Investigation of Increase in Roughness Due to Environmental Factors in Flexible Pavements Using Profile Data From Long-Term Pavement Performance Specific Pavement Studies 1 Experiment

CHAPTER 2. LITERATURE REVIEW AND MECHANISTIC-EMPIRICAL PAVEMENT DESIGN GUIDE (MEPDG) MODEL

This chapter begins with a summary of the results of a literature review of previous studies performed to investigate development of roughness in pavements. It concludes with a description of the models used in the MEPDG software to predict the increase in IRI.⁽³⁾

LITERATURE REVIEW

A literature review was performed that reviewed the results from previous studies that have been performed to investigate roughness development in pavements. A summary of the findings from the reviewed studies are presented in this section.

Ciavola and Mukherjee investigated roughness development trends at LTPP test sections.⁽⁴⁾ A construction number change at an LTPP test section indicated that maintenance or repairs had been performed at that section. The authors referred to the first change in construction number at a test section as the first intervention on the test section. The authors found that the average age of the pavement at the first intervention for AC pavements in the North Atlantic, North Central, Southern, and Western regions were 10.8, 15.5, 15.2, and 11.8 years, respectively. MIRIs of the AC pavements when the first intervention occurred in the North Atlantic, North Central, Southern, and Western regions were 105, 129, 91, and 107 inches/mi, respectively. They also studied the effect of traffic on IRI development by grouping the test sections into four groups based on their yearly equivalent single axle loads (ESALs). They saw no evidence that MIRI performance over time was sensitive to traffic loading in the four groups.⁽⁴⁾

Corley-Lay and Mastin used GPS sections that were flexible pavements in the PPDB to study the increase in MIRI over time.⁽⁵⁾ The purpose of this study was to determine whether the Highway Performance Monitoring System reassessment guidelines that require annual roughness measurements were justified.⁽⁶⁾ They fitted polynomial curves on the time-sequence MIRI data to evaluate changes in MIRI. Of the 189 sites analyzed, 88 sites were characterized as having little change in MIRI, with the change in MIRI being less than 15 inches/mi over a 3,000-day period. Several sites had slightly declining slopes, but this trend was considered to be within the range of test variability. Forty-one sites had a change in IRI of more than 50 inches/mi within 5,000 days, with half of these sites located in the WF zone.⁽⁵⁾

Stoffels et al. investigated the influence of moisture in the subgrade on roughness progression at 43 AC pavement sections that were included in the seasonal monitoring program (SMP) in the LTPP Program.⁽⁷⁾ The volumetric moisture content in the subgrade computed from the time domain reflectometer gauges located in the test sections were available in the PPDB. Moisture content values measured within 24 to 35 inches from the top of the subgrade were used for their analysis. The volumetric moisture content in the subgrade was used to develop a parameter called the moisture index. A mean value and a standard deviation was computed for this moisture index using the moisture index values computed over the period when moisture measurements were made on the subgrade. The power spectral density (PSD) function was used to decompose the longitudinal profile to different wavebands, and the change in roughness in each waveband was evaluated using the root mean square (RMS) slope of the waveband. The time dependent change of roughness in each waveband was statistically analyzed using the RMS slope with the soil information, freezing index (FI), freezing-thaw cycles, and moisture content of the subgrade. The analysis indicated that moisture in the subgrade significantly affected wavebands from 16 to 102 ft for freezing areas and 16 to 128 ft for nonfreezing areas. The wavebands that were most responsive to moisture were from 49 to 79 ft for freezing sites and 33 to 128 ft for nonfreezing sites. In the nonfreezing sites, the change in roughness increased with moisture variations and the moisture level. The moisture variations usually affected the longer wavebands. At the nonfreezing sites, the depth to the top of the subgrade was a significant factor affecting the increase in roughness, with deeper depth to subgrade decreasing the increase in roughness. For the freezing sites, moisture variations increased roughness progression, and the roughness increase was heavily influenced by the percentage of subgrade passing the 0.002-mm sieve.⁽⁷⁾

Von Quintus et al. developed distress-based models for predicting pavement smoothness of AC pavements and AC overlaid pavements.⁽⁸⁾ They concluded the initial pavement smoothness strongly influenced smoothness over time for new construction as well as for overlaid pavements. They found that transverse cracks influenced roughness progression for all AC pavements (i.e., overlaid and nonoverlaid). All severity levels of transverse cracks were found to have an effect on increasing the IRI for new pavements with relatively thin AC layers, but only the moderate and/or high severity levels were found to influence the IRI of AC overlays and deep strength flexible pavements (i.e., new pavements with ATB). Fatigue cracking caused an increase in IRI except for AC overlays on rigid pavements. All severity levels of fatigue cracking affected IRI. For relatively thin AC surfaces, the variation of rutting either measured by the coefficient of variation of rutting or standard deviation of rutting influenced IRI rather than the mean rut depth. The mean rut depth was found to be important for AC overlays of rigid pavements. For deep strength AC pavements and AC overlays of flexible pavements, rutting or variation of rutting did not influence IRI. Age affected IRI for new AC pavements over aggregate base and AC overlays of flexible pavements. For the other pavement types, the surface distresses were more important than the age and explained the increase in IRI. The authors defined a parameter called site factor, which depended on soil properties and climatic factors, and the site factor only influenced IRI progression of flexible pavements on aggregate base. Block cracking had a significant effect on IRI except on new pavements with an ATB and AC overlays of flexible pavements. Patching had a significant effect on IRI for new pavements with ATB and AC overlays of flexible as well as rigid pavements. The authors found that small high-severity patches dramatically increased IRI.⁽⁸⁾

Lu and Tolliver used LTPP data to investigate the effect of freeze-thaw cycles and wet days on development of roughness.⁽⁹⁾ They used an exponential function of pavement age to represent the increase in IRI at the test sections. The test sections were divided into environmental zones based on the number of freeze-thaw days and the number of wet days. A freeze-thaw day was defined as a day when the air temperature in the day changed from less than 32 °F to greater than 32 °F. A wet day was defined as a day when the amount of precipitation exceeded 0.01 inches. The authors used the following freeze-thaw regions: no freeze-thaw where the freeze-thaw days in a year was fewer than 70, medium freeze-thaw where the freeze-thaw days in a year were between 70 and 140, and severe freeze-thaw where the freeze-thaw days in a year ranged from 140 to 230. They used two precipitation regions classified as dry where the wet days were fewer than 100 days a year and wet where the wet days were greater than 100 days per year. These freeze-thaw classifications and wet day classifications resulted in six environmental zones. They found that IRI increased with increasing freeze-thaw cycles and that the IRI deterioration rates are greater in the wet region compared with dry regions. They found the lowest increase in IRI was obtained in the no-freeze-thaw dry region.⁽⁹⁾

Tighe et. al. studied the change in roughness at 65 overlaid test sections in Canada.⁽¹⁰⁾ IRIs of these test sections were determined before the overlay and after the overlay at annual intervals. Nine years of IRI data were available for these sites at the time of analysis. The climatic regions considered in this study were no- to low-freeze, wet high-freeze, and dry high-freeze. The subgrade at the sites was classified as fine or coarse, while the overlay thickness was divided into three categories (i.e., 0.8 to 2 inches, 2 to 4 inches, and 4 to 6.1 inches). The authors found that the overall national trend of roughness progression for the analyzed period was essentially linear, with an average starting IRI after overlay of 70 inches/mi and the average overall IRI after 9 years being 108 inches/mi. They found overlay thickness and climatic factors were the two major factors that had a significant effect on roughness progression, with subgrade type also having a substantial effect under certain conditions. The test sections in the dry and high-freeze zones showed little change in IRI over the evaluated period. IRIs of the test sections on fine-grained subgrade soils increased at a higher rate than the test sections that were on coarse-grained subgrades. The authors indicated the greatest increase in roughness would occur for the following combination of factors: thin overlay thickness, wet and low-freeze zone, fine-grained subgrade, and high traffic. The conditions that would minimize roughness progression would be a thick overlay, dry high-freeze zone, and coarse-grained subgrade. Observations and statistical analysis indicated that traffic effect on roughness progression was not significant in most cases. The authors indicated the lack of apparent traffic level effect might have been due to the boundary that was chosen to differentiate between high and low traffic levels (i.e., 200,000 ESALs per year), and in reality, the traffic on the sections might have only represented one level of traffic.⁽⁷⁾

Kutay investigated the roughness development of flexible pavements for different wavelengths using PSDs of longitudinal pavement profile data.⁽¹¹⁾ Data collected at SPS-1 test sections in the LTPP Program were used in this study. The change in roughness was evaluated using the change in the amplitude of the PSD of the longitudinal profile over different wavelengths. Eight wavelength ranges were used in this study. The results indicated that the AC thickness and base thickness influenced roughness development at most wavelengths, with thicker AC and base thicknesses resulting in a smaller increase in IRI. The plasticity index (PI) of the subgrade and the fines content in the subgrade affected roughness development for wavelengths shorter than 1.6 ft, with greater fines content and greater PI values resulting in greater increases in roughness. The mean summer temperature affected the roughness development for wavelengths up to 33 ft, with higher temperatures resulting in a greater increase in roughness. The annual precipitation affected roughness increase for wavelengths greater than 1.6 ft, with roughness development increasing with increasing level of precipitation. The roughness development decreased with increasing traffic levels for wavelengths less than 3.2 ft, while roughness development increased with increasing traffic levels for wavelengths greater than 3.2ft.⁽¹¹⁾

Paterson presented a model for predicting roughness progression in flexible pavements.⁽¹²⁾ He indicated roughness increase could not be directly related to traffic because each pavement was designed for the expected traffic, and consequently, pavements subjected to higher traffic would have greater initial structural strength compared with lower traffic pavements. In the model Paterson developed, the incremental changes in IRI were modeled through three components: structural effects, surface distress, and environment-age factors. He used data collected in Brazil for a United Nations project to develop the model. The data showed that road roughness developed through multiple mechanisms, and significant increases in roughness could occur even without structural weakness. Paterson indicated roughness progression followed a generally accelerating pattern, with the initial increase in roughness depending on traffic loading relative to pavement strength and environmental effects. The rate increased once surface defects such as cracking, potholes, and patching occurred. Factors that affected roughness changes were rut depth variations, pavement strength, cracking, and traffic in the structural deformation component; cracking, patching, and potholes on the surface defects component; and roughness and time in the environment-age component. Most of the increase in roughness in high-strength pavements was caused by nontraffic factors such as the environment.⁽¹²⁾

Puccinelli and Jackson studied the effect of deep frost penetration and freeze-thaw cycling on pavement performance.⁽¹³⁾ They defined three environmental zones for freezing: no freeze that had an FI of less than 90 °F days per year, a moderate freeze region that had an FI between 90 and 720 °F days per year, and a deep freeze region having an FI greater than 720 °F days per year. The authors developed performance models to evaluate the independent effect of freeze-thaw cycles and FI on pavement performance using data from the PPDB. They developed models to predict fatigue cracking, rutting, and roughness of flexible pavements using data from 510 test sections that included data from GPS experiments 1, 2, and 6 and SPS experiments 1 and 2. The predictions indicated significant differences existed among the different climatic regions.⁽¹³⁾

Chatti et al. used data collected at SPS-1 and SPS-8 projects in the LTPP Program to evaluate the effect of different structural, material, and environmental factors on pavement performance.⁽¹⁴⁾ Their analysis of pavement roughness data for the SPS-1 projects found that pavement roughness was affected by all experiment factors in the SPS-1 experiment (i.e., AC layer thickness, base type, base thickness, drainage, climatic zone, and subgrade type) but not at the same level. Pavements with an ATB base performed best in terms of roughness, and pavements with thicker bases had lesser increases in IRI. In general, pavements built on fine-grained soils showed greater increases in roughness compared with pavements built on coarse-grained soils, especially in the WF region. The authors also noted that the changes in roughness of the sections in the WF zone were significantly greater than those in the WNF zone. For undrained pavement sections, the change in IRI for pavements with an ATB base was less than the change in IRI for pavement with a DGAB base. The effect of drainage was only significant for DGAB sections. Based on their observations, the authors suggested that for pavements built on fine-grained soils, greater AC thicknesses and/or treated bases would help to reduce the rate of roughness progression. They also indicated that drainage appeared to be effective in reducing the rate of increase of roughness for sections with a DGAB base, especially for the sections located in the WF zone. For the SPS-8 experiment, the authors observed that pavements located in wet climates had a greater increase in IRI compared with pavements in dry climates. They also indicated that sections in the WF zone and sections built on active soils had greater changes in IRI.⁽¹⁴⁾

Martin studied the environmental contribution to the total roughness of sealed granular unbound pavements and AC pavements in Australia.⁽¹⁵⁾ For his study, he used the Australian Road Research Board Transport Research roughness progression model for sealed granular unbound pavements and the World Bank’s Highway Design and Maintenance Standard Model HDM-III for AC pavements. Both of these models predicted the portion of the change in IRI of a pavement that could be attributed to environmental effects. Martin used the results from these models to estimate the cost of road maintenance that could be attributed to environmental effects.⁽¹⁵⁾

Perera and Kohn performed a study to investigate roughness development of pavements using the data from the PPDB.⁽¹⁶⁾ For GPS-1 test sections, which were AC pavements on a granular base, the factors that had the strongest relationship to the increase in roughness were the percentage of material in the base passing through the No. 200 sieve, FI, PI of the subgrade, and pavement age.⁽¹⁶⁾

MEPDG MODEL FOR PREDICTING ROUGHNESS OF FLEXIBLE PAVEMENTS

The MEPDG describes the models that predict the increase in IRI that are used in the MEPDG software.⁽³⁾ The models that predict the increase in IRI use the premise that the increase in roughness is caused by occurrence of surface distress on the pavement. The MEPDG is published by the American Association of State Highway and Transportation Officials (AASHTO), and the MEPDG software is referred to as AASHTOWare^® Pavement ME Design.

Model in the 2008 Interim Edition of the MEPDG

The MEPDG interim guide released in 2008 listed the model shown in figure 1 to predict the increase in IRI for new AC pavements and AC overlays of existing flexible pavements.⁽¹⁷⁾

Figure 1. Equation. Formula for predicting the increase in IRI for new AC pavements and AC overlays of existing flexible pavements.⁽¹⁷⁾

Where:

IRI = predicted IRI (inches/mi).
IRI₀ = initial IRI after construction (inches/mi).
SF = site factor (see figure 2).
FC_Total = area of fatigue cracking (combined alligator, longitudinal, and reflection cracking in the wheelpath) expressed as a percentage of total lane area. All load-related cracks are combined on an area basis, with length of cracks multiplied by 1 ft to convert length to area.
TC = length of transverse cracking (including the reflection of transverse cracking in the existing AC pavement) (ft/mi).
RD = average rut depth (inches).

The site factor is calculated according to the equation shown in figure 2.

Figure 2. Equation. Formula for computing the SF.⁽¹⁷⁾

Where:

Age = pavement age (years).
PI = plasticity index of the soil (percent).
Precip = average annual precipitation (inches).
FI = average annual FI (°F days).

As seen in figure 1, the distresses that affect IRI are fatigue cracking, transverse cracking, and rutting. In addition to these distresses, the increase in IRI is also related to the interaction of pavement age with the PI of the subgrade, average annual precipitation, and average annual FI.

The model shown in figure 1 was developed using the data available in the PPDB and has been referred to as the globally calibrated model. Figure 3 shows the comparison of measured and predicted IRI values from this model.⁽¹⁷⁾ The coefficient of determination (R²) of the model was 0.56, and the standard error of estimate for this model was 18.9 inches/mi. As seen in figure 3, there can be a considerable error associated with the IRI values predicted from this model, with the error appearing to increase with the magnitude of IRI.

©AASHTO

Figure 3. Graph. Relationship between measured and predicted IRI from the MEPDG model for flexible pavements, ⁽¹⁷⁾ from Mechanistic-Empirical Pavement Design Guide: A Manual of Practice, 2015, by the American Association of State Highway and Transportation Officials, Washington, D.C. Used by permission.

An evaluation was performed to study the sensitivity of the parameters included in the site factor, (i.e., PI, average annual precipitation, and average annual FI) on the increase in IRI.

Figure 4 shows the increase in IRI predicted by the MEPDG model for flexible pavements due to the PI of the subgrade as a function of pavement age for four different PI values. As shown in figure 4, the impact of PI on the increase in IRI was negligible. For example, the increase in IRI after 20 years attributed to a PI value of 40 was only 0.25 inches/mi, which was a negligible increase in IRI.

Figure 4. Graph. Increase in IRI due to PI of subgrade predicted by the MEPDG model for four different PI values.

Figure 5 shows the increase in IRI predicted by the MEPDG model for flexible pavements due to average annual precipitation as a function of age for four different precipitation levels. As shown in figure 5, the impact of precipitation on the increase in IRI was negligible. For example, the increase in IRI after 20 years attributed to an annual average precipitation of 70 inches was only 0.17 inches/mi, which was a negligible increase in IRI.

Figure 5. Graph. Increase in IRI due to precipitation predicted by the MEPDG model for four different annual precipitation levels.

Figure 6 shows the increase in IRI predicted by the MEPDG model for flexible pavements due to the FI as a function of pavement age for four different FI values. As shown in figure 6, the impact of FI on the increase in IRI was negligible. For example, the increase in IRI after 20 years attributed to an average annual FI of 1,200 °F days was only 0.23 inches/mi, which was a negligible increase in IRI.

Figure 6. Graph. Increase in IRI due to FI predicted by the MEPDG model for four different average annual FI values.

Model in the 2015 Version of the MEPDG

The MEPDG manual of practice released in 2015 listed the model shown in figure 7 to predict the increase in IRI for new AC pavements and AC overlays of existing flexible pavements.⁽³⁾

Figure 7. Equation. Formula for predicting the increase in IRI for new AC pavements and AC overlays of existing flexible pavements.⁽³⁾

This equation is similar to the model that was presented in the interim edition.⁽¹⁷⁾ However, the 2015 version has a new equation for computing the SF.⁽³⁾ The equation presented in the 2015 version for computing the site factor is shown in figure 8.

Figure 8. Equation. Formula for computing SF.⁽³⁾

Where:

P₀₂ = percentage passing through the 0.02-mm sieve.
P₂₀₀ = percentage passing through the 0.075-mm sieve.

The first part of the equation for computing SF has age as a factor, while the second part does not have age. Therefore, the contribution to SF from the second part is constant, irrespective of the age of the pavement.

The sensitivity of the parameters in the first part of the SF equation to IRI with increasing pavement age was evaluated using the following procedure:

• Hold annual precipitation and FI constant at 30 inches and 500 °F days, respectively, and evaluate IRI changes for P₀₂ values of 20, 40, and 60 percent.
  
  
• Hold FI and P₀₂ constant at 500 °F days and 50 percent, respectively, and evaluate IRI changes for annual precipitation values of 20, 40, and 60 inches.
  
  
• Hold annual precipitation and P₀₂ constant at 30 inches and 50 percent, respectively, and evaluate IRI changes for FI values of 100, 500, and 900 °F days.

Figure 9 shows the impact on IRI of holding annual precipitation and FI constant at 30 inches and 500 °F days, respectively, and varying P₀₂ from 20 to 60 percent. Figure 10 shows the impact on IRI of holding FI and P₀₂ constant at 500 °F days and 50 percent, respectively, and varying the average annual precipitation from 20 to 60 inches. Figure 11 shows the impact on IRI of holding the average annual precipitation and P₀₂ constant at 30 inches and 50 percent, respectively, and varying the FI values from 100 to 900 °F days. Although variations in the parameters did affect IRI for all three cases, the changes in IRI at 20 years that were observed for all three cases due to the variation of each parameter were small. For all three cases, the difference in IRI at 20 years between the lowest and highest value of the parameter being varied was 3 inches/mi.

Figure 9. Graph. Change in IRI by fixing precipitation and FI and varying P₀₂.

Figure 10. Graph. Change in IRI by fixing FI and P₀₂ and varying average annual precipitation.

Figure 11. Graph. Change in IRI by fixing average annual precipitation and P₀₂ and varying average annual FI.

As indicated previously, the second term in the SF equation (figure 8) does not have age as a parameter and is therefore constant. Table 3 shows the IRI values corresponding to the second part of the SF equation for various values of precipitation, PI, and P₂₀₀. In this table, the effect of a parameter was studied by holding two parameters constant and then changing the value of the third parameter. As seen from the IRI values shown in table 3, the effect of parameters in the second term of the SF equation on IRI was negligible.

Table 3. Effect of parameters in the second term of the SF equation on IRI.