26 ■ Transportation Research Record 1709

Paper No. 00 - 1313

Field Evaluation of Geocell Use in

Wheel rutting in excess of 100 mm occurred within 7 days after re- and a moisture barrier that prevents water movement between layers.
construction of a section of Route 30 in Delaware County, Pennsylvania, There are currently many types of geosynthetics available on the
an urban roadway with high average daily traffic. The road was re- market. They can be divided into seven main categories: geotextiles,
constructed on a weak subgrade, and soil tests produced California bear- geogrids, geonets, geomembranes, geocells, geosynthetic clay liners,
ing ratio values between 1.5 and 6.5 percent. A geocell confinement system and geocomposites (4).
was used to stabilize the subgrade and solve this problem. The geocell One current approved method, used for subgrade stabilization in
used was constructed of high-density polyethylene. Each geocell panel Pennsylvania, involves undercutting unsuitable material and replac-
was fabricated using 60 strips 3.3 m long and 100 mm thick. The panels ing this material with subbase or select borrowed material. Class 4
were 2.4 m × 6.1 m when expanded and created a honeycomb-patterned geotextile (10-oz nonwoven needle-punched material) can also be
cellular confinement system. The geocell system was used in combination used as a separator between the subgrade and subbase to prevent
with other geosynthetics (geogrid or geotextiles, or both). The findings intrusion of the subgrade soil into the subbase and aggregate pene-
indicated that the geocell system used in this project performed very tration into the subgrade. These techniques have been proven to sta-
favorably. The roadway has not shown any pavement distress (alligator bilize the subgrade and increase pavement service life (5). Concrete,
cracking, rutting, etc.) during the 3-year evaluation period. To date, an expensive alternative, can also be used as a bridge over weak soil.
falling weight deflectometer data, initially collected after construction, While all these materials have generally functioned satisfactorily for
have since been collected annually for 2 years. Analysis suggests that the similar applications, each varies in price, ease of construction, and
geocell, in combination with a geosynthetic layer, provides a significant degree of satisfactory performance.
improvement to the pavement structure capacity when built over a weak Engineering District 6-0 in Pennsylvania encountered a serious
subgrade. The resilient modulus of the 150-mm-thick aggregate layer problem attempting to rehabilitate a section of Route 30 in Delaware
was improved almost twofold. There is a good possibility that this increase County, Pennsylvania. The section was located near the Blue Route
was caused by the aggregate confinement resulting from the geocell and (an urban roadway, in this case with an average daily traffic of 17,065,
the separation provided by another geosynthetic layer.
of which 3 percent were trucks) of Lancaster Ave., SR-476 of the Sec-
tion 630 project in the city of St. Davids. The subgrade was found
The misconception in conventional layered roadway designs is that to be extremely unstable, with a California bearing ratio (CBR) of
respective layers of various pavement components will remain un- 4.0 percent when testing was performed during excavation. Visual
changed over the existing subgrade throughout the service life inspection of the subgrade showed the soil to be gray and very fine.
of the pavement. However, changes in load and environment may With random amounts of stream pebbles and some 100 mm of rock,
weaken pavement system layers overtime (1). A common practice of it resembled sandy river silt. The dark gray color and smell of the soil
state department of transportation engineers when designing roads samples indicated the soil was aquatic, with a high level of organic
on weak subgrade soil is to include an extra amount of “sacrificial” material present. The soil was classified as loam to silt loam (in accor-
aggregate in addition to the amount required by standard design dance with Unified Soil Classification System) and A-4 in accordance
methods (2). Without this practice, a significant portion of the base- with AASHTO.
subbase course aggregate may be lost to the weak subgrade through A conventional pavement section was constructed with the follow-
aggregate penetration or subgrade soil pumping, thereby effectively ing cross section: 38-mm ID-2 hot-mix asphalt (HMA) wearing sur-
reducing the ability of the base-subbase course to distribute traffic- face, 50-mm ID-2 HMA base binder, 200-mm HMA base course [i.e.,
loading stresses (1, 2). In addition, the presence of water in a base- bituminous concrete base course (BCBC)], 150-mm 2A subbase, and
subbase course layer drastically changes the pattern by which surface a Class 4 nonwoven geotextile for separation. All pavement materials
loading pressures are distributed (3). meet Pennsylvania Department of Transportation (PennDOT) speci-
In pavement systems, various geosynthetics are used to provide one fications. Although the pavement design met site restrictions, it was
or more of the following: reinforcement that increases tensile strength unsuccessful. Failure of the reconstructed area was evident within
of a particular layer; strain energy absorption between pavement lay- 7 days, when wheel rutting in excess of 100 mm was observed.
ers; separation that maintains the integrity of particular layers by pre- The unstable subgrade material could not be removed, however,
venting intermixing; drainage/filtration that allows the water to flow, because of its close proximity to utilities. And the road could not be
thereby dissipating pore water pressure while limiting soil movement; raised to thicken the pavement section because of clearance con-
cerns regarding a railroad bridge. Using traditional road construc-
I. L. Al-Qadi, Charles Edward Via, Jr., Department of Civil and Environmental
Engineering, 200 Patton Hall, Virginia Polytechnic Institute and State University,
tion methods, the following options were considered: open the area
Blacksburg, VA 24061-0105. J. J. Hughes, Pennsylvania Department of Trans- up and allow the material to air dry, undercut the area and replace
portation, 1118 State Street, Harrisburg, PA 17120. the material with dry material, install additional subsurface drainage
Al-Qadi and Hughes Paper No. 00 - 1313 27

and allow the area to drain over time, or install a cement concrete base met the requirements of a Class 4 geotextile in accordance with Penn-
for the roadway bridging the areas in question. DOT specifications. The aggregate backfill used to fill the geocell is
The District 6-0 Construction Unit decided these methods were too type 2A (crusher run) in accordance with PennDOT specifications.
expensive, time consuming, or impractical. The shallow depth of the Construction started in June 1995.
utilities in the area prohibited undercutting or installing any addi- The basic construction plan involved reconstructing SR-30, Lan-
tional subsurface drainage. The time restraints for stage construction caster Ave., in two large patches, with each patch reconstructed as a
of the full-depth reconstruction would not allow the area to be opened designated left and right side. Each patch area was reconstructed over
for drying. The placement of a cement concrete base was possible, a continuous 36-h period on a weekend. These day-and-night opera-
but because of the expense of allowing access to a large number of tions took four weekends to complete. The first phase was from STA
businesses and the excessive time to reconstruct the roadway, this 46+50 to STA 44+50 RT (“RT” indicates “right side”) on SR-30. The
option was not considered feasible. bituminous material was milled off and the road was excavated. Two
A common problem in reconstructing urban roadways is existing soil samples were obtained and classified as AASHTO A-4. The
utilities, which restrict the depth of excavation for preparing the sub- CBR was found to be 3.6 percent.
grade for new road construction. The reconstruction of any roadway Because of its instability, the subgrade was then compacted with
can become complicated when an existing pavement is removed and a roller in static mode. A Class 4 geotextile was then placed over the
the subgrade is no longer confined. When the subgrade is exposed entire subgrade. The geotextile was covered with a high-strength,
to rainfall and construction traffic in an unconfined state, this further uniaxle geogrid from STA 45+28 to STA 46+03 RT. The geogrid
weakens the soil and makes it very difficult to restore its original was placed in the transverse direction across the geotextile and was
strength by compaction alone. The subgrade can be scarified and lapped by 600 mm. It was connected with steel tie wire because of
allowed to dry in the sun, then fine-graded and compacted. Any the unavailability of Bodkins connectors to mechanically attach the
unsuitable subgrade material can also be removed or undercut and geogrid sections together. The geogrid was placed in the transverse
replaced with dry earth if field conditions and time allow. The CBR direction so that the machine direction of the geogrid material, which
values during construction could actually be far lower as a result of has greater strength, would resist deformation of the subgrade caused
the disturbed condition of the subgrade soil, which would result in by construction equipment. The geocell was then placed. The geo-
an inadequate pavement design. cell depth was 100 mm and formed 2.4-m × 6.1-m panels when ex-
For the aforementioned reasons, a geocell confinement system was panded and staked down. All adjacent panels were connected together
considered a possible solution to overcome the weak subgrade con- using a Bostitch pneumatic staple. The heavy-duty stables were nec-
dition. The geocell system was recently introduced to strengthen the essary to maintain system continuity. The cells were then filled with
aggregate base layer. It has been reported to be lightweight and easy 2A subbase material. The geocell was filled by dumping the subbase
to install, and it can be backfilled with a variety of readily available material onto the adjacent existing pavement and pushing the subbase
aggregate. The geocell is thought to increase the structural capacity out onto the geocell with a small bulldozer. Once the geocell was
of the pavement without increasing the depth of the pavement struc- backfilled, triaxle dump trucks could back onto the filled cells and
ture and also reduce the stress on the subgrade resulting from traffic unload. The placement of the geosynthetics has not stopped the sub-
loading. The stress reduction is a function of design loading, the thick- grade from moving thus far. It did, however, allow heavy equipment
ness of each pavement layer, resilient modulus, and drainage capa- to traverse the subbase by distributing the point-load of the wheels
bility. The subbase should be designed to carry construction traffic over a larger area.
with minimum deflection and deformation. Compared with the con- On the following day, 200 mm of BCBC and 50 mm of ID-2 HMA
ventional alternative, the estimated savings from using the experi- binder were placed. No movement was observed during placement
mental geocell on this project was more than 45 percent according to of the first 100-mm lift of BCBC material. During placement of the
PennDOT analysis. second 100-mm lift of BCBC, there was very little movement. No
This research project evaluated the following: constructibility of movement of material was observed during placement of the 50 mm
a geocell system, its performance in stabilizing weak subgrade, and of ID-2.
its effectiveness in the field. To achieve the objectives of this pro- The following week, the pavement was removed from STA 43+75
ject, the road surface was visually inspected for base failure over a to STA 46+08 LT (“LT” indicates “left side”). The subgrade was
period of three years. Falling weight deflectometer (FWD) data were overexcavated 600 mm from STA 46+08 to STA 45+34 LT (CBR =
collected annually and analyzed. 6.5 percent in the undercut). The base thickness was increased in this
area. Class 4 geotextile was placed and covered with the following:
450 mm of AASHTO #1 modified (100-mm rock and No. 10 screen-
CONSTRUCTION AND FIELD INVESTIGATION ings), 150-mm 2A subbase, another layer of Class 4 geotextile, geo-
grid, and 100-mm-depth geocell filled with subbase. An additional
The geocell used on this project was constructed of high-density layer of geogrid was added and covered with 50 mm of subbase before
polyethylene with a carbon black content of 2 percent for stabiliza- being paved over with 200 mm of BCBC and 50 mm of ID-2. The first
tion against ultraviolet attack. Each geocell section was fabricated three rows of geocell were filled by hand to stiffen the leading edge.
using 60 strips 3.3 m long and 100 mm thick. The strips were ultra- The remainder of the subbase was then pushed slowly over the geocell
sonically spot-welded uniformly across each cell joint. The geocell with a small bulldozer to ensure the cells were completely filled. The
was used in combination with other geosynthetics such as geotextile geocell was manually held down during this phase of the backfilling
and geogrid. The geogrid used on this project was made from strips to ensure it did not float off the geotextile.
of polyester (Polyethylene Teraphthalate) that was heat-bonded at In the third week, the left side at the intersection of Lancaster
the joints (the product has been discontinued and is no longer avail- Avenue, SR-30, and Spring Mill Road, SR-320, was excavated. The
able). A high-strength woven geotextile and a nonwoven needle- subgrade soil exhibited the same color and weak characteristics as
punched geotextile were also used. The 10-oz nonwoven geotextile the previous reconstructed areas of roadway. The excavation started
28 Paper No. 00 - 1313 Transportation Research Record 1709

at STA 35+77 to STA 37+91 LT (CBR = 1.5 percent). This area also conducted over a period of two years. In this evaluation, the FWD
included a turning lane, which explains the extra width noted device dropped a calibrated mass of approximately 40 kN onto a
between some stations listed in Table 1. 300-mm-diameter plate in contact with the pavement surface. The
The right side was excavated the following week. The subgrade device recorded the magnitude of the applied load and the vertical
soil exhibited the same gray color. The excavation started at STA deflection response of the pavement surface at seven locations—the
35+75 to STA 38+50 RT (CBR = 4.6 percent). This area was also center of the loaded plate and six other locations offset from the loaded
undercut by 300 mm, covered with Class 4 geotextile, and back- axis (–300, 0, 300, 600, 900, 1200, and 1500 mm). In the analysis, the
filled with #1 modified, 300-mm depth. The #1 modified was then deflection at –300 mm was omitted.
covered with a 150-mm-thick layer of 2A subbase to smooth the Two types of analysis may be performed on FWD data. The sim-
top of the coarse-graded #1 modified aggregate. Then a layer of plest and most direct relies on computing a surface modulus (E0)
high-performance woven geotextile was placed. The woven geo- defined as the applied load divided by the measured axial deforma-
textile was then covered with 100-mm-thick geocell and backfilled tion. This value is analogous to a spring constant (kN/mm) and pro-
with 150-mm 2A subbase. The reinforced aggregate base was then vides a gross measure of the overall structural value of the pavement
covered with the plan thickness of HMA materials. Figure 1 provides system, including the subgrade. It requires that no assumptions be
details of each section. made relative to the thickness or elastic response of component layer
FWD data were initially collected on October 24, 1995, to estab- materials. However, the value is subject to modification in HMA-
lish a datum, and also on June 7, 1996, and June 2, 1997, at the test surfaced pavement because of the effects of temperature on the vis-
locations listed in Table 1. Each test location has different combi- coelasticity of the asphalt-bound materials, and it can be further
nations of geosynthetics or standard paving materials that were used influenced by the presence of an effective rigid layer underlying the
for comparison. pavement at some depth. In addition, the value is best suited only
for comparing measurements of pavement systems with the same
layer structure.
DATA COLLECTION AND ANALYSES A more sophisticated analysis is possible using various tech-
niques of backcalculation, which seek to match the observed pave-
Visual inspection of roadway over three years after reconstruction ment response to that returned by a mathematical model of layered,
has revealed no signs of pavement distress such as excessive rutting, linear-elastic half-space. The techniques generally rely upon vary-
cracking, or base failure. This indicates that the different combinations ing the linear-elastic moduli of the component material layers until
of geosynthetics are performing adequately. a satisfactory match to the observed surface deflection is achieved
In addition, an FWD analysis was used to perform a structural eval- (minimum error between measured and calculated deflections). The
uation of the reconstructed road. Three FWD measurements were techniques are also affected by the nonlinearity of the subgrade;

TABLE 1 Pavement Sections and FWD Test Locations

FIGURE 1 Pavement cross section.

thus, they may produce erroneous results if analysis is conducted 4000 were the weakest (on both sides). Both sections were originally
without the requisite skill and experience. concrete pavements overlaid with HMA. To quantitatively predict
FWD measurements were used to calculate the surface modulus the effect of geosynthetics, the subgrade resilient moduli of the sec-
and backcalculate the resilient modulus of the subgrade based on tions without geosynthetics were backcalculated. For the sections
known thicknesses and reasonably assumed resilient moduli of pave- with geosynthetics, the resilient modulus of the layer containing geo-
ment layers based on material testing and field experience. When cells, other geosynthetics, or both was varied until a reasonable sub-
considering the whole pavement system response in the analysis (sur- grade resilient modulus value was backcalculated. The backcalculated
face modulus), Stations 3375 and 4000 were the weakest. The con- modulus value was comparable to the one obtained for sections with-
crete section (Station 2400) had the highest surface modulus. The out geosynthetics. The assumed resilient moduli and measured layer
sections with 2A undercut showed greater surface moduli than sim- thickness for all layers are presented in Table 2a for the left-side
ilar designs without undercut. In general, because of the significant sections and in Table 2b for the right-side sections. The measured
variation in pavement system designs and because the method eval- deflections (from third FWD measurements) are shown in Figure 2.
uates the whole pavement system, the geosynthetic effectiveness The moduli values (backcalculated from all three FWD measured
could not be quantified based on surface moduli values. deflections) are presented in Figures 3 and 4.
ELMOD software was used for the backcalculation analysis. The To investigate the effect of geocell on the pavement performance,
analysis approach used in this software addresses the issues of tem- FWD measurements at Stations 2150 and 4450, 4550, and 4575 were
perature susceptibility of the HMA layer and the nonlinear behavior compared, as shown in Figure 3. E3 is significantly greater (almost
of the aggregate layer. The ELMOD program uses either the Ode- twice) for Stations 4450, 4550, and 4575 compared to station 2150.
mark-Boussinesq transformed section approach, or the “curve-fitting” E4 (the backcalculated value of subgrade) is in general greater for sta-
method normally used with numerical integration techniques, to back- tions 4450, 4550, and 4575, undercut #1 stone, than for Station 2150
calculate the resilient modulus of each layer. Odemark’s layer trans- and reached a similar value as 2A aggregate. However, when two
formation approach is used with Boussinesq’s equations to calculate layers of geogrids were used (Stations 3600 and 3601) instead of the
surface deflections in a homogeneous, isotropic, linear-elastic, and undercut and geotextile, no additional improvement was noted. In
semi-infinite space, and an iterative procedure is used to determine 1996 and 1997, the backcalculated subgrade resilient modulus was
those moduli that result in the same measured deflections. This even lower when geogrids were used.
approach has been found to be a good approximation of the general- Similarly, for the right side (Figure 4), a close comparison can be
ized Burmister equations when the HMA layer is thinner than the made between Stations 4425, 4550 and 2150. Results showed that
diameter of the loading plate and the modular ratio of two adjacent the inclusion of geocells, two layers of geogrid, and a geotextile has
layers (Ei /E i+1) is greater than 2 (6). In the analyses for this study, the almost doubled the resilient modulus of that layer. The backcalcu-
following resilient modulus assumptions, which were used through- lated subgrade resilient modulus is also higher, which would make
out the analyses, were made based on experience and data analysis the aforementioned results more conservative. In addition, compar-
of sections without geosynthetics: ID-2 HMA wearing surface is ing the performance of Patch 1 LT and Patch 2 RT indicates that
5175 kPa; base course HMA is 4830 kPa; asphalt-stabilized base geogrid had not provided a significant improvement to the system at
is 518 kPa; 2A aggregate is 193 kPa; concrete is 13.8 MPa; and the presence of the geocell.
undercut fill is 104 kPa. The difference between Stations 3725 and 3650 is the width of
Because of the different designs of the 12 sections, it was difficult the undercut. The responses of the two stations were similar, and the
to compare the measured deflections and to quantitatively associ- width of the undercut had no significant effect. However, the data
ate them with geocell/geosynthetic effectiveness. However, general show that the undercut in this case provides little improvement when
observations can be made. For instance, sections at Stations 3375 and geocell is used (compared to stations 4425 and 4550).
TABLE 2 Resilient Moduli: (a) Left-Side Pavement Layers, ( b) Right-Side Pavement Layers
FIGURE 2 Measured deflection, 1997: (a) left side, ( b) right side.
FIGURE 3 Backcalculated resilient modulus for left side: (a) 1995, (b) 1996. (continued on
next page)
Al-Qadi and Hughes Paper No. 00 - 1313 33

FIGURE 3 (continued ) Backcalculated resilient modulus for left side: (c) 1997.

FIGURE 4 Backcalculated resilient modulus for right side: (a) 1995. (continued on next
page)
FIGURE 4 (continued ) Backcalculated resilient modulus for right side: (b) 1996,
(c) 1997.
Al-Qadi and Hughes Paper No. 00 - 1313 35

CONCLUSIONS help in this study: Bruce Shelly, Lorraine Davies, Tom Leitzel, Dave
Wassel, John Pfander, and Dennis Brehm.
In all cases, it was difficult to isolate the effect of the geocell con-
finement system as it has been used in combination with geogrid,
geotextile, or both. However, it can be concluded that in sections REFERENCES
where 100-mm-thick geocells were used, the resilient modulus of the
aggregate layer increased by almost twofold due to the material con- 1. Al-Qadi, I. L., T. L. Brandon, and S. A. Bhutta. Geosynthetically Stabi-
finement. As a result of the aggregate confinement provided by the lized Flexible Pavement. Geosynthetics 1997, Long Beach, Calif., 1997,
pp. 647–662.
geocell and the subgrade-subbase separation provided by the geo- 2. Geotextile Design Examples, Contract DTFH-86-R-102. FHWA, U.S.
textile, it appears that a geotextile-geocell combination may provide Department of Transportation, 1989.
a significant improvement when used on top of a weak subgrade of 3. Cedergren, H. R. Seepage, Drainage, and Flow Nets, 3rd ed. John Wiley
heavily trafficked pavement. & Sons, Inc., New York, 1989.
4. Koerner, R. M. Designing with Geosynthetics, 3rd ed. Prentice Hall,
Englewood Cliffs, N.J., 1994.
5. Al-Qadi, I. L., and S. A. Bhutta. Designing Low-Volume Roads with
Geosynthetics. 7th International Conference on Low-Volume Roads. In
ACKNOWLEDGMENTS Transportation Research Record 1652, Vol. 2, TRB, National Research
Council, Washington, D.C., 1999, pp. 206–216.
The authors wish to acknowledge the support of the New Products 6. Ullidtz, P., and K. R. Peattie. Pavement Analysis by Programmable
Calculators. Transportation Engineering Journal of the ASCE, Vol. 106,
Evaluations and Research Unit of PennDOT, the Engineering District No. TE5, Sept. 1980.
6-0 Construction Unit’s SR-30 project staff, and Delaware County
Maintenance District 6-3. Special thanks go to the following for their Publication of this paper sponsored by Committee on Geosynthetics.