1 Calibration of the MEPDG for Flexible Pavement Design in

2 Arkansas
3
4
5 Word Count: 4114 (Manuscript: 3064; 2 Tables and 5 Figures: 1050)
6
7
8
9
10
11
12 Authors:
13
14 Kevin D. Hall (corresponding author)
15 Professor and Head
16 Department of Civil Engineering
17 University of Arkansas
18 4190 Bell Engineering Center, Fayetteville, AR 72701
19 (479) 575-8695
20 kdhall@uark.edu
21
22 Danny X. Xiao
23 Graduate Research Assistant
24 Department of Civil Engineering
25 University of Arkansas
26 4190 Bell Engineering Center, Fayetteville, AR 72701
27 (479)799-7686
28 xxiao@uark.edu
29
30 Kelvin C.P. Wang
31 Professor
32 Department of Civil Engineering
33 University of Arkansas
34 4190 Bell Engineering Center, Fayetteville, AR 72701
35 479-575-8425
36 kcw@uark.edu
37
38
39
40
41
42
43 Paper Prepared for Publication and Presentation at the
44 2011 Annual Meeting of the Transportation Research Board
45
46 November, 2010
47

TRB 2011 Annual Meeting Paper revised from original submittal.

 1
Hall, Xiao and Wang

1 ABSTRACT
2 The Mechanistic-Empirical Pavement Design Guide (MEPDG) contains pavement
3 performance prediction models calibrated primarily using data from the Long-Term
4 Pavement Performance (LTPP) program. Because of potential differences between ‘national’
5 and ‘local’ conditions states are reporting either a partial or full calibration of the MEPDG on
6 a local level. A national guideline for local calibration was developed under NCHRP Project
7 1-40B. Arkansas has invested heavily in efforts to implement the MEPDG. The initial local
8 calibration of flexible pavement models in the MEPDG for Arkansas is summarized. For the
9 current MEPDG, predicted distresses did not accurately reflect measured distresses,
10 particularly for longitudinal and transverse cracking. However, the pavement sections
11 available for this study are generally in good condition. Due to the lack of measured
12 transverse cracking, the transverse cracking model was not calibrated. The difference in
13 defining transverse cracking between the MEPDG and LTPP may be critical in terms of data
14 collection and identification. Thermal cracking should be specifically identified in a
15 transverse cracking survey to calibrate the transverse cracking model in MEPDG.
16 Calibration coefficients were optimized for the alligator cracking and longitudinal cracking
17 models. In general the alligator cracking and longitudinal cracking models are improved by
18 calibration. However a question remains regarding the suitability of the calibrated models for
19 routine design. In addition, the smoothness model (IRI) was not calibrated, since the
20 predicted IRI is a function of the other predicted distresses. A lack of data forced the use of
21 many default values in the MEPDG. It is recommended that additional sites in Arkansas be
22 established and a more robust data collection procedure be implemented for future calibration
23 efforts. The procedure for local calibration of the MEPDG using LTPP and PMS data in
24 Arkansas is established. Additional development of database software for data manipulation,
25 pre-processing, and quality control – currently underway in Arkansas – will significantly
26 streamline the calibration process.
27
28

TRB 2011 Annual Meeting Paper revised from original submittal.

 2
Hall, Xiao and Wang

1 INTRODUCTION
2 The Mechanistic-Empirical Pavement Design Guide (MEPDG) was produced in 2004
3 through National Cooperative Highway Research Program (NCHRP) Project 1-37A, and
4 subsequently delivered to the American Association of State Highway and Transportation
5 Officials (AASHTO) in 2008. Pavement performance prediction models contained in the
6 current MEPDG were calibrated primarily using data from the Long-Term Pavement
7 Performance (LTPP) program. Because of potential differences between ‘national’ and ‘local’
8 conditions – including climate, material properties, traffic patterns, construction and
9 maintenance activities – pavement performance predicted by the MEPDG should be
10 compared to and verified against local experience. Moreover, LTPP data from sites located in
11 some states (e.g. Arkansas) were not used in the national calibration; local calibration is
12 likely necessary for these locations.
13
14 States are reporting either a partial or full calibration of the MEPDG on a local level. Kang
15 and Adams (2007) calibrated the longitudinal and alligator fatigue cracking models for
16 Michigan, Ohio and Wisconsin. All models except top-down longitudinal cracking model
17 were validated for Montana (Von Quintus and Moulthrop, 2007). It was found that the
18 MEPDG over predicted total rutting because significant rutting was predicted in unbound
19 base and subgrade soil. Muthadi and Kim (2008) calibrated the rutting and bottom-up fatigue
20 cracking model for North Carolina using a spreadsheet-based approach. In an overview of
21 selected calibration studies, Von Quintus (2008) found that the measurement error of the
22 performance data has the greatest effect on the precision of MEPDG models. California
23 utilized data from accelerated pavement testing (APT) to calibrate its mechanistic empirical
24 pavement models (Ullidtz et al., 2008). Although data from APT could be ideal for model
25 calibration considering its advantages of controlled climate condition, precise loading, and
26 testing until pavement fails, most of states that do not have APT facilities can only rely on in-
27 service pavement sites. Texas was divided into five regions for the calibration of rutting
28 models (Banerjee, Aguiar-Moya and Prozzi, 2009). Washington selected two representative
29 calibration sections to calibrate all distress models (Li, Pierce and Uhlmeyer, 2009). A
30 national guideline for local calibration was also developed by NCHRP Project 1-40B (Von
31 Quintus, Darter and Mallela, 2009). Using Pavement Management Information System
32 (PMIS), MEPDG were verified for Iowa (Kim et al. 2010). Systematic difference was found
33 for rutting and cracking models.
34
35 Arkansas has invested heavily in efforts to implement the MEPDG. An initial sensitivity
36 analysis was conducted to determine the most significant parameters of the MEPDG (Hall
37 and Beam, 2005). Selected primary inputs required by the MEPDG, but not available through
38 traditional testing, were then analyzed – including hot-mix asphalt (HMA) dynamic modulus,
39 various aspects of the traffic load spectra, and the coefficient of thermal expansion of
40 Portland cement concrete (PCC) (Tran and Hall, 2007). In addition, a project aimed to
41 manage all data for the MEPDG was completed in which software, termed ‘PrepME’, was
42 developed to conveniently prepare data sets for MEPDG use (Wang et al, 2009). Currently, a
43 local calibration effort is progressing to allow the routine use of the MEPDG in Arkansas.
44

TRB 2011 Annual Meeting Paper revised from original submittal.

 3
Hall, Xiao and Wang

1 This paper presents the process of local calibration, including data retrieval, data quality
2 checks, validation, calibration, and verification. Problems and issues encountered during the
3 process are highlighted.
4
5 DATA PREPARATION
6 The MEPDG differs from many traditional pavement design methods in that the MEPDG
7 requires substantially more data related to climate, traffic, and materials. In Arkansas, two
8 sources are available for data necessary for local calibration: the Long-Term Pavement
9 Performance (LTPP) database and the Pavement Management System (PMS) maintained by
10 the Arkansas State Highway and Transportation Department (AHTD). Unfortunately,
11 Arkansas has relatively few LTPP sections. For flexible pavements, General Pavement
12 Studies (GPS) sites GPS-1 and GPS-2, and Specific Pavement Studies (SPS) sites SPS-1 and
13 SPS-8 may be used. The AHTD PMS contains pavement materials, construction, and
14 performance data for a variety of sites, focused primarily on sites constructed since 1996
15 when the Superpave HMA mixture design system was implemented in Arkansas.
16
17 Table 1 lists 26 sections available from both LTPP and PMS sources, categorized by HMA
18 thickness and base types. Eighty percent of the sections (20 sites) were randomly selected for
19 calibration efforts; twenty percent (6 sites) are preserved for subsequent validation.
20
21 TABLE 1 Experiment Matrixa,c
AC+ No. of
HMA thickness ACb+GBb AC+ATBb
CTBb sections
Thin (≤5 in.) 0113, 0803 0116, 0118, 0120, 0121, 0122 7
0114, 0804, 0115, 0117, 0119, 0123, 0124, 2042,
Intermediate 15
070079, R50084 3058, 3071, R80065, R50067 3048
R20149, 090001,
Thick (≥8 in.) 4
009948, 070018
No. of sections 10 14 2 26
22 a
Numbers in the table refer to job/project numbers for various asphalt pavements in Arkansas
23 b
AC: Asphalt Concrete; GB: Granular Base; ATB: Asphalt Treated Base; CTB: Cement Treated Base;
24 PCC: Portland Cement Concrete.
25 c
Gray (highlighted) pavement sections randomly selected for validation.
26
27 Figure 1 shows the locations of these sections. They are well distributed across the five
28 physiographic regions of Arkansas: Ozark plateaus, Arkansas River valley, Ouachita
29 Mountains, West gulf coastal plain, and the Mississippi river alluvial plain.
30
31 Data for the LTPP sections was obtained from the latest LTPP Standard Data Release 24,
32 which was released in January 2010. Data for the AHTD sites was taken from the PMS
33 database. MEPDG Version 1.100 was used to generate pavement performance predictions in
34 this study.
35
36 Traffic
37

TRB 2011 Annual Meeting Paper revised from original submittal.

 4
Hall, Xiao and Wang

1 The LTPP database contains sufficient traffic data, such as volume count, vehicle
2 classification and axle load distribution, to be used directly in the MEPDG. However, only
3 volume count and truck percentage are available from the Arkansas PMS. Therefore, results
4 from previous research were used to provide consistent data for all sections (Tran & Hall,
5 2007). Default values were used for monthly adjustment, hourly truck distribution, and
6 general traffic input (Level 3 input). Site specific vehicle class distribution data was used
7 whenever it was available (Level 1 input); otherwise, recommended values from MEPDG
8 were used according to Truck Traffic Classification (TTC) groups (Level 2 input). Statewide
9 axle load distribution factors from previous research were used in this study (Level 2 input).
10
11 The 26 sites cover different levels of traffic. The Average Annual Daily Truck Traffic
12 (AADTT) ranges from 10 to 10,475 vehicles per day. Traffic growth rate ranges from 0 to
13 6.9%. In terms of functional classification, these sites include rural interstates, rural major
14 arterials, minor arterials and major collectors.
15

05-3071 090001
R50067
009948
05-3058

R50084 05-0113~0124
R80065

05-3048
070018 05-0803,0804

R20149

070079

LTPP
05-2042 PMS
16
17 FIGURE 1 Locations of calibration sites.
18
19 Climate
20

TRB 2011 Annual Meeting Paper revised from original submittal.

 5
Hall, Xiao and Wang

1 By providing the GPS coordinates of each site, climate data was generated by interpolating
2 from nearby climate stations. Depth of water table was extracted from the National Water
3 Information System of the United States Geological Survey (USGS). Note that this is only a
4 rough estimation because sites with valid well-data may not be close to the pavement site.
5
6 Structure
7
8 Layer structures of LTPP sites are recorded in Section_Layer_Structure in
9 Administration.mdb. Information regarding HMA mixtures, such as gradation, binder type,
10 and volumetric properties are available in Inventory.mdb and Material_Test.mdb. Default
11 values were accepted for thermal properties. It should be noted that the as-built air voids was
12 assumed to be 8 percent for all sections according to Arkansas construction specifications
13 (Arkansas State Highway and Transportation Department, 2003). Only a limited amount of
14 information is available for base and subgrade properties, such as gradation, plasticity index,
15 liquid limit and stiffness/modulus. Therefore, MEPDG Level 3 default values were accepted
16 as long as the material type is accurately determined.
17
18 Performance Data
19
20 Five flexible pavement performance predictions are provided by the MEPDG: alligator
21 cracking, longitudinal cracking, transverse cracking, rutting, and International Roughness
22 Index (IRI). For LTPP sections, the corresponding measured performance data are recorded
23 in Monitoring.mdb. Similar to the national calibration (ARA, 2004), low, medium and high
24 severity alligator cracking were summed as ‘alligator cracking’ without adjustment; low,
25 medium and high severity in-wheelpath longitudinal cracking were added without adjustment
26 as ‘longitudinal cracking’; and low, medium and high severity transverse cracking were
27 summed as ‘transverse cracking’ using the same weighting function in the national
28 calibration.
29
30 Sections in the Arkansas PMS have hard-copy records of yearly manual distress surveys,
31 rutting measurements using straightedge method and some profile measurements. The LTPP
32 Distress Identification Manual was followed in all the manual distress surveys (Miller and
33 Bellinger, 2003). The records were interpreted manually according to the distresses listed in
34 LTPP database. Therefore, the performance data of LTPP sections and PMS sections are
35 somewhat comparable. However, a concern was noted. Crack length, as recorded on hard-
36 copy forms in the PMS, have been found to vary from the actual distance in the field. This
37 may be exacerbated by the shortened length of the PMS sections (100 ft) as compared to
38 LTPP sections (500 ft). A small error in the hard-copy forms may become significant when
39 extrapolated to feet-per-mile.
40
41 RESULTS ANALYSIS
42
43 Verification
44
45 After data preparation, the MEPDG was run with the national-default calibration coefficients.
46 Figure 2 shows the comparison of predicted and measured alligator cracking, longitudinal

TRB 2011 Annual Meeting Paper revised from original submittal.

 6
Hall, Xiao and Wang

1 cracking, transverse cracking, rutting and IRI. Figure 3 illustrates the difference between
2 predicted and actual distress as a function of HMA thickness, for longitudinal cracking and
3 rutting. All data is presented here; there is no attempt in this analysis to identify and remove
4 ‘outliers’ from the data sets.
5
6 It is observed that predicted distresses do not match well with measured distresses,
7 particularly for longitudinal and transverse cracking. However, it should be pointed out that
8 most of the data points group near the origin in Figures 2 (a), (b) and (c). For example 91
9 percent of measured alligator cracking is lower than 10 percent; in addition, 92 percent of
10 predicted longitudinal cracking is lower than 1000 ft/mi. In general, the pavement sections
11 available for this study are in good condition (on average, only 3.1% alligator cracking, 711
12 ft/mi longitudinal cracking, 126 ft/mi transverse cracking, 0.21 inches total rutting, and 70.9
13 in/mi for IRI). Additional observations related to the results follow.
14 • Fatigue Cracking: Both alligator cracking and longitudinal cracking predicted by
15 MEPDG are forms of fatigue cracking. Transfer functions are used to predict visual
16 cracking from mechanistic “damage” at the bottom and top of HMA layers. This
17 results in the HMA layer thickness becoming a significant factor influencing
18 performance predictions.
19 • Asphalt Treated Base (ATB): Although it is a type of base, ATB is not modeled as
20 “Treated Base” but as “Asphalt” (albeit with a reduced stiffness). Therefore, the
21 HMA layer in the sections with asphalt treated base becomes very thick in the
22 MEPDG, which reduces the stress and strain at the bottom and top of the HMA layer,
23 in turn reducing the predicted alligator cracking and longitudinal cracking.
24 • Transverse Cracking: In the MEPDG, transverse cracking is primarily related to
25 thermal cracking, caused by thermal stress in pavement. However, transverse
26 cracking in LTPP database and PMS are measured according to the LTPP Distress
27 Identification Manual, in which transverse cracking is defined as cracks that are
28 predominately perpendicular to pavement centerline (Miller & Bellinger, 2003). The
29 implementation of Performance-Graded (PG) binders for HMA in Arkansas has all-
30 but eliminated thermal cracking in flexible pavements; accordingly the MEPDG
31 predicts no thermal cracking for Arkansas climate, using a properly selected PG
32 binder. However, transverse cracking is recorded in distress surveys, suggesting
33 additional cracking mechanisms may be predominate in Arkansas.
34 • Rutting: Eighty percent of the pavement sections have 0.1 to 0.3 inches of rutting,
35 even for the sites older than 15 years. This suggests either: (a) rutting reached a
36 maximum of 0.3 inches by consolidation under traffic, without plastic failure; or (b)
37 rutting measurements (typically by straightedge) were recorded as a maximum of 0.3
38 inches regardless of the actual measurement.
39

TRB 2011 Annual Meeting Paper revised from original submittal.

 7
Hall, Xiao and Wang

40
N=301 7000
N=301
Predicted Alligator Cracking (%)
35
6000

Predicted Longitudinal Cracking

30
5000
25
4000
20
3000

(ft/mi)
15
2000
10
5 1000

0 0
0 10 20 30 40 0 2000 4000 6000
Measured Alligator Cracking (%) Measured Longitudinal Cracking (ft/mi)
1
2 (a) (b)
1 0.8
Predicted Transverse Cracking

N=301 0.7
Predicted Total Rutting (in)

0.8
0.6
0.6 0.5
(ft/mi)

0.4
0.4
0.3
0.2 0.2
0.1
0 N=293
0
0 1000 2000 3000
0.0 0.2 0.4 0.6 0.8
Measured Transverse Cracking (ft/mi)
Measured Total Rutting (in)
3
4 (c) (d)
140
120
Predicted IRI (in/mi)

100
80
60
40
20
N=193
0
0 30 60 90 120
5 Measured IRI (in/mi)
6 (e)
7 FIGURE 2 Verification of national calibrated model: (a) alligator cracking, (b)
8 longitudinal cracking, (c) transverse cracking, (d) total rutting, and (e) IRI.

TRB 2011 Annual Meeting Paper revised from original submittal.

 8
Hall, Xiao and Wang

1
10000 0.7
8000 N=301 N=293
0.6

Residual of Total Rutting (in)

6000 0.5
Residual of LC (ft/mi)

y = -67.775x + 438.5 0.4 y = -0.0172x + 0.2893

4000 R² = 0.0245 R² = 0.282
2000 0.3
0.2
0
0.1
-2000 0.0
-4000 -0.1
-6000 -0.2
-8000 -0.3
0 5 10 15 20 25 0 5 10 15 20 25
2 HMA Thickness (in) HMA Thickness (in)
3 (a) (b)
4 FIGURE 3 HMA thickness influences the prediction of (a) longitudinal cracking,
5 and (b) total rutting. (Note: Residual = Predicted – Measured)
6
7
8 Calibration
9
10 Generally, prediction models are calibrated by minimizing the sum of standard error (SSE)
11 between predicted and measured values:
N
12 SSE = ∑ ( predicted − measured ) 2
i =1 (1)
13 Due to the nature of the data, the transverse cracking model was not calibrated. In addition,
14 the smoothness model (IRI) was not calibrated, since the predicted IRI is a function of other
15 predicted distresses. The Solver function within Microsoft Excel was used to optimize the
16 coefficients in the alligator cracking and longitudinal cracking models. Iterative runs of the
17 MEPDG using discrete calibration coefficients were utilized to optimize the rutting model.
18 For this analysis, it was assumed that the national rutting model for granular base is the same
19 as it for Arkansas because rutting mainly occurs in the HMA layers and subgrade; hence, the
20 default coefficient for rutting in granular base was not adjusted. Table 2 lists the adjusted
21 calibration coefficients. Figure 4 provides a comparison of the predicted and measured
22 distresses before and after calibration.
23
24 Observations based on the calibration process and results follow.
25 • General: the alligator cracking and longitudinal cracking models are improved by
26 calibration. However a question remains regarding the suitability of the calibrated
27 models for routine design.
28 • Quality of Input Data: Many default values are used in MEPDG in this study because
29 these data are not available. There is a continuing concern that the quality of input
30 data reduces the accuracy of MEPDG. It is recommended that additional sites be
31 established and a more robust data collection procedure be implemented for future
32 calibration efforts.
33

TRB 2011 Annual Meeting Paper revised from original submittal.

 9
Hall, Xiao and Wang

1 TABLE 2 Summary of Calibration Factors

Calibration Factor Default Calibrated
Alligator cracking
C1 1 0.688
C2 1 0.294
C3 6000 6000
Longitudinal cracking
C1 7 3.016
C2 3.5 0.216
C3 0 0
C4 1000 1000
AC rutting
βr1 1 1.20
βr2 1 1
βr3 1 0.80
Base rutting
Bs1 1 1
Subgrade rutting
Bs1 1 0.50
2
25 25
Predicted Alligator Cracking (%)

Predicted Alligator Cracking (%)

N=229 N=229
20 20

15 15

10 10

5 5

0 0
0 10 20 0 5 10 15 20 25
Measured Alligator Cracking (%) Measured Alligator Cracking (%)
3
4 (a) (b)
5000 5000
N=225
Predicted Long. Cracking (ft/mi)

N=225 4500
Predicted Long.l Cracking (ft/mi)

4500
4000 4000
3500 3500
3000 3000
2500 2500
2000 2000
1500 1500
1000 1000
500 500
0 0
0 1000 2000 3000 4000 5000 0 1000 2000 3000 4000 5000
Measured Longitudinal Cracking (ft/mi) Measured Longitudinal Cracking (ft/mi)
5

TRB 2011 Annual Meeting Paper revised from original submittal.

 10
Hall, Xiao and Wang

1 (c) (d)
0.7 0.7
0.6
Predicted Total Rutting (in)

Predicted Total Rutting (in)

0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
N=225 N=225
0 0
0.0 0.2 0.4 0.6 0.0 0.2 0.4 0.6
Measured Total Rutting (in) Measured Total Rutting (in)
2
3 (e) (f)
4 FIGURE 4 Comparison of national (a, c, e) and calibrated (b, d, f) models.
5
6 Validation
7
8 The calibrated models are validated by running the MEPDG on the remaining six sites to
9 compare predicted and measured performance. One example (site 05-0113) is shown in
10 Figure 5. The five additional sites demonstrate similar results. It is clear that local calibration
11 reduced the difference between predicted and measured distress; additional efforts (sites, data)
12 will be necessary to further reduce this difference.
13
70 1
National National
0.9
60 Local Local
Measurement 0.8 Measurement
Alligator Cracking, %

50 Limit 0.7 Limit

Total Rutting, in

0.6
40
0.5
30 0.4
20 0.3
0.2
10 0.1
0 0
0 50 100 150 200 250 0 50 100 150 200 250
14 Month Month
15 FIGURE 5 Comparison of national and calibrated models on site 05-0113.
16
17 CONCLUSIONS
18 This paper summarizes the initial local calibration of flexible pavement models in the
19 MEPDG for the Arkansas. Conclusions from the effort follow.

TRB 2011 Annual Meeting Paper revised from original submittal.

 11
Hall, Xiao and Wang

1 • The availability and quality of design, materials, construction, and performance data
2 are critical for local calibration. It is likely that states like Arkansas will need to
3 establish additional calibration sites to supplement available LTPP and PMS data.
4 • The difference in defining transverse cracking between the MEPDG and LTPP may
5 be critical in terms of data collection and identification. Thermal cracking should be
6 specifically identified in a transverse cracking survey to calibrate the transverse
7 cracking model in MEPDG.
8 • Proper modeling of asphalt treated base is vital to producing realistic predictions of
9 alligator cracking and longitudinal cracking, due to the influence of total HMA
10 thickness on the damage predictions at the bottom and top of HMA layer.
11 • The procedure for local calibration of the MEPDG using LTPP and PMS data in
12 Arkansas is established. Additional development of database software for data
13 manipulation, pre-processing, and quality control – currently underway in Arkansas –
14 will significantly streamline the calibration process.
15
16 ACKNOWLEDGEMENTS
17 This paper was prepared under research project TRC-1003, “Local Calibration of MEPDG”,
18 sponsored by the Arkansas State Highway and Transportation Department and the Federal
19 Highway Administration. The author would like to thank Daniel Byram, Jacqueline Hou for
20 their work on this project. The opinions expressed in the paper are those of the authors, who
21 are responsible for the accuracy of the facts and data herein, and do not necessarily reflect the
22 official policies of the sponsoring agencies. This paper does not constitute a standard,
23 regulation, or specification.
24
25 REFERECNES
26 ARA Inc. (2004). Guide. Retrieved 4 15, 2009, from Mechanistic-Empirical Pavement
27 Design Guide: http://www.trb.org/mepdg/guide.htm
28 Arkansas State Highway and Transportation Department. (2003). Arkansas 2003 Standard
29 Specification for Highway Construction. Little Rock: Arkansas State Highway and
30 Transportation Department.
31 Banerjee, A., Aguiar-Moya, J., & Prozzi, J. A. (2009). Texas experience using LTPP for
32 calibration of the MEPDG permanent deformation models. Transportation Research
33 Record: Journal of the Transportation Research Board, No. 2094 , 12-20.
34 Hall, K. D., & Beam, S. (2005). Estimating the sensitivity of design input variables for rigid
35 pavement analysis with a mechanistic-empirical design guide . Transportation
36 Research Record: Journal of the Transportation Research Board , 65-73.
37 Kang, M., & Adams, T. M. (2007). Local calibration for fatigue cracking models used in the
38 Mechanistic-Empirical Pavement Design Guide. Proceedings of the 2007 Mid-
39 Continent Transportation Research Symposium. Ames, Iowa.
40 Kim, S., Ceylan, H., Gopalakrishnan, K., Smadi, O., Brakke, C., & Behnami, F. (2010).
41 Verification of Mechanistic-Empirical Pavement Design Guide (MEPDG)
42 Performance Predictions Using Pavement Management Information System (PMIS).
43 Transportation Research Board Annual Meeting (pp. CD-ROM). Washington D.C.:
44 Transportation Research Board.

TRB 2011 Annual Meeting Paper revised from original submittal.

 12
Hall, Xiao and Wang

1 Li, J., Pierce, L. M., & Uhlmeyer, J. (2009). Calibration of Flexible Pavement in
2 Mechanistic-Empirical Pavement Design Guide fr Washington State. Transportation
3 Research Record: Journal of the Transportation Research Board, No. 2095 , 73-83.
4 Miller, J. S., & Bellinger, W. Y. (2003). Distress Identification Manual for the Long-Term
5 Pavement Performance Program . Washington D.C.: Federal Highway
6 Administration.
7 Muthadi, N. R., & Kim, Y. R. (2008). Local Calibration of Mechanistic-Empirical Pavement
8 Design Guide for Flexible Pavement Design. Transportation Research Record:
9 Journal of the Transportation Research Board, No. 2087 , 131-141.
10 Tran, N. H. (2005). Characterizing and predicting dynamic modulus of hot-mix asphalt for
11 mechanistic-empirical design guide. Fayetteville, AR: University of Arkansas.
12 Tran, N., & Hall, K. D. (2007). Development and influence of statewide axle load spectra an
13 flexible pavement performance. Transportation Research Record: Journal of the
14 Transportation Research Board, No. 2037 , 106-114.
15 Ullidtz, P., Harvey, J., Tsai, B.-W., & Monismith, C. L. (2008). Calibration of Mechanistic-
16 Empirical Models for Flexible Pavements Using California Heavy Vehicle
17 Simulators. Transportation Research Record: Journal of the Transportation Research
18 Board , 20-28.
19 Von Quintus, H. (2008). Local Calibration of MEPDG- An Overview of Selected Studies.
20 Journal of the Association of Asphalt Paving Technologists , 935-974.
21 Von Quintus, H. L., Darter, M., & Mallela, J. (2009). Recommended Practice for Local
22 Calibration of the ME Pavement Design Guide. Round Rock, Texas: Applied
23 Research Associates, Inc.-Transportation Sector.
24 Von Quintus, H., & Moulthrop, J. (2007). Mechanistic-Empirical Pavement Design Guide
25 Flexible Pavement Performance Prediction Models. Helena, MT: Montana
26 Department of Transportation.
27 Wang, K. C., Li, Q., Hall, K. D., Nguyen, V., Gong, W., & Hou, Z. (2008). Database Support
28 for the New Mechanistic-Empirical Pavement Design Guide . Transportation
29 Research Record: Journal of the Transportation Research Board, No. 2087 , 109-
30 119.
31

TRB 2011 Annual Meeting Paper revised from original submittal.