LTPP Data Analysis: Influence of Design and Construction Features on the Response and Performance of New Flexible and Rigid Pavements (2005)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

155 CHAPTER 5 - ANALYSIS RESULTS FOR THE SPS-1 EXPERIMENT 5.1 INTRODUCTION The purpose of this chapter is to provide a summary of findings from the previous studies and the results of the various analyses conducted for the SPS-1 experiment on flexible pavements in this study. The performance and structural response indicators used in the analysis include fatigue (alligator) cracking, rutting, longitudinal cracking in the wheel path and outside the wheel path, transverse cracking, IRI, and deflections measured by sensors 1 and 7. In this chapter, results from site-level analyses will be followed by results from overall analysis, and results from apparent relationship between response and performance, leading to the summary of findings. Before presenting the results from analyses, a discussion on the effect of construction on the performance of SPS-1 pavements and a brief discussion of the performance of test sections at each site are presented. The analyses conducted include: • Site-level analyses (on performance measures): Evaluation of the consistency of the effects of design factors across sites. • A comprehensive overall analysis of the performance and response data, which includes the following methods of analysis: a) Extent of distress by experimental factors (Frequencies) b) Linear Discriminant Analysis (LDA) c) Binary Logistic Regression (BLR) d) Analysis of Variance (ANOVA) • An investigation of apparent relationships between response and performance at the site level and for the overall population.

156 5.2 PREVIOUS STUDIES This section summarizes the findings from the literature review of research reports that deal with pavement performance in the field. The review included FHWA/LTPP reports, NCHRP reports as well as additional literature, and was focused on research that has identified factors affecting pavement response and performance including roughness. The most relevant reports were found to be those from studies addressing the Long Term Pavement Performance (LTPP) experiments. 5.2.1 Summary of Findings The information obtained from the literature review has been used to identify various factors that have been shown in past research as having an effect on response and performance progression. 5.2.1.1 Factors Affecting Flexible Pavement Performance A study [1] entitled “Structural Factors for Flexible Pavements” was conducted using SPS-1 data. The summary of findings from this preliminary study is given below: Layer Thickness • The SPS-1 test sections with thick (178-mm (7-inch)) AC surface layers appear to be smoother and develop less fatigue cracking than those sections with thin (102-mm (4- inch)) surface layers. This confirms a similar finding from earlier studies. • In the SPS-1 experiment, AC surface thickness and the age of the project appear to influence the amount of fatigue cracking that occurs. The test sections that are younger and have thicker AC surface layers have the least fatigue cracking. Base Layer • Hot-mix asphalt (HMA) pavements with unbound aggregate base layers show greater rut depths than those sections with asphalt-treated base layers. This suggests that a portion of the rutting measured at the surface is a result of permanent deformations in the unbound aggregate base layer, which is consistent with a previous finding from analysis of the GPS test sections.

157 • The HMA pavements with unbound aggregate layers have slightly more fatigue cracking and higher IRI values than those sections with asphalt-treated base layers. • The test sections with coarse-grained soils, asphalt-treated base layers, permeable base layers, thicker bases, and thicker HMA layers were found to be smoother. • The test sections with permeable asphalt-treated base layers exhibit more fatigue cracking than those without permeable base layers. Subgrade • HMA pavements built over coarse-grained subgrade soils are smoother than pavements built over fine-grained subgrade soils. This is consistent with the finding in the SPS-2 JPCP: A stiffer foundation contributes to smoother pavements. • HMA pavements built over coarse-grained subgrade soils and in a no-freeze climate are smoother and stay smoother over a longer period of time than do those built over fine- grained subgrade soils in a freeze climate. HMA pavements built over fine-grained sub- grades and in a wet-freeze climate are substantially rougher than those built in other climates. • HMA pavements built over fine-grained subgrade soils have more fatigue cracking than those projects built over coarse-grained subgrade soils. • Subgrade soil type and, to a lesser degree, age are important to the amount of transverse cracking measured at each site. More transverse cracking has occurred on the HMA pavements built on fine-grained soils than on pavements built on coarse-grained soils. Another study [2]was conducted using SPS-1 data to identify the factors affecting pavement smoothness. A summary of main findings from this research is given below: • A significant difference between early age IRI of pavements placed on DGAB and ATB was observed. No significant difference in early age IRI was obtained on pavements placed on PATB when compared to other two base types. • The SPS-1 projects that showed the highest increase in IRI were located in Kansas, Iowa and Ohio, reasons being fatigue and transverse cracking for KS (20), transverse and longitudinal cracking (WP) for IA (19), and rutting for OH (39) site. Some of the test

158 sections in TX (48) are showing higher increase in IRI of over 10% within an approximate 6-month period, which is attributed to rutting. • Although the pavements in IA (19), KS (20), and OH (39) achieved a smooth pavement initially, many sections, including very thick sections had high increases in roughness during the initial life of the pavement. Achieving a smooth pavement initially does not guarantee that it will remain smooth even during the initial life. • Mix design problems in the AC, inadequate preparation of the subgrade prior to placing the pavement, or other construction problems can cause smoothly built pavements to have higher increase in roughness within a short time period. Two studies[3, 4] were conducted to investigate the effects of sub-drainage on the performance of asphalt and concrete pavements. The following is a summary of their findings for asphalt pavements: • Based on 7 years (on average) of SPS-1 data, those HMA sections built on permeable bases without edge drains were found to perform better than those with edge drains. • The ranking of performance in terms of IRI and cracking for various base types with all other design features matched is from poor to good performance: un-drained dense- graded aggregate bases, drained permeable asphalt-treated bases, and un-drained dense- graded asphalt-treated bases. • The results in terms of rutting for the above three sub-drainage designs were inconclusive. The results from the site level analysis for SPS-1 experiment are summarized in the next section.

159 5.3 EFFECT OF CONSTRUCTION ON PAVEMENT PERFORMANCE For the SPS-1 experiment, detailed construction guidelines were developed by LTPP (see Chapter 2) for the participating agencies to control variability in construction across sites. The lesser variability across sites, the lower the “noise”, and easier it is to determine the “pure” effects of the design factors on pavement performance. However, some deviations have occurred during construction of various sites and some of those issues were highlighted in the construction reports prepared by the participating agencies (see Chapter 3). In general, sections with serious construction deviations will perform poorer than those with normal construction conditions and inclusion of such sections in the analysis may distort (bias) the effects of design factors. Sections with deviations may be identified at least in three ways, based on: • Construction issues highlighted in construction reports for each site, • Unusual performance trend of individual sections, and • Unusual material properties of pavement layers. Depending on the nature of construction issues (construction quality) the performance of a pavement is affected. Minor issues (such as minor thickness deviation in base course) may not seriously impact initial performance where as major issues (such as HMA mix issues, compaction or drainage problems) have greater chances of affecting performance early in the life of a pavement. Hence, a construction deviation (poor quality) may be used to identify sections that may potentially show an abnormal performance. Also, some sections with serious early performance concerns were “de-assigned” from LTPP database (for example, 39-101 and 20- 101). In this study, any abnormality in early performance was used as an indicator to identify substandard sections. The performance of all the sections, over time, was observed for this purpose and those sections that had premature “failure” (within first 2 to 3 years of service life) were identified. It was observed that most of the identified sections are from a few sites (for example KS for fatigue cracking and TX for rutting in SPS-1) in the experiments, indicating consistent construction problems in those sites. In such cases, all sections from the identified sites were excluded from related analyses. Material properties (example: HMA mix properties) of pavement layers may also indicate non-compliance in construction. Limited material data are available in DataPave (Release 17). This data were also considered to explain the probable causes of unexpected performance wherever possible.

160 In order to further investigate the construction-related performance issues, each performance measure for all pavement sections in SPS-1 experiment was examined over time. This analysis helped minimize the bias, if any, in the results. The analysis is discussed next with illustrations. This section of the report is followed by site-wise performance summaries of the test sections. 5.3.1 Construction-related Issues A brief discussion of construction-related performance issues for each performance measure in SPS-1 is presented in this section of the report. Based on the time-series plots for all distress measures it was found that premature “failure” was predominantly observed in rutting for a relatively large number of pavement sections. Hence, this is presented first among all the performance measures. Rutting Figure 5-1 shows rutting in all the SPS-1 test sections over time. It can be observed that a considerable number of sections have noticeably high initial rutting. The premature rutting in these pavements can be further classified into two types based on the causes of rutting: • Mix-related rutting, and • Base layer rutting (this could be because of wet base and poor drainage or poor compaction in un-bound pavement layers). Therefore, the premature or early rutting in pavements was separated from the structural rutting in order to study the effects of structural factors on the long-term pavement performance. Table 5-1 is a summary of details regarding the pavement sections that exhibited early (pre- mature) rutting in the SPS-1 experiment. The causes of rutting were identified by using the transverse profile data available in the LTPP database based on the criteria developed in NCHRP Report 468 [5]. Figure 5-3 through Figure 5-6 show the average transverse profiles of some sections, from four of the SPS-1 sites, which showed premature rutting. HMA material-related data from the field cores were extracted from DataPave (Release 17.0). Unfortunately a limited amount of material data is available at this point in time; therefore, only a few of the mix-related properties could be calculated. The summary of the mix-related data for the sites that showed early rutting is given in Table 5-2. It can be seen from these asphalt mix properties that there is

161 high variation in the field air void content between these identified sites. High air voids in the pavement sections at the Kansas site KS (20) and very low air voids (high VFA) at the Texas site TX (48), are noticeable. The pavements built at these two sites have shown extensive cracking and rutting, respectively. To investigate the effect of structural factors on the rutting performance, the pavement sections were separated, as explained above, into two categories: (a) pavements with premature rutting, and (b) pavements which exhibited structural rutting. Figure 5-2 shows the rutting performance of sections with probable “structural” rutting. The effect of outliers (sections with premature rutting) on the rutting performance for SPS-1 pavements can be observed by comparing Figure 5-7 and Figure 5-8. An analysis of variance (ANOVA) was performed separately on two subsets of the data as well as on data from all the sections (superset of the two subsets). The first subset represents the pavement sections which have shown higher rutting at an early age (mix-related or material- related). The second subset includes the pavements which have shown normal rutting growth, and these pavements were assumed to be exhibiting “structural rutting”. Table 5-3 is the summary of results from ANOVA. Initially, only the main effects of structural factors were considered in the analysis by blocking the site. The results indicated that none of the structural factors has a significant effect on premature rutting. These results are reasonable, as it is expected that pavements will undergo accelerated rutting early in their service life (irrespective of the pavement structure) if the asphalt layer has mix-related issues or when the base has drainage-related issues. Next, ANOVA was performed by taking all the experimental factors and the results are shown in Table 5-5. The mean rut depths from both analyses by each experimental factor are shown in Table 5-4 and Table 5-6, respectively, to illustrate the effects. A brief discussion of the results from analysis of structural rutting is given below: HMA Thickness: Pavement sections with “thin” 102 mm (4-inch) HMA surface layer have undergone higher rutting compared to those with “thick” 178 mm (7-inch) HMA surface layer. However, this difference was not found to be of practical significance at this point in time. Base Type: The effect of base type on the structural rutting is not statistically significant effect, at this point. On average, sections with DGAB have shown slightly higher rutting than those built

162 with treated bases. A slight (0.05<p-value <0.1) interaction effect was observed between base type and subgrade type. Among pavements built with DGAB, those built on fine-grained soils have shown higher rutting than those built on coarse-grained soils. Base Thickness: Higher rutting occurred in sections with 203 mm (8-inch) base than those with 305 mm (12-inch) or 406 mm (16-inch) base. However, the effect is statistically marginally significant, is not of practical significance. Drainage: On average, sections with no drainage have shown slightly higher rutting than those with drainage. However, this effect of drainage on the structural rutting was not found to be statistically significant at this point in time. Subgrade Type: Rutting in sections built on fine-grained soils is comparable with rutting in sections built on coarse-grained soils. The effect of subgrade types was not found to be statistically significant for structural rutting. Climatic Zone: Effect of climate was found to be statistically significant. Pavements located in WNF zone have shown higher rutting than those located in WF zone. However, this effect is not of practical significance. Given the above findings, all subsequent statistical analyses on rutting performance, (presented later in this chapter) were performed on data from the sections with structural rutting. This assumes that only pavement sections which have exhibited structural rutting will capture the effects of design factors on the long-term rutting performance.

163 Table 5-1 Identified sites and sections with rutting problems Site Sections deleted due to extensive rutting Probable cause of rutting Comments Arizona, AZ (4) All 12 sections HMA and Base The rutting is either occurring due to HMA mix problems or base or both (from transverse profile) Kansas, KS (20) All 12 sections HMA and Base Some sections have shown HMA mix related rutting and others showed base problems (from transverse profile ) Michigan, MI (26) 113, 114, 119, 122 120 HMA The first four sections were deleted from the LTPP database just after construction Nebraska, NE (31) All 12 sections HMA From transverse profile Ohio, OH (39) 101, 102 Base From transverse profile Texas, TX (48) All 12 sections HMA Premature rutting in HMA layer [6] Virginia, VA (51) 113 Base Only one section has shown accelerated deterioration at this site due to base problems (from transverse profile) Total 56 sections Table 5-2 Average asphalt mixture properties in the field Superpave Specifications State Asphalt Content (%) Air Void Content (%) VMAa (%) VFAa (%) VMA 1 (%) VFA2 (%) 4 4.4 10.6 19.1 44.3 19 4.2 10.5 - - 20 4.1 15.3 21.4 29.1 26 5.0 6.5 - - 31 4.2 6.2 - - 39 6.6 11.2 - - 48 4.4 1.8 12.2 85.0 51 4.9 9.7 20.4 52.6 >14 65-75 Note: a Gsb values are missing therefore, these properties can not be calculated, 1 the minimum VMA requirement for nominal maximum size of 12.5 mm, 2 VFA requirements for traffic > 100 million ESALs (Source: Superpave Mix Design, SP-2)

164 0 5 10 15 20 25 30 35 0 1 2 3 4 5 6 7 8 9 10 11 Age (years) R ut D ep th (m m ) Figure 5-1 Rutting with time for SPS-1 pavements - All sections 0 5 10 15 20 25 30 35 0 1 2 3 4 5 6 7 8 9 10 11 Age (years) R ut D ep th (m m ) Figure 5-2 Rutting with time for SPS-1 pavements – Selected sections

165 -25 -20 -15 -10 -5 0 5 10 0 1000 2000 3000 4000 Distance from the edge (mm) R u t d e p t h ( m m ) Figure 5-3 Transverse profile for base rutting—Section 20-0102 -25 -20 -15 -10 -5 0 5 10 15 0 1000 2000 3000 4000 Distance from the edge (mm) R u t d e p t h ( m m ) Figure 5-4 Transverse profile for asphalt rutting— Section 31-0113 -14 -12 -10 -8 -6 -4 -2 0 2 4 0 1000 2000 3000 4000 Distance from the edge (mm) R u t d e p t h ( m m ) Figure 5-5 Transverse profile for (HMA + base) rutting— Section 39-0101 -20 -15 -10 -5 0 5 0 1000 2000 3000 4000 Distance from the edge (mm) R u t d e p t h ( m m ) Figure 5-6 Transverse profile for (Base) rutting— Section 51-0113

166 y = 3.6226x0.27 R2 = 0.3535 0 5 10 15 20 25 30 35 0 2 4 6 8 10 12 Age (years) R ut D ep th (m m ) Figure 5-7 Rutting growth with time for SPS-1 pavements – All sections y = 3.2717x0.2846 R2 = 0.3936 0 5 10 15 20 25 30 35 0 2 4 6 8 10 12 Age (years) R ut D ep th (m m ) Figure 5-8 Rutting growth with time for SPS-1 pavements – Selected sections

167 Table 5-3 Summary of p-values from ANOVA for determining the effect of main design factors on pavement rutting Rutting Type Design Factor Non-structural ruttinga Structural rutting Overall HMA thickness 0.71 0.074 0.20 Base type 0.20 0.51 0.017 Base thickness 0.99 0.08 0.195 Drainage 0.12 0.25 0.030 Site (blocked) 0.15 0.00 0.00 R 2=0.343 N=53 R2=0.55 N=159 R2=0.57 N=212 Note: a Mix-related or premature rutting in un-bound layers. Table 5-4 Summary of marginal means from ANOVA for determining the effect of main design factors on pavement rutting Rutting Type Design Factor Non-structural (mm) Structural rutting (mm) Overall (mm) 102 mm 9.0 5.3 6.1 HMA thickness 178 mm 10.0 4.9 5.8 DGAB 11.0 5.2 6.5 ATB 9.05 4.9 5.7 Base type ATB/DGAB 8.2 5.1 5.6 203 mm 10.0 5.6 6.3 305 mm 10.0 5.1 5.8 Base thickness 406 mm 9.05 5.0 5.8 N 11.0 5.3 6.3 Drainage Y 9.0 4.9 5.6 MSEa 0.206 0.062 0.10 Note: a MSE is in natural log. 1 inch = 25.4 mm

168 Table 5-5 Summary of p-values from ANOVA for determining the effect of experimental factors on pavement rutting Rutting Type Experimental Factor Non-structural ruttinga Structural rutting HMA thickness 0.16 0.043 Base type 0.94 0.54 Base thickness 0.76 0.09 Drainage 0.50 0.28 Subgrade 0.46 0.43 Zone 0.27 0.00 Traffic 0.000 0.013 R 2=0.552 N=53 R2=0.55 N=159 Note: a Mix-related or premature rutting in un-bound layers. Table 5-6 Summary of marginal means from ANOVA for determining the effect of experimental factors on pavement rutting Rutting Type Design Factor Non-structural rutting Structural rutting 102 mm 9.7 5.7 HMA thickness 178 mm 11.8 5.0 DGAB 10.7 5.4 ATB 10.7 5.2 Base type ATB/DGAB 10.7 5.3 203 mm 9.7 5.7 305 mm 10.7 5.0 Base thickness 406 mm 10.7 5.2 N 11.8 5.5 Drainage Y 10.7 5.1 F 13 5.2 Subgrade C 9.7 5.4 WF 6.5 4.9 WNF 13 5.6 DF 14.4 4.0 Zone DNF - 7.1 MSEa 0.137 0.091 Note: a MSE is in natural log. 1 inch = 25.4 mm

169 Fatigue Cracking It was observed that sections from the Kansas, KS (20), site exhibited the highest area of cracking at an early age compared to sections from other sites. The sections at KS (20) site had a wet subbase during construction (based on the construction report). Also, from the materials data (DataPave) it was found that the test sections at KS (20), on average, have “high” air void content in the HMA (see Table 5-2). These reasons could have caused the abnormally high cracking in the sections at this site. Figure 5-9 and Figure 5-10 are time-series plots of fatigue cracking for all the pavements sections, before and after exclusion of sections from the Kansas site, KS (20). All statistical analyses pertaining to fatigue cracking, presented in this chapter, were conducted without including data from sections at Kansas site, KS (20). Roughness and other Performance Measures Figure 5-11, through Figure 5-14 show the time-series plots for IRI, transverse cracking and longitudinal (WP and NWP) cracking, respectively. It can be observed that only a few sections have exhibited an abnormal performance. Exclusion of data from these sections was not considered necessary as their inclusion will not impact the results considerably. Therefore, all the pavement sections were included in the analyses of roughness, longitudinal cracking (WP and NWP) and transverse cracking. 5.3.2 Drainage-related issues In the above section, construction-related issues have been linked to the poor performance of some pavement sections. Some construction and/or maintenance related issues with respect to the in-pavement drainage were also identified in previous research [3, 4]. The in- pavement drainage for some of the SPS-1 flexible pavement sections was found to have some deviations from design.

170 0 50 100 150 200 250 300 350 400 0 1 2 3 4 5 6 7 8 9 10 11 Age (years) Fa tig ue C ra ck in g A re a (s q- m ) Figure 5-9 Fatigue cracking with time for SPS-1 pavements – All sections 0 50 100 150 200 250 300 350 400 0 2 4 6 8 10 12 Age (years) Fa tig ue C ra ck in g (s q- m ) Figure 5-10 Fatigue cracking with time for SPS-1 pavements – Selected sections

171 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0 1 2 3 4 5 6 7 8 9 10 11 Age (years) IR I ( m /k m ) Figure 5-11 IRI with time for SPS-1 pavements – All sections 0 10 20 30 40 50 60 70 80 0 1 2 3 4 5 6 7 8 9 10 11 Age (years) Tr an sv er se C ra ck in g (m ) Figure 5-12 Transverse cracking with time for SPS-1 pavements – All sections

172 0 50 100 150 200 250 300 0 1 2 3 4 5 6 7 8 9 10 11 Age (years) Lo ng . C ra ck in g- W P (m ) Figure 5-13 Longitudinal cracking-WP with time for SPS-1 pavements – All sections 0 50 100 150 200 250 300 350 0 1 2 3 4 5 6 7 8 9 10 11 Age (years) Lo ng . C ra ck in g- N W P (m ) Figure 5-14 Longitudinal cracking-NWP with time for SPS-1 pavements – All sections

173 All the drained sections of the SPS-1 experiment were video taped to assess the condition of the drainage in the project 1-34C [4]. A subjective assessment of the quality of the drainage functioning in each test section as “good” or “poor” was reported. The ratings assigned to each section are summarized in Table 5-7. The “poor” rating was indicative of; (i) buried lateral outlet, (ii) outlet fully blocked with silt, gravel or other debris (iii) longitudinal drains being fully blocked, or (iv) a considerable amount of stagnant water in the longitudinal drain. A “good” rating was given to the drainage if a reasonably sufficient flow of water was evident even if some amount of material was present in the drains. Hall et al [4] conducted preliminary analysis of the performance of SPS-1 test sections in light of their assessment of drainage, and a brief summary of their findings are presented below: • Undrained pavement sections built on DGAB may develop cracking, rutting and roughness more rapidly than drained sections built on ATB. • Undrained pavement sections built on ATB may develop roughness and cracking more slowly than those built with drained DGAB, while the un-drained sections may develop rutting more rapidly. • Undrained pavement sections built on ATB/DGAB may develop roughness and rutting more quickly than those on drained DGAB, while the undrained sections may develop cracking more slowly. • Also, among the drained sections, those with “good” rating for drainage performed better than undrained sections, while those with “poor” rating did not. However, the above trends were based only on the average performance and in no case, were the differences detected statistically significant. These preliminary findings (from Hall et al) should be considered during the interpretation/validation of the results (from this study) regarding the effect of drainage.

174 Table 5-7 Subjective ratings of drainage functioning at SPS-1 test sections based on video inspection results (source: [4]) Test Section 0101 0113 0102 0114 0103 0115 0104 0116 0105 0117 0106 0118 0107 0119 0108 0120 0109 0121 0110 0122 0111 0123 0112 0124 Base Type DGAB ATB ATB/DGAB PATB/DGAB ATB/PATB State Un-drained Drained AL (1) G1 G G G G G AZ (4) G G G G G ?2 AR (5) P3 P P P P P DE (10) G G G G G G FL (12) P P P P P P IA (19) P G P P G ? G P ? ? P KS (20) ? P ? ? ?*4 ? LA (22) ? P P ? ? ? MI (26) P ? ? P ? P ? MT (30) G G G G G G NE (31) G G G G G G NV (32) P P P P P P NM (35) P P ? P P P OH (39) G G G P ? G ? G G G OK (40) ?* ?* ?* ?* ?* ?* TX (48) P P P P P P VA (51) G G G ? G G G G G WI (55) ? ? ? ? ? ? 1G= Drainage function rated as good 2? = Drainage outlet not found 3P = Drainage function rated as poor 4?*= Camera could not be inserted

175 5.4 SPS-1 PROJECT PERFORMANCE SUMMARIES This section is a summary of the performance trends for each site within the SPS-1 experiment based on the latest year data. The performance summary for each site is based on the data available in the Release 17.0 of the DataPave. The severity levels for all types of cracking were combined to calculate its total magnitude. This descriptive summary is intended to help the reader gain an understanding of performance of test sections at each site. The performance of pavement sections regarding selected distresses is presented here, for each site. The identified distresses include fatigue cracking (sq-m), longitudinal cracking-WP (m), longitudinal cracking- NWP (m), transverse cracking (m), rutting (mm) and roughness (m/km). Additional details about each of the sites can be found in site-level summaries presented in Appendix A1 and performance data tables in Appendix A2. Alabama, AL (1) Performance data is available for 10 years (1994-2003) at this site. The ‘proposed’ traffic is 237 KESAL per year. Fatigue cracking is the dominant distress at this site. Sections 103, 104, 106, 107 and 112 have less than 10% (area) cracking while all other sections have fatigue cracking of range 10% to 15%. A wide range of longitudinal cracking-WP (between 5 m and 30 m) occurred on all the sections. Longitudinal cracking-NWP, between 80 m and 200 m, occurred on all the sections, by year 8. Transverse cracking, of range 15 m to 50 m, was observed in sections 101, 102, and 105 respectively. Sections 102 and 105 have shown 10 mm and 17 mm of rutting, respectively, while other sections have rutting between 6 mm to 9 mm. Sections 102 and 107 have IRI of 1.4 m/km and 1.7 m/km, respectively, while other sections have IRI less than 1.0 m/km. Arizona, AZ (4) The performance data is available for 10 years (1994-2003) at this site. The ‘proposed’ traffic is 214 KESAL per year. Less than 3% (of area) of fatigue cracking occurred in sections 113, 119 and 124, while less than 1% cracking occurred in other sections. Longitudinal cracking-WP is the dominant type of cracking with all the sections showing a cracking between 10 to 150 m. Longitudinal cracking-NWP of 150 m and 120 m occurred on sections 114 and 120, respectively; while in other sections longitudinal cracking-NWP is less than 50 m. Transverse

176 cracking of length 76 m and 45 m occurred on sections 113 and 121, while other sections have less than 30 m of transverse cracking. Rutting of 14 mm and 25 mm occurred on sections 114 and 119, respectively, after 6 years. In other sections rutting ranged from 3 mm to 9 mm. All sections except 113, 120 and 122 have IRI greater than 1 m/km while other sections have IRI less than 1.0 m/km. Arkansas, AR (5) The performance data is available for 9 years (1995-2003) for this site. The ‘proposed’ traffic is 385 KESAL per year. Fatigue cracking area ranged from 10% to 25% in sections 119, 120 and 121; whereas, all other sections have exhibited less than 10% of fatigue cracking area. All the sections exhibited longitudinal cracking-WP less than 10 m. All the sections have exhibited longitudinal cracking-NWP, which ranged from 140m to 280 m. Transverse cracking of 48 m was observed only in section 119, whereas all other sections have less than 20 m of cracking. Rutting between 5 mm to 9 mm was observed at the site. Sections 119 and 120 have IRI of about 1.7 m/km while other sections have IRI of about 1 m/km. Delaware, DE (10) The performance data is available for 7 years (1996-2003) for this site. The ‘proposed’ traffic is 309 KESAL per year. Fatigue cracking is the dominant distress at this site. Sections 101 and 102 have cracking of about 10% and 20%, while in other sections cracking was less than 10%. Longitudinal and transverse cracking did not occur on any of the sections. All sections except 102 have shown rutting of range 2 to 4 mm while, sections 102 has 7 mm of rut depth. IRI for all the test section is less than 1.0 m/km. Florida, FL (12) The performance data is available for 7 years (1996-2003) for this site. The ‘proposed’ traffic is 464 KESAL per year. Fatigue cracking less than 1% occurred in the sections. Longitudinal cracking-WP, less than 10 m, occurred in sections 107, 108, 110, and 112. Longitudinal cracking-NWP, less than 50 m was observed in sections 101,105,108 and 110. Transverse cracking was not observed on any of the sections. Rutting of about 4 mm occurred in all sections, except sections 103, 105, 110 and 111, which exhibited rutting of about 6 mm. All the test sections have shown less than 1 m/km of IRI.

177 Iowa, IA (19) The performance data is available for 9 years (1995-2003) at this site. The ‘proposed’ traffic is 132 KESAL per year. Fatigue cracking of 10% (area) was observed on section 102, and all other sections have fatigue cracking less than 2%. Longitudinal cracking-WP of range 50 m to 100m occurred in 102, 104, 105, and 107, while in other sections this cracking is less than 30 m. Longitudinal cracking-NWP of range 110 m to 295 m occurred at this site, in all the sections. Transverse cracking of 30 m to 70 m was also observed in sections 101 through 106, while in other sections this cracking is less than 20 m. Sections 107, 108, and 109 have rut depth of about 7 mm while other sections have rutting between 3 mm and 6 mm. Sections 101 and 102 have IRI values of 1.8 m/km and 2.5 m/km, while IRI in other sections ranged from 1.0 m/km to 1.6 m/km. Kansas, KS (20) The performance data is available for 8 years (1993-2001) at this site. The ‘proposed’ traffic is 228 KESAL per year. Fatigue cracking is the main distress type in all the sections. Fatigue cracking ranged from 5% to 20%. Longitudinal cracking-WP has not occurred at this site. Sections 103, 104, 105 and 110 have longitudinal cracking-NWP of 72 m to 189 m, while in other sections the this cracking is less than 20 m. Sections 103, 105 and 110 have transverse cracking less than 16 m. Sections 101, 102, 107 (data available for first 2 years only) have shown rut depth between 16 mm to 25 mm, whereas section 105 has exhibited 13 mm of rutting, after 3 years. Other sections have rutting less than 5 mm. Section 105 has an IRI of 2.7 m/km while sections 104, 109 through 112 have IRI less than 1.2 m/km. Other remaining sections have IRI between 1.5 and 2.0 m/km. Louisiana, LA (22) The cracking data is available for only 2 years (1997-1999) at this site. The ‘proposed’ traffic is 524 KESAL per year. The rutting data is available for 6 years (1998-2003) while roughness data is available only for one year (1997). No cracking occurred on any of the sections. Sections 119, 122 and 123 have rutting less than 4.0 mm while; other sections have rutting of about 5 mm. All sections have initial roughness less than 0.8 m/km.

178 Michigan, MI (26) The performance data is available for 7 years (1996-2003) for this site. The ‘proposed’ traffic is 189 KESAL per year. The performance data is only available for eight sections for this site, as four sections (112, 114, 119 and 122) have been ‘deassigned’ from LTPP due to construction issues. Fatigue cracking of 2 to 10% was observed on sections 115, 116, 117 and 124, while no cracking occurred in other sections. Longitudinal cracking-WP and transverse cracking has not occurred on any of the test sections. Longitudinal cracking-NWP was observed on all test sections. Section 116 has longitudinal cracking-NWP of 35 m while this cracking in other sections ranged from 120 m to 188 m. Sections 115 and 117 have rutting of 9 mm and 12 mm while other sections have rutting of about 6 mm. Sections 118 and 117 have IRI of 1.2 m/km and 1.4 m/km whereas other sections have IRI less than 1.0 m/km. Montana, MT (30) The performance data is available for 5 years (1998-2003) at this site. The ‘proposed’ traffic is 127 KESAL per year. Fatigue cracking was observed in sections 115, 117, 120, 121, and 122, with a range of 5% to 10%. Other sections have fatigue cracking of less than 5%. Longitudinal cracking-WP occurred only in sections 113, 114, 115, 118, and 124, with a range of 5 to 10 m. Longitudinal cracking-NWP occurred in all sections, with a range of 150 to 208 m. Transverse cracking were observed only in sections 113, 115 and 121, and this cracking was between 5 m to 10 m. Rutting of 8 mm occurred in sections 120 and 121, while it ranged from 3 to 5 mm in other sections. Sections 120 and 121 have IRI of about 1.5 m/km whereas other sections have IRI of range 0.8 m/km to 1.1 m/km. Nebraska, NE (31) The performance data is available for 7 years (1995-2002) at this site. The ‘proposed’ traffic is 113 KESAL per year. Fatigue cracking has just initiated only in sections 113 and 114. Longitudinal cracking-WP of 97 m was observed in section 113. Longitudinal cracking-NWP and transverse cracking did not occur in any of the sections. Rutting of 29 mm was observed on section 113 by year 5. Among other sections, 114, 115, 118, 123, and 124 have rutting between 11 mm and 15 mm, while others have rutting between 5 mm and 8 mm. Sections 113 has an IRI of 1.9 m/km and all other sections have IRI of about 1.0 m/km. All test sections have an initial IRI between 0.9 and 1.4 m/km.

179 Nevada, NV (32) The performance data was collected for 8 years (1996-2003) for this site. The ‘proposed’ traffic is 475 KESAL per year. Fatigue cracking area is less than 1% in all the test sections. No noticeable longitudinal cracking-WP was observed in any of the sections. More than 50 m of longitudinal cracking-NWP was observed only in section 103 and sections 101, 102, 104, 105 and 109 have less than 35 m of longitudinal cracking-NWP. Less than 15 m of transverse cracking occurred in sections 102, 107 and 109. Sections 101, 104, 107, and 108 have rut depth less than 4 mm while other sections have rut depth of about 5 mm. All sections except 102 have IRI less than 1 m/km. Section 102 has an IRI value of 1.4 m/km. New Mexico, NM (35) The performance data was collected for 8 years (1997-2003) at this site. The ‘proposed’ traffic is 149 KESAL per year. Fatigue cracking, less than 1% of area, occurred in sections 102 to 104. Sections 101, 103,105 and 107 have exhibited less than 45 m of longitudinal cracking-WP, while section 102 has 70 m of this cracking. Furthermore, less than 25 m of longitudinal cracking- NWP was observed only in sections 101, 102, and 107. No transverse cracking was observed at this site. All sections at this site have exhibited rut depth of about 7 mm. All sections except 101, 102, 103, 107 and 112 have IRI between 1.0 to 1.3 m/km while other sections have less than 1.0 m/km of IRI. Ohio, OH (39) The performance data was collected for 7 years (1996-2002) for this site. The ‘proposed’ traffic is 390 KESAL per year. At this site, cracking data for sections 101, 102,105 and 107 are only available for the initial year. These sections were ‘deassigned’ from the LTPP because of premature rutting. Fatigue cracking is the dominant cracking distress within all sections at this site. All sections, except 109, have fatigue cracking between 3 to 15% of area. All sections, except 104, 111, and 112, have between 175 to 245 m of longitudinal cracking-WP. Also, all sections except 109, 110, 111 and 112, have longitudinal cracking-NWP between 200 to 260 m. No transverse cracking was observed in any section at this site. Sections 101 and 102 had more than 10 mm of rut depth after only 1 year. Also sections 103 108, and 109 have exhibited rutting of about 10 mm. IRI more than 2 m/km was observed on sections 101 and 102 after only 1 year

180 of construction. All sections have IRI greater than 1.4 m/km, while sections 103 and 108 have IRI of 3 and 2 m/km, respectively. Oklahoma, OK (40) The performance data was collected for 6 years (1997-2003) for this site. The ‘proposed’ traffic is 281 KESAL per year. Less than 3% (area) of fatigue cracking was observed in all the sections. No longitudinal cracking-WP occurred on any section at this site. Longitudinal cracking-NWP between 20 m and 120 m was observed in all sections except, section 113, which has no longitudinal cracking-NWP. Transverse cracking less than 4 m was observed in sections 113, 115 and 121. Sections 113, 117 and 122 have more than 10 mm of rut depth, while the other sections have rutting between 4 to 8 mm. Sections 113 through 118 have IRI of about 1.1 m/km whereas, all other sections have less than 1.0 m/km of IRI. Texas, TX (48) At this site, performance data was collected for 5 years (1997-2002). The ‘proposed’ traffic is 360 KESAL per year. Fatigue cracking, less than 1% (area) and longitudinal cracking-WP (between 20 m and 50 m), was observed only in sections 113, 117, 118 and 122. Only sections 112 and 119 have exhibited longitudinal cracking-NWP of 54 m and 77 m, respectively. No transverse cracking was observed in any of the sections. Severe early rutting, between 10 mm to 18 mm, was observed in sections 114, 115, 116, 119, 123 and 124, after 1 year. Rutting in these sections progressed to about 14 mm to 26 mm, by year 4. All other sections have rutting less than 11 mm, by year 4. Sections 115, 116 and 119 have IRI between 1.2 and 1.8 m/km, after only 2 years and all other sections have IRI less than 1.0 m/km. Virginia, VA (51) The performance data was collected for 6 years (1995-2002) for this site. The ‘proposed’ traffic is 257 KESAL per year. More than 25% of fatigue cracking was observed only on sections 103 and 120. Longitudinal cracking-WP and transverse cracking has not occurred on any of the sections. Sections 114 and 120 have exhibited longitudinal cracking-NWP of 40 m and 4 m, respectively. Section 103 has rut depth of 12 mm just after 1 year and this progressed to 21 mm by year 5. All other sections have exhibited rutting ranging between 4 mm and 6 mm. All sections, except 113, have IRI less than 1.3 m/km. Section 113 has an IRI of 1.9 m/km.

181 Wisconsin, WI (55) The performance data was collected for only 4 years (1998-2002) at this site. The ‘proposed’ traffic is 161 KESAL per year. Fatigue cracking less than 5% was observed only on sections 113, 114, and 116. Longitudinal cracking-WP and transverse cracking has not occurred on any of the sections. A wide range (between 90 m and 300 m) of longitudinal cracking-NWP was observed on all the sections. Sections 120 and 121 have rutting of 4.0 mm while all other sections have rutting between 6 to 9 mm. All sections have IRI less than 1.3 m/km. 5.5 SITE-LEVEL ANALYSIS The site-level analysis deals with each SPS-1 project separately. The main advantage of this analysis is that it is unaffected by the variability between SPS-1 sites. For each site, the climatic conditions, subgrade type and traffic are the same. Construction conditions, material sources and surveys can also be considered the same for a given SPS-1 project. As described in Chapter 4, the site-level analyses consists of two types of comparisons: (i) Level-A— In this analysis all designs (0101 through 0112, or 0113 through 0124) at a given site are compared such that only one factor is held common within the sections of each group. For example in level-A analysis, the effects of HMA thickness (102 mm vs. 178 mm) were considered for all twelve sections within a site by ignoring base type & thickness, and drainage. (ii) Level-B—analysis- In this analyses, most of the factors are ‘controlled’ for comparisons. In level B analysis, the effect of HMA thickness (102 mm vs. 178 mm) was compared for only those sections for which all other factors are the same. The concepts of Performance Index (PI) and the relative performance were used in the site level analysis. These concepts were introduced in Chapter 4 of this report. The site-level analysis process is summarized in Figure 5-15.

182 Figure 5-15 Methodology for site level analysis (SPS-1) Site Level Analysis Level-A Comparisons Level-B Comparisons Effect of HMA Thickness 4” vs. 7” Effect of Drainage Yes vs. No 4” vs. 7” Effect of Base Type DGAB, ATB, ATB/DGAB, PATB/DGAB, PATB/ATB Effect of HMA Thickness 4” vs. 7” Controlling for other factors Effect of Base Type ATB, ATB/DGAB Controlling for other factors Effect of Base Thickness 12” vs. 16” Controlling for other factors Effects of design features and site factors

183 Each analysis was conducted separately for each performance measure. The pavement performance measures considered include: • Fatigue cracking • Rutting • Roughness (IRI) • Transverse cracking, and • Longitudinal cracking (WP and NWP) The PIs and relative performance ratios for the main design factors were calculated at all sites. Because the relative performance is a ratio, it can be used across the sites. The relative performance ratios for a given design factor from all eighteen states were used to test the significance of its effect. A two-level non-parametric (Wilcoxon Signed Ranks) test was done to evaluate the effect of HMA thickness and drainage, and a multiple level non-parametric comparison test was done to evaluate the effect of base type. The p-values from these tests are reported in the discussion of the results. To evaluate the interactive effect of the design factors with climatic zone and subgrade type, the average relative performance ratios within the same climatic zones and for both subgrade types were compared. The computed PIs and relative performance ratios for all the distresses are summarized and presented in Appendix A4. The following is a summary of the main findings from each method of analysis, categorized by design factor and performance measure.

184 5.5.1 Effects of design features on performance – Paired Comparisons at Level-A A summary of p-values obtained from the non-parametric tests on RPIs (for all 18 sites) from level-A analyses is in Table 5-8. It is important to note that the significance of a factor indicates consistency in its effect across all the sites but not necessarily a significance of its effect on the magnitude of distress. Table 5-8 Summary of p-values (non-parametric test) for Site Analysis - Level-A Performance Measures Longitudinal cracking Design Factor Fatigue cracking Rutting Roughness Transverse cracking WP NWP HMA thickness 0.013 0.408 0.009 0.050 0.311 0.368 Base type 0.033 0.529 0.000 0.079 0.599 0.883 Base thickness - - - - - - Drainage 0.047 0.056 0.020 0.040 0.035 0.028 The following is a summary of the main findings from paired comparisons at level-A categorized by design factor and performance measures: HMA Thickness The effect of HMA thickness is consistent on roughness (IRI), fatigue cracking, and transverse cracking, with 7-inch HMA pavements showing better performance. It is not consistent for longitudinal cracking (WP and NWP)) and rutting (see Table 5-8). • Fatigue Cracking: HMA thickness appears to have a consistent effect on fatigue cracking. Sections with 178 mm HMA thickness have consistently (across sites) performed better than those with 102 mm HMA thickness. Twelve sites show a positive effect, compared to five sites showing a negative effect and one site showing no effect. The effect of HMA thickness is less seen among the sites located in WF zone. Also, on average, the superior performance of 178 mm over 102 mm HMA sections can be seen more for sections on coarse-grained subgrade than for those on fine-grained subgrade. • Rutting: The effect of HMA thickness on rutting is not consistent. Nine sites show a positive (lesser rutting in sections with 178 mm HMA surface thickness) effect and the other nine show a negative effect, which shows no definitive trend of the effect across sites.

185 • Roughness (IRI): The effect of HMA thickness on IRI is consistent across sites. On average, sections with 178 mm HMA surface thickness have consistently performed better than those with 102 mm HMA thickness. This trend was observed in thirteen out of eighteen sites and two sites showed no effect. • Transverse Cracking: The effect of HMA thickness on transverse cracking is consistent across sites. Sections with 178 mm HMA thickness have consistently performed better than those with 102 mm HMA thickness. Eight out of eighteen sites showed less transverse cracking for 178 mm sections; seven sites showed no transverse cracking, while three sites showed more transverse cracking for 178 mm HMA sections. In terms of climatic zones, the positive effect of HMA thickness was observed in all zones except for WF. This could be attributed to the severe environmental conditions in WF zone where even thicker HMA may not be able to inhibit cracking. Averaging over all climatic zones, the effect of HMA thickness is essentially the same for both subgrade types. • Longitudinal Cracking-WP: The effect of HMA thickness on longitudinal cracking-WP is not consistent. Nine sites show a positive effect and the other nine sites show a negative effect, which shows no definitive trend of the effect across sites. • Longitudinal Cracking-NWP: HMA thickness seems to have no consistent effect on longitudinal cracking-NWP. Nine sites show a positive effect and the eight sites show a negative effect, indicating no definitive trend of the effect across sites. Effect of Base Type The effect of base type on fatigue cracking, roughness, and transverse cracking is consistent across sites, with sections on DGAB showing the worst performance and sections on ATB+PATB showing the best performance. • Fatigue cracking: Base type has a consistent effect, across sites, on fatigue cracking. Sections with DGAB have shown the most amount of cracking while sections with ATB+PATB have shown least amount of cracking. The order of performance from best to worst is as follows: (1) PATB+ATB, (2) ATB, (3) ATB+DGAB (4) PATB+DGAB and (5) DGAB. • Rutting: The effect of base type is not consistent. However, on average, sections with PATB+ATB have slightly lesser rutting compared to sections with DGAB.

186 • Roughness (IRI): The effect of base type on IRI is consistent across sites. Sections with DGAB have consistently shown the highest IRI-values while sections with ATB+PATB have shown the lowest values. The order of performance from best to worst is as follows: (1) PATB+ATB, (2) ATB+DGAB, (3) ATB, (4) PATB+DGAB, and (5) DGAB. • Transverse cracking: The effect of base type appears to be marginally consistent across sites (p=0.079). Sections with PATB+ATB have somewhat lesser cracking than sections with DGAB. • Longitudinal cracking-WP: The effect of base type is not consistent across sites. Nonetheless, on average, the permeable bases appear to be performing better. The worst performance was shown by the sections with DGAB. • Longitudinal cracking-NWP: The effect of base type is not consistent across sites. However, on average, the best performing bases are the permeable bases- PATB+ATB and PATB+DGAB, and the worst performing base is DGAB. Effect of Drainage The effect of drainage on all performance measures is consistent across sites, with drained pavements showing better performance than un-drained pavements (see Table 5-8). • Fatigue cracking: Drainage has a consistent effect across sites on fatigue cracking. Sections with drainage have performed better than those without drainage. Twelve sites show a positive effect (better performance of drained sections), compared to five sites that show a negative effect. Averaging over all climatic zones, the effect of drainage can be seen better for sections with fine-grained subgrade as opposed to coarse-grained subgrade. • Rutting: The effect of drainage on rutting is consistent. Sections with drainage have consistently performed better than those without drainage. Twelve sites show a positive effect, compared to four sites showing a negative effect and two sites showing no effect. • Roughness (IRI): Drainage has a consistent effect on roughness. Sections with drainage have consistently (across sites) performed better than those without drainage. In fifteen out of eighteen sites drained sections have shown a better performance, while reverse trend was found in only three sites. This effect is less seen among sections in DF zone and could be attributed to the fact that drainage is not as important in this zone.

187 • Transverse cracking: Drainage has somewhat consistent effect on transverse cracking. Sections with drainage have performed better than those without drainage in most of the sites. Seven sites show a positive effect, while no transverse cracking occurred in seven of the sites. Averaging over all climatic zones, the effect of drainage is better seen for sections built on fine-grained subgrade than for sections built on coarse-grained subgrade. • Longitudinal cracking-WP: The effect of drainage on longitudinal cracking-WP is consistent across sites. On average, sections with drainage have consistently performed better than those without drainage. Eleven sites show a positive effect, compared to three sites showing a negative effect and four sites showing no effect. Moreover, the effect is more prominent for sections on fine-grained subgrades as opposed to those on coarse-grained subgrades. • Longitudinal cracking-NWP: Drainage has a consistent effect on longitudinal cracking- NWP. On average, sections with drainage have consistently performed better than those without drainage. Thirteen sites show a positive effect, compared to three sites showing a negative effect and two sites showing no effect.

188 5.5.2 Effects of design features – Paired Comparisons at Level-B As explained in Chapter 4, level-B comparisons are more “controlled” compared to level- A comparisons. To study the consistency of the effect of HMA thickness across sites, nonparametric testing was performed on relative performance corresponding to 102 mm and 178 mm HMA thicknesses, within each base type. Similarly, to investigate the effect of base thickness, nonparametric testing was performed on relative performance corresponding to 305 mm and 406 mmbase thicknesses, within PATB/DGAB and ATB/PATB. Also, nonparametric testing was performed on relative performance corresponding to ATB and ATB/DGAB, within 203 mm and 305 mm base thicknesses. For each of these effects, the corresponding p-values are presented in Table 5-9. A brief summary of results from the level-B comparisons follows. Table 5-9 Summary of p-values (non-parametric test) for Site Analysis - Level-B Performance Measures Longitudinal cracking Design Factor Fatigue cracking Rutting Roughness Transverse cracking WP NWP HMA thickness 0.041 0.080 0.005 0.010 0.203 0.110 Base type 0.969 0.214 0.150 0.552 0.929 0.551 Base thickness 0.307 0.022 0.046 0.933 0.499 0.387 HMA Thickness The effect of HMA thickness can be examined for sections with different base types. Among sections with DGAB, it was observed that HMA thickness has a positive effect of on fatigue cracking, transverse cracking, and roughness (IRI). Also, on average, a positive effect of HMA thickness was observed on sections with ATB for fatigue cracking, but it is not consistent. The same effect was observed for rutting. This suggests that increasing HMA thickness is more effective in pavement sections with unbound bases as compared to those with treated bases. Base Type The effect of all five base types cannot be evaluated effectively for Level-B comparisons since the only base types that can be compared are ATB and ATB+DGAB. Nonetheless, the

189 results show that the difference in performance between these two base types is not statistically significant. Base Thickness The effect of base thickness can be better seen when using Level-B comparisons, mainly because it is a more secondary effect relative to HMA layer thickness and base type (treated versus untreated). Therefore it is more useful to look at this effect for two types of (permeable) bases: (1) DGAB and (2) ATB. The effect of base thickness was found to be consistent for rutting and roughness in pavement sections with DGAB. The effect of base thickness is not consistent for fatigue cracking and longitudinal cracking-NWP. However, on average, thicker (16-inch) bases have shown better performance than thinner (305 mm) bases. Finally, the effect of base thickness is not consistent for transverse cracking and longitudinal cracking-WP.

190 5.6 OVERALL ANALYSIS The results obtained from statistical analyses performed on the SPS-1 data are presented in this section. Both the performance and response variables were analyzed to study the effects of various design and site-factors on the pavement sections. Analyses were performed combining all data and is referred to as ‘Overall’ analyses. Analyses were also conducted in each climatic zone combining data from all sections within a zone as per the recommendation of the project panel. Linear Discriminant Analysis (LDA), Binary Logistic Regression (BLR), and Analysis of Variance (ANOVA) are the statistical methods that were employed for analyses. Before presenting the results from statistical analyses, the extent of distresses that occurred on the test sections is discussed. 5.6.1 Extent of Distress by Experimental Factor This section discusses the effect of the key experimental factors on performance through the relationship between the magnitude and relative occurrence of the observed distresses. Figure 5-16 through Figure 5-21 show the percentage of test sections that have exceeded various levels of distress for the key performance measures, categorized by experimental (design and site) factors. Note that the effect of climatic zone is only shown for the wet regions because of the limited number of sites in the dry regions (only four). The following is a brief interpretation of these figures: Fatigue cracking: Figure 5-16 indicates that about 70% of all test sections have shown some fatigue cracking, with about 10% of all test sections showing 20% or higher cracking by area. The effects of specific design and site factors are discussed below. a) HMA Thickness: About 75% of sections with thin HMA surface layer have shown some fatigue cracking as compared to about 65% of sections with thick HMA surface layer; the effect of HMA thickness tends to be larger for higher levels of fatigue cracking. b) Base Type: The difference in the percentage of test sections that have shown fatigue cracking between those with unbound (DGAB) and those with treated (ATB) bases is highest among all experimental factors (about 15%), with sections built on DGAB bases showing the highest percentages.

191 c) Base Thickness: The effect of base thickness on fatigue cracking was found to be insignificant. d) Drainage: The effect of drainage in terms of higher percentage of test sections showing fatigue cracking is more pronounced at the lower levels of fatigue; the effect becomes insignificant at the later stages of fatigue. This could mean that drainage is more effective in the early life of the pavement, and becomes less effective later in the pavement life. Also, the effect of drainage is slightly more visible for fine-grained than for coarse- grained subgrade soils [Figure 5-16 (d)]. e) Climate: There are consistently more sections in wet-freeze (WF) than wet-no-freeze (WNF) climate that have shown fatigue cracking exceeding various levels, with about 10% more sections in WF than in WNF climate. f) Subgrade Type: There are consistently about 15% more sections built on fine-grained than coarse-grained subgrade soils that have shown fatigue cracking exceeding various levels, and the effect of subgrade soils tends to be larger for higher levels of fatigue cracking. Rutting: Figure 5-17 indicates that about 60% of all test sections have shown rut depths higher than 0.25 inch (6.25 mm), and about 20% of all test sections showing rut depths higher than 0.5 inch (12.5 mm). The effects of specific design and site factors are discussed below. a) HMA Thickness: The effect of HMA thickness on rutting was found to be negligible. b) Base Type: There are about 10% to 15% more sections with unbound (DGAB) bases that have rut depths greater than 7.5 mm than those with treated (ATB) bases. This difference is relatively constant even at higher rut depths. c) Base Thickness: There is a slight effect of base thickness for sections that have rut depths that are less than 7.5 mm, with about 5% more sections with thinner (8 inch) bases than those with thicker (16 inch) bases. The effect becomes less apparent for rut depths greater than 7.5 mm. d) Drainage: There are consistently about 5% more sections without drainage than with drainage that have exceeded various rut depth levels. Also, the effect of drainage is slightly more noticeable at the higher rut depths and for fine-grained subgrade soils [Figure 5-17 (d)].

192 e) Climate: The effect of climate (within wet regions) on rutting appears to be more significant at rut depth higher than 7.5 mm, with about 10% more sections in wet-freeze than in wet-no-freeze climate exceeding various rut depths. f) Subgrade Type: There are consistently about 10% more sections built on fine-grained than coarse-grained subgrade soils that exceed various rut depths. Roughness (IRI): Figure 5-18 indicates that about 60% of all test sections have shown IRI values higher than 1 m/km, with about 20% of all test sections showing IRI values higher than 1.4 m/km. The effects of specific design and site factors are discussed below. a) HMA Thickness: The percentage of test sections with thin (102 mm) HMA surface layer that have exceeded an IRI of 1.2 m/km is about 40% as compared to about 20% for test sections with thick (7 inch) HMA surface layer. The percentage of test sections with thin (102 mm) HMA surface layer exceeding higher IRI levels is 5% to 10% more than that of test sections with thick (7 inch) HMA surface layer. b) Base Type: The percentage of test sections with unbound aggregate base (DGAB) that have exceeded an IRI of 1.2 m/km is about 40% as compared to about 20% for test sections with asphalt treated base (ATB). The percentage of test sections with a 203 mm base exceeding higher IRI levels is about 10% to 15% more than that of test sections with a 406 mm base thickness. c) Base Thickness: The effect of base thickness on roughness is more pronounced than for other performance measures, showing a percentage of test sections with a DGAB base exceeding 1.2 m/km and higher IRI levels that is about 10% to 15% more than that of test sections with an ATB base. d) Drainage: There are consistently about 5% more sections without drainage than with drainage that have exceeded various IRI levels. e) Climate: The effect of climate (within wet regions) on roughness appears to be the most significant, with about 20% to 30% more sections in wet-freeze than in wet-no-freeze climate exceeding 1.2m/km and higher IRI levels. f) Subgrade Type: The effect of subgrade type on roughness is more pronounced than for other performance measures, with about 15% to 30% more sections on fine-grained than on coarse-grained subgrade exceeding 1.2m/km and higher IRI levels.

193 Transverse cracking: Figure 5-19 indicates that about 40% of all test sections have shown some transverse cracking, with about 10% of all test sections showing 20m or higher length of transverse cracking. The effects of specific design and site factors are discussed below. a) HMA Thickness: Only a slight effect of HMA thickness was found on transverse cracking. b) Base Type: Base type appears to be a significant factor affecting transverse cracking. About 10% to 15% more test sections with unbound aggregate base (DGAB) than those built with an asphalt treated base (ATB) at various levels of transverse cracking. c) Base Thickness: Only a slight effect of base thickness was observed. d) Drainage: Only a slight effect of drainage on transverse cracking was observed. e) Climate: Climate seems to be a significant factor affecting transverse cracking. There are about 15% to 20% more test sections in WF zone than those built in WNF zone at various levels of transverse cracking. f) Subgrade Type: Subgrade soil type seems to have some effect on transverse cracking, in that, slightly higher proportion of sections built on fine-grained soils have shown cracking compared to those built on coarse-grained soil. Longitudinal cracking: Figure 5-20 and Figure 5-21 indicate that about 50% of all test sections have shown some longitudinal cracking-WP and about 75% of all test sections have shown some longitudinal cracking-NWP. The effects of experimental factors are discussed below. a) HMA thickness: HMA thickness appears have a negligible effect on longitudinal cracking. b) Base Type: There seems to be a slight effect of base type on longitudinal cracking, in that sections with ATB has shown lesser cracking than those with DGAB. c) Base Thickness: Only a slight effect of base thickness was observed. Sections with 406 mm base thickness have slightly less occurrence of cracking than those with 203 mm base. d) Drainage: There appears to be some positive effect of drainage on lower levels of longitudinal cracking-WP. However this effect was observed to be negligible for higher levels of cracking.

194 e) Climatic Zone: The effect of climatic zone (within wet regions) on longitudinal cracking is more pronounced than other effects, especially for longitudinal cracking-NWP. About 10% to 20% more test sections in wet-freeze than in wet-no-freeze climate exceed 100 m or more of longitudinal cracking-WP, and about 20% to 35% more test sections in wet- freeze than in wet-no-freeze climate exceed 100 m or more of longitudinal cracking- NWP. f) Subgrade Type: Only a slight positive effect of subgrade type was observed for longitudinal cracking-WP.

195 0% 20% 40% 60% 80% 100% 0% 5% 10% 15% 20% 25% % Fatigue Cracking P e r e c e n t o f t e s t s e c t i o n s 4 7 (a) HMA thickness 0% 20% 40% 60% 80% 100% 0% 5% 10% 15% 20% 25% % Fatigue Cracking P e r c e n t o f t e s t s e c t i o n s DGAB ATB ATB+DGAB (b) Base type 0% 20% 40% 60% 80% 0% 5% 10% 15% 20% 25% % Fatigue Cracking P e r c e n t o f t e s t s e c t i o n s 8 12 16 (c) Drainage 0% 20% 40% 60% 80% 0% 5% 10% 15% 20% 25% % Fatigue Cracking P e r e c e n t o f t e s t s e c t i o n s No Drainage Drainage (d) Drainage with fine-grained subgrade 0% 20% 40% 60% 80% 100% 0% 5% 10% 15% 20% 25% % Fatigue Cracking P e r e c e n t o f t e s t s e c t i o n s WF WNF e) Climatic zone 0% 20% 40% 60% 80% 100% 0% 5% 10% 15% 20% 25% % Fatigue Cracking P e r e c e n t o f t e s t s e c t i o n s Fine Coarse (f) Subgrade type Figure 5-16 Effect of experimental factors on fatigue cracking

196 0% 20% 40% 60% 80% 100% 0 2.5 5 7.5 10 12.5 15 Rut Depth (mm) P e r e c e n t o f t e s t s e c t i o n s 4 7 (a) HMA Thickness 0% 20% 40% 60% 80% 100% 0 2.5 5 7.5 10 12.5 15 Rut Depth (mm) P e r c e n t o f t e s t s e c t i o n s DGAB ATB ATB+DGAB (b) Base type 0% 20% 40% 60% 80% 100% 0 2.5 5 7.5 10 12.5 15 Rut Depth (mm) P e r c e n t o f t e s t s e c t i o n s 8 12 16 (c) Base thickness 0% 20% 40% 60% 80% 100% 0 2.5 5 7.5 10 12.5 15 Rut Depth (mm) P e r e c e n t o f t e s t s e c t i o n s No Drainage Drainage (d) Drainage with fine-grained subgrade 0% 20% 40% 60% 80% 100% 0 2.5 5 7.5 10 12.5 15 Rut Depth (mm) P e r e c e n t o f t e s t s e c t i o n s WF WNF (e) Climatic zone 0% 20% 40% 60% 80% 100% 0 2.5 5 7.5 10 12.5 15 Rut Depth (mm) P e r e c e n t o f t e s t s e c t i o n s Fine Coarse (f) Subgrade type Figure 5-17 Effect of experimental factors on rutting

197 0% 20% 40% 60% 80% 100% 0.6 0.8 1 1.2 1.4 1.6 1.8 2 IRI (m/km) P e r e c e n t o f t e s t s e c t i o n s 4 7 (a) HMA thickness 0% 20% 40% 60% 80% 100% 0.6 0.8 1 1.2 1.4 1.6 1.8 2 IRI (m/km) P e r c e n t o f t e s t s e c t i o n s DGAB ATB ATB+DGAB (b) Base type 0% 20% 40% 60% 80% 100% 0.6 0.8 1 1.2 1.4 1.6 1.8 2 IRI (m/km) P e r c e n t o f t e s t s e c t i o n s 8 12 16 (c) Base thickness 0% 20% 40% 60% 80% 100% 0.6 0.8 1 1.2 1.4 1.6 1.8 2 IRI (m/km) P e r e c e n t o f t e s t s e c t i o n s No Drainage Drainage (d) Drainage 0% 20% 40% 60% 80% 100% 0.6 0.8 1 1.2 1.4 1.6 1.8 2 IRI (m/km) P e r e c e n t o f t e s t s e c t i o n s WF WNF (e) Climatic zone 0% 20% 40% 60% 80% 100% 0.6 0.8 1 1.2 1.4 1.6 1.8 2 IRI (m/km) P e r e c e n t o f t e s t s e c t i o n s Fine Coarse (f) Subgrade type Figure 5-18 Effect of experimental factors on roughness

198 0% 20% 40% 60% 80% 100% 0 5 10 15 20 25 Transverse Cracking (m) P e r c e n t o f t e s t s e c t i o n s 4 7 (a) HMA thickness 0% 20% 40% 60% 80% 100% 0 5 10 15 20 25 Transverse Cracking (m) P e r c e n t o f s e c t i o n s DGAB ATB ATB+DGAB (b) Base type 0% 20% 40% 60% 80% 100% 0 5 10 15 20 25 Transverse Cracking (m) P e r c e n t o f s e c t i o n s 8 12 16 (c) Base thickness 0% 20% 40% 60% 80% 100% 0 5 10 15 20 25 Transverse Cracking (m) P e r c e n t o f t e s t s e c t i o n s No Drainage Drainage (d) Drainage 0% 20% 40% 60% 80% 100% 0 5 10 15 20 25 Transverse Cracking (m) P e r e c e n t o f t e s t s e c t i o n s WF WNF (e) Climatic zone 0% 20% 40% 60% 80% 100% 0 5 10 15 20 25 Transverse Cracking (m) P e r e c e n t o f t e s t s e c t i o n s Fine Coarse (f) Subgrade type Figure 5-19 Effect of experimental factors on transverse cracking

199 0% 20% 40% 60% 80% 100% 0 50 100 150 200 250 Longitudinal Cracking-WP (m) P e r e c e n t o f t e s t s e c t i o n s 4 7 (a) HMA thickness 0% 20% 40% 60% 80% 100% 0 50 100 150 200 250 Longitudinal Cracking-WP (m) P e r c e n t o f t e s t s e c t i o n s DGAB ATB ATB+DGAB (b) Base type 0% 20% 40% 60% 80% 100% 0 50 100 150 200 250 Longitudinal Cracking-WP (m) P e r c e n t o f t e s t s e c t i o n s 8 12 16 (c) Base thickness 0% 20% 40% 60% 80% 100% 0 50 100 150 200 250 300 Longitudinal Cracking-WP (m) P e r e c e n t o f t e s t s e c t i o n s No Drainage Drainage (d) Drainage 0% 20% 40% 60% 80% 100% 0 50 100 150 200 250 Longitudinal Cracking-WP (m) P e r e c e n t o f t e s t s e c t i o n s WF WNF (e) Climatic zone 0% 20% 40% 60% 80% 100% 0 50 100 150 200 250 Longitudinal Cracking-WP (m) P e r e c e n t o f t e s t s e c t i o n s Fine Coarse (f) Subgrade type Figure 5-20 Effect of site factors on longitudinal cracking-WP

200 0% 20% 40% 60% 80% 100% 0 50 100 150 200 250 Longitudinal Cracking-NWP (m) P e r e c e n t o f t e s t s e c t i o n s 4 7 (a) HMA thickness 0% 20% 40% 60% 80% 100% 0 50 100 150 200 250 Longitudinal Cracking-NWP (m) P e r c e n t o f t e s t s e c t i o n s DGAB ATB ATB+DGAB (b) Base type 0% 20% 40% 60% 80% 100% 0 50 100 150 200 250 Longitudinal Cracking-NWP (m) P e r c e n t o f t e s t s e c t i o n s 8 12 16 (c) Base thickness 0% 20% 40% 60% 80% 100% 0 50 100 150 200 250 Longitudinal Cracking-NWP (m) P e r e c e n t o f t e s t s e c t i o n s No Drainage Drainage (d) Drainage 0% 20% 40% 60% 80% 100% 0 50 100 150 200 250 Longitudinal Cracking-NWP (m) P e r e c e n t o f t e s t s e c t i o n s WF WNF (e) Climatic zone 0% 20% 40% 60% 80% 100% 0 50 100 150 200 250 Longitudinal Cracking-NWP (m) P e r e c e n t o f t e s t s e c t i o n s Fine Coarse (f) Subgrade type Figure 5-21 Effect of site factors on longitudinal cracking-NWP

201 5.6.2 Frequency-based methods Two frequency-based methods were used- Linear Discriminant Analysis and Binary Logistic Regression (details in Chapter 4). The results from these analyses are as follows: Discriminant Analysis In this analysis, two mutually exclusive groups were defined as follows: • Alligator, transverse and longitudinal cracking: Cracked versus non-cracked. • Rutting: Rut depth < 7 mm versus rut depth > 7mm • Roughness: IRI < 1.4 m/km versus IRI>1.4 m/km This analysis was intended to identify the experimental factors which help in discriminating the cracked versus non-cracked pavement sections. As most of the pavements in the SPS-1 experiment have not shown a high level of distress, this analysis will help in finding the significant design and site factors contributing to occurrence of distress. In order to include the effect of traffic and pavement age, these were considered as covariate in this analysis. Table 5-10 summarizes the results of network level analysis. The performance measures were defined as dichotomous variables and all design and site factors were used as independent variables. The following summarizes the results from this analysis: • Fatigue cracking: The effects of drainage condition and base type were found to discriminate between cracked and non-cracked sections. Test sections without drainage built on unbound (DGAB) bases are more likely to crack. • Rutting: The effects of drainage condition, subgrade soil, base thickness were found to discriminate between sections having rut depths greater or less than 7mm. Test sections without drainage with thinner bases and built on fine-grained subgrade soils in wet zones are more likely to exhibit severe rutting. • Roughness: The effects of climatic zone, subgrade soil and base thickness were found to discriminate between sections having IRI greater or less than 1.4 m/km. Test sections with thinner bases built on fine-grained subgrade soil and in wet freeze zone are more likely to exhibit higher roughness. Sections with higher initial roughness are more likely to become rougher with age.

202 • Transverse cracking: The effect of base type to a lesser degree was found to discriminate between cracked and non-cracked sections. Test sections in wet freeze zone with unbound (DGAB) bases are more likely to crack. Also, older sections are more likely to crack. • Longitudinal cracking: The effect of climatic zone was found to discriminate between cracked and non-cracked sections (inside the wheel path). Test sections built in wet-freeze zone are more likely to crack outside the wheel path. Also, older sections are more likely to crack (in and outside the wheel path). Table 5-10 Summary of p-values from LDA for determining the effect of experimental factors on pavement performance measures Performance Measures Longitudinal cracking Design Factor Fatigue cracking Rutting Roughness Transverse cracking WP NWP HMA thickness 0.39 1.000 0.370 0.320 0.88 0.310 Base type 0.098 0.517 0.250 0.139 0.77 0.690 Base thickness 0.92 0.077 0.076 0.736 0.19 0.421 Drainage 0.09 0.056 0.370 1.000 0.47 0.310 Subgrade type 0.045 0.011 0.000 0.177 0.184 0.001 Climatic Zone 0.392 0.578 0.000 0.417 0.002 1.000

203 Logistic Regression The binary logistic regression model was used to model the probability of occurrence for the various performance measures. This method requires fewer assumptions than discriminant analysis and even when the assumptions required for discriminant analysis are not satisfied, it performs well. The overall models for each of the performance measures were found to be significant. The results using the maximum likelihood method are summarized in Tables 5-18 and 5-19. • Fatigue cracking: HMA Thickness— Thin (102 mm) pavement sections have a slightly higher probability of cracking than thick (178 mm) sections when all other variables are held constant. Base Type— Pavement sections with unbound (DGAB) base have a significantly higher probability of cracking than those with bound (ATB/DGAB) bases. Drainage— Pavement sections with no drainage have a higher probability than those sections with drainage. Climatic Zone— Pavement sections in freeze zones have a significantly higher probability of cracking than those in the no-freeze environments. • Rutting: Drainage— Pavement sections with no drainage have a slightly higher probability of rutting (rut depth > 7 mm) than those sections with drainage. Subgrade Type— Pavement sections built on fine-grained subgrade soils have a significantly higher probability of rutting than those sections built on coarse-grained subgrade soils. Climatic Zone— Pavement sections in WNF zones have a significantly higher probability of rutting than those in the WF environments.

204 Table 5-11 Summary of p-values from BLR for determining the effect of experimental factors on pavement performance measures (Wet zones) Performance Measures Longitudinal cracking Design Factor Fatigue cracking Rutting Roughness Transverse cracking WP NWP HMA thickness 0.160 (1.8) 0.833 (1.1) 0.068 (3.7) 0.493 (0.6) 0.360 (1.5) 0.31 (0.7) Base type 0.024 (2.4) 0.972 (1.0) 0.006 (33) 0.711 (1.2) 0.437 (1.7) 0.396 (1.7) Base thickness 0.420 (1.7) 0.212 (2.5) 0.038 (14) 0.632 (1.3) 0.410 (1.8) 0.733 (0.8) Drainage 0.045 (2.8) 0.124 (2.2) 0.278 (2.4) 0.316 (2.7) 0.40 (0.6) 0.837 (0.9) Subgrade type 0.960 (1.0) 0.015 (3.4) 0.000 (571) 0.345 (0.005) 0.000 (22) 0.009 (0.34) Climatic Zone 0.088 (2.2) 0.098 (.42) 0.000 (420) 0.316 (10) 0.976 (1.0) 0.73 (0.862) Note: The values in parenthesis are odds ratios Table 5-12 Summary of p-values from BLR for determining the effect of experimental factors on pavement performance measures (All zones) Performance Measures Longitudinal cracking Design Factor Fatigue cracking Rutting Roughness Transverse cracking WP NWP HMA thickness 0.19 (1.5) 0.98 (1.0) 0.068 (3.7) 0.224 (0.5) 0.527 (1.25) 0.34 (0.7) Base type 0.013 (2.2) 0.98 (1.0) 0.006 (33) 0.043 (3.2) 0.55 (1.4) 0.22 (2.0) Base thickness 0.81(0.8) 0.4 (2.3) 0.038 (14) 0.974 (1.0) 0.376 (1.6) 0.77 (1.2) Drainage 0.009 (3.1) 0.22 (1.7) 0.278 (2.4) 0.473 (1.6) 0.700 (1.2) 0.297 (1.5) Subgrade type 0.073 (0.5) 0.87 (1.1) 0.000 (571) 0.006 (0.001) 0.076 (2.2) 0.019 (0.42) Climatic Zone WF-DNF WNF-DNF DF-DNF WF-WNF 0.000 0.000 (12) 0.002 (9) 0.000 (23) (1.4) 0.747 0.72 (0.83) 0.611(1.4) (0.6) 0.001 - - - (420) 0.019 - - - (17) 0.254 - - - (0.76) 0.002 0.003 (5.1) 0.003 (6.4) 0.000 (27) (0.8) Note: Very high value of odds ratio is caused by the un-balanced data or too few sections in one of the categories.

205 • Roughness: HMA Thickness— Thin (102 mm) pavement sections have a higher probability of showing higher roughness (IRI > 1.4 m/km) than thick (7 inch) sections. Base Type— Pavement sections with unbound (DGAB) base have a significantly higher probability of showing higher roughness than those with bound (ATB/DGAB) bases. Base Thickness— Pavement sections with thin bases have a significantly higher probability of showing higher roughness than those with thick bases. Subgrade Type— Pavement sections built on fine-grained subgrade soils have a significantly higher probability of showing higher roughness than those sections built on coarse-grained subgrade soils. Climatic Zone— Pavement sections built in wet-freeze zone have a significantly higher probability of showing higher roughness than those sections built in wet-no-freeze zone. • Transverse cracking: Climatic Zone— Pavement sections built in wet-freeze zone have a higher probability of cracking than those sections built in wet-no-freeze zone. Also, older pavement sections have a significantly higher probability of cracking. • Longitudinal cracking: Subgrade Type— Pavement sections built on fine-grained subgrade soils have a higher probability of longitudinal cracking in the wheel path than those sections built on coarse- grained subgrade soils. Climatic Zone— Pavement sections built in freeze zone have a significantly higher probability of cracking (outside the wheel path) than those sections built in no-freeze zone. Also, older pavement sections have a significantly higher probability of cracking.

206 5.6.3 Analysis of Variance Several analyses of variance (ANOVA) were conducted for each of the performance measures and response indicators. The first ANOVA was targeted at determining the significance of only the main structural design factors considered in the experiment. This was achieved by blocking the site factor (to neutralize the effects of subgrade type, climatic conditions, traffic, age, and construction variability) as well as accounting for the variability in target layer thicknesses. The main structural design factors included in the ANOVA are listed below: • HMA thickness (102 mm versus 178 mm) • Base type (DGAB, ATB or DGAB+ATB) • Base thickness (8 inch, 12 inch or 16 inch) • Drainage condition (with versus without Permeable ATB) To meet the assumptions of ANOVA, the dependent variables (performance measures) had to be transformed using the natural logarithm (see Chapter 4). This was particularly relevant for all cracking distresses because of the large number of zeroes in those populations. A negative consequence from this is that the number of sections used in the analysis is reduced. 5.6.3.1 Effect of Design Factors on Pavement Performance The results from this analysis are summarized in Table 5-13 and indicate that the most significant design factor is the base type, which has a significant effect, statistically as well as operationally, on all performance measures. The ∆IRI (which is the change in IRI between initial and latest value) is also significantly affected by base thickness. The initial roughness is significantly affected by all the design factors except for drainage. Also, the effect of drainage condition on rutting, and the effect of base thickness on longitudinal cracking-NWP and change in roughness are statistically significant. For investigating the mean difference between the levels of design factors, the marginal means (predicted cell means from the model) were transformed back to the original scale of the distress. These conversions were necessary in order to find out the practical/operational mean difference. The marginal means were back transformed using the properties of lognormal distribution. A random variable X is considered to have a lognormal distribution if Y=ln (X) has

207 a normal probability distribution, where ln (X) is the natural logarithm to the base e. Equations (5-1) and (5-2) are used to calculate the mean and variance of a random variable X.    += 2 2 1exp yyx σµµ (5-1) ( )[ ]1exp 222 −= yxx σµσ (5-2) where µy and σy2 are the mean and the variance of lognormal distribution. The marginal means of performance measures (for which natural logarithmic transformation was necessary to meet the ANOVA assumptions) were estimated by using equation 5-1. The mean squared error (MSE) was considered as the “best” estimate of the variance for lognormal distribution in all analyses. Table 5-14 shows the back transformed marginal means for all levels of design factors in the SPS-1 Experiment. The following discussion summarizes the effect of key design factors on performance: • Effect of base type: The effect of base type was found to be significant for all performance measures except for rutting. Pavement sections with dense-graded aggregate bases (DGAB) have shown the worst performance for all distresses while those with asphalt treated bases (ATB) have shown the best performance. Sections built with DGAB have shown significantly (operationally and statistically) higher fatigue cracking compared to those built with ATB. On average higher rutting was observed on pavement sections built with DGAB than those constructed with ATB. In the case of other distresses (change in roughness, transverse cracking, and longitudinal cracking) the difference in the performance of sections built on DGAB and sections built on ATB were only found to be statistically significant (i.e., they are not of operational significance at this point in time). • Effect of HMA thickness: In general, thin [102 mm (4-inch)] pavement sections were built rougher than thick [178 mm (7-inch)] pavements. On an average, thin pavements [102 mm (4-inch)] have shown slightly higher fatigue cracking and rutting than thick [178 mm (7- inch)] pavements. However, this effect was found to be of marginal statistical significant. • Effect of base thickness: Sections with thicker bases [305 mm (12-inch) and 406 mm (16- inch)] were built smoother compared to those with thinner base [203 mm (8-inch)]. Also,

208 sections with thinner base [203 mm (8-inch)] have shown more change in roughness than those with thicker base [305 mm (12-inch) and 406 mm (16-inch)]. However, this change in roughness is not of practical significance. On an average, pavement sections with 203 mm (8-inch) base have shown more rut depths than those with 305 mm (12-inch) and 406 mm (16-inch) thick bases. However, this effect is not of practical significance. More longitudinal cracking-NWP occurred in sections built with 203 mm (8-inch) base compared to those with 406 mm (16-inch) base. • Effect of drainage condition: On average, pavement sections with drainage have shown slightly lower rutting than those without drainage; however this difference in performance is not statistically significant.

209 Table 5-13 Summary of p-values from ANOVA for determining the effect of main design factors on pavement performance measures—Overall Performance Measures IRI Longitudinal cracking Design Factor Fatigue cracking Rut1 depth ∆IRI IRIo Transverse cracking WP NWP HMA Thickness 0.163 0.074 0.870 0.006 0.758 0.737 0.787 Base Type 0.000* 0.510 0.004 0.000 0.016 0.079 0.031 Base Thickness 0.951 0.080 0.027 0.028 0.697 0.488 0.008* Drainage 0.347 0.250 0.293 0.160 0.544 0.645 0.874 Site (blocked) 0.000 0.000 0.000 0.000 0.000 0.000 0.000 R 2=0.712 N=133 R2=0.55 N=159 R2=0.624 N=163 R2=0.603 N=212 R2=0.630 N=75 R2=0.72 N=97 R2=0.770 N=140 Note: The model considered for this analysis only has main effects for all design factors. * Also shows operational/practical significance, 1 Structural rutting only Table 5-14 Summary of marginal means from ANOVA for determining the effect of main design factors on pavement performance measures—Overall Average Performance IRI Longitudinal cracking Design Factor Fatigue cracking (sq-m) Rut depth (mm) ∆IRI (m/km) IRIo (m/km) Transverse cracking (m) WP (m) NWP (m) 102 mm 10.5 (1.6) 5.3 (1.76) 0.45 (-0.86) 0.837 1.5 (-0.204) 8.2 (1.43) 21 (2.6) HMA Thickness 178 mm 7.6 (1.28) 4.9 (1.7) 0.45 (-0.85) 0.786 1.7 (-.12) 7.6 (1.346) 22.6 (2.68) DGAB 17.4 (2.1) 5.2 (1.82) 0.5 (-0.77) 0.87 2.8 (0.40) 9.8 (1.6) 26.8 (2.85) ATB 6.3 (1.04) 4.9 (1.69) 0.4 (-0.975) 0.79 1.2 (-0.48) 5.4 (1.0) 18 (2.45) Base Type ATB/DGAB 7.0 (1.12) 5.1 (1.67) 0.47 (-0.825) 0.776 1.3 (-0.38) 9.4 (1.56) 21.5 (2.63) 203 mm 9.2 (1.46) 5.6 (1.78) 0.51 (-0.74) 0.843 1.8 (-0.054) 9 (1.51) 34.4 (3.1) 305 mm 8.5 (1.39) 5.1 (1.7) 0.45 (-0.87) 0.79 1.4 (-0.313) 6.4 (1.18) 25.5 (2.8) Base Thickness 406 mm 9.3 (1.47) 5.0 (1.71) 0.42 (-0.94) 0.80 1.7 (-0.12) 8.7 (1.48) 12.6 (2.2) N 10 (1.55) 5.3 (1.79) 0.47 (-0.82) 0.838 1.8 (-0.06) 8.5 (1.46) 21.5 (2.63) Drainage Y 8 (1.32) 4.6 (1.67) 0.44 (-0.89) 0.785 1.4 (-0.264) 7.4 (1.32) 22 (2.65) MSE (1.51) (0.06) (0.13) - (1.3) (1.4) (0.87) Note: Values in parenthesis are the lognormal marginal mean values.

210 The ANOVA was conducted for the design factors within each climatic zone as per the project panel recommendations. However, this analysis suffers from the lack of data within zones, especially within “Dry” zones, where only 2 sites each are available for DF and DNF zones. The results of ANOVA for “Wet” zones are more reliable as 8 and 6 sites are available within WF and WNF zones, respectively. The following discussion summarizes the effect of key design factors on performance in WF climatic zone (see Table 5-15 and Table 5-16): • Effect of HMA thickness: In general, thin (102 mm) pavement sections were built rougher than thick (7-inch) pavements. On average, thin pavements (102 mm) have shown slightly more fatigue cracking and rutting than thick (178 mm) pavements. However, this effect was not found to be statistically significant. • Effect of base type: The effect of base type was found to be significant for all performance measures except for transverse and longitudinal cracking. In the case of distresses that are significantly affected by base type, pavement sections with dense-graded aggregate bases (DGAB) have shown the worst performance while those with asphalt treated bases (ATB) have shown the best performance. Sections built with DGAB have shown significantly (operationally and statistically) more fatigue cracking, rutting, and change in roughness compared to those built with ATB. • Effect of base thickness: Sections with 305 mm (12-inch) bases were built smoother compared to those with 203 mm (8-inch) base. Also, sections with 203 mm (8-inch) base have shown significantly (practically and statistically) higher change in roughness than those with 406 mm base. However, the change in roughness was not found to be practically significant between sections with 8-inch (406 mm) base and those with 305 mm (12-inch) base. On average, pavement sections with 203 mm (8-inch) base have shown more rut depth than those with 305 mm (12-inch) and 406 mm (16-inch) thick bases. However, this effect was not found to be statistically significant. More longitudinal cracking-NWP occurred in sections built with 203 mm base compared to those with 406 mm base. • Effect of drainage condition: Pavement sections with drainage have shown less rutting than those without drainage. This difference in performance was found to be both statistically and practically significant.

211 Table 5-15 Summary of p-values from ANOVA for determining the effect of design factors on flexible pavement performance—WF Zone Performance Measures IRI Longitudinal cracking Design Factor Fatigue cracking Rut Depth ∆IRI IRIo Transverse cracking WP NWP HMA thickness 0.745 0.688 0.277 0.133 0.560 0.893 0.762 Base type 0.004* 0.001* 0.076* 0.012 0.128 0.232 0.400 Base thickness 0.832 0.504 0.040* 0.084 0.278 0.873 0.069* Drainage 0.674 0.012* 0.874 0.003* 0.359 0.813 0.885 Site (blocked) 0.000 0.000 0.000 0.000 0.000 0.000 0.000 R 2=0.638 N=58 R2=0.631 N=92 R2=0.620 N=76 R2=0.628 N=92 R2=0.71 N=31 R2=0.695 N=36 R2=0.813 N=60 * Also shows operational/practical significance Table 5-16 Summary of marginal means from ANOVA for determining the effect of main design factors on pavement performance measures— WF Zone Average Performance IRI Longitudinal cracking Design Factor Fatigue cracking (sq-m) Rut depth (mm) ∆IRI (m/km) IRIo (m/km) Transverse cracking (m) WP (m) NWP (m) 102 mm 17.6 (2.3) 6.1 (1.76) 0.50 (-0.773) 0.891 8.7 (1.92) 24.5 (2.26) 29.3 (3.22) HMA Thickness 178 mm 15.9 (2.2) 5.9 (1.73) 0.55 (-0.673 0.844 7.4 (1.76) 23.1 (2.2) 31.75 (3.3) DGAB 34.4 (3.0) 7.5 (1.96) 0.55 (-0.665) 0.932 11.0 (2.15) 23.1 (2.2) 26.5 (3.12) ATB 10.3 (1.8) 5.7 (1.68) 0.45 (-0.865) 0.833 5.7 (1.5) 11.5 (1.5) 33.4 (3.35) Base Type ATB/DGAB 14.0 (2.1) 5.2 (1.6) 0.56 (-0.64) 0.84 8.3 (1.87) 47.5 (2.92) 30.2 (3.25) 203 mm 17.1 (2.3) 6.4 (1.8) 0.61 (-0.561) 0.91 11.5 (2.2) 28.5 (2.41) 38.8 (3.5) 305 mm 18.8 (2.4) 5.8 (1.7) 0.55 (-0.673) 0.833 8.3 (1.87) 23.6 (2.22) 33.0 (3.34) Base Thickness 406 mm 15.4 (2.2) 6.0 (1.74) 0.42 (-0.94) 0.86 5.7 (1.49) 19.7 (2.04) 21.3 (2.9) N 18.1 (2.4) 6.8 (1.86) 0.53 (-0.714) 0.924 9.4 (2.0) 22 (2.15) 30.5 (3.26) Drainage Y 15.4 (2.2) 5.4 (1.63) 0.52 (-0.732) 0.811 6.9 (1.69) 25.5 (2.3) 29.6 (3.23) MSE (1.073) (0.113) (0.139) (0.493) (1.881) (0.316) Note: Values in parenthesis are the lognormal marginal mean values.

212 The following discussion summarizes the effect of key design factors on performance in WNF climatic zone (see Table 5-17 and Table 5-18): • Effect of base type: The effect of base type was found to be significant only for the change in roughness. Pavement sections with dense-graded aggregate bases (DGAB) have shown higher change in roughness compared to those with asphalt treated bases (ATB). However, difference was not practically significant. • Effect of HMA thickness: On average, thin pavements (102 mm) have shown slightly more fatigue cracking compared to thick (178 mm) pavements. This effect was found to be both statistically and practically significant. • Effect of base thickness: The effect of base thickness on various performance measures was found to be statistically insignificant. However, on average, higher change in roughness was observed in sections with 203 mm base compared to those with 305 mm or 406 mm base. • Effect of drainage condition: Pavement sections with drainage have shown less rutting than those without drainage. This difference in performance was found to be both statistically and practically significant. The fractional factorial design for the SPS-1 experiment calls for a tradeoff between selecting the number of “runs” and testing all possible interactions. Therefore, all possible two- way interactions were considered in the analysis. An ANOVA was run with two-way interaction effects between the main structural design factors. No significant interaction effect was detected.

213 Table 5-17 Summary of p-values from ANOVA for determining the effect of design factors on flexible pavement performance—WNF Zone Performance Measures IRI Longitudinal cracking Design Factor Fatigue cracking Rut Depth ∆IRI IRIo Transverse cracking WP NWP HMA thickness 0.077* 0.576 0.948 0.141 0.383 0.759 0.532 Base type 0.545 0.547 0.065* 0.117 0.470 0.803 0.110 Base thickness 0.703 0.476 0.144 0.559 0.806 0.937 0.265 Drainage 0.298 0.031* 0.725 0.032 0.306 0.760 0.142 Site (blocked) 0.000 0.000 0.008 0.000 0.276 0.037 0.000 R 2=0.834 N=36 R2=0.561 N=72 R2=0.503 N=49 R2=0.680 N=72 R2=0.965 N=14 R2=0.662 N=24 R2=0.579 N=45 * Also shows operational/practical significance Table 5-18 Summary of marginal means from ANOVA for determining the effect of main design factors on pavement performance measures— WNF Zone Average Performance IRI Longitudinal cracking Design Factor Fatigue cracking (sq-m) Rut depth (mm) ∆IRI (m/km) IRIo (m/km) Transverse cracking (m) WP (m) NWP (m) 102 mm 6.4 (1.2) 6.2 (1.78) 0.34 (-1.157) 0.82 0.3 (-1.412) 3.5 (0.562) 13.7 (1.96) HMA Thickness 178 mm 2.9 (0.43) 5.9 (1.74) 0.33 (-1.165) 0.785 1.0 (-0.363) 4.1 (0.73) 17.5 (2.2) DGAB 5.5 (1.05) 5.9 (1.744) 0.40 (-0.989) 0.84 0.6 (-0.872) 3.7 (0.614) 19.7 (2.32) ATB 3.2 (0.524) 6.4 (1.815) 0.29 (-1.3) 0.81 1.6 (0.143) 2.8 (0.356) 8.5 (1.48) Base Type ATB/DGAB 4.4 (0.84) 5.8 (1.712) 0.32 (-1.2) 0.764 0.2 (-1.93) 5.2 (0.97) 21.8 (2.42) 203 mm 5.8 (1.104) 6.4 (1.82) 0.39 (-1) 0.82 0.4 (-1.171) 4.5 (0.812) 26.6 (2.62) 305 mm 4.7 (0.89) 5.8 (1.72) 0.31 (-1.23) 0.793 0.5 (-0.982) 3.6 (0.603) 14.7 (2.03) Base Thickness 406 mm 2.9 (0.42) 5.9 (1.73) 0.30 (-1.26) 0.79 0.8 (-0.51) 3.3 (0.523) 9.6 (1.6) N 3.3 (0.54) 6.6 (1.85) 0.57 (-1.136) 0.833 0.9 (-0.385) 3.4 (0.54) 11.1 (1.75) Drainage Y 5.8 (1.1) 5.2 (1.6) 0.53 (-1.2) 0.772 0.3 (-1.391) 4.2 (0.752) 21.3 (2.4) MSE (1.3) (0.09) (0.146) (0.647) (1.37) (1.321) Note: Values in parenthesis are the lognormal marginal mean values.

214 5.6.3.2 Effect of Site Factors on Pavement Performance The third ANOVA was targeted at determining the significance of subgrade type and climatic zone. Traffic, age and variability in target layer thicknesses were considered as covariates. The subgrade type and climatic zone were included as main factors in addition to the structural design factors. For fatigue cracking, the analysis was run with and without the Kansas (20) data, since the test sections in Kansas (20) have a large amount of fatigue cracking and the project is known to have had construction problems with a wet subbase and variable densities. The analysis for rutting was also done with and without the Texas (48) data since rutting for these sections is believed to be due to the asphalt mix [6]. The results from this analysis are summarized in Tables 5-19 and 5-20, and generally show lower R2 values than those in Table 5-13. This may be partially due to the effects of variations in environmental conditions within a given climatic zone and variations in material properties within different pavement layers among sites. In addition, construction and material variability is not accounted for in this analysis, since the site factor is not blocked. The results seem to indicate that the effect of the climatic zone is significant for all performance measures and that the effect of subgrade type can be significant for most of them. However, caution must be exercised in interpreting these results given the unbalanced nature of the design with respect to climatic zone: there are only two projects in each of the dry zones, Dry-Freeze (DF) and Dry- No-Freeze (DNF), as opposed to eight projects in the Wet-Freeze (WNF) and six projects in the Wet-No-Freeze (WF) zones. Finally, ANOVA was conducted by considering the main and interaction (two-way) effects for all six experimental factors. The results of this analysis are summarized in Tables 5- 21 and 5-22. The conclusions are based on the main effects when interaction between site factors is not significant.

215 Table 5-19 Summary of p-values from ANOVA for determining the effect of site factors on pavement performance measures (Main effects only) Performance Measures IRI Longitudinal cracking Site Factor Fatigue cracking Rut Depth ∆IRI IRIo Transverse cracking WP NWP Subgrade type 0.680 0.432 0.000* 0.067 0.020 0.015 0.013 Climatic Zone 0.000* 0.000* 0.000* 0.001 0.000* 0.000 0.000 Traffic Level 0.000 0.024 0.083 - 0.000 0.437 0.000 Age 0.068 0.013 0.091 - 0.000 0.050 0.009 R 2=0.288 N=124 R2=0.305 N=159 R2=0.401 N=163 R2=0.215 N=212 R2=0.674 N=67 R2=0.434 N=95 R2=0.534 N=134 Note: The model considered for this analysis has main effects for all six experiment factors. KS (20) was not considered for analysis of all cracking measures, whereas, rut depth analysis was conducted without TX (48). * Also shows operational/practical significance. Table 5-20 Summary of marginal means from ANOVA for determining the effect of site factors on pavement performance measures (Main effects only) Performance Measures IRI Longitudinal cracking Site Factor Fatigue cracking Rut Depth ∆IRI IRIo Transverse cracking WP NWP F 15.4 (1.16) 5.2 (1.6) 0.59 (-0.62) 0.81 1.2 (-0.38) 44 (2.5) 20.8 (2.2) Subgrade type C 18.5 (1.34) 5.4 (1.65) 0.44 (-0.92) 0.76 4.9 (1) 13.2 (1.3) 42.0 (2.9) WF 53.4 (2.4) 4.9 (1.56) 0.59 (-0.628) 0.86 13.2 (2) 26.5 (2.0) 139 (4.1) WNF 24.0 (1.6) 5.6 (1.68) 0.336 (-1.186) 0.797 1.5 (-0.15) 16.1 (1.5) 25.4 (2.4) DF 26.5 (1.7) 4.0 (1.34) 0.58 (-0.641) 0.784 3.2 (0.6) 5.92 (0.5) 56.5 (3.2) Climatic Zone DNF 4.5 (-0.08) 7.1 (1.92) 0.596 (-0.613) 0.695 0.4 (-1.5) 177 (3.9) 3.12 (0.3) MSE (3.156) (0.091) (0.19) (1.161) (2.556) (1.668)

216 Table 5-21 Summary of p-values from ANOVA for determining the effect of site factors on pavement performance measures (With interaction effects) Performance Measures IRI Longitudinal cracking Site Factor Fatigue cracking Rut Depth ∆IRI IRIo Transverse cracking WP NWP Subgrade type 0.626 0.886 0.018 0.653 0.247 0.480 0.191 Climatic Zone 0.049 0.007* 0.000* 0.003 0.004* 0.002 0.000* Subgrade*Zone 0.000* 0.257 0.562 0.000* 0.092 0.000* 0.496 Traffic Level 0.031 0.028 0.150 - 0.197 0.655 0.000 Age 0.068 0.025 0.565 - 0.077 0.014 0.202 R 2=0.575 N=124 R2=0.47 N=159 R2=0.525 N=163 R2=0.435 N=212 R2=0.931 N=67 R2=0.755 N=95 R2=0.648 N=134 Note: The model considered for this analysis has main effects for all six experiment factors and all possible two-way interactions between them. * Also shows operational/practical significance. Table 5-22 Summary of marginal means from ANOVA for determining the effect of site factors on pavement performance measures (Interaction effects only) Performance Measures IRI Longitudinal cracking Site Factor Fatigue cracking Rut Depth ∆IRI IRIo Transverse cracking WP NWP Subgrade type - - - - - - - Climatic Zone - - - - - - - F 22.4 (1.7) - - 0.95 - 60.8 (3.1) - WF C 67 (2.8) - - 0.76 - 3.0 (0.1) - F 135 (3.5) - - 0.78 - 27.3 (2.3) - Subgrade* Zone WNF C 2.5 (-0.5) - - 0.83 - 5.5 (0.7) - MSE (2.816) (2.015) Note: The cell means are only given when interaction is significant. For main effects see Table 5-20 for marginal means.

217 In case of significant interaction between site factors, the interpretation of results are based on the comparison of cell means, i.e., the mean performance of sections corresponding to each subgrade type should be compared within each climatic zone. The following discussion summarizes the effect of climatic zone and subgrade type on the key performance measures: • Fatigue cracking: More fatigue cracking was observed on sections located in “wet” climates. The interaction effect between subgrade soil type and climatic zone is statistically significant (see Table 5-21); therefore the conclusions are based on the interaction effect. More cracking was observed in pavement sections built on fine-grained subgrade especially in WNF zone. Among the sections located in WNF zone, the difference between the mean cracking of sections built on fine-grained and sections built on coarse-grained soil is also practically significant. • Structural Rutting: Rutting was higher among sections located in “wet” climate and was generally higher for pavement sections on fine-grained subgrade. Both of these effects statistically and practically significant. There were high rut depths observed in the Dry-No- Freeze (DNF) zone; however, it is believed that this is more related to the asphalt mix as opposed to structural rutting. • Roughness: Both subgrade type and climatic zone are very significant factors affecting roughness growth (see Table 5-19). The pavements constructed on fine-grained soils have shown higher changes in roughness than those constructed on coarse-grained soils. Also, pavements located in the WF zone have shown higher change in roughness than those located in WNF zone. These effects were found to be statistically and practically significant (see Table 5-20). The effect of subgrade soil appears to be mainly caused by the initial roughness being significantly higher in sites with fine-grained subgrade. The initial IRI (IRIo) was found to be associated with future roughness, especially among sections built on fine-grained soils and among sections located in “wet” climate. • Transverse cracking: The effects of subgrade type and climate on transverse cracking are significant. More cracking was observed in sections built on coarse-grained soil compared to those built on fine-grained soil. However, the magnitude of cracking at this point in time is too low to conclude on the effect. More cracking occurred in sections located in WF zone

218 compared to those located in other zones, and this effect was found to be statistically and practically significant. • Longitudinal cracking: As the interaction effect between subgrade type and climatic zone is significant for longitudinal cracking-WP, the conclusions are based on comparing cell means for sections built on each subgrade type within each climatic zone. It was found that among pavements located in WF zone, those constructed on fine-grained soils have shown significantly more cracking than those constructed on coarse-grained soils. The effects of subgrade type and climate were significant in the case of longitudinal cracking-NWP. Significantly more cracking was observed in the sections built on coarse-grained soil compared to those built on fine-grained soil. Also pavements located in “freeze” climate have shown significantly more cracking compared to those in “no-freeze” climate. Given the unbalanced design of the experiment with respect to climatic zone, a one-way ANOVA was performed to investigate the effects of subgrade type (fine-grained versus coarse- grained soils) and climatic zones (wet versus dry, freeze versus no-freeze), one at a time. The p- values and mean performances by site factors, from this analysis, are summarized in Tables 5-23 and 5-24, respectively. To indicate the direction of effects for site factors, the “+” and “-“ signs are also reported along with the p-values. The “+” indicates that, within a factor, the first level is exhibiting more distress than the second level, while “-“ indicates otherwise. For example, in the case of the effect of subgrade on fatigue cracking, the “+” indicates more cracking in pavements constructed on fine-grained soils (first level for subgrade) compared to those constructed on coarse-grained soils (second level for subgrade). The p-values indicate that subgrade type appears to be significantly affecting fatigue cracking, rut depth, roughness, transverse and longitudinal cracking-WP. The pavements built on fine- grained subgrade have shown higher distress than those constructed on coarse-grained subgrade. The effect was found to be practically significant in the case fatigue cracking, rutting, change in roughness and transverse cracking. The effects of site factors by performance measure are listed below:

219 • Fatigue Cracking: Climate appears to be significantly affecting fatigue performance. Pavements located in “wet” or “freeze” climate have exhibited significantly higher amount of fatigue cracking than those located in “dry” or “no-freeze” climate, respectively. This effect was found to be practically significant. • Rut Depth: On average, rutting appears to be higher in wet climate. Also pavements located in DNF zone were found to have significantly more rutting compared to those located in DF zone. However, it is believed that this is more related to the asphalt mix as opposed to structural rutting, as mentioned before at the beginning of this chapter. • Roughness: Significantly higher growth in roughness was observed for pavements located in WF zone compared to those located in WNF zone. This effect was found to be practically significant. • Transverse Cracking: It was found that pavements located in WF zone have exhibited significantly higher transverse cracking than those located in WNF zone. This effect was found to be practically significant. • Longitudinal Cracking-WP: Significantly more longitudinal cracking-WP was observed in pavements located in WF zone compared to those located in WNF zone. Also, significantly more longitudinal cracking-WP was exhibited by the pavements located in DNF zone compared to those located in DF. In DNF zone, longitudinal cracking-WP and rutting is mainly contributed by sections in the Arizona, AZ (4), site, where HMA-related issues are believed to be causing the distresses. • Longitudinal Cracking-NWP: Significantly more longitudinal cracking-NWP was exhibited by the pavements located in WF zone compared to those located in other zones. Also, more cracking was observed in pavements located in DF zone compared to those built in DNF zone. These effects indicate that this distress could be related to “freeze” environment. It should be noted that the data from the four projects in the dry climatic zones show negative trends in several performance measures. This may be in part due to the lower number of projects in these zones.

220 Table 5-23 Summary of p-values from one-way ANOVA for determining the effect of site factors on pavement performance measures Performance Measures IRI Longitudinal cracking Site Factor Fatigue cracking Rut depth ∆IRI IRIo Transverse cracking WP NWP Subgrade Type Fine vs. Coarse 0.03 (+)* 0.002 (+)* 0.00 (+)* 0.011 (+)* 0.016 (+)* 0.001 (+) 0.26 (-) Climatic Zone Wet vs. Dry Freeze vs. No Freeze WF vs. WNF DF vs. DNF 0.021 (+)* 0.011 (+)* 0.063 (+)* 0.054 (+)* 0.087 (+) 0.001 (-) 0.893 (-) 0.000 (-) 0.596 (-) 0.000 (+)* 0.000 (+)* 0.281 (-) 0.000 (+)* 0.010 (+)* 0.030 (+)* 0.231 (+) 0.005 (+)* 0.06 (+)* 0.00 (+)* 0.055 (-) 0.919 (+) 0.038 (-) 0.096 (+) 0.000 (-) 0.040 (+) 0.000 (+) 0.000 (+) 0.001 (+) * Also shows operational/practical significance Table 5-24 Summary of marginal means from one-way ANOVA for determining the effect of site factors on pavement performance measures Performance Measures IRI Longitudinal cracking Site Factor Fatigue cracking Rut depth ∆IRI IRIo Transverse cracking WP NWP Fine 54.2 5.8 0.613 0.86 18.97 64.5 163.0 Subgrade Type Coarse 24.2 4.8 0.451 0.79 6.70 17.4 113.0 Wet 41.7 5.4 0.524 0.85 15.10 39.7 150.5 Dry 17.4 4.7 0.552 0.73 04.83 38.1 74.0 Freeze 45.6 4.8 0.616 0.85 13.57 24.5 189.7 No Freeze 18.4 5.8 0.425 0.78 06.20 57.7 27.1 WF 53.7 5.2 0.635 0.88 24.3 31.3 244.7 WNF 24.5 5.6 0.360 0.81 2.80 14.8 29.4 DF 26.2 4.0 0.480 0.76 2.25 4.10 83.1 Climatic Zone DNF 07.8 6.8 0.572 0.70 6.00 108.8 14.7

221 5.6.3.3 Effect of Design Factors on Pavement Performance (univariate) based on standard deviate As explained before, the experiment design and the performance of the test sections have rendered the SPS-1 experiment “unbalanced”. Fourteen out of eighteen sites, in the experiment are located in “wet” climate, of which eight are in the WF zone. In addition, all 24 unique designs were not built in every soil-climate combination. Furthermore, non-occurrence of distresses in a considerable number of sections contributed to the unbalance. This could be a reason for insignificance of interaction effects between the design and site factors from multivariate analyses presented above. In light of the above concerns, a simplified analysis considering one design factor at a time (univariate) was performed using one-way ANOVA (as in the case of analysis of the effects for site factors). The performance of test sections was not found to be consistent across sites indicating the influence of site conditions (see Chapter 3). The site conditions that could have contributed to this variation in performance are traffic, age, construction quality, measurement variability, and material properties, apart from the experimental site factors (i.e. subgrade and environment). In order to separate the “true” effects of the experimental factors, this “noise” had to be nullified. The standard deviate for each performance measure was calculated, within each site, for all the sections using equation 5-3. This measure indicates the relative performance of a design compared to the other designs. As this measure was calculated for each section, considering one site at a time, it indicates the relative standing of the section compared to other sections. It thus helps nullify the variation in performance (due to site conditions) among sites, as the sections are weighed with respect to companion sections in each site.    −= σ µxDeviateStd . (5-3) The above approach of using the standard deviate is similar to blocking of the site factors performed in the multivariate analysis. One-way analysis of variance (univariate) was performed on the standard deviates of the sections to study the effects of each design factor by taking one design factor at a time. The analyses were performed on data from all sections and also on subsets of data stratified by different subgrade types, climates and combinations of these. This helps identify the effects of design factors under different site conditions.

222 In the SPS-1 experiment, HMA thickness and drainage have two levels (i.e. 102 mm vs. 178 mm and drainage vs. no drainage). But for base thickness and base type, three levels (203 mm vs. 305 mm vs. 406 mm and DGAB vs. ATB vs. ATB/DGAB) are present. Moreover, 406 mm base thickness was provided only for drained sections and ATB/DGAB was built only for un-drained sections making the design unbalanced. Therefore, for studying the effects of base thickness and base type, analyses were done separately among drained sections and among un-drained sections. To see the “pure” effect of each design factor, comparisons of standard deviates were also made between the levels of each design factor while controlling the other factors, as in the case of level-B analyses (site-level). The results from this analysis are in Appendix A5. The effects of design factors, based on the above-mentioned analyses, on each performance measure are discussed next. Fatigue Cracking The effects of the design and site factors, in terms of standard deviate, are shown in Figure 5-22. In addition, the summary of p-values corresponding to the analyses performed to study the effects of design factors on fatigue cracking is shown in Table 5-25. The mean area (m2) of fatigue cracking (PI) corresponding to each comparison presented in Table 5-25 is shown in Table 5-26. Though the univariate analyses were performed on standard deviates, the mean cracking was used to identify operationally significant effects. The effects of design factors on fatigue cracking, based on this analysis, are presented below: HMA thickness: The effect of HMA surface thickness is statistically and operationally significant, especially among sections located in WNF zone. Sections built with “thin” [102 mm (4-inch)] HMA surface have exhibited higher fatigue cracking than those built with “thick” [178 mm (7-inch)] HMA surface. On average, among sections built on fine-grained soils, higher fatigue cracking was observed on “thin” [102 mm (4-inch)] sections compared to “thick” [178 mm (7-inch)] sections. However, this effect is not significant. Similar trend was found among sections built on coarse-grained soils and the effect is statistically and operationally significant. Base thickness: The effect of base thickness is marginal among sections built on fine-grained subgrade soil, in that sections with thick 406 mm (16-inch) permeable base have exhibited lesser

223 cracking than those with 305 mm (12-inch) or 203 mm (8-inch) base thickness. Also among sections located in WF zone the effect of base thickness on alligator cracking is statistically and operationally significant. Sections built with 406 mm (16-inch) permeable base have exhibited lesser cracking than those with 305 mm (12-inch) or 203 mm (8-inch) base. Base Type: The effect of base type (unbound versus treated base) is statistically and operationally significant, with ATB giving the “best” performance and DGAB showing the “worst” performance. This effect is more prominent among sections built on fine-grained soils compared to sections built on coarse-grained soils. Also, this effect is more noticeable among sections located in WF zone. Drainage: The effect of drainage is significant (statistically and operationally) among those in WF zone and built on fine-grained subgrade soils. Sections with drainage have lesser cracking than those without drainage. This effect is more prominent for the sections constructed with DGAB than those with ATB. This shows that drainage is more effective if provided with DGAB than when with ATB. The interaction effects among the experimental factors, on fatigue cracking, are reported below: In general, “thin” sections with DGAB on fine-subgrade soils have exhibited the most alligator cracking while “thick” sections with ATB on coarse-grained subgrade soils have exhibited the least alligator cracking. Among un-drained pavements, on average, an increase in HMA surface thickness from 102 mm (4-inch) to 178 mm (7-inch) has a slightly higher effect on fatigue cracking for pavements with DGAB than for pavements with ATB. Among sections located in the WF zone, those with DGAB have shown the highest amount of cracking while those with ATB have the least. In addition, among pavements located in WF zone, those with 406 mm (16-inch) drained base have less fatigue cracking than others. These effects were found to be statistically and practically significant. Among pavements with DGAB and built on fine-grained soils, those with drained base have lesser fatigue cracking than others. Also, among sections with drainage and built on fine-grained soils, those with 406 mm base have lesser cracking. These effects were found to be statistically and practically significant.

224 -1.3 -0.8 -0.3 0.2 0.7 1.2 10 2 m m 17 8 m m 20 3 m m (N D ) 30 5 m m (N D ) 20 3 m m (D ) 30 5 m m (D ) 40 6 m m (D ) D G A B A TB A TB /D G A B PA TB /D G A B A TB /P A TB N D D Design Factors M ea n st d. d ev ia te (a) Overall -1.3 -0.8 -0.3 0.2 0.7 1.2 10 2 m m 17 8 m m 20 3 m m (N D ) 30 5 m m (N D ) 20 3 m m (D ) 30 5 m m (D ) 40 6 m m (D ) D G A B A TB A TB /D G A B PA TB /D G A B A TB /P A TB N D D Design Factors M ea n st d. d ev ia te Fine Coarse (b) By subgrade type -1.3 -0.8 -0.3 0.2 0.7 1.2 10 2 m m 17 8 m m 20 3 m m (N D ) 30 5 m m (N D ) 20 3 m m (D ) 30 5 m m (D ) 40 6 m m (D ) D G A B A TB A TB /D G A B PA TB /D G A B A TB /P A TB N D D Design Factors M ea n st d. d ev ia te WF WNF (c) By zone type Figure 5-22 Effect of design factors on fatigue cracking (1 inch = 25.4 mm)

225 Table 5-25 Summary of p-values for comparisons of standard deviates— Fatigue cracking By subgrade By climatic zone By subgrade and zone WF WNF DF DNF Design Factor Comparison Overall Fine Coarse WF WNF DF DNF F C F C F C F C HMA thickness 102 mm vs. 178 mm 0.003 0.167 0.008 0.890 0.002 0.102 0.213 0.900 0.900 0.160 0.005 Overall 203 mm vs. 305 mm vs. 406 mm 0.211 0.086 0.890 0.043 0.410 0.63 0.265 0.070 0.040 0.850 0.420 ND 203 mm vs. 305 mm 0.737 0.802 0.480 0.271 0.168 0.181 0.028 0.817 0.062 0.336 0.374 Base thickness D 203 mm vs. 305 mm vs. 406 mm 0.381 0.020 0.512 0.137 0.609 0.977 0.159 0.060 0.523 0.207 0.587 Overall DGAB vs. ATB vs. ATB/DGAB 0.000 0.004 0.011 0.003 0.057 0.390 0.500 0.060 0.050 0.150 0.350 ND DGAB vs. ATB vs. ATB/DGAB 0.002 0.003 0.207 0.003 0.140 0.881 0.449 0.024 0.095 0.254 0.557 D DGAB vs. ATB 0.001 0.058 0.008 0.027 0.157 0.101 0.313 0.186 0.060 0.270 0.680 Base type All Bases DGAB vs. ATB vs. ATB/DGAB vs. DGAB/PATB vs. ATB/PATB 0.000 0.000 0.036 0.000 0.165 0.522 0.290 0.003 0.060 0.270 0.680 Overall Drainage vs. No-Drainage 0.111 0.058 0.610 0.038 0.884 0.740 0.180 0.050 0.330 0.870 0.750 DGAB Drainage vs. No-Drainage 0.070 0.010 0.770 0.030 0.720 0.370 0.220 0.040 0.300 0.430 0.840 Drainage ATB Drainage vs. No-Drainage 0.160 0.580 0.200 0.040 0.190 0.500 0.250 0.170 0.150 0.240 0.500 N 188 80 108 80 60 24 24 44 36 24 36 Note: Shaded cells show statistically significant at 90% or higher level of confidence.

226 Table 5-26 Summary of means of PI for fatigue cracking By subgrade and zone By subgrade By climatic zone WF WNF DF DNF Design Factors Comparison Overall Fine Coarse WF WNF DF DNF F C F C F C F C 102 mm 14.6 19.2 11.8 10.7 19.8 24.3 3.4 5.5 15.2 47.0 1.6 24.3 0.2 6.5 HMA thickness 178 mm 9.4 10.3 8.7 11.2 8.4 13.0 2.3 7.4 15.5 20.3 0.6 13.0 0.0 4.5 203 mm 13.4 17.2 10.9 11.8 16.5 21.9 1.4 6.9 16.4 40.1 0.8 21.9 0.0 2.9 305 mm 11.9 14.8 9.9 12.4 13.5 14.3 3.6 8.4 16.5 31.8 1.2 14.3 0.2 7.0 Overall 406 mm 8.8 7.8 9.5 5.6 9.7 21.5 4.2 2.0 9.9 22.1 1.4 21.5 0.0 8.5 203 mm 13.4 22.3 7.5 8.7 22.1 18.0 1.1 8.3 9.1 54.4 0.6 18.0 0.0 2.2 ND 305 mm 9.8 10.5 9.4 14.0 8.1 5.9 5.9 8.3 19.6 18.7 1.0 5.9 0.4 11.5 203 mm 13.4 8.8 16.2 17.0 8.2 27.9 1.9 4.5 27.4 18.7 1.2 27.9 0.1 3.8 305 mm 14.9 21.4 10.5 10.1 21.6 26.9 0.2 8.5 11.8 51.5 1.6 26.9 0.0 0.4 Base thickness D 406 mm 8.8 7.8 9.5 5.6 9.7 21.5 4.2 2.0 9.9 22.1 1.4 21.5 0.0 8.5 DGAB 18.8 23.4 16.1 18.5 22.5 26.0 3.2 7.7 27.2 53.8 1.7 26.0 0.2 6.3 ATB 6.8 7.7 6.2 5.8 6.8 14.7 2.7 5.3 6.4 15.8 0.8 14.7 0.0 5.4 Overall ATB/DGAB 8.8 13.4 5.7 8.2 11.5 10.3 2.0 8.2 8.3 28.0 0.5 10.3 0.0 4.1 DGAB 20.6 34.3 13.0 20.0 30.6 13.9 3.8 11.3 25.9 74.3 1.4 13.9 0.5 7.2 ATB 6.4 6.0 6.7 7.7 3.2 11.5 4.7 6.9 8.9 7.3 0.4 11.5 0.1 9.3 ND ATB/DGAB 8.8 13.4 5.7 8.2 11.5 10.3 2.0 8.2 8.3 28.0 0.5 10.3 0.0 4.1 DGAB 17.7 16.9 18.1 17.6 17.1 34.0 2.8 5.9 28.0 40.1 1.8 34.0 0.0 5.7 Base type D ATB 7.2 8.9 5.9 4.3 9.1 16.8 1.4 4.0 4.7 21.4 1.0 16.8 0.0 2.8 ND 11.6 16.4 8.5 11.3 15.1 11.9 3.5 8.3 14.4 36.5 0.8 11.9 0.2 6.8 Overall D 12.3 12.7 12.0 10.6 13.1 25.4 2.1 4.8 16.3 30.7 1.4 25.4 0.0 4.2 ND 20.6 34.3 13.0 20.0 30.6 13.9 3.8 11.3 25.9 74.3 1.4 13.9 0.5 7.2 DGAB D 17.7 16.9 18.1 17.6 17.1 34.0 2.8 5.9 28.0 40.1 1.8 34.0 0.0 5.7 ND 6.4 6.0 6.7 7.7 3.2 11.5 4.7 6.9 8.9 7.3 0.4 11.5 0.1 9.3 Drainage ATB D 7.2 8.9 5.9 4.3 9.1 16.8 1.4 4.0 4.7 21.4 1.0 16.8 0.0 2.8

227 Structural Rutting The effects of the design and site factors, in terms of standard deviate, are shown in Figure 5-23. The summary of p-values corresponding to the analyses performed to study the effects of design factors on structural rutting is presented in Table 5-27. The mean rut depth (PI), in mm, corresponding to each comparison presented in Table 5-27 are shown in Table 5-28. The effects of design factors on rutting, based on this analysis, are presented below: HMA thickness: Among sections built on coarse-grained soils, the sections built with “thin” [102 mm (4-inch)] HMA surface have exhibited higher rut depths than those built with “thick” [178 mm (7-inch)] HMA surface. This effect is statistically significant but not operationally significant, at this point in time. Thus increasing HMA thickness from 102 mm to 178 mm may be more effective in retarding rutting in the case of sections with coarse-grained soils than in the case of sections with fine-grained soils. On average, sections built on fine-grained soils have slightly higher rutting than those with coarse-grained soils. Base thickness: The effect of base thickness is significant (statistical and operational) among sections located in WNF zone where higher rutting was observed for the sections built with 203 mm (8-inch) thick base than for those built on 406 mm (16-inch) thick base. In addition, this effect seems to be more apparent for the sections built on coarse-grained subgrade soils. Base Type: In general, the effect of base type (unbound versus treated base) is not statistically significant. However on average, sections built on ATB have shown the better performance than those sections built on DGAB. This effect (DGAB vs. ATB) is more prominent among sections located in WF zone and built on fine-grained soils. Drainage: In general, the effect of drainage is statistically significant with un-drained sections showing higher rutting than those with drainage. However, this effect is not operationally significant. This effect is significant (statistical and operational) among sections located in WNF zone and built on fine-grained soils. Also the effect is significant (statistical and operational) among sections in WF zone and built on coarse-grained soils. The results suggest that drainage

228 may be more effective in inhibiting rutting for pavements on fine-grained soils, when located in WNF zone. The interaction effects among the experimental factors, on structural rutting, are reported below: A marginal effect of drainage was observed on pavements built with ATB and on fine- grained soils. Also, among drained pavements located in WF zone, those with DGAB have shown higher rutting than those with ATB. Furthermore, among sections located in WF zone and built with ATB, those with drainage have shown significantly less amount of rutting than those without drainage. Both of the above effects were found to be statistically significant and are of operational significance. Among un-drained sections located in WNF zone, those with 305 mm (12-inch) base thickness have less amount of rutting than those with 203 mm (8-inch) base thickness. This effect was found to be statistically significance and is practically meaningful. For sections built on DGAB and located in WNF zone, those with drainage have shown slightly lesser rutting than those without drainage. The effect was not found to be statistically significant.

229 -1.0 -0.8 -0.5 -0.3 0.0 0.3 0.5 0.8 1.0 10 2 m m 17 8 m m 20 3 m m (N D ) 30 5 m m (N D ) 20 3 m m (D ) 30 5 m m (D ) 40 6 m m (D ) D G A B A TB A TB /D G A B PA TB /D G A B A TB /P A TB N D D Design Factors M ea n st d. d ev ia te (a) Overall -1.0 -0.8 -0.5 -0.3 0.0 0.3 0.5 0.8 1.0 10 2 m m 17 8 m m 20 3 m m (N D ) 30 5 m m (N D ) 20 3 m m (D ) 30 5 m m (D ) 40 6 m m (D ) D G A B A TB A TB /D G A B PA TB /D G A B A TB /P A TB N D D Design Factors M ea n st d. d ev ia te Fine Coarse (b) By subgrade type -1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 10 2 m m 17 8 m m 20 3 m m (N D ) 30 5 m m (N D ) 20 3 m m (D ) 30 5 m m (D ) 40 6 m m (D ) D G A B A TB A TB /D G A B PA TB /D G A B A TB /P A TB N D D Design Factors M ea n st d. d ev ia te WF WNF (c) By zone type Figure 5-23 Effect of design factors on structural rutting (1 inch = 25.4 mm)

230 Table 5-27 Summary of p-values for comparisons of standard deviates— Structural rutting By subgrade By climatic zone By subgrade and zone WF WNF DF DNF Design Factor Comparison Overall Fine Coarse WF WNF DF DNF F C F C F C F C HMA thickness 102 mm vs. 178 mm 0.458 0.210 0.0270 0.580 0.501 0.001 0.078 0.420 0.900 0.900 0.300 Overall 203 mm vs. 305 mm vs. 406 mm 0.270 0.850 0.240 0.560 0.046 0.390 0.620 0.600 0.200 0.300 0.002 ND 203 mm vs. 305 mm 0.080 0.193 0.245 0.947 0.049 0.241 0.745 0.036 0.159 0.990 0.001 Base thickness D 203 mm vs. 305 mm vs. 406 mm 0.675 0.438 0.695 0.907 0.283 0.669 0.712 0.784 0.496 0.358 0.126 Overall DGAB vs. ATB vs. ATB/DGAB 0.180 0.225 0.230 0.520 0.003 0.770 0.370 0.004 0.170 0.002 0.440 ND DGAB vs. ATB vs. ATB/DGAB 0.579 0.879 0.497 0.931 0.240 0.632 0.161 0.660 0.527 0.095 0.823 D DGAB vs. ATB 0.332 0.110 0.692 0.044 0.070 0.471 0.487 0.000 0.400 0.156 0.308 Base type All Bases DGAB vs. ATB vs. ATB/DGAB vs. DGAB/PATB vs. ATB/PATB 0.140 0.274 0.368 0.214 0.004 0.323 0.409 0.017 0.053 0.030 0.578 Overall Drainage vs. No-Drainage 0.028 0.110 0.140 0.095 0.002 0.069 0.980 0.700 0.007 0.004 0.170 DGAB Drainage vs. No-Drainage 0.810 0.760 0.860 0.940 0.100 0.180 0.710 0.590 0.370 0.110 0.550 Drainage ATB Drainage vs. No-Drainage 0.030 0.100 0.170 0.010 0.250 0.120 0.270 0.290 0.006 0.580 0.310 N 161 77 84 66 59 24 12 30 36 35 24 Note: Shaded cells show statistically significant at 90% or higher level of confidence.

231 Table 5-28 Summary of means of PI for structural rutting By subgrade and zone By subgrade By climatic zone WF WNF DF DNF Design Factors Comparison Overall Fine Coarse WF WNF DF DNF F C F C F C F C 102 mm 5.4 5.7 5.1 5.2 5.8 4.7 6.3 5.3 5.1 5.9 5.7 4.7 6.3 HMA thickness 178 mm 5.2 5.8 4.6 5.4 5.5 3.3 6.7 5.7 5.1 5.7 5.1 3.3 6.7 203 mm 5.5 5.9 5.2 5.4 6.1 4.0 6.5 5.8 5.1 5.7 6.7 4.0 6.5 305 mm 5.2 5.7 4.7 5.2 5.4 3.8 6.4 5.1 5.4 6.0 4.5 3.8 6.4 Overall 406 mm 5.1 5.7 4.5 5.0 4.9 4.7 6.7 5.6 4.4 5.3 4.3 4.7 6.7 203 mm 5.9 6.3 5.5 5.7 6.8 3.9 6.5 6.3 5.3 6.2 7.5 3.9 6.5 ND 305 mm 5.1 5.5 4.6 5.2 5.4 3.2 6.4 4.2 5.9 6.2 4.2 3.2 6.4 203 mm 5.0 5.3 4.8 4.9 5.3 4.0 6.4 5.0 4.9 5.1 5.6 4.0 6.4 305 mm 5.3 6.0 4.7 5.3 5.4 4.7 6.3 6.2 4.5 5.7 5.0 4.7 6.3 Base thickness D 406 mm 5.1 5.7 4.5 5.0 4.9 4.7 6.7 5.6 4.4 5.3 4.3 4.7 6.7 DGAB 5.4 6.1 4.7 5.5 5.3 4.3 6.5 6.8 4.7 5.4 5.2 4.3 6.5 ATB 5.1 5.4 4.9 5.0 5.6 3.7 6.5 4.7 5.2 5.7 5.4 3.7 6.5 Overall ATB/DGAB 5.6 6.0 5.2 5.4 6.5 3.8 6.1 5.1 5.8 6.9 5.8 3.8 6.1 DGAB 5.3 6.0 4.9 5.2 5.9 3.5 6.5 5.6 5.1 6.0 5.8 3.5 6.5 ATB 5.4 5.7 5.1 5.6 5.8 3.3 6.8 5.3 5.9 5.7 5.8 3.3 6.8 ND ATB/DGAB 5.6 6.0 5.2 5.4 6.5 3.8 6.1 5.1 5.8 6.9 5.8 3.8 6.1 DGAB 5.4 6.2 4.6 5.7 5.0 4.9 6.6 7.1 4.4 5.1 4.7 4.9 6.6 Base type D ATB 4.9 5.2 4.7 4.5 5.4 4.1 6.3 4.2 4.8 5.6 5.2 4.1 6.3 ND 5.5 5.9 5.1 5.4 6.1 3.5 6.5 5.2 5.6 6.2 5.8 3.5 6.5 Overall D 5.1 5.7 4.7 5.1 5.2 4.5 6.5 5.7 4.6 5.4 5.0 4.5 6.5 ND 5.3 6.0 4.9 5.2 5.9 3.5 6.5 5.6 5.1 6.0 5.8 3.5 6.5 DGAB D 5.4 6.2 4.6 5.7 5.0 4.9 6.6 7.1 4.4 5.1 4.7 4.9 6.6 ND 5.4 5.7 5.1 5.6 5.8 3.3 6.8 5.3 5.9 5.7 5.8 3.3 6.8 Drainage ATB D 4.9 5.2 4.7 4.5 5.4 4.1 6.3 4.2 4.8 5.6 5.2 4.1 6.3

232 Roughness The effects of the design and site factors, in terms of standard deviate, are shown in Figure 5-24. The summary of p-values corresponding to the analyses performed to study the effects of design factors on roughness is shown in Table 5-28. The mean PI corresponding to each comparison presented in Table 5-28 is shown in Table 5-29. The effects of design factors on roughness, based on this analysis, are presented below: HMA thickness: In general, the effect of HMA surface thickness is statistically significant. Sections built with “thin” [102 mm (4-inch)] HMA surface have exhibited higher change in roughness than those built with “thick” [178 mm (7-inch)] HMA surface. This effect is more prominent among sections built on fine-grained soils. Base thickness: On the whole, the effect of base thickness is statistically and operationally significant. Sections with “thick” [406 mm (16-inch)] permeable base have exhibited the least change in roughness whereas sections with “thin” [203 mm (8-inch)] base thickness have shown the highest change in roughness. This effect is more significant among sections located in “wet” climate than among sections located in “dry” climate. Base Type: In general, the effect of base type is statistically and operationally significant. Sections built with DGAB have exhibited higher change in roughness than those built with ATB. This effect is more prominent among sections built on fine-grained soils. Drainage: By and large, the effect of drainage is only statistically significant i.e. it is not of practical significance at this point in time. Sections without drainage have exhibited higher change in roughness than those built with drainage. This effect is significant (statistical and operational) among sections built on fine-grained soils and located in WF zone. This effect is more prominent for sections with DGAB. This suggests that drainage is more effective for pavements with DGAB on fine-grained soils, especially when in WF zone.

233 The interaction effects among the experimental factors, on the change in roughness, are reported below: Also for un-drained pavements built on fine-grained soils, the effect of base type is significant, in that pavements with ATB have significantly lower ∆IRI. Furthermore, the effect of drainage for sections with DGAB and built on fine-grained soils, is significant. The above effects were found to be statistically significant and are of practical significance. For un-drained pavements built on coarse-grained soils, an increase in base thickness from 203 mm (8-inch) to 305 mm (12-inch) has a marginally significant effect, in that sections with thicker base have lower ∆IRI. However, this effect is not of practical significance at this point in time. It should be noted that, in general, pavements built on fine-grained soils have shown higher ∆IRI than those built on coarse-grained soils, especially among sections in WF zone. Also, the change in roughness among sections located in WF zone is higher than those in WNF zone.

234 This page is intentionally left blank.

235 -1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 10 2 m m 17 8 m m 20 3 m m (N D ) 30 5 m m (N D ) 20 3 m m (D ) 30 5 m m (D ) 40 6 m m (D ) D G A B A TB A TB /D G A B PA TB /D G A B A TB /P A TB N D D Design Factors M ea n st d. d ev ia te (a) Overall -1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 10 2 m m 17 8 m m 20 3 m m (N D ) 30 5 m m (N D ) 20 3 m m (D ) 30 5 m m (D ) 40 6 m m (D ) D G A B A TB A TB /D G A B PA TB /D G A B A TB /P A TB N D D Design Factors M ea n st d. d ev ia te Fine Coarse (b) By subgrade type -1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 10 2 m m 17 8 m m 20 3 m m (N D ) 30 5 m m (N D ) 20 3 m m (D ) 30 5 m m (D ) 40 6 m m (D ) D G A B A TB A TB /D G A B PA TB /D G A B A TB /P A TB N D D Design Factors M ea n st d. d ev ia te WF WNF (c) By zone type Figure 5-24 Effect of design factors on change in IRI (1 inch = 25.4 mm)

236 Table 5-29 Summary of p-values for comparisons of standard deviates— Change in IRI By subgrade By climatic zone By subgrade and zone WF WNF DF DNF Design Factor Comparison Overall Fine Coarse WF WNF DF DNF F C F C F C F C HMA thickness 102 mm vs. 178 mm 0.047 0.008 0.800 0.540 0.220 0.038 0.660 0.140 0.400 0.160 0.670 Overall 203 mm vs. 305 mm vs. 406 mm 0.008 0.120 0.070 0.045 0.028 0.900 0.340 0.320 0.069 0.200 0.145 ND 203 mm vs. 305 mm 0.019 0.128 0.076 0.083 0.068 0.861 0.540 0.372 0.056 0.242 0.185 Base thickness D 203 mm vs. 305 mm vs. 406 mm 0.445 0.898 0.498 0.501 0.265 0.884 0.133 0.287 0.381 0.448 0.629 Overall DGAB vs. ATB vs. ATB/DGAB 0.023 0.036 0.220 0.100 0.340 0.180 0.810 0.110 0.740 0.190 0.180 ND DGAB vs. ATB vs. ATB/DGAB 0.028 0.030 0.252 0.193 0.418 0.205 0.992 0.086 0.997 0.446 0.416 D DGAB vs. ATB 0.212 0.087 0.754 0.150 0.691 0.484 0.750 0.136 0.443 0.236 0.601 Base type All Bases DGAB vs. ATB vs. ATB/DGAB vs. DGAB/PATB vs. ATB/PATB 0.002 0.001 0.331 0.032 0.423 0.331 0.744 0.003 0.934 0.528 0.223 Overall Drainage vs. No-Drainage 0.006 0.011 0.170 0.028 0.163 0.920 0.160 0.007 0.834 0.900 0.055 DGAB Drainage vs. No-Drainage 0.014 0.009 0.300 0.060 0.340 0.520 0.180 0.001 0.830 0.900 0.23 Drainage ATB Drainage vs. No-Drainage 0.420 0.120 0.970 0.260 0.720 0.280 0.610 0.240 0.580 0.630 0.800 N 200 92 108 92 60 24 24 56 36 24 36 Note: Shaded cells show statistically significant at 90% or higher level of confidence.

237 Table 5-30 Summary of means of PI for change in IRI By subgrade and zone By subgrade By climatic zone WF WNF DF DNF Design Factors Comparison Overall Fine Coarse WF WNF DF DNF F C F C F C F C 102 mm 0.27 0.44 0.14 0.38 0.13 0.18 0.31 0.52 0.17 0.20 0.08 0.18 0.50 0.11 HMA thickness 178 mm 0.22 0.32 0.14 0.37 0.07 0.04 0.24 0.45 0.24 0.05 0.08 0.04 0.29 0.19 203 mm 0.30 0.48 0.16 0.46 0.16 0.10 0.29 0.59 0.26 0.24 0.11 0.10 0.45 0.14 305 mm 0.22 0.34 0.12 0.35 0.05 0.14 0.23 0.47 0.17 0.03 0.06 0.14 0.34 0.12 Overall 406 mm 0.18 0.25 0.11 0.23 0.06 0.08 0.33 0.28 0.15 0.08 0.05 0.08 0.41 0.24 203 mm 0.35 0.59 0.16 0.56 0.15 0.09 0.34 0.77 0.24 0.21 0.11 0.09 0.51 0.17 ND 305 mm 0.24 0.38 0.12 0.37 0.05 0.15 0.29 0.52 0.13 0.03 0.07 0.15 0.43 0.15 203 mm 0.22 0.29 0.17 0.29 0.17 0.11 0.22 0.29 0.30 0.28 0.09 0.11 0.34 0.10 305 mm 0.21 0.29 0.13 0.34 0.06 0.12 0.14 0.41 0.22 0.04 0.06 0.12 0.20 0.07 Base thickness D 406 mm 0.18 0.25 0.11 0.23 0.06 0.08 0.33 0.28 0.15 0.08 0.05 0.08 0.41 0.24 DGAB 0.31 0.49 0.16 0.47 0.14 0.19 0.27 0.63 0.24 0.24 0.07 0.19 0.41 0.12 ATB 0.18 0.27 0.11 0.28 0.06 0.05 0.27 0.34 0.18 0.05 0.06 0.05 0.38 0.15 Overall ATB/DGAB 0.26 0.38 0.15 0.39 0.10 0.07 0.30 0.51 0.19 0.04 0.14 0.07 0.39 0.21 DGAB 0.41 0.73 0.16 0.66 0.15 0.27 0.34 1.02 0.18 0.26 0.07 0.27 0.53 0.14 ATB 0.23 0.37 0.10 0.36 0.06 0.02 0.32 0.47 0.18 0.06 0.06 0.02 0.50 0.14 ND ATB/DGAB 0.26 0.38 0.15 0.39 0.10 0.07 0.30 0.51 0.19 0.04 0.14 0.07 0.39 0.21 DGAB 0.25 0.35 0.16 0.35 0.14 0.13 0.22 0.40 0.28 0.23 0.08 0.13 0.33 0.11 Base type D ATB 0.16 0.21 0.11 0.22 0.05 0.08 0.23 0.25 0.17 0.04 0.06 0.08 0.31 0.16 ND 0.30 0.48 0.14 0.46 0.10 0.12 0.32 0.64 0.19 0.12 0.09 0.12 0.47 0.16 Overall D 0.20 0.28 0.14 0.29 0.10 0.10 0.23 0.33 0.22 0.14 0.07 0.10 0.32 0.14 ND 0.41 0.73 0.16 0.66 0.15 0.27 0.34 1.02 0.18 0.26 0.07 0.27 0.53 0.14 DGAB D 0.25 0.35 0.16 0.35 0.14 0.13 0.22 0.40 0.28 0.23 0.08 0.13 0.33 0.11 ND 0.23 0.37 0.10 0.36 0.06 0.02 0.32 0.47 0.18 0.06 0.06 0.02 0.50 0.14 Drainage ATB D 0.16 0.21 0.11 0.22 0.05 0.08 0.23 0.25 0.17 0.04 0.06 0.08 0.31 0.16

238 Transverse Cracking The effects of the design and site factors, in terms of standard deviate, are shown in Figure 5-25. The summary of p-values corresponding to the analyses performed to study the effects of design factors on transverse cracking is presented in Table 5-31. The mean PI corresponding to each comparison presented in Table 5-31 is shown in Table 5-32. The effects of design factors on transverse cracking, based on this analysis, are presented below: HMA thickness: The effect of HMA surface thickness is not significant. However, on an average, sections with “thin” HMA surface have slightly higher cracking than sections with “thick” HMA layer. Base thickness: The effect of base thickness is marginally significant among sections located in WF zone, especially among sections built on fine-grained soils. Sections built with 406 mm base have shown the least cracking while sections with 203 mm or 305 mm base have shown the highest cracking. Base Type: On the whole, the effect of base type is statistically significant. Sections with ATB have exhibited the least cracking while sections with DGAB have shown the highest cracking. However, this effect is not operationally significant at this point in time. Drainage: In general, sections with un-drained sections showing higher cracking than those with drainage. In addition, this effect is significant (statistically and operationally) among sections built on fine-grained subgrade and located in WF zone. On the whole, at this point in time, sections in WF zone have shown higher cracking than those located in WNF zone indicating that transverse cracking is associated with low temperatures. Also, among drained pavements built on coarse-grained soils, those with ATB performed better than those with DGAB. However, among pavements with DGAB and built on fine-grained soils, those with drainage have shown significantly less transverse cracking than those without drainage. These effects were statistically significant and are of practical importance.

239 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 10 2 m m 17 8 m m 20 3 m m (N D ) 30 5 m m (N D ) 20 3 m m (D ) 30 5 m m (D ) 40 6 m m (D ) D G A B A TB A TB /D G A B PA TB /D G A B A TB /P A TB N D D Design Factors M ea n st d. d ev ia te (a) Overall -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 10 2 m m 17 8 m m 20 3 m m (N D ) 30 5 m m (N D ) 20 3 m m (D ) 30 5 m m (D ) 40 6 m m (D ) D G A B A TB A TB /D G A B PA TB /D G A B A TB /P A TB N D D Design Factors M ea n st d. d ev ia te Fine Coarse (b) By subgrade type -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 10 2 m m 17 8 m m 20 3 m m (N D ) 30 5 m m (N D ) 20 3 m m (D ) 30 5 m m (D ) 40 6 m m (D ) D G A B A TB A TB /D G A B PA TB /D G A B A TB /P A TB N D D Design Factors M ea n st d. d ev ia te WF WNF (c) By zone type Figure 5-25 Effect of design factors on transverse cracking (1 inch = 25.4 mm)

240 Table 5-31 Summary of p-values for comparisons of standard deviates— Transverse cracking By subgrade By climatic zone By subgrade and zone WF WNF DF DNF Design Factor Comparison Overall Fine Coarse WF WNF DF DNF F C F C F C F C HMA thickness 102 mm vs. 178 mm 0.100 0.170 0.330 0.610 0.310 0.310 0.097 0.850 0.630 0.180 0.940 Overall 203 mm vs. 305 mm vs. 406 mm 0.480 0.410 0.255 0.342 0.038 0.405 0.880 0.100 0.700 0.200 0.240 ND 203 mm vs. 305 mm 0.391 0.865 0.168 0.083 0.022 0.830 0.840 0.013 0.713 0.123 0.135 Base thickness D 203 mm vs. 305 mm vs. 406 mm 0.814 0.745 0.698 0.681 0.992 0.508 0.688 0.545 0.591 0.828 0.368 Overall DGAB vs. ATB vs. ATB/DGAB 0.008 0.240 0.012 0.353 0.240 0.025 0.013 0.660 0.120 0.345 0.370 ND DGAB vs. ATB vs. ATB/DGAB 0.172 0.215 0.618 0.770 0.824 0.370 0.040 0.914 0.235 0.548 0.644 D DGAB vs. ATB 0.003 0.630 0.001 0.240 0.474 0.040 0.049 0.374 0.120 0.794 0.318 Base type All Bases DGAB vs. ATB vs. ATB/DGAB vs. DGAB/PATB vs. ATB/PATB 0.011 0.006 0.037 0.542 0.285 0.107 0.010 0.167 0.185 0.329 0.434 Overall Drainage vs. No-Drainage 0.089 0.004 0.910 0.330 0.037 0.440 0.530 0.008 0.330 0.140 0.160 DGAB Drainage vs. No-Drainage 0.210 0.008 0.520 0.760 0.170 0.610 0.160 0.160 0.370 0.160 0.850 Drainage ATB Drainage vs. No-Drainage 0.070 0.110 0.190 0.050 0.250 0.710 0.850 0.070 0.260 0.810 0.240 N 120 48 72 24 48 24 24 12 12 24 24 Note: Shaded cells show statistically significant at 90% or higher level of confidence.

241 Table 5-32 Summary of means of PI for transverse cracking By subgrade and zone By subgrade By climatic zone WF WNF DF DNF Design Factors Comparison Overall Fine Coarse WF WNF DF DNF F C F C F C F C 102 mm 4.5 6.1 3.4 12.5 1.4 1.5 5.7 18.9 6.2 2.8 0.0 1.5 0.0 11.4 HMA thickness 178 mm 3.7 5.7 2.4 14.5 0.6 0.8 2.1 20.7 8.2 1.0 0.3 0.8 0.0 4.1 203 mm 4.0 5.0 3.4 11.2 1.9 1.1 4.2 13.2 9.1 3.4 0.3 1.1 0.0 8.4 305 mm 4.7 8.1 2.4 18.6 0.4 0.9 3.1 30.9 6.3 0.8 0.0 0.9 0.0 6.1 Overall 406 mm 2.9 2.5 3.1 6.6 0.3 2.0 5.1 8.8 4.3 0.6 0.1 2.0 0.0 10.3 203 mm 4.7 7.1 3.1 11.5 2.8 0.7 5.7 18.1 4.8 5.1 0.5 0.7 0.0 11.4 ND 305 mm 5.6 11.2 1.8 24.4 0.4 1.1 1.7 43.4 5.5 0.7 0.0 1.1 0.0 3.4 203 mm 3.1 2.0 3.8 10.8 0.5 1.7 1.9 5.9 15.6 1.0 0.0 1.7 0.0 3.7 305 mm 3.3 3.5 3.2 9.9 0.5 0.5 5.1 12.1 7.7 0.9 0.0 0.5 0.0 10.2 Base thickness D 406 mm 2.9 2.5 3.1 6.6 0.3 2.0 5.1 8.8 4.3 0.6 0.1 2.0 0.0 10.3 DGAB 5.4 5.2 5.5 15.1 0.7 2.2 8.3 18.3 11.9 1.3 0.0 2.2 0.0 16.6 ATB 2.5 4.7 1.1 10.5 0.5 0.4 0.8 17.4 3.6 0.7 0.3 0.4 0.0 1.6 Overall ATB/DGAB 4.8 10.6 1.0 17.0 3.1 0.4 0.5 29.8 4.2 6.3 0.0 0.4 0.0 0.9 DGAB 7.0 9.8 5.1 21.1 1.0 1.8 10.0 35.3 6.9 1.9 0.0 1.8 0.1 20.0 ATB 3.6 7.1 1.4 15.8 0.7 0.5 0.7 27.2 4.4 0.6 0.8 0.5 0.0 1.3 ND ATB/DGAB 4.8 10.6 1.0 17.0 3.1 0.4 0.5 29.8 4.2 6.3 0.0 0.4 0.0 0.9 DGAB 4.3 2.2 5.8 11.1 0.5 2.5 7.2 7.0 15.3 0.9 0.0 2.5 0.0 14.3 Base type D ATB 1.8 3.1 0.9 7.0 0.4 0.4 0.9 10.9 3.1 0.7 0.0 0.4 0.0 1.8 ND 5.1 9.1 2.5 17.9 1.6 0.9 3.7 30.7 5.2 2.9 0.3 0.9 0.0 7.4 Overall D 3.1 2.6 3.4 9.1 0.4 1.4 4.0 8.9 9.2 0.8 0.0 1.4 0.0 8.1 ND 7.0 9.8 5.1 21.1 1.0 1.8 10.0 35.3 6.9 1.9 0.0 1.8 0.1 20.0 DGAB D 4.3 2.2 5.8 11.1 0.5 2.5 7.2 7.0 15.3 0.9 0.0 2.5 0.0 14.3 ND 3.6 7.1 1.4 15.8 0.7 0.5 0.7 27.2 4.4 0.6 0.8 0.5 0.0 1.3 Drainage ATB D 1.8 3.1 0.9 7.0 0.4 0.4 0.9 10.9 3.1 0.7 0.0 0.4 0.0 1.8

242 Longitudinal Cracking- WP The effects of the design and site factors, in terms of standard deviate, are shown in Figure 5-26. The summary of p-values corresponding to the analyses performed to study the effects of design factors on longitudinal cracking-WP is presented in Table 5-33. The mean PI corresponding to each comparison presented in Table 5-33 is shown in Table 5-34. The effects of design factors on longitudinal cracking-WP, based on this analysis, are presented below: HMA thickness: The effect of HMA thickness on longitudinal cracking-WP is inconclusive. Sections with 102 mm HMA surface layer and sections with 178 mm HMA surface layer have shown comparable levels of longitudinal cracking-WP. Base thickness: The effect of base thickness on longitudinal cracking-WP is inconclusive. In general, all sections have shown comparable performance. Base Type: The effect of base type on longitudinal cracking-WP is inconclusive. However, on average, sections built on ATB have exhibited least cracking compared to other sections, especially among sections built on fine-grained soils. Drainage: In general, the effect of drainage is statistically significant with un-drained sections showing higher cracking than those with drainage. However, this effect is not operationally significant. This effect is statistically and operationally significant among sections built on fine- grained soils, especially among sections located in WF zone. In addition, drainage seems to be more effective for sections with DGAB. On the whole, at this point in time, sections in WF zone have exhibited much higher cracking than those in other climatic zones. Among pavements built on fine-grained soils, those built with DGAB have shown higher longitudinal cracking-WP and those built with ATB have shown the least longitudinal cracking-WP. This main effect of base type was statistically and operationally significance. Also among pavements built on fine-grained soils, drainage has a significant effect on longitudinal cracking and this effect is more pronounced (significant) among pavements built with DGAB. This effect is statistically significant and is of practical importance.

243 -1.30 -0.80 -0.30 0.20 0.70 1.20 10 2 m m 17 8 m m 20 3 m m (N D ) 30 5 m m (N D ) 20 3 m m (D ) 30 5 m m (D ) 40 6 m m (D ) D G A B A TB A TB /D G A B PA TB /D G A B A TB /P A TB N D D Design Factors M ea n st d. d ev ia te (a) Overall -1.30 -0.80 -0.30 0.20 0.70 1.20 10 2 m m 17 8 m m 20 3 m m (N D ) 30 5 m m (N D ) 20 3 m m (D ) 30 5 m m (D ) 40 6 m m (D ) D G A B A TB A TB /D G A B PA TB /D G A B A TB /P A TB N D D Design Factors M ea n st d. d ev ia te Fine Coarse (b) By subgrade type -1.30 -0.80 -0.30 0.20 0.70 1.20 10 2 m m 17 8 m m 20 3 m m (N D ) 30 5 m m (N D ) 20 3 m m (D ) 30 5 m m (D ) 40 6 m m (D ) D G A B A TB A TB /D G A B PA TB /D G A B A TB /P A TB N D D Design Factors M ea n st d. d ev ia te WF WNF (c) By zone type Figure 5-26 Effect of design factors on longitudinal cracking-WP (1 inch = 25.4 mm)

244 Table 5-33 Summary of p-values for comparisons of standard deviates— Longitudinal cracking-WP By subgrade By climatic zone By subgrade and zone WF WNF DF DNF Design Factor Comparison Overall Fine Coarse WF WNF DF DNF F C F C F C F C HMA thickness 102 mm vs. 178 mm 0.851 0.420 0.630 0.850 0.560 0.127 0.425 0.440 0.400 .730 0.290 Overall 203 mm vs. 305 mm vs. 406 mm 0.451 0.237 0.813 0.690 0.620 0.685 0.780 0.512 0.570 0.770 0.790 ND 203 mm vs. 305 mm 0.703 0.720 0.379 0.787 0.499 0.550 0.653 0.754 0.969 0.934 0.325 Base thickness D 203 mm vs. 305 mm vs. 406 mm 0594 0.362 0.839 0.803 0.558 0.248 0.929 0.480 0.497 0.364 0.439 Overall DGAB vs. ATB vs. ATB/DGAB 0.281 0.067 0.990 0.410 0.920 0.470 0.230 0.093 0.680 0.790 0.620 ND DGAB vs. ATB vs. ATB/DGAB 0.172 0.045 0.923 0.240 0.890 0.321 0.256 0.193 0.097 0.501 0.718 D DGAB vs. ATB 0.668 0.052 0.757 0.990 0.745 0.354 0.481 0.148 0.451 0.321 0.650 Base type All Bases DGAB vs. ATB vs. ATB/DGAB vs. DGAB/PATB vs. ATB/PATB 0.058 0.000 0.988 0.305 0.753 0.085 0.349 0.030 0.698 0.181 0.893 Overall Drainage vs. No-Drainage 0.049 0.001 0.815 0.297 0.220 0.045 0.650 0.052 0.290 0.063 0.943 DGAB Drainage vs. No-Drainage 0.020 0.001 0.850 0.130 0.340 0.080 0.710 0.080 0.450 0.110 0.510 Drainage ATB Drainage vs. No-Drainage 0.650 0.090 0.630 0.700 0.360 0.190 0.180 0.510 0.480 0.170 0.950 N 152 68 84 44 60 24 24 32 12 24 36 Note: Shaded cells show statistically significant at 90% or higher level of confidence.

245 Table 5-34 Summary of means of PI for longitudinal cracking-WP By subgrade and zone By subgrade By climatic zone WF WNF DF DNF Design Factors Comparison Overall Fine Coarse WF WNF DF DNF F C F C F C F C 102 mm 13.9 17.3 11.2 22.3 2.8 0.9 39.5 30.7 1.5 2.8 2.8 0.9 12.7 66.4HMA thickness 178 mm 14.3 17.9 11.3 24.6 2.7 2.6 35.4 31.4 5.4 5.7 0.7 2.6 4.4 66.4 203 mm 13.8 18.0 10.5 23.2 2.5 1.7 37.9 32.5 0.8 3.5 1.9 1.7 12.3 63.5 305 mm 14.3 18.5 10.8 24.7 3.7 2.2 33.0 31.3 6.3 5.7 2.3 2.2 8.2 57.8Overall 406 mm 14.5 14.7 14.3 21.3 1.3 0.6 47.6 27.5 2.7 2.7 0.3 0.6 0.0 95.2 203 mm 14.3 21.0 9.2 26.5 3.0 1.9 34.9 37.4 1.0 4.5 1.9 1.9 15.9 54.0ND 305 mm 15.0 22.9 8.6 27.8 5.2 3.7 27.5 37.8 1.0 7.6 3.6 3.7 13.7 41.3 203 mm 12.9 13.6 12.4 18.5 1.8 1.4 42.3 25.7 0.6 1.9 1.8 1.4 6.8 77.8 305 mm 13.2 12.2 14.0 20.6 1.3 0.0 41.3 22.6 14.3 2.7 0.4 0.0 0.0 82.6 Base thickness D 406 mm 14.5 14.7 14.3 21.3 1.3 0.6 47.6 27.5 2.7 2.7 0.3 0.6 0.0 95.2 DGAB 15.3 19.8 11.7 25.2 2.1 2.5 44.1 35.0 1.6 3.8 0.9 2.5 15.3 72.9 ATB 12.0 14.0 10.3 19.5 2.5 1.0 31.5 24.1 5.9 5.2 0.7 1.0 1.2 61.7Overall ATB/DGAB 16.8 22.2 12.5 30.9 5.2 1.9 35.8 42.5 1.8 3.1 6.7 1.9 9.8 61.8 DGAB 16.9 27.7 9.2 32.2 3.2 5.5 39.5 47.9 0.9 5.5 1.6 5.5 31.5 47.5 ATB 10.7 17.1 5.2 20.1 3.9 1.0 18.3 26.7 0.3 9.7 0.1 1.0 3.1 33.5ND ATB/DGAB 16.8 22.2 12.5 30.9 5.2 1.9 35.8 42.5 1.8 3.1 6.7 1.9 9.8 61.8 DGAB 14.2 15.2 13.5 21.4 1.3 0.4 47.2 28.6 2.1 2.7 0.4 0.4 4.5 89.9 Base type D ATB 12.8 11.9 13.7 19.1 1.6 0.9 40.3 22.3 9.7 2.1 1.2 0.9 0.0 80.5 ND 14.7 22.0 8.9 27.2 4.1 2.8 31.2 37.6 1.0 6.1 2.8 2.8 14.8 47.6Overall D 13.5 13.5 13.6 20.2 1.5 0.7 43.7 25.3 5.9 2.4 0.8 0.7 2.3 85.2 ND 16.9 27.7 9.2 32.2 3.2 5.5 39.5 47.9 0.9 5.5 1.6 5.5 31.5 47.5DGAB D 14.2 15.2 13.5 21.4 1.3 0.4 47.2 28.6 2.1 2.7 0.4 0.4 4.5 89.9 ND 10.7 17.1 5.2 20.1 3.9 1.0 18.3 26.7 0.3 9.7 0.1 1.0 3.1 33.5 Drainage ATB D 12.8 11.9 13.7 19.1 1.6 0.9 40.3 22.3 9.7 2.1 1.2 0.9 0.0 80.5

246 Longitudinal Cracking- NWP The effects of the design and site factors, in terms of standard deviate, are shown in Figure 5-27. The summary of p-values corresponding to the analyses performed to study the effects of design factors on longitudinal cracking-NWP is presented in Table 5-35. The mean PI corresponding to each comparison presented in Table 5-35 is shown in Table 5-36. The effects of design factors on longitudinal cracking-NWP, based on this analysis, are presented below: HMA thickness: The effect of HMA surface thickness is not significant. Comparable amount of cracking occurred in sections with “thin” and “thick” HMA surface. Base thickness: The effect of base thickness is not significant. On average, sections with 406 mm base have shown slightly lesser cracking than other sections. Base Type: The effect of base type is not significant. Comparable amount of cracking occurred in all sections, irrespective of base type. Drainage: In general, sections with un-drained sections showing higher cracking than those with drainage. However, this effect is not significant. Also, this effect is more apparent among sections located in WF zone. On the whole, at this point in time, it seems that longitudinal cracking-NWP is not a “structural” distress. It may be more affected by climate. It may be noted that the amount of longitudinal cracking-NWP is higher among sections located in “freeze” climate.

247 This page is intentionally left blank.

248 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 10 2 m m 17 8 m m 20 3 m m (N D ) 30 5 m m (N D ) 20 3 m m (D ) 30 5 m m (D ) 40 6 m m (D ) D G A B A TB A TB /D G A B PA TB /D G A B A TB /P A TB N D D Design Factors M ea n st d. d ev ia te (a) Overall -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 10 2 m m 17 8 m m 20 3 m m (N D ) 30 5 m m (N D ) 20 3 m m (D ) 30 5 m m (D ) 40 6 m m (D ) D G A B A TB A TB /D G A B PA TB /D G A B A TB /P A TB N D D Design Factors M ea n st d. d ev ia te Fine Coarse (b) By subgrade type -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 10 2 m m 17 8 m m 20 3 m m (N D ) 30 5 m m (N D ) 20 3 m m (D ) 30 5 m m (D ) 40 6 m m (D ) D G A B A TB A TB /D G A B PA TB /D G A B A TB /P A TB N D D Design Factors M ea n st d. d ev ia te WF WNF (c) By zone type Figure 5-27 Effect of design factors on longitudinal cracking-NWP (1 inch = 25.4 mm)

249 Table 5-35 Summary of p-values for comparisons of standard deviates— Longitudinal cracking-NWP By subgrade By climatic zone By subgrade and zone WF WNF DF DNF Design Factor Comparison Overall Fine Coarse WF WNF DF DNF F C F C F C F C HMA thickness 102 mm vs. 178 mm 0.304 0.510 0.440 0.430 0.490 0.120 0.230 0.250 0.940 0.460 0.790 Overall 203 mm vs. 305 mm vs. 406 mm 0.220 0.540 0.350 0.490 0.610 0.300 0.320 0.620 0.780 0.840 0.250 ND 203 mm vs. 305 mm 0.545 0.733 0.615 0.831 0.594 0.096 0.440 0.617 0.366 0.942 0.388 Base thickness D 203 mm vs. 305 mm vs. 406 mm 0.467 0.513 0.513 0.523 0.377 0.919 0.584 0.146 0.988 0.415 0.090 Overall DGAB vs. ATB vs. ATB/DGAB 0.280 0.390 0.420 0.620 0.500 0.760 0.042 0.960 0.430 0.630 0.650 ND DGAB vs. ATB vs. ATB/DGAB 0.524 0.278 0.702 0.963 0.621 0.857 0.014 0.740 0.650 0.679 0.327 D DGAB vs. ATB 0.196 0.268 0.451 0.792 0.540 0.569 0.180 0.840 0.877 0.456 0.711 Base type All Bases DGAB vs. ATB vs. ATB/DGAB vs. DGAB/PATB vs. ATB/PATB 0.235 0.157 0.743 0.540 0.701 0.465 0.052 0.587 0.764 0.743 0.311 Overall Drainage vs. No-Drainage 0.104 0.120 0.420 0.082 0.400 0.077 0.360 0.150 0.320 0.540 0.110 DGAB Drainage vs. No-Drainage 0.220 0.190 0.660 0.330 0.880 0.380 0.330 0.280 0.790 0.580 0.420 Drainage ATB Drainage vs. No-Drainage 0.210 0.090 0.780 0.240 0.370 0.130 0.007 0.250 0.670 0.430 0.084 N 184 76 108 76 60 24 24 40 36 24 36 Note: Shaded cells show statistically significant at 90% or higher level of confidence.

250 Table 5-36 Summary of means of PI for longitudinal cracking-NWP By subgrade and zone By subgrade By climatic zone WF WNF DF DNF Design Factors Comparison Overall Fine Coarse WF WNF DF DNF F C F C F C F C 102 mm 47.7 35.9 55.8 78.0 14.9 77.2 6.7 59.9 97.2 13.7 15.7 77.2 4.1 9.3 HMA thickness 178 mm 49.0 40.8 54.9 77.7 15.6 84.1 4.3 66.6 90.5 15.7 15.5 84.1 0.8 7.8 203 mm 48.9 37.4 56.4 79.9 16.8 84.2 4.0 65.2 93.6 14.9 18.1 84.2 4.2 3.8 305 mm 49.0 40.8 55.1 79.1 13.7 77.9 9.1 66.8 93.9 13.4 13.9 77.9 1.7 16.6Overall 406 mm 45.4 35.0 53.5 70.5 15.1 78.6 0.0 52.6 94.3 17.3 13.6 78.6 0.0 0.1 203 mm 52.4 47.7 55.5 88.0 13.9 89.3 4.6 81.2 94.8 19.9 9.9 89.3 3.0 6.3 ND 305 mm 46.9 35.9 54.6 77.9 9.8 78.1 10.1 61.1 96.6 10.4 9.4 78.1 2.8 17.4 203 mm 43.6 20.4 57.7 66.7 21.2 76.6 3.1 36.6 91.8 7.4 30.3 76.6 6.1 0.0 305 mm 52.2 47.4 55.8 80.8 19.6 77.7 7.7 74.0 89.8 18.0 20.7 77.7 0.0 15.5 Base thickness D 406 mm 45.4 35.0 53.5 70.5 15.1 78.6 0.0 52.6 94.3 17.3 13.6 78.6 0.0 0.1 DGAB 47.8 34.2 56.6 77.0 16.3 79.5 10.2 58.5 94.3 14.9 17.2 79.5 4.6 15.7 ATB 48.9 41.6 54.4 77.9 15.4 80.2 2.7 65.7 93.4 16.2 14.8 80.2 0.7 4.7 Overall ATB/DGAB 48.3 39.6 54.6 79.5 12.3 85.0 0.7 67.2 93.7 10.3 13.5 85.0 1.3 0.0 DGAB 47.0 28.8 57.2 81.9 11.1 81.0 15.6 56.8 98.7 12.4 10.2 81.0 5.5 25.7 ATB 53.1 52.7 53.3 86.6 12.2 85.2 5.9 80.5 94.7 22.7 5.2 85.2 1.8 9.9 ND ATB/DGAB 48.3 39.6 54.6 79.5 12.3 85.0 0.7 67.2 93.7 10.3 13.5 85.0 1.3 0.0 DGAB 48.3 37.0 56.2 74.4 19.8 78.5 6.6 59.1 91.4 16.6 21.9 78.5 4.1 9.1 Base type D ATB 46.1 33.8 55.1 71.9 17.5 76.8 0.6 55.0 92.6 11.9 21.2 76.8 0.0 1.3 ND 49.6 41.6 55.0 82.8 11.9 83.7 7.4 70.6 95.7 15.2 9.7 83.7 2.9 11.9Overall D 47.2 35.4 55.7 73.1 18.6 77.6 3.6 56.9 92.0 14.2 21.6 77.6 2.0 5.2 ND 47.0 28.8 57.2 81.9 11.1 81.0 15.6 56.8 98.7 12.4 10.2 81.0 5.5 25.7DGAB D 48.3 37.0 56.2 74.4 19.8 78.5 6.6 59.1 91.4 16.6 21.9 78.5 4.1 9.1 ND 53.1 52.7 53.3 86.6 12.2 85.2 5.9 80.5 94.7 22.7 5.2 85.2 1.8 9.9 Drainage ATB D 46.1 33.8 55.1 71.9 17.5 76.8 0.6 55.0 92.6 11.9 21.2 76.8 0.0 1.3

251 5.6.4 Effect of Experimental Factors on Pavement Response This section of the report is a discussion of the results from analyses of FWD data (pavement response) of the SPS-1 sections. Three parameters were chosen for analyses−peak deflection under FWD load (d0), far-sensor deflection (d6), and AREA. The peak deflection under FWD load is indicative of the “overall capacity” of the pavement structure while the far- sensor deflection is illustrative of the subgrade “strength”. The AREA is the area under the first three feet of the deflection basin. The computational details regarding the AREA can be found in the reference “LTPP Data Analysis: Feasibility of Using FWD Deflection Data to Characterize Pavement Construction Quality”, NCHRP Project 20-50 [7], by Richard N. Stubstad, October 2002. The AREA is indicative of stiffness of the upper layers of the pavement relative to the stiffness of the underlying layers. Higher the AREA higher is the stiffness of upper layers in relation to underlying layers. An ANOVA was conducted with the peak deflection under the FWD load plate (do), the far sensor deflection at 60 inches from the FWD load (d6) and the AREA as the dependent variables. All the response parameters have been calculated using the initial deflections of the test sections. It should be noted that the pavement surface temperature at the time of testing was taken as a covariate along with the age at the time of testing and variability in the HMA and base layer thicknesses. The natural logarithmic transformation has been applied to the three response indicators to fulfill the ANOVA assumptions. The results from ANOVA are summarized in Tables 5-37 and 5-38. The brief discussion of the results is given below: Peak Deflection under FWD Load (d0) When only design factors were considered by blocking the site effects, interactions between HMA thickness and base type (p=0.043), base thickness and base type (p=0.000), base type and drainage (p=0.000), have shown significant effects on the peak deflection. Among the pavement sections built on DGAB, those with 102 mm HMA thickness have shown higher peak deflection than those with 178 mm HMA thickness. Also as expected, thicker bases for each base type have helped in reducing the peak deflection. However, the reduction in peak deflection was more significant in the case of ATB and ATB/DGAB base

252 types. Furthermore, among pavement sections built on DGAB, those with drainage have shown lesser peak deflections than those without drainage. As mentioned before, DGAB with drainage refers to PATB/DGAB with drainage. It was assumed in the experiment, to study the effect of drainage among sections with DGAB that PATB has the same structural “strength” as DGAB. This assumption appears to be invalid from above result. When all the site and design factors were considered simultaneously along with the two- way interactions between the main factors, the interaction between subgrade soil and climatic zone (p=0.005) was found to have a very significant effect on the peak deflection. Among the pavement sections located in “wet” climates, those built on fine-grained subgrade soils have a significantly higher peak deflection (d0) as compared to those built on coarse-grained subgrade soils. This effect is more prominent on pavements located in WNF zone than those located in WF zone. Far Sensor Deflection (d6) When only design factors were considered by blocking the site effects, the main effects of base type (p=0.000), base thickness (p=0.005) and drainage (p=0.012) have significant effects on the far-sensor deflection (60-inches away from the center of the load). Pavement sections built on DGAB bases have shown higher far-sensor deflections than those built on other base types. Pavement sections constructed on 203 mm bases have also shown significantly higher far-sensor deflections than those built on 305 mm or 406 mm bases. Furthermore, sections built with drainage have lesser far-sensor deflections than those without drainage. This effect can be attributed to PATB as the effect of drainage and the effect of PATB cannot be separated (confounded). When all the site and design factors were considered simultaneously along with the two-way interactions between the main factors, an interaction between subgrade type and climatic zone (p=0.000) was found to have a significant effect on the far-sensor deflections. Among pavement sections located in “wet” climate, those built on fine-grained subgrade soils have higher deflections than those built on coarse-grained soils. This effect is significant among sections located in WNF zone. The ANOVA results also show that HMA thickness and pavement mid depth temperature do not have a significant affect on the far-sensor deflection. The results seem

253 reasonable, as this deflection (d6) represent the subgrade strength, which is independent of the HMA thickness and pavement temperature. AREA When only design factors are considered by blocking the site effects, the interactions between HMA thickness and base type (p=0.002), base thickness and base type (p=0.03), and, drainage and base type (p=0.000) have shown significant effects on AREA. Among pavement sections built on DGAB, those with “thin” HMA surface layer have significantly lower AREA values compared to those with “thick” HMA surface layer, implying that the upper layers of these pavement sections are “less stiff”. Also, the increase in HMA thickness from 102 to 178 mm on ATB does not significantly increase AREA. For sections built on DGAB, increasing base thickness from 8 to 12 inches has not shown a significant effect on AREA; however a two-fold increase in base thickness (from 8 to16 inch) has shown a significant increase in AREA. Also, base thickness does not seem to have a significant effect on AREA in pavement sections with ATB bases. Among the pavement sections constructed on DGAB, those with drainage have a significantly different AREA compared to those without drainage; test sections with drainage have higher AREA, implying higher stiffness. This indicates that the structural capacity of the PATB layer is somewhat higher than that of the DGAB. When all the site and design factors were considered simultaneously along with the two- way interactions between the main factors, the interaction between subgrade type and climatic zone (p=0.000) was found to have a very significant effect on AREA. Among the pavement sections located in WNF zone, those built on fine-grained subgrade soils have significantly higher AREA values than those built on coarse-grained soils. However, in the case of sections located in WF zone, this effect is not significant indicating that AREA could be independent of the subgrade soil type.

254 Table 5-37 Summary of p-values from ANOVA for determining the effect of design factors on flexible pavement response — Overall Performance Measures Design Factor Peak Deflection (do) Far Deflection (d6) AREA HMA thickness 0.000 0.560 0.000 Base type 0.000 0.000 0.000 Base thickness 0.000 0.005 0.214 Drainage 0.590 0.012 0.000 Mid depth temperature 0.000 0.738 0.000 Site (blocked) 0.000 0.000 0.000 R 2=0.884 N=210 R2=0.864 N=210 R2=0.854 N=210 Table 5-38 Summary of p-values from ANOVA for determining the effect of site factors on flexible pavement response — Overall Performance Measures Design Factor Peak Deflection (do) Far Deflection (d6) AREA Subgrade 0.000 0.000 0.353 Zone 0.000 0.000 0.000 Subgrade*Zone 0.005 0.005 0.005 Mid depth temperature 0.000 0.495 0.000 R 2=0.865 N=210 R2=0.658 N=210 R2=0.682 N=210

255 5.7 APPARENT RELATIONSHIP BETWEEN RESPONSE AND PERFORMANCE In this section of the report the observations regarding apparent relationships between flexible pavement response (FWD testing) and performance are presented. The usefulness of such relationships can be divided into two categories: • Explanatory: To provide an explanatory information for a given performance trend. For example, a relationship between AC rutting and the farthest sensor deflection would indicate that rutting is related to the subgrade soil. • Predictive: To provide a predictive capability of the future level of a given performance measure. For example, the initial high average deflection of a section may explain its future cracking and rutting (due to subgrade) performance. Explanatory relationships were established using multiple regressions on data from all the test sections in the experiment. Predictive relationships were established based on bivariate correlation analyses at the site level, and using scatter plots on data from all sections. The DLR data were used for predictive relationships regarding the instrumented sections in Ohio. 5.7.1 Overall Analysis—Explanatory Relationships In this section, the entire population of the SPS-1 experiment was used to seek apparent explanatory relationships between response and performance. This analysis was done irrespective of the experimental design matrix layout, since pavement response should reflect the effects of the various structural designs. In other words, the analysis spans over all the SPS-1 sections, as opposed to it being restricted to individual structural designs. The spatial variability of the deflections and deflection-based indices (within a section) was considered by taking the 95th percentile within each section. As deflection on all sections was measured during different seasons and times, the impact of temperature and moisture conditions cannot be ignored. Additionally the deflection and deflection-based parameters (SCI, BDI etc.,) are influenced by variety of factors, such as: • Asphalt temperature (at mid depth) • Thickness of asphalt layer • The layer moduli of various layers and overall pavement structure • Subgrade strength • Apparent stiff layer depth

256 • Pavement distresses etc. To consider the effect of various variables on the response at the same time, the multiple linear regression technique was used. The pavement response parameters (surface deflection (d0), SCI and BDI) were taken as dependent variables and all other variables (temperature, asphalt thickness, subgrade strength and pavement distresses) were considered as independent variables. As expected, the surface deflection is significantly correlated with the asphalt layer thickness, mid-depth asphalt temperature, and deflection at the outer most sensors that represents the subgrade strength. Furthermore, fatigue cracking, longitudinal cracking-WP and transverse cracking in all the sections have shown statistically significant relationships with the surface deflection. The results of multiple regression analyses within each zone have also shown that, on average, fatigue and transverse cracking has a significant positive effect on the surface deflection. An example of such multiple regression models is given by the following equation: LCWPTCACdAgeacHTod 001.0004.0001.0)6ln(602.0031.0085.017.0694.3)ln( ++++−−+= (R2=0.856, SE=0.269) where: ln is the natural logarithm T is the mid-depth asphalt concrete temperature (C) Hac is the HMA layer thickness Age is the age of the pavement section at the time of FWD testing AC is alligator cracking (sqm) at the time of testing TC is transverse cracking length (m) LCNWP is longitudinal cracking not in the wheel path (m) The sensitivity analysis of the regression model for overall SPS-1 database was performed to observe the explanatory relationships between various independent variables with the surface deflection (d0) under the FWD load plate. The following conclusions can be made from these results: • The effect of asphalt thickness on the measured surface deflection (d0) is very significant (p=0.000). The thicker the HMA layer, the lower the deflection will be. • Mid-depth asphalt temperature at the time of testing has a significant positive effect on d0 (p=0.000). • The age of the pavement has indicated a negative effect on d0 (p=0.000). Aging effect on HMA pavement may cause the stiffening of asphalt thus may reduce the deflections.

257 • The higher the “subgrade” deflection, d6 (deflection at the outer most sensor, or 60 inches in this case), the higher d0 will be (p=0.000). • Fatigue cracking (p=0.000), longitudinal cracking-WP (p=0.006) and longitudinal cracking- NWP (p=0.012) have a significant positive effect on d0; i.e., higher cracking will cause an increase in d0. • Similarly, transverse cracking has a significantly positive effect on d0 (p=0.001). 5.7.2 Site Level Analyses— Predictive Relationships This section summarizes the findings regarding predictive relationships between initial response (FWD deflection or deflection-based indices) and future pavement performance (cracking, rutting and roughness), at the site level. Various deflection-based indices [7, 8] were calculated based on the individual deflection basins for each section; these indices include: • AREA (the area under first three feet of deflection basin), • SCI─ Surface Curvature Index, (d0 – d12), • BDI─ Base Damage Index, [8](d12 – d24), • d36 ─ (d0-d36), • do (peak deflection under the load), • d6 (farthest deflection at 60 inches away from the load), • ES (effective stiffness of upper (bound) layer), and • Eg (subgrade modulus calculated from surface deflection at 36 inches from the load). Bivariate correlation analyses between response parameters (deflections or deflection basin parameters) and performance (cracking, rutting and roughness) were conducted for all the states within SPS-1 experiment. The latest performance for each section within the SPS-1 experiment was used in these analyses. The effect of temperature on the measured deflection was taken into account by applying a temperature correction [9] . It is to be noted that for a site age, traffic, construction, material properties and environment are the same and thus this provides a good opportunity for seeking apparent relationships. Figure 5-28 and Figure 5-29 are examples of bivariate relationships between SCI and AREA with fatigue cracking. The site in the state of Kansas (20) was chosen for this example because of high extent of cracking at the site. It can be seen that for the sections in this site, initial SCI and initial AREA have a slight association (ρ = 0.4) with the future fatigue cracking,

258 in that higher the SCI or lower the AREA, higher is the cracking. Similarly, Figure 5-31 is the relationship between BDI and future rutting for the same site. In this case, BDI has a strong correlation (ρ = 0.77) with future rutting i.e. sections that had higher initial BDI have higher rutting at a later stage. Also, Figure 5-30 shows the variation in future roughness (IRI) as a function of BDI. In this case, BDI has a correlation (ρ = 0.42) with the future roughness of all the pavement sections for this site (KS (20)). Figure 5-32 and Figure 5-33 show relationships of roughness and rut depth with BDI and AREA for the sections in the site of OH (39). Strong correlations (ρ = 0.85 each) were observed between future roughness and rut depth. Sections that had higher initial BDI had higher future roughness, and sections with lower AREA had higher future rutting.

259 y = 1.7875x + 126.76 R2 = 0.1661 0 50 100 150 200 250 300 350 400 1 10 100 1000 SCI Fa tig ue C ra ck in g (s qm ) Figure 5-28 Fatigue cracking and SCI relationship─ State (20) Kansas y = -14.815x + 548.98 R2 = 0.143 0 100 200 300 400 0 10 20 30 AREA Fa tig ue C ra ck in g (s qm ) Figure 5-29 Fatigue cracking and AF relationship─ State (20) Kansas y = 0.0096x + 1.0706 R2 = 0.178 0.0 0.5 1.0 1.5 2.0 2.5 3.0 10 100 BDI IR I ( m /k m ) Figure 5-30 Roughness and BDI relationship─ State (20) Kansas y = 0.2283x + 3.8001 R2 = 0.6012 0 5 10 15 20 25 30 35 10 100 BDI R ut D ep th (m m ) Figure 5-31 Rut depth and BDI relationship─ State (20) Kansas y = 0.0176x + 1.4784 R2 = 0.7185 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 10 100 1000 BDI IR I ( m /k m ) Figure 5-32 Roughness and BDI relationship─ State (39) Ohio y = -1.2739x + 41.542 R2 = 0.7329 0 5 10 15 20 25 10 15 20 25 30 AREA Ru t D ep th (m m ) Figure 5-33 Rut depth and AF relationship─ State (39) Ohio

260 Tables 5-39 through 5-41 are summaries of correlation coefficients from the bivariate analyses for three performances (fatigue cracking, rutting and roughness) and various deflection parameters between all the sites. The results show that fifteen out of seventeen sites have a consistent trend of relationship between AREA, BDI, do and future cracking. Also, fourteen out of seventeen sites have a positive association between SCI and fatigue cracking. On an average, AREA, SCI and BDI have reasonable associations with fatigue cracking for all the sites in SPS-1 experiment. Sections that have higher fatigue cracking had higher initial SCI or BDI, and lower AREA. The deflection basin parameters do not have a consistent association with rutting across the sites (see Table 5-40). This inconsistency may be explained in light of different rutting mechanisms for flexible pavements i.e., structural or asphalt mix rutting. Consistent trends were observed only between BDI and future roughness across most of the sites (15 out of 17 sites) in the SPS-1 experiment (see Table 5-41). Sections that had higher BDI have higher roughness. Apparent relationships between AREA and various performance measures (fatigue cracking, rutting and roughness) were found to be significant within sites that have shown considerable distress. Higher AREA means stiffer upper layers of a pavement. Sections that had higher AREA exhibited lesser cracking, rutting and roughness. Based on the magnitude of correlation coefficients, it was also found that sections that had stiffer bound layers are more likely to exhibit cracking than (structural) rutting.

261 Table 5-39 Summary of correlations for deflections and DBPs with fatigue cracking State Area ES/Esg ES Esg d0 d6 SCI BDI Zone SG 31 Nebraska -0.79 -0.66 -0.64 -0.37 0.90 0.43 0.94 0.91 WF F 26 Michigan 0.48 0.59 0.42 0.21 -0.42 -0.03 -0.46 -0.45 WF F 19 Iowa -0.45 -0.11 -0.09 0.05 0.41 0.15 0.55 0.49 WF F 20 Kansas -0.38 -0.26 -0.25 -0.14 0.33 -0.01 0.41 0.30 WF F 39 Ohio -0.66 -0.44 -0.44 -0.43 0.58 0.20 0.63 0.63 WF F 55 Wisconsin -0.46 -0.38 -0.37 -0.02 0.38 0.04 0.51 0.22 WF C 10 Delaware -0.93 -0.78 -0.65 -0.04 0.72 -0.03 0.93 0.73 WF C 5 Arkansas -0.07 -0.02 -0.19 -0.54 0.34 0.59 0.17 0.34 WF C 51 Virginia -0.72 -0.57 -0.58 -0.32 0.75 0.04 0.79 0.75 WNF F 1 Alabama -0.79 -0.68 -0.64 -0.08 0.72 -0.27 0.75 0.73 WNF F 48 Texas -0.48 -0.33 -0.57 -0.49 0.74 0.65 0.78 0.58 WNF C 40 Oklahoma -0.47 -0.43 -0.59 -0.16 0.25 -0.05 0.28 0.28 WNF C 12 Florida -0.40 -0.37 -0.40 0.14 0.50 -0.12 0.46 0.55 WNF C 30 Montana -0.36 -0.31 -0.58 -0.74 0.53 0.83 0.34 0.42 DF C 32 Nevada -0.49 -0.355 -0.31 0.10 0.38 -0.17 0.41 0.29 DF C 35 New Mexico 0.19 0.22 0.33 -0.14 0.31 -0.01 -0.14 0.60 DNF F 4 Arizona -0.13 -0.22 0.06 0.53 -0.10 -0.55 -0.03 -0.03 DNF C (-) ρ 15 15 14 12 2 9 3 2 (+) ρ 2 2 3 5 15 8 14 15 Mean -0.41 -0.30 -0.32 -0.14 0.43 0.10 0.43 0.43 Std 0.36 0.34 0.34 0.32 0.33 0.35 0.38 0.33 CoV 0.89 1.12 1.03 2.21 0.76 3.55 0.89 0.76 Note: The SPS-1 sections in State 22 (Louisiana) are young and have not shown any significant distress therefore, are not included in this analysis.

262 Table 5-40 Summary of correlations for deflections and DBPs with rut depth State Area ES/Esg ES Esg d0 d6 SCI BDI Zone SG 31 Nebraska -0.45 -0.48 -0.24 0.29 0.28 0.06 0.41 0.29 WF F 26 Michigan 0.32 0.53 0.41 0.10 -0.14 -0.04 -0.16 -0.14 WF F 19 Iowa -0.56 -0.43 -0.41 -0.22 0.40 0.13 0.43 0.46 WF F 20 Kansas -0.80 -0.55 -0.59 -0.23 0.76 0.11 0.82 0.78 WF F 39 Ohio -0.86 -0.88 -0.88 -0.72 0.79 0.51 0.75 0.76 WF F 55 Wisconsin 0.26 0.31 0.37 0.37 -0.60 -0.45 -0.48 -0.65 WF C 10 Delaware -0.72 -0.55 -0.62 -0.25 0.66 0.37 0.73 0.61 WF C 5 Arkansas 0.37 0.51 0.51 0.03 -0.11 -0.02 -0.20 -0.03 WF C 51 Virginia -0.58 -0.40 -0.39 -0.22 0.69 0.02 0.78 0.68 WNF F 1 Alabama -0.51 -0.34 -0.26 0.23 0.63 -0.30 0.73 0.69 WNF F 48 Texas 0.65 0.62 0.80 0.37 -0.67 -0.28 -0.65 -0.62 WNF C 40 Oklahoma 0.02 0.23 -0.15 -0.43 0.60 0.67 0.54 0.55 WNF C 12 Florida 0.56 0.60 0.40 -0.62 -0.33 0.66 -0.44 -0.39 WNF C 30 Montana -0.62 -0.66 -0.70 -0.59 0.68 0.56 0.60 0.63 DF C 32 Nevada 0.05 -0.002 -0.03 -0.31 0.03 0.13 -0.10 0.25 DF C 35 New Mexico 0.35 0.27 0.31 -0.24 -0.46 0.25 -0.47 -0.09 DNF F 4 Arizona -0.19 -0.31 -0.33 -0.20 0.06 0.02 0.05 0.04 DNF C (-) ρ 9 10 11 11 6 5 7 6 (+) ρ 8 7 6 6 11 12 10 11 Mean -0.16 -0.09 -0.11 -0.16 0.19 0.14 0.20 0.22 Std 0.51 0.50 0.49 0.34 0.51 0.33 0.52 0.48 CoV 3.22 5.46 4.65 2.20 2.61 2.33 2.66 2.13 Note: The SPS-1 sections in State 22 (Louisiana) are young and have not shown any significant distress therefore, are not included in this analysis.

263 Table 5-41 Summary of correlations for deflections and DBPs with IRI State Area ES/Esg ES Esg d0 d6 SCI BDI Zone SG 31 Nebraska -0.61 -0.44 -0.54 -0.43 0.66 0.49 0.62 0.72 WF F 26 Michigan -0.75 -0.71 -0.78 -0.82 0.78 0.74 0.73 0.76 WF F 19 Iowa -0.53 -0.31 -0.28 0.04 0.18 -0.12 0.30 0.25 WF F 20 Kansas -0.34 -0.31 -0.38 -0.38 0.40 -0.16 0.27 0.42 WF F 39 Ohio -0.79 -0.65 -0.65 -0.58 0.84 0.37 0.85 0.85 WF F 55 Wisconsin 0.37 0.45 0.50 0.23 -0.54 -0.04 -0.46 -0.54 WF C 10 Delaware -0.71 -0.60 -0.68 -0.36 0.73 0.28 0.68 0.76 WF C 5 Arkansas -0.02 -0.03 -0.17 -0.41 0.21 0.42 0.05 0.21 WF C 51 Virginia -0.70 -0.50 -0.55 -0.38 0.78 0.06 0.83 0.77 WNF F 1 Alabama -0.47 -0.32 -0.31 -0.31 0.69 0.23 0.66 0.69 WNF F 48 Texas 0.38 0.44 0.36 -0.17 -0.11 0.15 -0.20 -0.09 WNF C 40 Oklahoma 0.45 0.49 0.25 -0.35 0.17 0.56 0.10 0.10 WNF C 12 Florida -0.31 -0.31 -0.42 0.14 0.47 -0.15 0.38 0.51 WNF C 30 Montana -0.55 -0.61 -0.69 -0.62 0.63 0.62 0.50 0.55 DF C 32 Nevada -0.47 -0.229 -0.20 -0.16 0.71 0.44 0.72 0.68 DF C 35 New Mexico 0.39 0.50 0.49 -0.31 -0.02 0.33 -0.35 0.29 DNF F 4 Arizona -0.56 -0.46 -0.47 -0.08 0.63 0.05 0.64 0.65 DNF C (-) ρ 13 13 13 14 3 4 3 2 (+) ρ 4 4 4 3 14 13 14 15 Mean -0.31 -0.21 -0.27 -0.29 0.43 0.25 0.37 0.45 Std 0.44 0.42 0.42 0.27 0.39 0.28 0.41 0.37 CoV 1.44 2.00 1.57 0.93 0.91 1.12 1.11 0.83 Note: The SPS-1 sections in State 22 (Louisiana) are young and have not shown any significant distress therefore, are not included in this analysis.

264 5.7.3 Overall Analyses— Predictive Relationships This section summarizes the findings of apparent relationships between initial response (FWD deflection or deflection base indices) and future pavement performance (cracking, rutting and roughness), based on data from all the test sections in the SPS-1 experiment. The relationships were explored using bivariate scatter plots between selected response parameters and performance measures for all the pavements in the experiment. Though the sites differ in age, traffic, climate, and materials this analysis are intended to use the wealth of data from all the sections in the experiment. Moreover, the variation in age of the sites may not be very critical at this point in time as no definitive trends were observed between pavement age and performance (see Figure 5-34). Also, it is assumed in this analysis that deflection basin parameters (pavement response) will “characterize” the structural features such as HMA surface thickness, base type and base thickness. In other words, pavement response was assumed to be strongly correlated with the structural capacity of the pavement. In order to account for the effects of subgrade type and climate, relationships were explored for different subgrade soil types (fine- and coarse-grained soils) and climates (WF, WNF, DF and DNF). Figure 5-35 shows a scatter plot between SCI (from initial response) and fatigue cracking. Among pavements constructed on fine-grained soils, ones with higher initial SCI have more cracking, especially in WNF zone. Similarly, from Figure 5-36 it seems that stiffer pavements (higher AREA) on fine-grained soils, especially if located in WF climatic zone, have higher fatigue cracking. Higher longitudinal cracking-WP was observed for the pavement sections with higher initial AREA, especially among pavements located in WNF climatic zone (see Figure 5-37). The observation may imply that pavements with stiffer structural layers are more likely to exhibit this distress. No apparent relation was observed between AREA and longitudinal cracking-NWP (see Figure 5-38). The distress was found to be independent of the structural capacity (AREA)

265 of various pavements. Thus the probable cause of this distress type may be the environment and not the loading (traffic). Figure 5-39 shows a scatter plot between AREA and transverse cracking. No apparent trend was observed in the plot. This could imply that this distress type is not load-related and probably caused by the environment. The apparent relationship between initial BDI and rutting is shown in Figure 5-40. It seems that among pavements constructed on fine-grained soils and located in WF zone, those with higher BDI experienced higher rutting. It can also be observed that some pavements with less BDI (stronger structure) have experienced higher rutting as compared to pavements with high BDI. These pavements could have experienced mix-related rutting (not structural rutting). Figure 5-41 is the scatter plot between BDI and latest IRI. It seems that among pavements constructed on fine-grained soils and located in WF zone, those with higher BDI developed higher roughness.

266 Age (years) F at ig u e C ra ck in g (s q -m ) 1086420 350 300 250 200 150 100 50 0 SG C F (a) Fatigue cracking Age (years) L on g. C ra ck in g- W P ( m ) 1086420 250 200 150 100 50 0 SG C F (b) Long. cracking-WP Age (years) L on g. C ra ck in g- N W P ( m ) 1086420 300 250 200 150 100 50 0 SG C F (c) Long. cracking-NWP Age (years) T ra n sv er se C ra ck in g (m ) 1086420 90 80 70 60 50 40 30 20 10 0 SG C F (d) Transverse cracking Age (years) R u t d ep th ( m m ) 1086420 30 25 20 15 10 5 0 Zone DF DNF WF WNF (e) Rut depth Age (years) IR I (m / km ) 1086420 4.5 4.0 3.5 3.0 2.5 2.0 1.5 1.0 0.5 Zone DF DNF WF WNF (f) Roughness Figure 5-34 Relationships between age and different performance measures

267 SCI F at ig u e C ra ck in g (s q -m ) 100010010 350 300 250 200 150 100 50 0 SG C F (a) Effect by subgrade soil type SCI F at ig u e C ra ck in g (s q -m ) 100010010 350 300 250 200 150 100 50 0 Zone DF DNF WF WNF (b) Effect by climatic zone Figure 5-35 Apparent relationships between SCI and fatigue cracking

268 Area Factor F at ig u e C ra ck in g ( sq -m ) 3025201510 350 300 250 200 150 100 50 0 SG C F (a) Effect by subgrade soil type Area Factor F at ig u e C ra ck in g ( sq -m ) 3025201510 350 300 250 200 150 100 50 0 Zone DF DNF WF WNF (b) Effect by climatic zone Figure 5-36 Apparent relationships between AREA and fatigue cracking

269 Area Factor L C _W P ( m ) 3025201510 250 200 150 100 50 0 SG C F (a) Effect by subgrade soil type Area Factor L C _W P ( m ) 3025201510 250 200 150 100 50 0 Zone DF DNF WF WNF (b) Effect by climatic zone Figure 5-37 Apparent relationships between AREA and longitudinal cracking-WP

270 Area Factor L C _N W P ( m ) 3025201510 300 250 200 150 100 50 0 SG C F (a) Effect by subgrade soil type Area Factor L C _N W P ( m ) 3025201510 300 250 200 150 100 50 0 Zone DF DNF WF WNF (b) Effect by climatic zone Figure 5-38 Apparent relationships between AREA and longitudinal cracking-NWP

271 Area Factor T ra n sv er se C ra ck in g ( m ) 3025201510 90 80 70 60 50 40 30 20 10 0 SG C F (a) Effect by subgrade soil type Area Factor T ra n sv er se C ra ck in g ( m ) 3025201510 90 80 70 60 50 40 30 20 10 0 Zone DF DNF WF WNF (b) Effect by climatic zone Figure 5-39 Apparent relationships between AREA and transverse cracking

272 BDI R u t d ep th ( m m ) 100101 30 25 20 15 10 5 0 SG C F (a) Effect by subgrade soil type BDI R u t d ep th ( m m ) 100101 30 25 20 15 10 5 0 Zone DF DNF WF WNF (b) Effect by climatic zone Figure 5-40 Apparent relationships between BDI and rut depth

273 BDI IR I (m / k m ) 100101 4.5 4.0 3.5 3.0 2.5 2.0 1.5 1.0 0.5 SG C F (a) Effect by subgrade soil type BDI IR I (m / k m ) 100101 4.5 4.0 3.5 3.0 2.5 2.0 1.5 1.0 0.5 Zone DF DNF WF WNF (b) Effect by climatic zone Figure 5-41 Apparent relationships between BDI and IRI

274 5.7.4 Dynamic Load Response for OH (39) test sections This section presents the summary of findings from the analysis of Dynamic Load Response (DLR) data from the instrumented flexible pavement sections in the state of Ohio. These sections were instrumented with strain gauges, pressure cells and LVDTs to measure the pavement “response”. The SHRP experiment in Ohio targeted four core sections (see Chapter 2) for the installation of sensors to monitor dynamic pavement response during controlled vehicle tests. The main objective in this project was to study the response-performance relationship by using the measured dynamic load response and actual observed performance of the sections, in the SPS-1 experiment. Therefore, an attempt was made to relate the observed performance of these instrumented sections with measured responses (strains and surface deflections in HMA surface layer, and stress at top of subgrade) by means of bivariate scatter plots. The bivariate relationships between measured responses and observed performances are shown in Figure 5-42 and Figure 5-43. However, this finding is limited to these four instrumented sections. The measured longitudinal strain (initial value) is “strongly” associated with future fatigue cracking, and the vertical stress at the top of the subgrade is “strongly” associated with future rutting. It should be noted that the pavement sections in OH (39) site exhibited premature rutting due to very wet and soft subgrade soil. Other observations regarding the dynamic load response of the instrumented test sections are summarized below: • In general, the strain in the longitudinal direction is higher than the strain in the transverse direction; this is consistent with the mechanistic analysis for flexible pavements. • Sections with higher strain values have poor fatigue performance. These results are in agreement with the mechanistic-empirical design predictions as fatigue cracking in the flexible pavements is generally considered to be related to the initial tensile strain at the bottom of the HMA layer (bottom up cracking). • The sections that exhibited high measured stress at the top of the subgrade and high surface deflection have shown poor rut performance. • The sections that exhibited high measured stress and deflection have higher roughness.

275 y = 0.3994x0.8073 R2 = 0.8998 0 5 10 15 20 25 30 35 0 50 100 150 200 250 Long. Strains (micro-strains) Fa tig ue C ra ck in g (s qm ) (a) Relationship between strain and fatigue cracking y = 0.5961x1.1511 R2 = 0.9269 0 2 4 6 8 10 12 14 16 0 5 10 15 20 Stress at top of subgrade (psi) Ru t D ep th (m m ) (b) Relationship between stress and rutting y = 1.3002x0.5544 R2 = 0.7777 0 2 4 6 8 10 12 14 16 0 20 40 60 80 Surface Deflection (mils) Ru t D ep th (m m ) (c) Relationship between deflection and rutting Figure 5-42 Relationship between measured responses and observed performances— Fatigue cracking and rutting

276 y = 0.7727x0.3798 R2 = 0.7164 0.0 0.5 1.0 1.5 2.0 2.5 0 5 10 15 20 Stress at top of subgrade (psi) IR I ( m /k m ) (d) Relationship between stress and roughness y = 1.0807x0.1565 R2 = 0.4399 0.0 0.5 1.0 1.5 2.0 2.5 0 20 40 60 80 Surface Deflection (mils) IR I ( m /k m ) (e) Relationship between deflection and roughness Figure 5-43 Relationship between measured responses and observed performances— Roughness

277 5.8 SYNTHESIS OF RESULTS FROM ANALYSES This section of the report summarizes the findings from various analyses performed on SPS-1 data. The methods employed in this study were explained in Chapter 4 and the results obtained from these analyses were presented above in this chapter. Broadly two types of analyses were employed−magnitude-based and frequency-based. The magnitude-based analyses that were used are one-way (univariate) and multivariate ANOVA. These methods are used for comparison of means. The frequency-based analyses that were used are Binary Logistic Regression (BLR) and Linear Discriminant Analysis (LDA). These methods help identify the factors that significantly contribute to the occurrence of a distress based on the likelihood of occurrence or non-occurrence of distresses. In site-level analyses, the performance of pavements within each site was compared. The results from site- level analysis were used to ascertain the consistency of the effects of experimental factors across all sites. The magnitude-based methods, though powerful, are more appropriate for analyses of distresses which have both high occurrence and magnitude (for example: fatigue cracking, roughness, and rutting). On the other hand, the frequency-based methods are more suitable when the occurrence of a distress is fairly high (for example: transverse cracking) but magnitude is low. An attempt has been made to summarize the above said effects of design and site features on the performance and response measures. The results were interpreted in light of the type of analysis, and occurrence and extent of distress. ANOVA being the most “powerful” among the methods was given higher importance. However, the results from this analysis may not be reliable in case of limited (low occurrence of distress) or unbalanced data. Therefore, in these cases, the effects of design features, on the occurrence of distresses were investigated using BLR and LDA. All results need to be interpreted in light of the experiment design, occurrence and extent of distresses, and analyses methods used. A “weak” effect at this point in time may become a “medium” or “strong” effect in the long term. Hence, all the conclusions are based on “mid- term” performance of the ongoing SPS-1 experiment. The synthesis of results is presented next for each performance measure separately.

278 5.8.1 Effects of structural factors for flexible pavements — SPS-1 experiment This section is subdivided into three parts: (i) pavement performance, (ii) pavement response, and (iii) relationship between response and performance. The structural factors include HMA thickness, base thickness, base type, and drainage. The experiment also includes studying the secondary effects of site factors, namely subgrade type and climatic zones. 5.8.1.1 Effect of Design and Site Factors on Pavement Performance The effects of the experimental factors on each performance measure are discussed below, one performance measure at a time. Fatigue Cracking All the experimental factors were found to be affecting fatigue cracking, though not at the same level. On the whole, pavements with “thin” 102 mm (4-inch) HMA surface layer have shown more fatigue cracking than those with “thick” 178 mm (7-inch) HMA surface layer. Also pavements constructed with only dense-graded aggregate base (DGAB) have shown more fatigue cracking than those with dense-graded asphalt treated base over unbound aggregate base (ATB/DGAB) and those with ATB base only, with the latter base type showing the best performance. The effects of HMA surface thickness and base type were found to be statistically and practically significant. The main effect of base thickness was found to be statistically insignificant. However, on average, pavements with 406 mm (16-inch) base thickness have shown slightly better fatigue performance than those with 203 mm (8-inch) or 305 mm (12-inch) base thickness. It should be noted that only pavement sections with drainage have a 406 mm (16-inch) base thickness according to the SPS-1 experiment design; therefore, it is unclear whether this effect is caused by the increased base thickness or by drainage provided with the permeable asphalt treated base (PATB). In this regard, the frequency-based analyses did show that pavements with drainage have significantly lower chances of cracking than those without drainage. In general, pavement sections built on fine-grained soils have more fatigue cracking than those built on coarse-grained soils. Also pavements located WF zone have shown more fatigue cracking than those located in WNF zone. These effects were found to be statistically significant and are of practical significance.

279 Among un-drained pavements, on average, an increase in HMA surface thickness from 102 mm (4-inch) to 178 mm (7-inch) has a slightly higher effect on fatigue cracking for pavements with DGAB than for pavements with ATB. The above effect of HMA surface thickness is more significant for sections built on coarse-grained soils. On the other hand, among pavements built on fine-grained soils, the effect of drainage is seen only in those sections with DGAB; i.e., those with drainage have less fatigue cracking than those without drainage. Also among drained pavements built on fine-grained soils, those with 203 mm (8-inch) base have more cracking than those with 305 mm (12-inch) or 406 mm (16-inch) base. These effects were found to be statistically and practically significant. Hence, for pavements built on fine-grained soils, thicker base helps improve fatigue performance for drained pavements while drainage helps improve fatigue performance for those with DGAB. The main effect of HMA thickness, discussed above, is mainly seen among sections located in WNF zone. The effect is of practical and statistical significance. This may be an indication that an increase of HMA thickness from 102 mm (4-inch) to 178 mm (7-inch) is not sufficient in resisting fatigue cracking for pavements in WF zone as compared to WNF zone. Among sections located in the WF zone, those with DGAB have shown the highest amount of cracking while those with ATB have the least cracking. In addition, those with 406 mm (16-inch) drained base have the least amount of fatigue cracking. These effects were found to be statistically and practically significant. This suggests that among pavements located in WF zone, “thick” 406 mm (16-inch) treated bases with drainage are less prone to cracking. The effects of HMA thickness and base thickness discussed above imply that, among sections located in WF zone, an increase in base thickness to 406 mm (with drainage) has a greater impact than an increase in HMA thickness from 102 mm (4-inch) to 178 mm (7-inch), suggesting that a thicker base and drainage helps in reducing frost effects. Structural Rutting The extent of structural rutting among the test sections in the SPS-1 experiment is 6.5 mm, on average, with a standard deviation of 2.4 mm. Their average age is about 7 years with a range between 4.5 and 10 years. The amount of rutting for the majority of these sections is within the normal range at this point in time. Therefore, the results at this point may only show initial trends and may not be of much practical significance.

280 Marginal main effects of drainage, HMA thickness, and base thickness on structural rutting were observed. Pavements with “thin” [102 mm (4-inch)] HMA surface layer have shown slightly more rutting than those with “thick” [178 mm (7-inch)] HMA surface layer. Also, on average, pavements with 406 mm (16-inch) drained base have shown somewhat better rut performance than those with 203 mm (8-inch) and 305 mm (12-inch) base. However, these effects of HMA surface thickness and base thickness were not found to be statistically significant. Pavements with drainage have less rutting than those without drainage. The effect of drainage on structural rutting was found to be statistically significant; however the effect is not of practical significance at this point in time. In general, pavement sections built on fine-grained subgrade have shown more rutting than those built on coarse-grained subgrade. This effect is statistically significant and appears to be of practical significance. On the other hand, there is no apparent effect of climate (WF vs. WNF) on structural rutting. Among the pavements built on coarse-grained soils, those with 178 mm (7-inch) HMA surface have shown slightly less rutting than those with 102 mm (4-inch) HMA surface. This effect was statistically significant; however it is not operationally meaningful at this point. The above suggests that for sections built on fine-grained soils an increase in HMA thickness from 102 mm (4-inch) to 178 mm (7-inch) may not be sufficient in reducing the amount of rutting. On the other hand, among pavements built on fine-grained soils, a marginal positive effect of drainage is seen in sections with ATB. Among drained pavements located in WF zone, those with DGAB have shown more rutting than those with ATB. Also, among sections located in WF zone and built with ATB, those with drainage have shown significantly less rutting than those without drainage. Both of these effects were found to be statistically significant and are of operational significance. This implies that, among pavements located in WF zone, those with ATB and drainage perform better than those with other combinations of base type and drainage. Among un-drained sections located in WNF zone, those with 305 mm (12-inch) base have less rutting than those with 203 mm (8-inch) base. This effect was found to be statistically significant and of practically significance. For sections built on DGAB and located in WNF zone, those with drainage have shown slightly less rutting than those without drainage. The effect was found to be marginally significant. These early trends imply that the importance of

281 drainage among pavements with DGAB is considerable in improving rut performance among sections located in WNF zone. On the other hand an increase in base thickness from 203 mm (8- inch) to 305 mm (12-inch) improves rut performance for un-drained sections, irrespective of base type. Roughness All the experimental factors were found to be affecting roughness, though not at the same level. Pavements with “thin” [102 mm (4-inch)] HMA surface layer have higher change in IRI (∆IRI) than those with “thick” [178 mm (7-inch)] HMA surface layer. This effect was found to be statistically significant but is not of practical significance at this point in time. Also, pavements constructed with DGAB have higher ∆IRI than those with ATB/DGAB and ATB, while pavements with ATB have the best performance for roughness. Pavements with thicker bases have lower ∆IRI. Also pavements with drainage have lower ∆IRI than un-drained pavements. The above main effects of base thickness, base type and drainage were found to be statistically significant and are of practical significance. In general, pavements built on fine-grained soils have shown higher ∆IRI than those built on coarse-grained soils, especially among sections in WF zone. Also, the change in roughness among sections located in WF zone is significantly higher than those in WNF zone. These effects were found to be statistically significant and are of practical significance. Among pavements built on fine-grained soils, an increase in HMA thickness from 102 mm (4-inch) to 178 mm (7-inch) has a significant positive effect on change in roughness. Also for un-drained pavements, those with ATB have significantly lower ∆IRI than those with DGAB. Finally the effect of drainage is significant only for sections with DGAB. The above effects were found to be statistically significant and are of practical significance. These effects suggest that, for pavements built on fine-grained soils, higher HMA thickness and/or treated base will help inhibit the increase in roughness. Also, drainage appears to be more effective in preventing an increase in roughness for sections with DGAB, especially among those located in WF zone. For un-drained pavements built on coarse-grained soils, an increase in base thickness from 203 mm (8-inch) to 305 mm (12-inch) has a marginally significant effect, in that sections with thicker base have lower ∆IRI. However, this effect is not of practical significance at this point in time.

282 Transverse Cracking The effect of base thickness on transverse cracking is insignificant, at this point. Pavements constructed with DGAB have more transverse cracking than those with ATB/DGAB and ATB, while pavements with ATB have shown the least amount of cracking. The effect was found to be statistically significant; however it is not of practical significance at this point in time. Slightly more cracking was observed on pavements with “thin” [102 mm (4-inch)] HMA surface layer. Also, pavements with drainage have shown slightly less cracking than un-drained pavements. However, these effects were not found to be statistically significant. In general, pavements built on fine-grained soils have shown more transverse cracking than those built on coarse-grained soils. This effect was found to be statistically significant and is of practical significance. Pavements located in WF zone have shown significantly more transverse cracking than those located in WNF zone. This main effect of climatic zone was found to be statistically significant and is of practical significance. This confirms that transverse cracking occurs mainly in freezing environment. Among drained pavements built on coarse-grained soils, those with ATB performed better than those with DGAB. Also, among pavements with DGAB and built on fine-grained soils, those with drainage have shown significantly less transverse cracking than those without drainage. These effects were statistically significant and appear to be of practical significance. Longitudinal Cracking-WP The effects of HMA and base thickness on longitudinal cracking-WP are insignificant at this point in time. Pavements with drainage have shown less cracking than un-drained pavements. The main effect of drainage was found to be statistically significant, but is not of practical significance at this point. In general, pavements built on fine-grained soils have shown more longitudinal cracking- WP than those built on coarse-grained soils. This effect is of statistical and practical significance. Also, on average pavements in WF zone have shown higher levels of longitudinal cracking-WP than those in WNF, especially among pavements built on fine-grained subgrade. This effect was found to be marginally significant.

283 Among pavements built on fine-grained soils, those built with DGAB have shown more longitudinal cracking-WP, and those built with ATB have shown the least amount of cracking. This main effect of base type was statistically and operationally significant. Also among pavements built on fine-grained soils, drainage has a significant effect on longitudinal cracking, and this effect is more pronounced among pavements built with DGAB. This effect was statistically significant and is of practical significance. This trend implies that if a pavement on fine-grained subgrade is constructed with a DGAB base, better performance (in terms of longitudinal cracking-WP) can be achieved by providing drainage. These effects are seen in both WF and. WNF zones. Longitudinal Cracking-NWP The effects of HMA thickness, base thickness, and base type on longitudinal cracking- NWP are insignificant at this point in time. Pavements with drainage have shown slightly less cracking than un-drained pavements. However, the effect of drainage was found to be only marginally significant. The effect of subgrade type was not found to be statistically significant. In general, more longitudinal cracking-NWP was observed among sections located in “freeze” climate compared to those in “no-freeze” climate. This main effect of climatic zone is statistically significant and is of practical significance. Also, the effect of drainage is more pronounced (with marginal statistical significance) among pavements located in “freeze” climate. However, this effect is not of practical significance. These initial trends indicate that longitudinal cracking-NWP is caused by “freeze” climate (frost effects), and that pavements without drainage may be more prone to it. 5.8.1.2 Effect of Design and Site Factors on Pavement Response Three pavement response parameters were chosen for ANOVA−peak deflection under FWD load (d0), far-sensor deflection (d6), and AREA. All the response parameters have been calculated using the initial deflections of the test sections. Also, the pavement surface temperature at the time of testing was taken as a covariate along with the age at the time of testing and variability in the HMA and base layer thicknesses. The natural logarithmic

284 transformation has been applied to the three response indicators to fulfill the ANOVA assumptions. The following discussion summarizes the effects of design and site factors on each of the response parameters. Peak Deflection under FWD Load (d0) The interactions between HMA thickness and base type, base thickness and base type, base type and drainage, have significant effects on the peak deflection (d0). Among the pavement sections built on DGAB, those with 102 mm (4-inch) HMA thickness have higher d0 than those with 178 mm (7-inch) HMA thickness. Also as expected, thicker bases for each base type have lower d0. However, this effect was more significant in the case of sections with treated bases (ATB or ATB/DGAB). Furthermore, pavement sections with PATB/DGAB have lower d0 than those with DGAB. The interaction between subgrade soil and climatic zone was found to have a very significant effect on d0. Test sections built on fine-grained soils have shown significantly higher d0 as compared to those built on coarse-grained soils. This effect is more prominent on pavements located in WNF zone. Far Sensor Deflection (d6) The effects of base type, base thickness and drainage have significant effects on the far- sensor deflection (d6). HMA thickness and pavement mid depth temperature do not have a significant effect on d6. The interaction between subgrade soil type and climatic zone was found to have a significant effect on d6. Test sections built on fine-grained soils have shown significantly higher d6 as compared to those built on coarse-grained soils. This effect is more prominent on pavements located in WNF zone. Pavement sections built with DGAB have shown higher far-sensor deflections than those built on other base types. Pavements constructed on 203 mm bases have also shown significantly higher far-sensor deflections than those built on 12-inch (203 mm) or 406 mm (16- inch) bases. Furthermore, pavement sections with PATB/DGAB have lower d6 than those with DGAB. These effects of the design factors on d6 are based on statistical analyses only, and may or may not be of practical importance.

285 AREA The interactions between HMA thickness and base type, base thickness and base type, and, drainage and base type have significant effects on the AREA parameter. Among pavement sections built on DGAB, those with “thin” HMA surface layer have lower AREA values compared to those with “thick” HMA surface layer, implying that the upper layers of these pavement sections are “less stiff”. The increase in HMA thickness from 102 mm (4-inch) to 178 mm (7-inch) on ATB does not significantly increase the AREA value. For sections built on DGAB, increasing base thickness from 203 mm (8-inch) to 305 mm (12-inch) has not shown a significant effect on AREA; however a two-fold increase in base thickness [from 8 to16 inch (203 to 406 mm)] has shown a significant increase in AREA. Also, base thickness does not seem to have a significant effect on AREA in pavement sections with ATB bases. Furthermore, pavement sections with PATB/DGAB have higher AREA values than those with DGAB. This indicates that the structural capacity of the PATB layer is somewhat higher than that of the DGAB. Among the pavement sections located in WNF zone, those built on fine-grained subgrade soils have significantly higher AREA values than those built on coarse-grained soils. However, in the case of sections located in WF zone, this effect is not significant indicating that AREA could be independent of the subgrade soil type. A simplified summary of results from all analyses is given in Table 5-42. The summary is only meant to give an overall assessment of the effects. The reader is strongly recommended to read the following write-up for a better understanding of all the effects. It is important to note that a “strong”, “medium” or “weak” effect should only be interpreted in terms of the difference in effects at the various levels of a factor. As an example, a “strong” effect of HMA thickness and a “strong” effect of subgrade soil type should not be interpreted as HMA thickness and subgrade type having the same strength of effect. A black circle indicates a “strong” effect (significant); a grey circle indicates a “medium” effect, and a white circle indicates a “weak” effect. Operational significance was determined only for “strong” or “medium” effects. It should be noted that an effect can be statistically significant (meaning that it is not a coincidence) but may not be operationally/ practically significant, at this point in time.

286 Table 5-42 ‘Simplified’ summary of effects of design and site factors for flexible pavements Performance Measures Response Measures Longitudinal cracking Design Factor Fatigue cracking Rutting Roughness Transverse cracking WP NWP Peak deflection do Peak deflection d6 Area Factor HMA thickness Base type Base thickness Drainage Climatic Zone Subgrade type Note: This table is solely for the purpose of summarizing some of the effects in a ‘simple’ format. The reader is urged to read relevant text in the report for a better understanding. Symbol Description Strong Effect (Main effect exists) Medium Effect (Interaction effect) Weak Effect

287 5.8.1.3 Apparent Relationship between Response and Performance Two types of relations between flexible pavement response under (FWD testing) and performance were explored for the SPS-1 pavement sections—explanatory and predictive. Explanatory relationships were established using multiple regressions on data from all the test sections in the experiment. Predictive relationships were established based on bivariate correlation analyses at the site level, and using scatter plots on data from all sections. The dynamic load response (DLR) data from instrumented sections in Ohio were used for predictive relationships. The salient findings are briefly presented below: Overall Analysis— Explanatory Relationship A regression model was developed considering the peak deflection (d0) as the dependent variable and variables such as temperature, asphalt thickness, subgrade strength and performance measures as independent variables. The observations based on the regression model are as follows: • Pavements with “thick” [178 mm (7-inch)] HMA surface layer were observed (with statistical significance) to have significantly lower deflections than those with “thin” [102 mm (4-inch)] HMA surface layer. • Mid-depth temperature of the HMA layer, at the time of testing, has a statistically significant effect on d0. Irrespective of design features, pavement deflections (d0) measured at higher temperatures is greater than those at lower temperatures. • Older pavements have slightly lower deflections (d0) compared to younger pavements, which could be due to stiffening (aging) of the asphalt. • Pavements with “weaker” subgrade (higher d6) have significantly higher d0 (with statistical significance). • Pavements with more cracking (fatigue cracking or longitudinal cracking) have a significantly higher d0 (with statistical significance), compared to those with less cracking. Site Level Analysis— Predictive Relationships This section summarizes the findings regarding the predictive relationships between initial response (FWD deflection or deflection basin indices) and future pavement performance

288 (fatigue cracking, rutting and roughness) at the site level. The data for sections from LA (22) were excluded from these analyses, as performance data for the sections are available for just one year. • On average, AREA, SCI and BDI have shown reasonable correlations with fatigue cracking for sections in most of the sites in the SPS-1 experiment. In most of the sites, pavements with higher initial SCI or BDI, or lower initial AREA were found to have higher fatigue cracking. • Consistent trends were observed between BDI and future IRI for the various sites in the SPS-1 experiment. In most of the sites, pavements with higher initial BDI were found to have higher IRI. • The deflection basin parameters have not shown a consistent relationship with rut depth for the various sites in the SPS-1 experiment. Overall Analysis— Predictive Relationships Relationships were explored between initial response (FWD deflection basin indices) and pavement performance (cracking, rutting and roughness), using bivariate scatter plots between selected response parameters and performance measures for all pavement sections in the experiment. The main observations based on these relationships are listed below: • Among pavements constructed on fine-grained soils, ones with higher SCI have shown more fatigue cracking, especially in WNF zone. Also, stiffer pavements (higher AREA) on fine-grained soils have shown more fatigue cracking, especially if located in WF climatic zone. • Higher longitudinal cracking-WP was observed for the pavement sections with higher AREA especially among pavements located in WNF climatic zone. • No apparent relation was observed between AREA and longitudinal cracking-NWP, implying that this distress could be independent of the pavement structural capacity. • No apparent trend was observed between AREA and transverse cracking. This could imply that this distress type is not load-related. • Among pavements constructed on fine-grained soils and located in WF zone, those with higher BDI experienced slightly higher rutting. It was also observed that some pavements with lower BDI (stronger structure) have experienced higher rutting as

289 compared to pavements with high BDI. These pavements could have experienced mix-related rutting (not structural rutting). • Among pavements constructed on fine-grained soils and located in WF zone, those with higher BDI developed slightly higher roughness over time. Dynamic Load Response for OH (39) test sections This section of the report presents the summary of findings from the analysis of measured Dynamic Load Response (DLR) data from the instrumented flexible pavement sections in the state of Ohio. The observations from the analysis of these instrumented sections are summarized below: • In general, the strains in the longitudinal direction are higher than the strains in the transverse direction; this is consistent with the results from mechanistic analysis of flexible pavements. • The sections that were observed to have higher initial strain values have shown worse fatigue performance. These results are in agreement with the mechanistic-empirical design predictions that fatigue cracking in flexible pavements is related to the initial tensile strain at the bottom of the HMA layer (bottom up cracking). • The sections that were observed to have high initial stress at the top of the subgrade layer and those that were observed to have high initial surface deflection under the load have shown poor rut performance.

290 This page is intentionally left blank.