Available online at www.sciencedirect.

com

ScienceDirect
International Journal of Pavement Research and Technology 9 (2016) 393–402
www.elsevier.com/locate/ijprt

Highway subsurface assessment using pavement surface distress

and roughness data
Muhammad Mubaraki
Jazan University, Saudi Arabia

Received 4 May 2016; received in revised form 8 October 2016; accepted 9 October 2016
Available online 15 October 2016

Abstract

In this paper, pavement sections of the highway connecting Jeddah to Jazan were selected and analyzed to investigate the relationship
between International Roughness Index (IRI) and pavement damage including cracking, rutting, and raveling. The Ministry of Trans-
port (MOT) of Saudi Arabia has been collecting pavement condition data using the Road Surface Tester (RST) vehicle. The MOT mea-
sures roughness (ROU), rutting (RUT), cracking (CRA), and raveling (RAV). Roughness measurements are calculated in terms of the
International Roughness Index (IRI). In the present three relationships, including CRA versus IRI, RUT versus IRI, and RAV versus
IRI have been developed. The models relating to three types of distresses under study, and model relating IRI to ride quality have been
studied in the present work. The results of the analysis indicate that a significant relationship exists between IRI and cracking, and IRI
and raveling at 95% confidence level. The results also show that the rutting did not show a significant relationship to IRI values. It can be
concluded from the results that cracking and raveling may possibly be described as ride quality distresses, whereas the rutting distress
may be described as non-ride quality distress. The results indicate that while statistically significant relationships exist between IRI and
both cracking and rutting, these relationships are not strong enough for IRI to be used as a surrogate measure for pavement condition.
Ó 2016 Chinese Society of Pavement Engineering. Production and hosting by Elsevier B.V. This is an open access article under the CC BY-NC-ND
license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Keywords: Pavement; Cracking; Rutting; Raveling; Roughness; Jeddah

1. Introduction ment current conditions, to identify projects for mainte-

nance, preservation and rehabilitation, to help prioritize
Pavement evaluation is a process by which field surveys projects and allocate budgets, and to help determine fund-
and testing are carried out to characterize the condition of ing needs. In addition, the collection of performance data
an existing pavement structure, both structurally and func- on a pavement network over time provides a valuable tool
tionally. The structural condition of a pavement refers to for tracking pavement performance as well as a mechanism
its ability to support the current and future traffic loadings, for developing performance models that can be used to pre-
whereas the functional condition refers to its ability to pro- dict future conditions (both with and without the applica-
vide a safe, smooth, and a quiet riding surface for the trav- tion of treatments) [13].
eling public [6]. Network-level evaluations are conducted Visual condition surveys are generally carried out to
on the pavement sections within the network of pavements determine the type of distress, its severity, its extent and
for which the agency is responsible, with the aim to docu- location. There are many methods for conducting
pavement condition surveys that are adopted by different,
agencies involved. Pavement condition index (PCI) ranges
E-mail address: [email protected] from zero to 100, where 100 represents an excellent
Peer review under responsibility of Chinese Society of Pavement
pavement condition. The PCI values are calculated based
Engineering.

http://dx.doi.org/10.1016/j.ijprt.2016.10.001
1996-6814/Ó 2016 Chinese Society of Pavement Engineering. Production and hosting by Elsevier B.V.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
394 M. Mubaraki / International Journal of Pavement Research and Technology 9 (2016) 393–402

on pavement distress type, severity and quantity collected through horizontal profile. It gives an idea about the riding
through visual inspection. To collect the appropriate dis- quality and represents the user’s opinion of the pavement
tress data an inspector is sent to a particular pavement sec- surface. ASTM E867 defines pavement roughness as ‘‘the
tion to record the existing distress types, severities and deviation of the pavement surface from a true planar sur-
quantities. The information is collected by actually walking face with characteristic dimensions that affect vehicle
through the section. The procedure provides detailed infor- dynamics, ride quality dynamic load and drainage [16].
mation on the pavement section condition. However, con- The IRI (International Roughness Index) was initiated
sidering the size of the city pavement network the by the international roughness experiments. IRI is a stan-
procedure is tedious, time consuming and very costly. dard roughness index adopted as a reference scale for all
Roughness is primarily a measure of riding quality of roughness devices and equipment. The IRI is based on sim-
the road pavement surface. Pavement roughness is intrinsi- ulation of roughness response of standard quarter car at
cally related to pavement serviceability, which is a measure speed of about 80 km/h. IRI model comprises of a series
of physical characteristics of pavement surface. Roughness of differential equations which relate the motions of a sim-
measurement systems that are currently used can be ulated quarter-car to the road profile. The IRI is the accu-
grouped into profilometric, vehicle response and subjective mulation of the motion between the sprung and the length
evaluation. Profilometric methods are the most accurate of the profile. It is measured in meter per kilometer or mil-
and best suited for detailed analysis. Roughness measure- limeter per meter [16].
ments are calculated in terms of the International Rough- Pavement roughness is one of the most important indi-
ness Index (IRI). The IRI is calculated over equally cators of pavement performance directly reflecting pave-
spaced intervals along the road profile, the IRI computa- ment serviceability to the road users. In recent years,
tion method converts the longitudinal and vertical profile some American states and Canadian provinces have used
data into a vehicle motion response using a mathematical IRI in their business plan as an objective measure of their
model. The IRI value is expressed as the units of displace- pavement network conditions [6,19].
ment over units of length. Roughness measurements are For example, the U.S. Federal Highways Administra-
performed at speeds between 40 and 50 km per hour. Thus, tion (FHWA) uses IRI as a performance measure for
roughness measurements can be collected in a relatively describing and monitoring the pavement condition of its
short time without excessive cost [12]. National Highway System. The States of Kansas and
Washington use IRI to describe the condition of their net-
2. Literature review work in terms of percentages of miles in IRI rating cate-
gories. Because IRI is a geographically- transferable,
2.1. Background repeatable and time-stable measure, its measurement at
network level has become a routine practice for many road
Pavement condition data are an essential component of agencies in recent years. At the network level, roughness is
any pavement management system, and is required to measured on an annual or biennial basis as a part of pave-
determine the existing pavement condition, to set up the ment evaluation that is critical to formulating maintenance
pavement maintenance needs, and to plan for future need and rehabilitation priorities.
[20]. Pavement management system (PMS) is increasing In 1982, IRI was proposed in Brazil by the World Bank
being used for efficient management of highways at all as a standard statistic to correlate and to calibrate rough-
levels of government. PMS strategies can offer assistance ness measurements [16]. The International Roughness
at two levels: the network level and the project level. Net- Index (IRI) was selected in Riyadh City as a standard pro-
work level information provides management with broad- cedure to collect roughness data. The Automatic Road
based data about the entire system. Information about Analyzer (ARAN) was used for roughness measurements.
planning purposes and financial analysis is often provided A standard speed of 40 ± 5 kmph for all street classes
by the network level, whereas the project level information was used. Generally, the critical lane (right or middle)
includes details about engineering design, construction and was selected for roughness measurements. The critical lane
cost accounting. Consequently, the data required for each is defined as the lane in a given street with utility cut patch-
level differ considerably. Pavement condition evaluation ing or high traffic [18].
methods include four main practical techniques [6]: (1) dis- An attempt was made to relate the pavement condition
tress survey, (2) roughness, (3) structural capacity, and (4) of an asphalt pavement to its roughness [4]. The study
skid resistance. involved varied roadway pavement sections from the
North Atlantic region in the United States and Canada.
2.2. Roughness evaluation The study introduces a new method for estimating roadway
condition, as conveyed by the Pavement Condition Index
In the 1970s the World Bank sponsored several large (PCI), using the International Roughness Index (IRI)
scale research programs aimed at developing cost effective which is strictly a direct assessment of pavement condition.
maintenance alternatives for roadway pavements. Rough- The study assessed the applicability of IRI as a predictor
ness is a measure of variation along the vertical axis variable of PCI. Furthermore, the study developed
 M. Mubaraki / International Journal of Pavement Research and Technology 9 (2016) 393–402 395

parametric models of pavement condition in terms of ously, different distress types differently impact IRI. It was
roughness using sample IRI and associated PCI data using found that severe potholes, patching, and rutting have the
readily available pavement rating software. A transformed highest correlation to IRI, while the stripping and corruga-
linear regression model predicts pavement condition for the tion have less correlation. Also, cracking, alligator crack-
given roughness. The results from this study confirms the ing, and bleeding are least related to IRI. Thus, the study
acceptability of the International Roughness Index as a shows that pavement distress and pavement roughness
predictor variable of the PCI. have commonly causal relationship affecting one another
Further; an analysis of variance confirms the existence in both directions.
of strong relationships between both the variables. The cor- Effect of asphalt concrete raveling on pavement rough-
relation between pavement distress and IRI was examined ness has been studied by Hozayen [8]. The author presented
for the data from several sites in the San Francisco Bay an approach to investigate the relationship between aggre-
area [10]. The goal of the research was to develop a surface gate raveling of asphalt pavements and surface roughness
roughness model that in turn could be used to estimate as a means for assessing pavement riding quality. Pavement
vehicle operating costs for the streets in the San Francisco distresses have been surveyed on various highway classes
Bay area. A correlation between IRI and pavement distress (primary, feeder and secondary) of the rural network
was developed based on 39 observations measured at 15 m together with the corresponding pavement condition index.
intervals on a 152.4 m test section. A linear relationship Results indicated that regression relationships (including
was developed between IRI and a composite Pavement power, exponential, and polynomial models) could be
Condition Index (PCI) using the type and severity of dis- established between aggregate raveling and pavement
tress observed for each segment of the roadway. The model roughness with a correlation factor ranging between 0.8
had an adjusted R2 of 0.52 suggesting that just over half of and 0.92 depending on highway class and location. Hall
the variance in IRI could be predicted with the aggregate et al. [7] studied the effect of various maintenance activities,
pavement condition index. Al-Mansour et al. [1] studied including slurry seal, chip seal, crack seal, and thin overlays
the effect of crack sealing, chip seal, and sand seal on on pavement roughness. Based upon a statistical analysis,
roughness in flexible pavements used in interstate and state they reported that the effect of chip seals, crack seals, and
highways. They reported low benefits in roughness reduc- slurry seals were not significant compared to a control sec-
tion due to 19 maintenance activities in the case of new tion which did not receive a maintenance treatment. How-
pavements and increased benefit in the roughness reduction ever, thin overlays were found to reduce pavement
of maintenance applied to aged pavements. roughness significantly. In this study, no improvement in
In 1999, the Taiwan Highway Bureau imported ARAN IRI was considered for pavements undergoing chip seals,
(Automated Road Analyzer), a specially equipped car that slurry seals, and crack seals, while a roughness reduction
measured pavement roughness as IRI (International resulting in a restored IRI level of 63 inch/mile was
Roughness Index). Since then, the traditional visual inspec- assumed following the application of an overlay. Although
tion performed at network level had the tendency to be the rate of change of IRI for overlays is higher than the
replaced by automatic inspection devices. A study aimed new pavement IRI deterioration, the same rate was consid-
to investigate and analyze the relationship between IRI ered for simplicity of calculation in this study.
and pavement distress was conducted by Karim and Lee A study in the United States investigated the interaction
[9]. The authors evaluated the applicability of IRI to be between pavement surface roughness and distress types
treated as representations of pavement performance to [15]. The study involved 462 pavement sections from
overcome the need for visual data collection. The visual thirty-seven projects in the state of Michigan. The pave-
inspection consumes a substantial amount of time and ment sections were analyzed to investigate the interaction
money. It also avoids the effects of unreliable and inconsis- between pavement surface roughness and distress. The
tence results that creep in due to personal factors in deter- Michigan Department of Transportation (MDOT) used
mining index value. A total of 125 road sections with a two measures of pavement performance in managing its
length of 1 km were used in the collection of data. More- pavement network: the Distress Index (DI), and the Ride
over the study aims to reinforce the concept that IRI Quality Index (RQI). The DI equals the sum of the distress
may completely reflect pavement distress conditions, and points along the project over the section length. The DI
there is a relationship between IRI and certain pavement scale starts at zero for a perfect pavement and it increases
distress types. The analytic process was based on back without a limit. MDOT categorizes DI into three levels:
propagation neural network methodology. The results of Low; <20, Medium; 20–40, and High; >40. A study was
the analysis prove that rapid measurement of IRI can be conducted to investigate a relationship between some pave-
used, and simplifies the woks of traditional visual inspec- ment distresses and ride quality by Aultman and Jackson
tion by a road inspector. The effects of distinct distress [2]. The Pavement Maintenance Management System used
types and their extent on IRI have been analyzed during in the United States Army Corp of Engineers (PAVER)
this study. The correlation coefficient between IRI and sorts five distresses as ride quality related types. PAVER
the distress variables reached 0.944, which shows that IRI describes the ride quality by riding in a standard size auto-
reflect pavement distress conditions to a large extent. Obvi- mobile over the pavement section at the posted speed
396 M. Mubaraki / International Journal of Pavement Research and Technology 9 (2016) 393–402

which is similar to the description of the roughness defini- considered. The network was trained with 90% of the data
tion. According to the PAVER visual inspection proce- and tested with remaining 10% data. Results of R2 and
dure, the inspector should take into account the vehicle MSE showed that the neural network performed highly sig-
vibration and reduction in comfort in order to evaluate nificantly in both training and testing phases. Finally, the
the severity levels for these types of distresses. The list of performance of the ANN model was compared with that
the five roughness related distresses in PAVER system of linear and nonlinear regression models. The mean abso-
includes: lute error (MAE) for the ANN model is around 18% less
than that for the linear model and 11% less than that for

Bumps and sags. the nonlinear model. MARE values are also 12.5% lower

Corrugation. in the case of ANN modeling, indicating that the ANN

Railroad Crossing. model yields a better forecast of road roughness for a given

Swells. set of distress parameters.

Shoving. Meegoda et al. [11] defined IRI as an average rectified
slope (ARS) which summarizes the ratio of the accumu-
There are other distress types not related to roughness lated suspension motion to the distance traveled. This
such as alligator, slippage, edge and block cracking, pol- value is obtained from a mathematical model of a standard
ished aggregate, bleeding and rutting. Ningyuan et al. quarter car traversing a measured profile at a speed of
[15] used a large field- measured IRI dataset collected in 50 mph or 80 km/h. The LTPP database provides the nec-
the summer and fall of 2001 on a total of 650 km highway essary pavement performance data, which contains detailed
to investigate the relationships between IRI, rutting and information about average IRI. A typical test section, was
pavement cracking The study objective was to determine used to collect information on seven modules: Inventory,
if these relationships were consistent enough to allow Maintenance, Monitoring (Deflection, Distresses, and Pro-
IRI, the more easily measured quantity, to be used as a sur- file), Rehabilitation, Materials Testing, Traffic, and Cli-
rogate for the others or it is necessary to collect all three matic. The compatibility of existing data sets must be
measures in pavement management programs. The results considered since sections with different ages and treatment
indicated that while statistically significant relationships methods were included in the study. The data were adapted
exist between IRI and both cracking and rutting, these rela- to fit a regression model between the number of years
tionships are not strong enough for IRI to be used as a sur- elapsed since the last overlay and the corresponding
rogate measure for pavement condition. It was concluded changes in IRI values. It was concluded that the pavement
that IRI, while appropriate for measuring ride ability, is near the end of service life starts to deteriorate quickly.
not an appropriate for measuring cracking or rutting. Elghriany et al. [5] found pavement roughness may have
Perera and Kohn [17] reported that, for pavement sec- a significant impact on traffic safety, but its effect on the
tions with IRI greater than 97 inch/mile before applying overall safety problem needs to be studied and quantified.
an overlay, the IRI after placing the overlay was reduced The study investigated the relationship between Interna-
to between 52 and 76 inch/mile. The authors reported that tional Roughness Index (IRI) and rates of crashes by look-
IRI values would be less than 64 inch/mile after the appli- ing into the traffic safety performance over time under
cation of an overlay when pre-overlay IRI values of less changing pavement conditions (change of IRI). Specifi-
than 97 inch/mile were present. Thus, for roughness predic- cally, the proposed model provides insight into the effect
tion of pavement following rehabilitation, an IRI level of of roadway (rigid pavement) surface conditions on traffic
63 inch/mile was assumed in this study. Maintenance repre- safety in order to develop a better understanding of the
sents pavement improvement activities which are per- problem. Data on a geographic information system (GIS)
formed when the pavement is in a structurally sound platform from different sources in the state of Ohio were
condition. Chandra et al. [3] studied the relationships compiled and analyzed using a statistical analysis
between pavement roughness and distress parameters like approach. Different model formulations were examined,
potholes, raveling, rut depth, cracked areas, and patch- and a quadratic relationship was found to be the most
work. The pavement distress data collected on four effective in linking the crash rates with pavement rough-
national highways in India using a network survey vehicle ness. The findings of this study by Elghriany et al. [5]
(NSV) are used to develop linear and nonlinear regression may serve as a good reference that can assist the state
models between roughness and distress parameters. Analy- and local transportation agencies in roadway maintenance
sis of variance of these models indicated that nonlinear decision making. From the review of available literature, it
relation is better than a linear model. R2 value, root mean is clear that the IRI is an acceptable predictor variable of
square error (RMSE), and mean absolute relative error pavement conditions. Some studies suggest that the Inter-
(MARE) also supported nonlinear models. An artificial national Roughness Index can be used to set maintenance
neural network (ANN), which is an advanced technique needs and priorities without visual inspections, others con-
of modeling, was also used in the study to model pavement sidered an IRI measurement at the network level as a part
roughness with distress parameters. A network with five of pavement evaluation that is critical to formulating main-
input nodes, 15 hidden nodes, and one output node was tenance and rehabilitation priorities. The studies reviewed,
 M. Mubaraki / International Journal of Pavement Research and Technology 9 (2016) 393–402 397

however, did not agree on type of distresses that mostly 4.3. IRI and distress model
affect pavement roughness.
The modeling procedure used in the present analysis is
3. Objectives based on statistical regression. The basic form of the
regression equation is as follows:
The main objective of this study is to study the relation- y ¼ b0 þ b1X;
ship between pavement condition evaluation and rough-
ness measurements and examine the effect of pavement where Y: dependent variable (response), X: independent
distress types on pavement roughness for the entire Jazan variable, and b0, b1: regression coefficients.
road network. To conduct the analysis, a spreadsheet program was
used to examine and tabulate the data, draw charts
(graphs) that describe the trend of the collected data, build
4. Methodology
models and to find out correlations and the relationship
between pavement roughness (IRI) and types of distress.
4.1. Study methodology
The IRI and distress model is constructed in order to pre-
dict types of distress in terms of a distress behavior as a
The methodology used in the present study consisted of
function of pavement roughness (IRI). The procedure
four main steps. The first step was to collect the required
involved the use of regression techniques to construct mod-
data for a selected sample of pavement sections from High-
els linking IRI and distress values. The most suitable model
way No. 5 which connects Jeddah with Jazan. The data
was then selected based on the results of statistical tests.
include information on three types of distress, including
The process of constructing the regression model was based
cracking, rutting and raveling, and the International
on three types of regression functions: linear, logarithm
Roughness Index for each included pavement section.
nonlinear, and quadratic nonlinear. The nonlinear relation
The second step of the methodology was to investigate
is transformed to linear regression by calculating new
the correlation between IRI and three collected types of
parameter called IRIa, where a is a constant. The regres-
distress data. The third step was to examine the effect of
sion equation is in the form: Distress ¼ b0 þ b1 IRIa. In
the aggregation of all distress types on the International
this equation the regression coefficients are b0 and b1.
Roughness Index measurements and the effect of each dis-
tress individually. The fourth step was to draw a conclusion
of the analysis and results obtained from the collected data 4.4. Adequacy of IRI and distress model
with respect of significance of the three distress types on
IRI, and ride quality related distresses. The idea of the After the regression equations are constructed, a process
study methodology can be expressed in Fig. 1. of testing was used. The aim of the testing process was to
find out the best model that represents the given data.
The analysis procedure was based on three main statistical
4.2. Correlation of IRI and types of distress regression tests, which are:
The Pearson correlation analysis was conducted to esti- 1. Test of hypothesis (T-test).
mate the correlation factor and correlation hypothesis and 2. Coefficient of multiple determination (R2).
to find out whether a significant statistical relation between 3. Analysisof Variance Test (ANOVA – Test).
IRI and types of distress exists. The statistical hypothesis
test displays a p-value to determine the significant effect The T-test of a hypothesis is incorporated to study the
of a distress on IRI values [16]. relationship between each of the regression coefficients

Effect of Pavement Distress Types on Pavement Roughness

Correlation Factors Analysis

Significant Not Significant

Ride Quality Distress No Ride Quality Distress

Develop Models for Each Distress No models

Fig. 1. Flowchart for Pavement roughness and pavement distress types.

398 M. Mubaraki / International Journal of Pavement Research and Technology 9 (2016) 393–402

(b0, b1) and the response variable at a certain degree of 4.6. IRI and distress density model
confidence (assumed 95% in our study). Hence the test
was conducted for all of the regression coefficients in the In this part, the regression technique was used to con-
regression model. The coefficient of multiple determina- struct models linking IRI and all three distresses under
tions (R2) was used to illustrate the adequacy of a fitted study, including cracking (CRA), rutting (RUT), and rav-
regression model. The R2 indicates the proportion of the eling (RAV).
response variable variance explained by the regression The linear regression model would be in the form:
model. The coefficient takes values from 0 (no correlation IRI ¼ b0 þ b1CRA þ b2RUT þ b3RAV
exists between variables) to 1.0 (perfect correlation between
variables). The Analysis Of Variance test (ANOVA) is a where b0, b1. . ., bn are regression coefficients.
powerful tool used to investigate the relationship between CRA = Cracking Distress Density.
the response variable (distress in this study) and the inde- RUT = Rutting Distress Density.
pendent variables (IRI). The ANOVA test extends the t- RAV = Raveling Distress Density.
test for a more general null hypothesis to estimate the qual- This model studied the effect of three distresses on pave-
ity of the regression line. That is, to determine whether the ment roughness (IRI). This model was then used to confirm
variation in distress readings is dependent on the IRI read- the earlier results in determining the ride quality types of
ings. This provides the significant evidence to accept or distress. However, this model was modified to generate
reject the regression model [14]. the most fitting relationship between IRI and the ride qual-
ity distress types. The effect of the pavement roughness on
4.5. Ride quality related distresses three distress types was studied and analyzed. Two regres-
sion models were developed and investigated; first model
Pavement roughness measurements are primarily a mea- was to relate the IRI index to all the three types of distress.
sure of the riding quality of roadway pavement surface. The second regression model was to relate the IRI index to
Usually pavement roughness is related to pavement ser- types of distresses that are considered as ride quality dis-
viceability. Pavement serviceability is a measure of the tresses. This model was investigated and interpreted
physical characteristics of the pavement surface. Therefore, through many statistical tests. The T-test of hypothesis
roughness is related to the opinion of the roadway users. It was utilized to study the relationship between each of the
has a significant effect on vehicle operating cost, safety, regression coefficients (b0, b1) and the response variable
comfort and speed of travel. In this section, an attempt at 95% degree of confidence. The ANOVA was used to esti-
was made to investigate pavement distresses to determine mate the quality of the regression line, that is, whether the
the riding quality related to different type of distresses. variation in IRI is dependent on the distress data. Finally,
The effect of all pavement distresses on the IRI reading the coefficient of multiple determinations (R2) was used to
was investigated in the present study. The density of each illustrate the sufficiency of a fitted regression model [10].
distress type was plotted against the respected IRI values.
The plot was then utilized to check the trend of the rela- 4.7. Site selection and data collection
tionship. The correlation test was applied to find the corre-
lation coefficient between IRI and the distress density. The The highway considered in the present study connects
Pearson correlation coefficient was used to measure the North West of Saudi Arabia with South West of Saudi
degree of linearity between distress density and IRI. The Arabia, and is more than 2000 km in length. The study
correlation coefficient assumes a value of 1 for maximum involves the analysis of roughness and distress data only
correlation. Positive correlation coefficient means that the from pavement sections between Jeddah city and Jazan
distress density tends to increase with increasing IRI. The city. The distance between Jeddah and Jazan is 800 km.
hypothesis test displays p-value that determines the signif- The selected sections have been determined to be having
icant effect of IRI on each distress type. The statistical anal- same factors affecting the performance of the pavement
ysis is then carried out at a confidence level of 95%. including maintenance and evaluation team. All pavement
Therefore, if the value is below the 5% significance level, information collection is performed by ARAN which has
then there is a sufficient evidence that there is a statistically been recommended by the World Bank. ARAN system
significant relationship between IRI and distress density. obtains the roughness index value and rutting depth of
The statistical regression techniques have been utilized to each pavement section by using a laser roughness measur-
achieve an appropriate model relating to the International ing device and supersonic rutting measuring device which
Roughness Index and ride quality distress types. The are mounted in ARAN. The pavement distress data are
ANOVA test was also used to investigate the relationship. obtained based on the pavement image obtained from the
The ANOVA test was extended to estimate the quality of image retrieval system. All the data are integrated with
the regression coefficients in the regression model. The p- the software package for each kilometer and stored in the
values indicate at 95% confidence the significant evidence Pavement Management System (PMS) database. All the
to accept or reject the regression coefficient of each distress data including pavement distress and roughness were col-
density in the regression model [2]. lected by the Ministry of Transport (MOT) in years of
 M. Mubaraki / International Journal of Pavement Research and Technology 9 (2016) 393–402 399

2009, 2011, and 2013. The study uses the MOT data and patching density was statistically significant at 5% signifi-
deploys them to conduct the methodology of the study. cance level. The result of the statistical test does not con-
The pavement distress and roughness data were available form to the previous results that suggest that cracks
for 400 km for each km interval. Detailed distress data distress is a ride quality type of distress at all severity levels.
were available in the form of distress type, distress density,
and as well as measurement of roughness. 5.2. Rutting vs IRI

5. Data analysis The relationship between IRI and rutting density is

presented in Fig. 3. The graph indicates a positive rela-
5.1. Cracking vs.IRI tionship between the rutting and IRI values. There is no
definite trend in the linearity of the relation. The correla-
Fig. 2 presents the graph showing the relationship tion coefficient of IRI and rutting density is 0.464 and
between IRI and cracking density. It can be seen from p-value is 0.151. The value of correlation coefficient of
Fig. 2 that the IRI never exceeded 5 for most data points. 0.464 indicates that the relationship between the percent
This may be attributed to the fact that data collecting van of rutting and IRI cannot be expressed adequately by a
drivers are instructed to avoid driving through heavily linear equation. For this reason, a logarithm nonlinear
deformed segments of the road. Also, for a cracking dis- relationship was developed. The fitted regression model
tress between 4% and 6%, the IRI values were clustered was as follows; Rutting = 8.2 + 2.5 log IRI. The p-value
around ‘‘1”. This irregularity most likely represents the seg- is 0.15 which indicates that the relationship between IRI
ments of the road with a very good IRI but had longitudi- and rutting density is statistically significant at 5%
nal cracks between the road and its paved shoulders. The significance level.
plot shown in Fig. 2 verifies the positive relationship
between the cracks and the IRI values. However, the graph 5.3. Raveling vs IRI
does not indicate a definite linearity of the relation. The
correlation coefficient of IRI and patching density is 0.5 Fig. 4 presents a relationship between IRI and raveling
and the p-value is equal to 0.019. The positive sign of the distress density. The raveling data were obtained from 50
coefficient indicates a positive relation between cracking different pavement sections. The graph shown in Fig. 4
density and IRI. The value of correlation factor indicates indicates a positive relationship between raveling and
that about 50% of the IRI and crack density observations IRI. The value of correlation factor indicates that about
were distinguished by the linear relation. The best represen- 45% of the IRI and raveling density observation were
tation for the cracking data was given by logarithmic described by the linear relation. The best model formula
formula: is as follows: Raveling = 28.2 + 9.00 log IRI. In the
hypothesis test, the p-value was found to be 0.006, which
Cracking ¼ 8:6 þ 2:7 log IRI
indicates that there is a sufficient evidence at 95% confi-
In the hypothesis test, the p-value was found to be 0.019 dence level of the relationship between IRI and raveling
thus indicating that the relationship between IRI and density.

Cracking vs IRI

40 y = 2.7326ln(x) + 8.5471
R² = 0.4002.
35
Percent of Cracking

30

25

20

15

10

0
0 1 2 3 4 5 6
IRI

Fig. 2. IRI and Cracking Distress Density relationship.

400 M. Mubaraki / International Journal of Pavement Research and Technology 9 (2016) 393–402

Rutting vs IRI

60

50
Percent of Rutting y = 2.4985ln(x) + 8.1702
40 R² = 0.3979

30

20

10

0
0 1 2 3 4 5
-10
IRI
Fig. 3. IRI and Rutting Density relationship.

Raveling vs IRI

80
y = 9.003ln(x) + 28.504
70 R² = 0.4003

60
Percent of Raveling

50

40

30

20

10

0
0 0.5 1 1.5 2 2.5 3 3.5 4
IRI

Fig. 4. IRI and Raveling Density relationship.

5.4. Model-1: relates IRI to all apparent distress IRI. However, the p-value of the rutting (RUT)
coefficient is greater than the 0.05 level of significance.
IRI ¼ 4:498 þ 0:0096CRA þ 0:0083RUT þ 0:0067RAV Therefore, there is a significant evidence to exclude the
pothole and rutting in predicting the roughness index.
Statistics of T-test of the regression coefficients. The p-values for the ANOVA test for the model are
Predictor coefficient T P equal to zero, indicating no significant amount of varia-
tion in the IRI response variables by the regression
Constant 4.498 161.73 0.000 model. The data reflect sufficient evidence of
CRA 0.0096 21.11 0.000 dependency-between-IRI-and-types of distress. The coeffi-
RUT 0.0083 5.01 0.07863 cient of multiple determinations (R2) equals 28%, which
RAV 0.0067 14.13 0.000 means that 28 percent of the IRI value are represented
and explained by the regression model. It also indicates
that while-statistically-significant-relationships exist
The p-values for most regression coefficients are either between IRI and some distress types, the relationship is
equal or close to zero, indicating significant relationship not strong enough for distressed type stone used as a
between the distress types and the predicted values of predictable measure for roughness condition.
 M. Mubaraki / International Journal of Pavement Research and Technology 9 (2016) 393–402 401

5.5. Model-2: relates IRI to ride quality distresses results of the present research indicated the riding quality
related types of distress. A statistical summary of the
The distresses that showed a significant relationship to results obtained in the analysis conducted herein is shown
the IRI index were chosen to be ride quality types dis- in Table 1. The-correlation-hypothesis test results shown
tresses. These distresses include cracks distress (CRA), in Table 1 indicate that the p-value is either zero or close
and raveling distress (RAV). The regression model devel- to zero for cracking, rutting and raveling. This is, however,
oped is as follows: not true for rutting. Therefore, it can be concluded with
IRI ¼ 3:58 þ 0:0077CRA þ 0:0054RAV 95% confidence that a significant relationship exists
between IRI and cracking and IRI and raveling. Further-
Statistics of T-test of the regression coefficients. more, the low correlation factors for all the three types of
Predictor coefficient T P distress showed that no definite conclusion could be drawn
about the linearity of the relationships. The best represen-
Constant 4.498 161.73 0.000 tation of the relationship was obtained using a nonlinear
CRA 0.0096 21.11 0.000 regression equation. In this study, it was found that the
RAV 0.0067 14.13 0.000 logarithmic equation shows reasonable fitting for the data.
The regression coefficient test computes p-values for each
regression coefficient. The regression coefficients for crack-
The p-values for most regression coefficients are equal to
ing and raveling are equal to zero. This indicates a signifi-
zero, indicating a significant relationship between the dis-
cant linear relationship between IRI and different
tress types and the predicted values of IRI. The p-values
regression coefficients. The result of both correlation and
for ANOVA-test is equal to zero, indicating no significant
regression coefficient tests indicates at 95% confidence that
amount of variation in the IRI response variables by the
cracking, and raveling are ride quality type distress. How-
regression model. The data reflect sufficient evidence of
ever, it is not realistic to conclude that IRI may completely
dependency relationship between IRI and distress values.
reflect pavement distress conditions. It is clear from Table 1
It is clear that the F-value for the second model is greater
that the rutting pointed toward no significant relationship
than that of the previous model indicating improvement
to IRI values.
of the second model over the first one. The coefficient of
multiple determinations (R2) is equal to 30.5%, which
means that 30.5 percent of the IRI values are represented 7. Conclusion
and explained by the regression model. However, the-
relationships are not strong enough for distress types to Pavement sections from highway connecting Jeddah
be used to predict pavement roughness condition. with Jazan was selected in this study to evaluate the rela-
tionships of three types of distresses including cracking,
6. Discussion rutting, and raveling to IRI values. Based on the results,
the statistical models that relate IRI values to the distress
The effect of all the distress types that appears in the density values were established. The results indicate that
study segment on the IRI reading was investigated. The while statistically significant relationships exist between

Table 1
IRI vs. distress density statistical summary.
Distress type Correlation factor Correlation hypothesis test Best presentation
Cracking r ¼ 0:500 P ¼ 0:019 Cracking ¼ 8:6 þ 2:7 log IRI
Rutting r ¼ 0:464 P ¼ 0:151 Rutting ¼ 8:2 þ 2:5 log IRI
Raveling r ¼ 0:450 P ¼ 0:006 Raveling ¼ 28:2 þ 9:00 log IRI
Regression model equations
Model 1: IRI ¼ 4:498 þ 0:0096CRA þ 0:0083RUT þ 0:0067RAV
Model 2:IRI ¼ 3:58 þ 0:0077CRA þ 0:0054RAV
Regression coefficients test (T-test)
Distress type Hypothesis test (Model 1) Hypothesis test (Model 2)
Cracking P ¼ 0:000 P ¼ 0:000
Rutting P ¼ 0:078 –
Raveling P ¼ 0:000 P ¼ 0:000
ANOVA
P ¼ 0 Model 1
P ¼ 0 Model 2
Regression model adequacy
R2 ¼ 28% Model 1
R2 ¼ 30:5% Model 2
402 M. Mubaraki / International Journal of Pavement Research and Technology 9 (2016) 393–402

IRI and the distresses under the study, these relationships Francisco Bay Area Transportation Research Record 1816, Trans-
are not strong enough for IRI to be used as a surrogate portation Research Board, Washington, DC, 2002.
[5] A. Elghriany, Ping. Yi, Peng. Liu, Yu. Quan, Investigation of the
measure for pavement condition. It is concluded that effect of pavement roughness on crash rates for rigid pavement, J.
IRI, while appropriate for measuring ride ability, is not Transport. Safety Security 8 (2) (2016).
appropriate for measuring cracking, rutting, and raveling. [6] R. Hass, W. Hudson, J. Zaniewnski, Modern Pavement Manage-
This result is in conformity with the previous studies ment, Krieger Publishing, 1994.
detailed in the earlier sections of this paper. [7] K.T. Hall, C.E. Correa, A.L. Simpson, LTPP data analysis:
effectiveness of maintenance and rehabilitation options, NCHRP
The results of the present study clearly suggest that the Project (2002) 20–50.
pavement roughness is not strongly enough to be used as [8] A. Hozayen, Effect of asphalt concrete raveling on pavement
an appropriate measure of the pavement deformation, roughness, Annual Conference of the Canadian Society for Civil
and any increase in the pavement roughness accelerates Engineering, Montréal, Québec, Canada, 2002.
the pavement deterioration. Similarly, any pavement dis- [9] Chatti Karim, D. Lee, Development of a preventive maintenance
strategy for minimizing roughness-related pavement damage, J.
tresses will also result in deterioration of the pavement Transport. Res. Board, 2001, The 80th Transportation Research
roughness index value. An increase in roughness leads to Board Annual Meeting, Washington, D.C..
higher dynamic axle loads, which in turn can lead to a tan- [10] J. Lin, J. Yau, L. Hsiao, Correlation analysis Between International
gible acceleration in pavement distress. Therefore, a dis- Roughness Index (IRI) and pavement Distress by Neural Network, J.
tinct practical advantage of the present research is that it Transport. Res. Board (2003), The 82nd Transportation Research
Board Annual Meeting, Washington, D.C..
will greatly assist in determining maintenance needs by [11] J. Meegoda, S. Gao, Roughness progression model for asphalt
defining the types of distress that are most likely to be pavements using long-term pavement performance data, J. Transp.
encountered. The results of the present research can pro- Eng. (2014) 04014037, http://dxdoi.org/10.1061/(ASCE)TE.1943-
vide valuable aid in determining the effective timing of 5436.0000682.
applying the maintenance needs by defining the time after [12] M. Mubaraki, Identification of pavement distress types and pavement
condition evaluation based on network level inspection for Jazan City
which the rate of deterioration will increase drastically. road network, J. Eng. Res. (TJER) 11 (1) (2014) 44–54.
The results of the present research can be beneficially uti- [13] M. Mubaraki, A methodology for project-level maintenance for
lized by the municipalities in setting up priority index for urban road, J. Transport-ICE 168 (3) (2015) 239–255.
the budget and for devising strategies for efficient mainte- [14] Myers, Keying E. Ye, Probability and Statistics for Engineers and
nance programs. The present research has the potential Scientists, 2011.
[15] Ningyuan Li, T. Kazmierowski, B. Sharma, Verification of network
to provide impetus for modeling in the field of pavement level pavement roughness measurements, J. Transport. Res. Board
maintenance. (2001), The 80th Transportation Research Board Annual Meeting,
Washington, D.C..
[16] K. Park, K. Lee, T. Natacha, Applicability of the international
References roughness index as a predictor of asphalt pavement condition, J.
Transport. Res. Board (2004), The 83rd Transportation Research
[1] A. Al-Mansour, K.C. Sinha, T. Kuczek, Effects of routine mainte- Board Annual Meeting, Washington, D.C..
nance on flexible pavement condition, J. Transport. Eng. ASCE 120 [17] R.W. Perera, S.D. Kohn, Ride quality performance of asphalt
(1) (1994). concrete pavements subjected to different rehabilitation strategies, in:
[2] L. Aultman, E. Jackson, Model relating pavement quality measures, ASCE Proceedings of 2006 Airfield and Highway Pavements
J. Transport. Res. Board (2004), The 83rd Transportation Research Specialty Conference, Atlanta, Georgia, April 30–May 3, 2006.
Board Annual Meeting, Washington, D.C.. [18] RPMMS, Riyadh Pavement Maintenance Management System,
[3] S. Chandra, C. Sekhar, A. Bharti, B. Kangadurai, Relationship Riyadh Municipality, 2005.
between Pavement Roughness and Distress Parameters for Indian [19] M.Y. Shahin, S.D. Kohn, Pavement Maintenance Management for
Highways, J. Transport. Eng. 139 (2013). Road And Parking Lots, 2002.
[4] S.A. Dewan, R.E. Smith, Estimating IRI from Pavement Distresses to [20] Tahar El-Korchi, Norman Wittels, Visual Appearance of surface
Calculate Vehicle Operating Costs for the Cities and Counties of San Distress in Pavement, TRB No. 1260 Washington, D.C., 2004.