Geotextiles and Geomembranes 43 (2015) 162e170

Contents lists available at ScienceDirect

Geotextiles and Geomembranes

journal homepage: www.elsevier.com/locate/geotexmem

Laboratory evaluation of flexible pavement structures containing

geocomposite drainage layers using light weight deflectometer
Jean-Pascal Bilodeau a, *, Guy Dore

a, 1, Catherine Savoie b, 2
a
Civil Engineering Department, Universit
e Laval, Quebec, QC G1V 0A6, Canada
b
Service de l'exploitation des CS de la Capitale Nationale, Minist
ere des Transports du Qu

ebec, 5353, boulevard Pierre-Bertrand, Qu

ebec, QC G2K1 M1,
Canada

a r t i c l e i n f o a b s t r a c t

Article history: The behaviour of flexible pavement structures in wet environments is significantly influenced by the
Received 9 October 2014 infiltration of excess water. The use of geocomposite drainage layers is an alternative to promote rapid
Received in revised form removal of any infiltration of excess water. Using four small-scale flexible pavement test sections, this
28 January 2015
laboratory study focused on measuring the effectiveness of the different configurations of geocomposite
Accepted 1 February 2015
Available online 18 February 2015
drainage layers in terms of reduction in water content and improvement of mechanical properties during
a drainage period. The experiment allowed measuring the significant effects of each drainage configu-
ration on pavement structural strength in the short-term and long-term drainage behaviour of each
Keywords:
Geosynthetics
experimental pavement structure. The shallow drainage blanket, as well as the vertical drain, showed a
Flexible pavements more pronounced effect in the top layer of the structures, while the deep drainage blanket showed a
Drainage limited effect. From the results obtained, a drainage configuration combining a shallow drainage blanket
Lightweight deflectometer with a vertical drain may be optimal.
© 2015 Elsevier Ltd. All rights reserved.

1. Introduction
Flexible pavement performance is significantly influenced by
climatic conditions, especially in northern environments where
freeze-thaw cycles and abundant precipitations are frequent (Dore

and Zubeck, 2009). In pavement design, water is recognized as one
of the key parameters that needs to be controlled in order to ensure
good performance (Uthus, 2007; Uthus et al., 2006; Witczak et al.,
2000; Richter, 2006; Lebeau, 2006; Dore
and Zubeck, 2009;
Bilodeau and Dore
, 2012; Savoie et al., 2012; Ekblad, 2007;
Carrera et al., 2009), and modern pavement geometry and design
techniques are focused on providing adequate drainage to elimi-
nate water underneath the pavement surface. However, to a certain
extent, the presence of moisture in the pores of unbound granular
materials and soils can have a positive effect. This is due to a
confining stress, referred to as matric suction, generated under the
unsaturated state at the airewater interface (Fredlund and
* Corresponding author. Tel.: þ1 418 656 2131x7242; fax: þ1 418 656 2928.
E-mail addresses: jean-pascal.bilodeau@gci.ulaval.ca (J.-P. Bilodeau), guy.dore@

), catherine.savoie@mtq.gouv.qc.ca (C. Savoie).
gci.ulaval.ca (G. Dore
1
Tel.: þ1 418 656 2203; fax: þ1 418 656 2928.
2
Tel.: þ1 418 643 0095x229.
Rahardjo, 1993; Lebeau, 2006). The matric suction increases with
a decrease of the degree of saturation. Geomaterial mechanical
properties are influenced to varying degrees by this stress param-
eter. An increase of matric suction is generally associated with an
increase of stiffness and bearing capacity (Witczak et al., 2000;
Bilodeau et al., 2010; Arnold et al., 2002; Bilodeau and Dore
,
2012; Theyse, 2002; Uthus, 2007). Water infiltration through
pavement surface produces excess water in pavement geomaterials

and
due to subhorizontal flow, capillary rise and frost suction (Dore
Zubeck, 2009; Lebeau, 2006; Swanson, 1985; Brandl, 2001). As a
result, their mechanical properties may decrease as the environ-
ment changes through the yearly cycle (Dore
and Zubeck, 2009;
Huang, 2004). Because of that, the rapid removal of excess water is
vital to ensure good behaviour of soils and materials in the layered
pavement system.
Geotextile drainage layers have been used over the last decades
to remove excess water from the pavement structures. These layers
typically have a hydraulic conductivity significantly higher than
geomaterials. They are typically used in the pavement system as
drainage blankets at various depths in the pavement layers or as
vertical drains at the pavement edge. Good results were generally
obtained in previous experimental or modelling projects where
geotextile layers were used to improve the drainage of pavement

http://dx.doi.org/10.1016/j.geotexmem.2015.02.002
0266-1144/© 2015 Elsevier Ltd. All rights reserved.
 J.-P. Bilodeau et al. / Geotextiles and Geomembranes 43 (2015) 162e170
test sections (Bilodeau et al., 2014; Lebeau, 2006; Lafleur and
Savard, 1995; Collins et al., 2005; Christopher et al., 2000; Al-Qadi
and Appea, 2003; Loulizi et al., 1999; Perkins et al., 2005; Scott
et al., 1998; Evans et al., 2002; Bahador et al., 2013). The work done
by Christopher et al. (2000), Evans et al. (2002) and Scott et al.
(1998) showed that geotextile drainage blankets placed in the
lower part of the pavement structures can effectively remove
excess water. This conclusion is partly supported by the work done
by Collins et al. (2005), who concluded that geotextile may
contribute to base materials stiffening over time for thinner pave-
ment sections, while there were no significant effects on thicker
pavement sections. Geotextile used as vertical drains on the pave-
ment sides also effectively reduced water infiltration from the
shoulders (Blond and Mlynarek, 1999; Blond et al., 2000) and
would correspond to the best pavement configuration regarding
pavement structural response following a saturation event
(Bilodeau et al. 2014). However, limited data are found on experi-
mental research work done in controlled laboratory environments.
Therefore, this research aims to quantify the mechanical properties
and drainage capacity of laboratory small-scale flexible pavements
constructed in the laboratory with geosynthetic drainage layers. It
is proposed to use the light weight deflectometer (LWD) to measure
the surface deflections and mechanical properties of the pavement
structure.
2. Materials and methods
Four experimental small-scale flexible pavement sections were
built in the laboratory. One of the test sections is the control section
without any drainage layer. The other three sections were built
with three different drainage configurations: drainage blanket be-
tween the base and the subbase (DB-B-SB), drainage blanket be-
tween the subbase and the subgrade (DB-SB-SG) and a vertical
drain (VD). The 7 mm thick vertical drains are non-woven geo-
textiles assembled by the needling process and made of poly-
propylene. The filter opening size (FOS) was 150 mm and the weight
was 900 g m2. The materials used for the drainage blankets are
made of non-woven polypropylene fibers assembled by the
needling process. Each blanket consisted of a drainage layer and
perforated drainage tubes positioned every 300 mm (20 mm
external diameter), both inserted between filtration layers (FOS of
120 mm). The weight of the drainage blanket product is 400 g m2
and its total thickness (excluding the drainage tubes) is 3.9 mm. For
the test conditions, the transmissivity is 2  104 and
5  104 m2 s1 for the vertical drain and the drainage blanket,
respectively.
The sections were built inside a circular container having a
height of 1.3 m and a diameter of 1.2 m as shown in Fig. 1. A 200 mm
open-graded drainage layer (OGDL) was placed at the bottom of the
container and connected to a drainage line. A geotextile was placed
at the top of the OGDL prior to the construction of the other
pavement layers. The pavement structures that were tested for the
four configurations consisted of a four-layer system: 500 mm of
subgrade (SM), 300 mm of subbase (SP), 200 mm of base (GW) and
50 mm of dense graded asphalt concrete (classified as EB10S, a
surfacing asphalt concrete used in Quebec with a nominal maximal
particle diameter of 10 mm). This structure is similar to full scale
regional and local pavement structures encountered in Quebec
built for low traffic conditions and medium subgrade soil frost
sensitivity. For example, a local pavement designed for 1  105
ESALs (Equivalent Single Axle Loads) on a silty sand soil (SM) (with
a fines content lower than 30%) would consist of 60 mm of asphalt
concrete, 200 mm of granular base and 300 mm of granular sub-
base. As Quebec road network includes many kilometers of low
volume regional and local roads, the tested pavement structure in
163
the laboratory are of great interest. Fig. 2 presents the grain-size
distribution of the materials used for the test sections. This figure
also summarizes the main gradation parameters, such as the uni-
formity coefficient Cu, the mean diameter d50 and the fine particles
percentage %F. According to USCS soil classification, the base, sub-
base and subgrade soil are classified as GW, SP and SM respectively.
The hydraulic conductivity of the three materials was measured in a
previous study (Savoie et al., 2012) and is equal to 9  106,
5  105 and 2  106 m s1 for the base, subbase and subgrade
materials, respectively.
For each experimental section, the subgrade, subbase and base
were compacted approximately at 90, 95 and 98% of their respec-
tive modified optimum proctor. In order to verify the compaction
level of the layer during construction, the mass and average water
content of the soil placed in the tank was measured, as well as the
thickness of each layer. This was done by weighing the mass of soil
inside each bucket added in the tank, by measuring the water
content of each compacted layer using the microwave method and
by measuring the average depth difference before the soil was
added and after the soil was compacted. A slope of 2% was profiled
at the top of each layer between the centre and the wall of the
container. Most of the time, the water contents of the compacted
layers were slightly less than the optimum moisture content. After
compaction of the asphalt concrete, nine drainage holes were
drilled inside this layer to ensure adequate air circulation. As shown
in Fig. 1, the test sections were equipped with individual volumetric
water content transducers at the centre of each layer. The volu-
metric water content qV is expressed as
VW
qV ¼ ¼ n  Sr (1)
VT
in which VW is the water volume, VT is the total volume, n is the
porosity and Sr is the degree of saturation. Referring to Fig. 1, the
DB-B-SB was positioned along the horizontal plane in which the
line aeb is included, the DB-SB-SG was positioned along the hori-
zontal plane in which the line ced is included and the VD was
positioned along the vertical plane aebecedeeef. Because of the
particular container geometry and configuration, when a drainage
blanket was used, a vertical drain was used from the blanket level to
the bottom of the subgrade layer in order to make the configuration
work. For full-scale pavements, the drainage blanket would be able
to drain at the pavement sides. In the container, as water evacua-
tion is located at the bottom of the container, the drainage blankets
were connected to vertical drains to ensure that the blanket system
worked properly.
Following construction of the test sections, each of them was
saturated using a water reservoir connected to the bottom drainage
line. The water level was gradually increased using a water head of
6 inches until the water level reached the bottom of the asphalt
concrete layer. This was verified using the piezometer (Fig. 1).
Measurements of the mechanical behaviour of the experimental
small-scale pavement structures were made when saturation had
completed. The mechanical behaviour was measured by a light
weight deflectometer (LWD) equipped with a loading plate having
a diameter of 150 mm. The plate diameter was selected in order to
ensure minimized stress levels at the tank edges during LWD
loading. Previous studies (Boutet et al., 2010; Nazzal et al., 2004)
showed that the selected configuration is adequate for this
container, as the zone of significant stress does not reach the tank
wall. A mass of 20 kg was used and the drop height was adjusted in
order to obtain a contact stress of approximately 560 kPa. A rub-
bermat was used between the asphalt concrete surface and the
LWD as recommended by the manufacturer. It was demonstrated
by Nazzal et al. (2004) that the depth at which the LWD induces
164 J.-P. Bilodeau et al. / Geotextiles and Geomembranes 43 (2015) 162e170

Fig. 1. Scheme of the experimental small-scale flexible pavements built inside a laboratory tank; a) Side view; b) Top view.

significant vertical stresses could reach approximately 1.5e2 times drainage times as presented in Table 1. The time interval between
the loading plate diameter. Therefore, for this study, it corresponds each measurement was shorter at the beginning of the testing, but
to the interface layer between the base course and subbase mate- gradually increased with time during the testing period, which
rials at a depth of approximately 225e300 mm. As shown in Fig. 1, lasted 10 days. During this period, the water contents were
the LWD tests were performed at three different locations around recorded every minute.
the pavement centre. At each of these locations, 5 LWD drops were
made: two to ensure good seating of the loading plate and three to
3. Experimental results
measure the surface deflection. An average value of three drops for
one location was used to establish the mean surface deflection for a
The results associated with the construction of the experimental
particular location and an average overall deflection value was
sections are presented in Table 2. The results are presented in terms
calculated using the deflection from the three locations. Therefore,
of compaction level (%) obtained for each layer, determined by the
a LWD test consists of measuring a total of 9 considered deflection
ratio of the obtained dry density to the maximum dry density.. The
values obtained from three locations around the pavement centre.
soil and granular materials porosity n obtained for each layer and
The deflection behaviour was first measured at the saturated state
pavement configuration is also provided. A summary of the main
following the saturation of the tank. Using the surface deflection,
soil characterization data, which are the maximum dry density
the surface modulus Et for a drainage time t can be obtained with
(rdmax), optimum water content (wopt) and density of solid particles
the Boussinesq solution (Ullidtz, 1998) expressed as
(rs) is provided in Table 2. In addition, the obtained OGDL porosity

was 40%. Fig. 3 presents the evolution of the volumetric water
2 1  m2 st a
Et ¼ (2) content at the centre of the tank and at the centre of each layer as a
dt function of time. This figure clearly shows the volumetric differ-
ences between each soil layer, as the volumetric water contents are
in which a is the loading plate radius, dt is the surface deflection
not of the same magnitude. Moreover, the drainage capacity of each
for a drainage time t, st is the surface contact stress obtained for a
material can also be appreciated as the subbase and base material
LWD drop at a drainage time t and m is the Poisson ratio (0.35).
volumetric water contents change more significantly than the silty
Following the initial measurement of the surface modulus for the
sand. Typically, the volumetric water content of the subgrade soil
saturated state, the pore water in the pavement layers was
rapidly reaches a stable condition in comparison with the granular
allowed to drain by removing the water reservoir at the bottom of
layers. Fig. 4 presents a closer look at the volumetric water content
the tank. The surface modulus values measured at various
data recorded for the first 80e160 min of drainage in order to study
the early drainage rate of each of the experimental section. This
analysis is presented for the base and subbase layer.

Table 1
LWD measurements schedule.

Time Frequency of LWD measurements

Day 1 Every hour

Day 2 and 3 Every 2 h
Day 4 and 5 Three times a day
Day 6 and 7 Two times a day
Day 8, 9 and 10 One time a day
Fig. 2. Grain-size distribution of the pavements materials.
 J.-P. Bilodeau et al. / Geotextiles and Geomembranes 43 (2015) 162e170 165

Table 2 the average of three measurements obtained from LWD drops. It

Characteristics of the experimental pavements. can be observed that a relationship between relative surface
Characteristic Pavement layera Pavement configuration modulus and time for each experimental section is obtained by
Reference VD DB-B-SB DB-SB-SG
averaging the results from the three different measurement loca-
tions. Differences in the drainage configurations influence the re-
% Compaction Subgrade 89 85 90 89
covery of the mechanical properties of the experimental pavements
Subbase 95 95 95 95
Base 98 95 98 98 as observed for each tank configuration. A power model was fitted
to the experimental data in order to analyze the effect of the geo-
Porosity n (%) Subgrade 25 28 24 25
Subbase 32 32 32 32
composite drainage layers on the recovery of the mechanical
Base 17 20 17 17 properties of the small-scale laboratory pavements. The power
model is expressed as
a
Base: Maximum dry density rdmax ¼ 2230 kg m3; optimum water content
wopt ¼ 6%; Grain density rs ¼ 2640 kg m3.Subbase: Maximum dry density
rdmax ¼ 1920 kg m3; optimum water content wopt ¼ 11.5%; Grain density ER ¼ At B (4)
rs ¼ 2690 kg m3.Subgrade: Maximum dry density rdmax ¼ 2230 kg m3; optimum
water content wopt ¼ 4.6%; Grain density rs ¼ 2640 kg m3. in which t is the drainage time and A and B are regression param-
eters. The power model was selected on the basis of the best fit
Typical results of the mechanical characterization of the criteria, quantified by the coefficient of determination (R2). The
experimental small-scale pavement sections using the LWD are parameter B is associated with the relative rate of recovery of the
presented in Fig. 5a for the reference section. The results are pre- mechanical properties of the experimental pavements as a function
sented as relative surface modulus, which corresponds to the sur- of time. The results of the data modelled with the power model are
face modulus at a drainage time t (Et) normalized by the surface presented in Fig. 5b. The model parameters and the obtained co-
modulus measured at a drainage time of t ¼ 0 hours (E0). The efficients of determination R2 are summarized in Table 3.
relative surface modulus is defined as

Et
ER ¼
E0
in which ER is the relative surface modulus (no units). The plots
presented in Fig. 5a are the average results obtained from the three
tests locations and the four pavement configurations. It should be
remembered that each relative modulus point considered is also
4. Results analysis
(3)
4.1. Effect of geocomposite drainage layer on drainage of the
experimental pavements
When looking at the drainage graphs in Fig. 3, it can be observed
that the critical drainage period before which most of the

Fig. 3. a) Volumetric water content according to time for the reference section; b) Volumetric water content according to time for the DB-SB-SG section; c) Volumetric water
content according to time for the VD section; d) Volumetric water content according to time for the DB-B-SB section.
166 J.-P. Bilodeau et al. / Geotextiles and Geomembranes 43 (2015) 162e170

Fig. 4. Early drainage behaviour of the experimental pavements; a) Granular base; b) Granular subbase layers.

volumetric water content variations occur is approximately
1000 min. Before that critical time, the volumetric water content
changes rapidly and reaches values near the residual volumetric
water content. This decreasing rate is due to matric suction
developing in the soil pores at the unsaturated state. For the base
and subbase layers, the residual water contents reach values of
volumetric water content around 0.1. However, when it comes to
the subgrade soil, the residual water contents are more variable
from one experimental section to another. Based on the experi-
mental results, the residual water content at the end of the drainage
period appears to be influenced by the presence and the configu-
ration of the drainage systems. For the base layer, the DB-SB-SG is
located far from the layer and no effect was observed in terms of
residual water content in comparison with the reference section.
However, the VD configuration allows obtaining slightly lower re-
sidual water content (0.09), while the DB-B-SB significantly reduces
the residual water content (0.075). For the subbase, the drainage
configurations do not have a significant impact on the magnitude of
the residual water content, which is always in the range of 0.06. In
order to study the long-term drainage behaviour of the experi-
mental pavements, the relationship between volumetric water
content qV and time was approximated using a power model in
Fig. 3. The simplified equation is stated as
qV ¼ at b (5)
in which t is the time and a and b are regression parameters, b
representing the drainage rate. The a and b parameters obtained
are summarized in Table 4 with the associated coefficients of
determination R2. The R2 values, ranging from 0.87 to 0.999 (with
an average value of 0.967), suggest that a good agreement is
observed between the collected data set and the values predicted
by the selected model. As the data presented in Table 4 suggests, for
the long-term drainage behaviour, the drainage configuration does
not have a significant effect on the evolution rate of qV as a function
of time (b) in the subgrade. For the subbase, the use of drainage
layers does not have a significant effect on the volumetric water
content measured at the centre of the layer, as b values slightly
higher or slighter lower than the reference section are observed.
The DB-B-SB configuration appears to slightly increase the b value
in the subbase, while the VD b value is almost equal to the value of
the reference tank. For the long-term drainage behaviour, the more
pronounced effect is measured in the base layer, as b values of
1.4e1.5 times the value of the reference section are noticed.
The data presented in Fig. 4 allow comparing the drainage rate
early in the drainage period for the base and the subbase layers. The
graphs present the evolution of the volumetric water content as a
function of time for the first 80 (base) to 160 (subbase) minutes of
the drainage period. When looking at the drainage slopes observed
at the beginning of the drainage period for the base layer, it is
observed that the drainage configuration influences the evolution
of the volumetric water content. The shallow drainage blanket

Table 3
Power model parameters for the relationship of ER versus time.

Pavement A B R2 Area under Relative area

configuration curve (h1) under curve

Reference 1.1603 0.03651 0.971 342 1

VD 1.1438 0.0448 0.955 350 1.02
Fig. 5. a) Average ER versus time relationships measured for the four drainage con-
DB-B-SB 1.0838 0.0634 0.977 362 1.06
figurations; b) Modelled ER versus time relationships for the four experimental
DB-SB-SG 1.174 0.03618 0.787 346 1.01
sections.
 J.-P. Bilodeau et al. / Geotextiles and Geomembranes 43 (2015) 162e170 167

Table 4 surface modulus than the reference at a specific drainage time.

Power model parameters for the relationship of qV versus time.
Pavement configuration Pavement layer a b
Reference Subgrade 0.3802 0.01211
Subbase 0.167 0.1148
Base 0.2343 0.08838
VD Subgrade 0.2695 0.0316
Subbase 0.1691 0.1104
Base 0.3137 0.1266
DB-SB-SG Subgrade 0.3856 0.0052
Subbase 0.1742 0.1067
Base 0.3991 0.1368
DB-B-SB Subgrade 0.409 0.0126
Subbase 0.2246 0.1381
Base 0.2418 0.1252
appears to be the most effective technique for the base, as its
volumetric water content variation is the most important for the
considered period and as the water content at 80 minutes is the
lowest. This is in good agreement with the proximity of the geo-
textile layer according to the granular base layer and with the fact
that the length of the drainage flow path for the excess water is
shorter (maximum 200 mm) in the vertical direction. For the period
considered, in the base layer, the VD presents drainage behaviour
similar to that of the reference section. This can be explained by the
fact that, as the volumetric water contents are measured at the
mid-depth of the layer at the centre of the tank, the VD is located at
a radial distance of 0.6 m from the transducer. This represents a
longer flow path in the horizontal axis for the excess water and the
influence of the drain on water located at the centre of the tank may
be limited. Based on the analysis of the moisture regime in the base
layer, the use of the DB-SB-SG presents the less performing
behaviour in the early drainage period. Indeed, the total volumetric
water content variation over the considered period is the lowest
and the volumetric water content at the end of the considered
period is the highest. As the deep drainage blanket is located far
from the base layer, its effect can be considered limited. Concerning
the effect of the drainage layer on the early drainage behaviour in
the subbase (0 to 160 min), the VD and DB-SB-SG show similar
trends, which reveal a faster decrease of the volumetric water
content when compared to the reference section. However, the
reference section shows a better behaviour than the DB-B-SB sec-
tion regarding the early drainage behaviour in the subbase.
Nevertheless, at 160 min, every experimental tank reached the
same volumetric water content in the subbase layer (approximately
0.1).
4.2. Effect of geocomposite drainage layer on pavement structure
mechanical recuperation
The results presented in Fig. 5a summarize the deflection data
collected for each test section during the drainage period. The re-
sults exposed in Table 3, with R2 ranging from 0.787 to 0.977,
suggest that the use of the power model adequately represents the
mechanical recovery phenomenon of the experimental pavement
structures. It can be observed that for each tank configuration, there
is an increasing trend with slight scattering for the relative surface
modulus ER. This slight scattering is interpreted as the effect of
small laboratory room temperature variations. Some cases, the VD
and the DB-B-SB, show higher ER for a given drainage time. Higher
ER at a specific time is associated with a higher surface modulus.
When a comparison is made with the reference pavement struc-
ture, a higher ER value means that the experimental section
equipped with a specific drainage configuration presents a higher
Fig. 5b presents the modelled experimental data collected for the
R2 four small-scale experimental pavements. The use of the power
0.870
model for the analysis of the mechanical recovery of the experi-
0.996 mental pavements allows obtaining information on the effect of the
0.992 geotextile drainage layers used at various positions within the
0.884 structures. It can be noticed that the relative surface modulus,
0.996 which quantifies the mechanical recovery rate of a given drainage
0.988 configuration relative to the reference section, is higher for each of
0.930 the experimental pavements built with a geotextile layer. There-
0.999 fore, the small-scale experimental pavements including geo-
0.997 composite drainage recover their mechanical properties faster than
0.988 the reference section. For example, according to the power models,
0.966 the time necessary to reach a relative surface modulus of 1.4 is 172,
0.994 93, 56 and 127 h for the reference, VD, DB-B-SB and DB-SB-SG,
respectively. The analysis of the regression parameters of the po-
wer models in Table 3 reveals that the recovery rate of the relative
surface modulus, characterized by the parameter B, may reach a
value of 1.74 times the reference tank B value for the DB-B-SB
section (1.74 ¼ 0.0634/0.03651; recovery rate B in Table 3). Based
on the B value, the VD also appears to be an efficient drainage
technique.
Table 3 also presents the calculated surface area under the curve
for the power models fitted to each of the tested pavement con-
figurations. The relative surface area under the curve, which is
expressed as the area under the curve for a given test configuration
divided by the area under the curve for the reference test section, is
also presented. These results quantify the total benefice which can
be expected for a saturation event of the pavement materials. For
the purpose of this study, the drainage period was monitored for
250 h from the weakest bearing capacity state, which was when the
water level was raised at the top the base layer. Dore
and Zubeck
(2009) proposed a relationship between surface deflection and
the number of equivalent axle loads N (ESAL) that can be applied
before rehabilitation. This relationship and its stiffness equivalent
derived from the Boussinesq solution are expressed:
dmax ¼ 537:08N 0:2694 (6)
E ¼ 274:48N0:2694 (7)
in which dmax is the maximum surface deflection under the load
(mm) and E is the surface stiffness (MPa). The calculation of the
relative surface area allows obtaining a direct comparison of the
surface stiffness of a pavement structure built with drainage geo-
synthetics to a reference pavement structure. For the drainage
period considered, a limited effect ranging from 1% (DB-SB-SG) to
6% (DB-B-SB) can be obtained from the use of specific geosynthetics
drainage configurations. Given the existing relationship between
surface stiffness (E) and allowable traffic (N), this analysis suggests
that the pavement life may be increased from 1 to 6% when
drainage geosynthetics are used. These benefices also have to be
weighted with the total time during the yearly cycle a pavement
structure is submitted to a near saturation event and a following
drainage period. Therefore, the expected benefice could be lower
and may vary according to climatic regions.
A similar study was conducted at an outdoor research facility
(Bilodeau et al., 2014) with the same materials as the ones used for
this indoor laboratory study. The analysis of Bilodeau et al. (2014)
was focused on the expected structural benefice during a ten
days drainage period following an initial raise of the water table to
the top of the subgrade soil. Full-scale pavement structures built in
a 30 m long, 9 m wide and 2 m deep concrete pit with water table
control were used to perform the study. It was concluded that the
168 J.-P. Bilodeau et al. / Geotextiles and Geomembranes 43 (2015) 162e170
VD section provided the better mechanical recovery of the entire
pavement structure. This supports the finding of this study, as the
VD also provides a structural benefice. The DB-B-SB pavement
section tested by Bilodeau et al. (2014) showed limited effects,
mostly due to two reasons: 1. the water table was not raised higher
than the top of the subgrade soil and; 2. as the pavement surface
condition was pretty good for the outdoor tests, the water content
variations in the granular materials was limited. These are inter-
preted as the main reasons explaining the limited effects of the DB-
B-SB in the study of Bilodeau et al. (2014). However, as the exper-
iments described in this study and the experimental work of
Bilodeau et al. (2014) are basically the same, but at different scale,
the previous findings and conclusions satisfyingly support and
validate the results of this study.
4.3. Combined interpretation of drainage and mechanical recovery
The mechanical performance of the VD and DB-B-SB tanks is in
good agreement with the analysis performed on the volumetric
water content data. Fig. 6a and b presents the relationship between
the ER value and the volumetric water content and matric suction in
the base layer. This choice was made because the LWD stress bulb
affects mostly the asphalt concrete layer and the granular base. The
matric suction was estimated using the approach suggested by Co ^ te


and Konrad (2003). This model is defined
qV ¼ qS for j < ja
 l
j (8)
qV ¼ qS a for j  ja
j behaviour of the subbase. The drainage data in the base layer
in which qV is the volumetric water content, qS is the saturated
volumetric water content, j is the matric suction, ja (kPa) is the air-
entry value and l is the pore-size distribution index. The value of ja
and l are estimated using the material properties with
LOGðja Þ ¼ 3:92  5:19nf (9)
l ¼ 0:385  0:021  n0:65
f  SSF (10)
in which nf (decimal) is the fine fraction porosity and SSF (m2 g1) is
the fine particles specific surface (determined with the methylene
blue test). The fine fraction porosity is calculated using
Fig. 6. a) ER versus volumetric water content in the granular base for the tested experimental pavements; b) ER versus matric suction in the granular base for the tested exper-
imental pavements.
n
nf ¼ (11)
n þ ð1  nÞ%F
in which n is the soil porosity (decimal) and %F (decimal) is the fine
particles content. In order to use this model and to calculate the
corresponding matric suction from volumetric water content, the
data presented in Table 2 were used. The values of l and ja were
determined using the volumetric properties of the base layer for
each test section (n and %F), as well as with a SSF value of
3.57 m2 g1 determined using the methylene blue test on the fine
particles.
It can be observed in Fig. 6a and b that the increase of the average
surface modulus according to time is significantly related to the
moisture conditions (volumetric water content and matric suction)
in the granular base layer. A decrease of the volumetric water con-
tent, as well as an increase of the matric suction, in the unbound base
layer is associated with an increase of ER. The rapid drainage of
excess water (Fig. 3) in the base layer for the DB-B-SB is also observed
in Fig. 6, as lower volumetric water contents and higher matric
suctions are observed for this configuration. For a given value of ER,
the experimental pavement with the shallow drainage blanket
shows lower water content and higher matric suction, as these
values are reached earlier in the drainage process due to faster
drainage conditions (Fig. 3). The VD does not show a significant ef-
fect when compared to the reference section, as these mechanical
response measurements are mostly associated with the pavement
base (Figs. 3 and 4). As a matter of fact, it was discussed that the effect
of the VD on the water contents in the base layer is not significant.
Nevertheless, it has a more pronounced impact on the drainage
collected for the DB-SB-SG experimental section suggested that this
technique was less efficient when compared to the reference. This is
also shown in Fig. 6, as higher water contents and lower matric
suction were measured. For the drainage data collected, this is
associated with the reach of a given ER value later in the drainage
process. However, the fact that the VD and DB-SB-SG allow a slightly
faster recovery (Fig. 5) in comparison with the reference section
suggests that the drainage in the subbase (Figs. 3 and 4) has a
noticeable, but lower, impact on the ER in comparison with the effect
of the base layer. This is in good agreement with the fact that the
stress bulb may reach depths somewhere between the lower base
and upper subbase. However, the effect of the pavement base ap-
pears to be much more significant than the pavement subbase.
 J.-P. Bilodeau et al. / Geotextiles and Geomembranes 43 (2015) 162e170
4.4. Summary
It was shown that the DB-B-SB presents the lowest residual
volumetric water content in the granular base over the long-term,
but also early in the drainage period. It is also one of the experi-
mental pavements with higher b values for the base and the
subbase when the long-term drainage behaviour was analyzed
using a power model. These findings are also in good agreement
with the fact that the LWD stress impulse affects mostly the
granular base. The VD section also has a significant positive effect
on the recovery of the mechanical properties of the experimental
small-scale flexible pavement. This is in agreement with the fact
that the VD effect is recognized for the long-term drainage
behaviour of the granular base. This effect was less obvious when
the early drainage data were analyzed. However, the early
drainage data revealed that the effect of the VD in the subbase
layer was as good and significant as the effect of the DB-SB-SG.
Therefore, on the basis of the data collected on the small-scale
experimental flexible pavements tested in laboratory controlled
conditions, it appears that shallow drainage blankets used in
combination with vertical drains have a significant effect on the
drainage behaviour and, consequently, on the recovery of the
mechanical properties of the structures.
5. Conclusion
An experimental study was conducted in laboratory in order to
assess the effect of geocomposite drainage layers on small-scale
pavement structures. The effectiveness of the drainage layers was
measured in terms of evolution of volumetric water content, as well
as the recovery of the pavement structural strength measured with
a light weight deflectometer. The experimental pavement struc-
tures were built with geocomposite drainage layers placed at
various locations in the pavement system. Each experimental
pavement was saturated with water followed by a drainage period
during which water content and surface deflections were moni-
tored. Following the analysis of the collected data, the following
conclusions were made:
 Over the long-term or the short-term, the residual water con-
tent and the drainage rate in the base layer were improved
when a drainage blanket between the base and the subbase was
used, while deeper drainage layers had a limited effect;
 Subbase layer residual water content was not significantly
influenced by the use of geocomposite drainage layers;
 Long-term drainage rate was slightly increased for the granular
pavement layers when a shallow drainage blanket was used;
 Drainage layers had enhanced the recovery of the mechanical
properties of the pavement system;
 The rate of the increase of the surface modulus can reach 1.74
times the one obtained for the control section when the
drainage blanket between the base and the subbase is
considered;
 Based on the recovery of the surface modulus during the
drainage period, the ranking of the drainage configurations,
from the most effective to the less effective, are as follows: 1) the
drainage blanket between the base and the subbase, 2) the
vertical drain, and 3) the drainage blanket between the subbase
and the subgrade;
 Based on the observed results of pavement structural response
and the measured effect of each drainage configuration, the
combination of a shallow drainage blanket and a vertical drain
would be optimal for drainage and recovery of flexible pave-
ment structural strength.
169
Acknowledgements
The research work presented in this paper is part of the NSERC
industrial research Chair on the interaction between heavy loads,
climate and pavements. The authors wish to thank the Natural
Sciences and Engineering Research Council of Canada (NSERC) and
the industrial partners for their financial support for this research.
Special thanks to Texel crew for providing materials and technical
advice for the project. The authors are grateful to the research as-
sociates, graduate students, technicians and research engineers
who also contributed to this study.
References
Al-Qadi, I.L., Appea, A.K., 2003. Eight-year field performance of secondary road
incorporating geosynthetics at subgrade-base interface. Transport. Res. Rec. J.
Transport. Res. Board 1849 (1), 212e220.
Arnold, G., Dawson, A., Hughes, D., Wermeister, S., Robinson, D., 2002. Serviceability
design of granular pavement materials. In: Proceedings of BCR2A. A.A. Balkema,
Netherlands, pp. 957e966.
Bahador, M., Evans, T.M., Gabr, M.A., 2013. Modeling effect of geocomposite
drainage layers on moisture distribution and plastic deformation of road sec-
tions. J. Geotech. Geoenviron. Eng. 139, 1407e1418.
Bilodeau, J.-P., Dore
, G., Perier, L., 2014. Falling weight deflectometer analysis of
flexible pavement structure built with geotextile drainage layers. Can. J. Civil
Eng. 41, 540e549.
Bilodeau, J.-P., Dore
, G., Pierre, P., 2010. Optimisation de la granulome
trie des mate
riaux
granulaires de fondation des chausse
es. Rev. Can. Ge
nie Civil 37, 1350e1361.
Bilodeau, J.-P., Dore
, G., 2012. Resilient modulus water sensitivity of compacted
unbound granular materials used as pavement base. Int. J. Pavement Eng. 13 (5),
459e471.

Mlynarek, J., 1999. Comportement observe
apre
s deux anne
es de suivi de
Blond, E.,
performances (1996-1998) du projet De
monstration du comportement et
estimation du gain e
conomique lie
s a
 l'utilisation de ge
osynthe
tiques comme

parateurs, filtres et drains en milieu routier municipal, rue Martineau a
se  Saint-
Hyacinthe. Rapport final. SAGEOS, Saint-Hyacinthe, p. 34.

Mlynarek, J., Brodeur, M., Chartrand, L., 2000. Fonctionnalite
Blond, E.,
et rentabilite


de drains et se
parateurs ge
osynthe
tiques en milieu routier municipal sur la rue
Martineau a  Saint-Hyacinthe. SAGEOS, Saint-Hyacinthe, p. 70.
Boutet, M., Dore
, G., Bilodeau, J.-P., Pierre, P., 2010. Development of models for the
interpretation of the dynamic cone penetrometer data. Int. J. Pavement Eng. 12
(3), 201e214.
Brandl, H., 2001. Freezing-thawing behavior of soils and other granular materials e
influence of compaction. In: Geotechnics for Roads, Rail Tracks and Earth. A.A.
Balkema, Rotterdam, The Netherlands, pp. 141e164.
Carrera, A., Dawson, A.R., Steger, J., 2009. State of the art of materials' sensitivity to
moisture change. ERA-Net Road Research report P2R2C2, 38 p.
Christopher, B.R., Hayden, S.A., Zhao, A., 2000. Roadway Base and Subgrade
Geocom-posite Drainage Layers. Test. Perform. Geosyn. Subsurf. Drain. 35e51.
Collins, B.M., Mahoney, J.P., Holtz, R.D., 2005. FWD analysis of pavement sections
with geotextile separators. In: Advances in Pavement Engineering. American
Society of Civil Engineers, pp. 1e14.
^ te

, J., Konrad, J.-M., 2003. Assessment of the hydraulic characteristics of unsat-
Co
urated base-course materials : a practical method for pavement engineers. Can.
Geotech. J. 40 (1), 121e136.
Dore
, G., Zubeck, H., 2009. Cold Regions Pavement Engineering. McGraw-Hill, NY.
Ekblad, J., 2007. Influence of Water on Coarse Granular Road Material Properties.
PhD Dissertation. Royal Institute of Technology, Stockholm, Sweden.
Evans, M.D., Henry, K.S., Hayden, S.A., Reese, M., 2002. The Use of Geocomposite
Drainage Layers to Mitigate Frost Heave in Soils. ASCE, Anchorage.
Fredlund, D.G., Rahardjo, H., 1993. Soil Mechanics for Unsaturated Soils. Wiley,
Interscience Publications, Toronto, Ont.
Huang, Y.H., 2004. Pavement Analysis and Design. Pearson Prentice Hall, New
Jersey.
Lafleur, J., Savard, Y., 1995. Efficiency of geosynthetic lateral drainage in northern
climates. Transport. Res. Rec. J. Transport. Res. Board 1534, 12e18.
Lebeau, M., 2006. De
veloppement d'une me
thodologie de se
lection des mate
riaux
de fondation routie re pour contrer les effets du de
gel. PhD dissertation. Uni-
versite
Laval, Que
bec, Canada.
Loulizi, A., Al-Qadi, I.L., Bhutta, S.A., Flintsch, G.W., 1999. Evaluation of geosynthetics
used as separators. Transport. Res. Rec. J. Transport. Res. Board 1687 (1), 104e111.
Nazzal, M., Abu-Farsakh, M., Alshibli, K., Mohamad, L., 2004. Evaluating the po-
tential use of a LFWD for characterizing pavement layers and subgrades. In:
Proceedings of GeoTrans 2004: Geotechnical Engineering for Transportation
Projects, Los Angeles, CA, 27e31 July 2004, pp. 915e924.
Perkins, S.W., Bowders, J.J., Christopher, B.R., Berg, R.R., 2005. Geosynthetic rein-
forcement for pavement systems: US perspectives. In: Proceedings of the Ses-
sions of the Geo-Frontiers 2005 Conference, Austin, TX, USA.
Richter, C.A., 2006. Seasonal Variations in the Moduli of Unbound Pavement Layers.
Federal Highway Administration, report FHWA-HRT-04-079, 283 p.
170 J.-P. Bilodeau et al. / Geotextiles and Geomembranes 43 (2015) 162e170

, G., Bilodeau, J.-P., Fachon, J., 2012. Effect of geosynthetic drainage
Savoie, C., Dore
layers on the recovery rate of pavement surface modulus. In: Miura, et al. (Eds.),
Proceedings of the 2nd International Conference on Transportation Geo-
technics. Hokkaido, Japan, 10e12 September 2012.
Scott, A.H., Christopher, B.R., Humphrey, D.N., Fetton, C., Dunn, P.A.J., 1998. Instru-
mentation of Reinforcement, Separation and Drainage Geosynthetic: Test Sec-
tions used in the Reconstruction of a Highway in Maine. ASCE, Dulut.
Swanson, H., 1985. Literature Review on Frost Heaving. Report No. CDOH-DTP-R-85-
1. Colorado Department of Highways, Denver, Colorado.
Theyse, H.L., 2002. Stiffness, Strength and Performance of Unbound Aggregate
Material: Application of South African HVS and Laboratory Results to California
Flexible Pavements. Report for the California Pavement Research Program.
University of California, Pavement Research Center.
Ullidtz, P., 1998. Modelling Pavement Response and Performance. Polyteknisk For-
lag, Lyngby, Denmark.
Uthus, L., 2007. Deformation Properties of Unbound Granular Materials. PhD
dissertation. University of Trondheim, Trondheim, Norway.
Uthus, L., Hermansson, A., Horvli, I., Hoff, I., 2006. A study on the influence of water
and fines on the deformation properties and frost heave of unbound aggregates.
In: Cold Regions Engineering e Current Practices in Cold Regions Engineering,
American Society of Civil Engineers.
Witczak, M.W., Andrei, D., Houston, W.N.. 2000. Development of the 2002 Guide for
the Design of New and Rehabilitated Pavement Structures. Inter Team Technical
Report (Seasonal 1), Resilient Modulus as Function of Soil Moisture e Summary
of Predictive Models, College of Engineering and Applied Sciences, Department
of Civil and Environmental Engineering, Arizona State University.