webster SL. Expedient road construction
over soft soils. U.S. Army Corps of Engineers; 2001. p. 107. Final
Report No. ERDC/GSL TR-01-7.
Steward J, Williamson R, Mohney J. Guidelines for use of fabrics in
construction and maintenance of low-volume roads. USDA, Forest
Service, Portland, OR. Also Reprinted as Report No. FHWA-TS-78-205,
Federal Highway Administration, Washington, D.C; 1977.
Tingle JS, Webster SL. Corps of engineers design of geosynthetic-
reinforced unpaved roads. Transp Res Rec 2003;1849:193–201.
United States Corps of Engineers. Use of geogrids in pavement
construction, ETL 1110-1-189. Washington, D.C.: Department of the
Army, U.S. Army Corps of Engineers; 2003. p. 37.
Watn A, Eiksund G, Jenner C, Rathmayer H. Geosynthetic reinforcement
for pavement systems: European perspectives. In: Proceedings: Geo-
Frontiers 2005, ASCE geotechnical special publications 130-
142. Reston, VA: American Society of Civil Engineers; 2005. p.
3019–37.