U.S. Department of Transportation
Federal Highway Administration
1200 New Jersey Avenue, SE
Washington, DC 20590
202-366-4000


Skip to content
Facebook iconYouTube iconTwitter iconFlickr iconLinkedInInstagram

Federal Highway Administration Research and Technology
Coordinating, Developing, and Delivering Highway Transportation Innovations

Report
This report is an archived publication and may contain dated technical, contact, and link information
Publication Number: FHWA-HRT-06-121
Date: November 2006

Long-Term Pavement Performance (LTPP) Data Analysis Support: National Pooled Fund Study Tpf-5(013)

Chapter 6. Environmental Performance Comparisons

One of the main objectives of this study was to compare pavement performance in various climatic settings. The environmental regions of interest were loosely defined as (1) southern freeze/northern no-freeze, (2) northern freeze, and (3) southern no-freeze regions. These regions were intended to capture pavement performance in areas with moderate frost depth (with multiple FTCs), deep sustained frost depth, and very little frost depth to provide a means of comparing the effect of FTC compared with the contribution of deep frost penetration (represented by FI) on pavement deterioration. As such, the following groupings were established for this study:

The deep-freeze regions were established to represent typical climatic conditions of the northern freeze zones where deep sustained frost penetration was presumed to occur. The moderate-freeze regions correspond to the southern freeze/northern no-freeze zone, which implied a multiple FTC climate. The region was established to represent the southern no-freeze region where frost penetration and multiple FTCs are very minimal.

While LTPP has defined wet/dry regions in terms of annual precipitation—508 mm/year (20 inches/year) and freeze/no-freeze regions as a function of FI (83.3 degree-Celsius days), the limits of the regions listed above have not been previously established; therefore, an investigation of FI and FTC relationships was performed on the analysis dataset to define limits for each of the scenarios.

Figure 41 provides the relationship between FI and FTC based on data from all test sections evaluated in the study. The no-freeze region was limited to areas with an FI less than 50 degree-Celsius days while the moderate-freeze was defined with FI between 50 and 400 degree Celsius days. The deep-freeze region consists of locations exhibiting an FI greater than 400 degree-Celsius days. These limits were established based on the natural break points inherent in the dataset, while also considering the LTPP-defined criteria for the freeze-, no-freeze regions (i.e., the moderate-freeze region must be centered on the LTPP freeze/no-freeze cutoff and include data points in both zones). The geographic boundaries for each of the regions were used to verify that selected FI limits were logical. The geographic location of each test section is plotted in figure 42.

Figure 41. Scatter plot. Regional FI and FTCs values.

View Alternate Text

Figure 42. Map. Geographic locations of climatic regions.

View Alternate Text
 

The majority of test sections classified as no-freeze is located in the Southern States with a few remaining sites located along the Pacific coast of the United States. The deep-freeze region consists of sites located in Canada and the northern United States. The band of sites in the moderate-freeze region is located across the middle of the United States running from northern Texas in the south to southern Washington in the north. These geographic locations are logical and consistent with the expected climates across the continent.

It should be noted that the following statement from the Task Order Proposal Request (TOPR) description of work indicates an inference that there are fewer FTCs in the northern freeze zone as compared with the moderate-freeze zone:

Near the southern edge of the freeze/no-freeze zone, pavements are subjected to multiple freeze-thaw cycles each year. Farther north, pavements may be subjected to few freeze-thaw cycles, but the depth of frost penetration is generally greater.

By inspection of figure 43, FTC values do not decrease in the deep-freeze regions (i.e., relatively high FI locations). Rather, the amount of FTC is approximately equivalent in the moderate- and deep-freeze regions; therefore, deep frost depths and multiple freeze-thaw cycles are not necessarily mutually exclusive factors and pavements do exist in areas that are classified as deep-freeze that also experience a large number of FTC.

Figure 43. Scatter plot. Relationship between FI and FTCs.

View Alternate Text

The above analysis was used to define the limits of deep-freeze, moderate-freeze, and no-freeze regions based on FI. To represent these climatic scenarios in the performance models, input values for each had to be established. Tables 20 and 21 provide the explanatory variable (i.e., input) values used to represent each of the regions for flexible and rigid pavements respectively. Nonenvironmental variables such as ACTHICK and logarithm ESAL divided by structural number (LESN) were held constant for each of the regions; therefore, performance predictions were made for consistent pavement structures exposed to similar traffic loading. This allowed the climatic effects on pavement performance to be compared without confounding the results by varying other contributing factors. The values for these nonclimatic factors were selected as the median value present in the analysis dataset. The scenarios listed in tables 19 and 20 incorporate constant BASE and SG types (DGAB and FINE, respectively). Predictions made using these two layer types are presented in the following text. Performance estimates were also established for other base/subgrade combinations, and key findings from those combinations are provided in the following discussion. This includes predictions for flexible pavement structures with overlay layers as well.

Table 20. Overview of climatic scenarios for flexible pavements.
ScenariosACTHICKBASESGLESNEXPFIFTCPRECIPCI
Deep-Freeze, Wet Region (low FTC) 6.5DGABFINE1.02G1688801140205
Moderate-Freeze, Wet Region (high FTC) 6.5DGABFINE1.02G11371301140645
No-Freeze, Wet Region 6.5DGABFINE1.02G1101011401300
Deep-Freeze, Dry Region (low FTC) 6.5DGABFINE1.02G168880380205
Moderate-Freeze, Dry Region (high FTC) 6.5DGABFINE1.02G1137130380645
 
Table 21. Overview of climatic scenarios for rigid pavements.
ScenariosFCDBASESGLEDTEXPFIFTCPRECIPCI
Deep-Freeze, Wet Region (low FTC) 29.5DGABFINE0.59S2688801140205
Moderate-Freeze, Wet Region (high FTC) 29.5DGABFINE0.59S21371301140645
No-Freeze, Wet Region 29.5DGABFINE0.59S2101011401300
Deep-Freeze, Dry Region (low FTC) 29.5DGABFINE0.59S268880380205
Moderate-Freeze, Dry Region (high FTC) 29.5DGABFINE0.59S2137130380645
 

Table 22 gives process details used to select environmental values for each climatic region shown in tables 20 and 21. With the exception of FTC, the values selected for all environmental variables were the median value present in the region. For example, the median FI of all test sections in the no-freeze region (i.e., 0<FI<50) was found to be 10 degree-Celsius days. Similarly, the CI value of 205 degree-Celsius days was determined to be the median CI value present for test sections in the deep-freeze region (FI>400).

Because the aim of the study is to compare the trade-off in pavement deterioration between deep frost and multiple FTCs, the percentile used to select FTC values was different for each region. In the deep-freeze region, one of the lowest FTC values present (10th percentile) was selected while one of the highest values was identified for the moderate-freeze region (90th percentile). This approach was used to isolate the contribution of FI from FTC in the deep-freeze region and allow the contribution of FTC to be investigated in the moderate-freeze region. As a caveat, it is apparent that pavements do exist in areas with both deep-frost penetration and multiple FTCs as can be seen in figure 43.

Table 22. Details on selection of environmental variables.
VariablePercentile CriteriaValue
Deep-Freeze (FI) 50th, deep-freeze (>400 FI)688
Moderate-Freeze (FI) 50th, moderate-freeze (50<FI<400)137
No-Freeze (FI) 50th, no-freeze (<50 FI)10
Low FTC 10th, FTC values in deep-freeze (>400 FI)80
High FTC 90th, FTC values in mod-freeze (50<FI<400)130
No-Freeze FTC 50th, FTC values in no-freeze (<50 FI)10
Dry (PRECIP) 50th, dry region (<508mm)380
Wet (PRECIP) 50th, wet region (>508mm)1140
Deep-Freeze (CI) 50th, deep-freeze (>400 FI)205
Moderate-Freeze (CI) 50th, moderate-freeze (50<FI<400)645
No-Freeze (CI) 50th, no-freeze (<50 FI)1300

When making comparisons, a determination must be made as to whether performance differences in the prediction of each scenario are statistically significant. The relatively small differences in performance can be confounded by the error inherent in the models. These differences cannot be attributed solely to the contribution of climatic setting. Rather, it is likely that the variability in the model is the source of the differences. Confidence intervals were used to determine if differences in performance were statistically significant. A confidence interval of 95 percent was selected for this study.

Using a lower confidence interval (i.e., 90 percent) would possibly lead to additional significant differences between the scenarios. Scenario departures that are marginally insignificant at the 95 percent confidence interval would likely be significant at the 90 percent level. The 95 percent level was used to provide highly significant differences, and it can also indicate the significance of variant performance. One can use the 74 confidence intervals provided to determine which marginal differences could be significant at a lower confidence interval.

In general, performance comparisons were made at a pavement age of 20 years for flexible pavements. This timeframe allows deterioration to accumulate to a sufficient level to make performance comparisons, and it is well within the range of pavement ages in the dataset used to develop the models. Approximately 80 out of 520 test sections in the flexible dataset were monitored at ages greater than 20 years. As such, there are sufficient quantities of observed data past 20 years represented in the models, and predictions made at an age of 20 years are not extrapolations from the dataset. In addition, the objective of the analysis was to study long term differences in performance so that financial effects of climate can be evaluated through life cycle cost analysis. The 20-year basis will provide a means of accomplishing this.

Comparisons were made at 20 years for the majority of the rigid pavement performance measures as well. Approximately 50 of the 275 test sections were monitored at ages greater than 20 years. Some of the scenarios evaluated initiated surface distress at after 20-year ages. To make comparisons at a point where all scenarios have initiated distress, the reference age was changed to 25 years for both TC and LC surface distress comparisons. Even at 25 years, the observations were not based on extrapolated predictions because approximately 29 of the 275 test sections were monitored at ages greater than 25 years. This reference age was not extended to 30 years because only six test sections were monitored at greater ages, which is a relatively limited number compared to the number of explanatory variables incorporated in the models.

Because transformations were incorporated into the model development process, it is essential that comparisons were made using the transformed scale. Mean predictions and confidence intervals were established based on transformed values; therefore, comparisons must be made in the same scale used to develop the estimates to be accurate. As such, all of the following prediction and confidence interval figures use a transformed scale (i.e., natural logarithm). Results can be converted to a linear scale if the model output is used solely as a predictive tool and the confidence intervals are not used for comparison.

PAVEMENT ROUGHNESS COMPARISONS FOR FLEXIBLE PAVEMENTS

Mean pavement roughness predictions were made for each of the scenarios defined above, and they are provided in figure 44 as a function of age. These predictions were established using an initial IRI of 1 m/km (63.3 inches/mi) at a pavement age of 1 year. Mean predicted values at 20 years along with the upper and lower limits of the 95 percent confidence interval are provided in figure 45 for each climatic region.

Figure 44 makes it apparent that pavement roughness accumulates at a more rapid rate in the wet deep-freeze and dry deep-freeze regions compared to the other regions, based on mean prediction values; however, these differences are not significant at 95 percent confidence and pavement age of 20 years as indicated in figure 45. The variability in the model confounds the differences observed in the mean predictions. The same observations can be made for all other base/subgrade combinations for both overlay and nonoverlay pavement structures.

Figure 44. Scatter chart. Mean predicted flexible pavement IRI values for each climatic region (BASE=DGAB/SG=FINE).

View Alternate Text
1 m/km = 5.28 ft/mi

Figure 45. Bar chart. Predicted flexible pavement IRI values at 20 years for each climatic region (BASE=DGAB/SG=FINE).

View Alternate Text
1 m/km = 5.28 ft/mi
 

As documented by Kameyama et al.(16) considerable increases in roughness during the freezing season do exist in areas of moderate and deep frost penetration; however, the increased roughness subsides by the middle of the following summer season. Based on the results in figure 45, any adverse long-term effects on flexible pavements from these seasonal cycles observed in the freezing areas are confounded by other variability.

PAVEMENT ROUGHNESS COMPARISONS FOR RIGID PAVEMENTS

Figure 46 provides mean rigid pavement roughness predictions as a function of age for each of the scenarios listed in table 20. Performance comparisons can be made using figure 47, which provides 95-percent-confidence interval information at a pavement age of 20 years for each of scenario.

Figure 46. Scatter graph. Mean predicted rigid pavement IRI values for each climatic region (BASE=DGAB/SG=FINE).

View Alternate Text
1 m/km = 5.28 ft/mi

Figure 47. Bar chart. Predicted rigid pavement IRI values at 20 years for each climatic region (BASE=DGAB/SG=FINE).

View Alternate Text
1 m/km = 5.28 ft/mi

Predicted pavement roughness at 20 years in areas of deep frost penetration is significantly different than in areas experiencing multiple FTCs. This can be seen in both the dry and wet climates. The wet no-freeze region exhibits roughness values that are significantly lower than that of the moderate- and deep-freeze wet regions. Rigid pavements in relatively wet climates and in areas exposed to frost penetration accumulate roughness more rapidly than wet climates in areas with no frost.

Similarly, the same observations for other base/subgrade combinations can be made. One exception is prediction values for pavements with ROCK/STONE subgrade type. Variability with this subgrade type results in insignificant performance differences between the dry multiple freeze-thaw climate and the wet no-freeze region.

RUT DEPTH COMPARISONS FOR FLEXIBLE PAVEMENTS

Mean predictions for the progression of rut depth with pavement age can be found in figure 48 for each of the climatic regions. The 20-year confidence interval results are presented in figure 49. Both of these graphs were established for DGAB base and FINE subgrade.

The following conclusions can be drawn for figure 49: While multiple FTCs and deep sustained frost penetration do not result in changes in rutting performance in wet climates (at a 95 percent confidence interval), pavements in both of these climates accumulate larger amounts of rutting compared to the no-freeze climates; conversely, in the dry region, multiple FTCs contribute to more rutting deterioration than both the deep-frost and no-freeze zones. The contribution of deep-frost penetration in relatively dry climates on rutting is not significantly different than that of the no-freeze wet region.

Figure 48. Scatter graph. Mean predicted flexible pavement RUT values for each climatic region (BASE=DGAB/SG=FINE).

View Alternate Text
1 mm = 0.04 inch

Figure 49. Bar chart. Predicted flexible pavement RUT values at 20 years for each climatic region (BASE=DGAB/SG=FINE).

View Alternate Text
1 mm = 0.04 inch
 

Rutting predictions for different base and subgrade combinations as well as overlay structures were also evaluated. In general, the observations discussed above for DGAB base, FINE subgrade, and nonoverlay structures can be made for other combinations. However, ROCK/STONE subgrade as well as LCB base types exhibit relatively large variability in both overlay and nonoverlay structures. This results in any performance differences between climates being statistically insignificant.

Rutting accumulation in pavements with PATB base type is not significantly different in areas exposed to deep-frost penetration and multiple FTCs (regardless of precipitation level). These accumulations are, however, significantly larger than the no-freeze wet region.

For pavements with NONBIT base, the multiple FTC climate exhibits significantly larger quantities of rutting as compared with the deep-frost penetration and no-freeze regions. Differences between the contribution of deep-frost penetration and no-freeze wet climates on rutting development are confounded by the variability in the model.

FATIGUE AND WHEELPATH CRACKING SURFACE DISTRESS COMPARISONS FOR FLEXIBLE PAVEMENTS

Mean predictions of FWPC progression for pavements with FINE subgrade and DGAB base are provided in figure 50. These predictions were generated by combining logistic analysis with regression models. As can be seen, variations in crack initiation timing as well as the rate of distress accumulation (after initiation) contribute to FWPC performance differences.

Confidence interval data for FWPC are illustrated in figure 51. The contribution of deep sustained frost significantly outweighs the effect of multiple FTCs on accumulation of FWPC. This holds true in both the wet and dry regions. The no-freeze region exhibits the largest FWPC values which, statistically speaking, are larger than FWPC predictions in areas experiencing multiple FTCs (regardless of the amount of precipitation).

Figure 50. Chart. Mean predicted flexible pavement FWPC values for each climatic region (BASE=DGAB/SG=FINE).

View Alternate Text
 

The same general trends are apparent for different base/subgrade pavement structure combinations. Increased variability in some combinations results in performance differences to be less than the error band of the model. This is the case for ROCK/STONE subgrade and LCB base types in both pavement structure groups. To a lesser extent, variability for the NONE base type results in insignificant differences between the deep-frost regions and the multiple FTC climates. Predicted accumulation in the no-freeze wet region is significantly larger than the multiple FTC climates (both precipitation levels).

TRANSVERSE CRACKING SURFACE DISTRESS COMPARISONS FOR FLEXIBLE PAVEMENTS

Figure 52 shows transverse cracking performance predictions for each climatic region in for DGAB base and FINE subgrade. Transverse cracking initiation ranges between 6 and 12 years, initiating earlier in the deep-freeze regions that experience colder minimum temperatures in the winter months. Conversely, pavements in the no-freeze regions, with mild winter temperatures, initiate transverse cracking much later.

Figure 51. Bar chart. Predicted flexible pavement FWPC values at 20 years for each climatic region (BASE=DGAB/SG=FINE).

View Alternate Text

Figure 52. Scatter chart. Mean-predicted flexible pavement TC values for each climatic region (BASE=DGAB/SG=FINE)

View Alternate Text
 

Figure 53 provides mean TC predictions accumulated at a pavement age of 20 years along with the upper and lower limits of the confidence interval. Pavements exposed to deep sustained frost penetration exhibit quantities of transverse cracking that are significantly larger than pavements in the multiple FTC climates. In addition, pavements exposed to climatic conditions that are representative of the southern reaches of the no-freeze region accumulate less transverse cracking (at 95 percent confidence) compared to areas with both deep sustained frost and multiple FTCs.

Figure 53. Bar chart. Predicted flexible pavement TC values at 20 years for each climatic region (BASE=DGAB/SG=FINE).

View Alternate Text
 

Changes in subgrade, base, and pavement structure types do alter the absolute magnitude of predicted TC. In some cases, the predicted TC deduct value reaches the equivalent of 100 on a natural logarithmic scale at a pavement age less than 20 years. Because the model is predicted natural logarithm of TC plus 0.1, a deduct value of 100 corresponds to the natural logarithm of 100.1 or 4.606. For example, if the predicted TC value for both of the deep-freeze regions reaches this maximum at age 16 while the other climatic scenarios do not reach the maximum value within 20 years, comparing confidence intervals at age 20 for this case would not account for the regions that reached the maximum value 4 years earlier. The deep-freeze regions were capped at 4.606 for those 4 years, while the other regions accumulated additional quantities of TC. To counter this, confidence intervals for all regions would be compared at the earliest age that one region reached the maximum value. For this example, age 16 would be used to make comparisons.

In general terms, comparisons between the various climatic scenarios are similar to those made for DGAB base, FINE subgrade, and nonoverlay pavement structure types. Key exceptions are noted below:

LONGITUDINAL CRACKING SURFACE DISTRESS COMPARISONS FOR RIGID PAVEMENTS

As mentioned previously, comparisons of rigid surface distress predictions were made at 25 years because of crack initiation ages greater than 20 years. Figure 54 presents the mean predicted LC values for each climatic scenario. As can be seen, distress initiation occurred at 25 years in the wet no-freeze region. Figure 55 provides the confidence intervals at 25 years.

Figure 54. Scatter graph. Mean predicted rigid pavement LC values for each climatic region (BASE=DGAB/SG=FINE).

View Alternate Text

Figure 55. Bar chart. Predicted rigid pavement LC values at 25 years for each climatic region (BASE=DGAB/SG=FINE).

View Alternate Text
 

Comparing the deep-freeze region to the moderate-freeze region for both wet and dry climates shows no significant difference in LC performance. On the other hand, the nofreeze wet region at 25 years exhibits significantly lower LC accumulation than the other regions.

The same observations can be made when evaluating longitudinal cracking for different subgrade and base combinations.

TRANSVERSE CRACKING SURFACE DISTRESS COMPARISONS FOR RIGID PAVEMENTS

Figure 56 shows transverse cracking predictions in for each of the climatic scenarios. Figure 57 shows confidence intervals. Crack initiation varies significantly with the moderate-freeze dry region initiating first and the deep-freeze wet region initiating 11 years after construction.

Figure 56. Scatter Graph. Mean predicted rigid pavement TC values for each climatic region (BASE=DGAB/SG=FINE).

View Alternate Text

Figure 57. Bar chart. Predicted rigid pavement TC values at 25 years for each climatic region (BASE=DGAB/SG=FINE).

View Alternate Text
 

For wet environments, the contribution of deep frost and multiple FTCs did not result in significant transverse crack accumulation differences as compared to the no-freeze region. In addition, transverse cracking differences between the deep-freeze and moderate-freeze climates in dry environments are insignificant; however, the effect of the wet no-freeze region resulted in significantly larger quantities of transverse cracking than the deep-freeze dry region at 25 years.

These observations hold true for the majority of base/subgrade combinations. Rigid pavement structures with PATB and asphalt-treated base (ATB) type categories exhibit considerable variability resulting in all performance differences to be insignificant and confounded by model error.

TRANSVERSE JOINT FAULTING COMPARISONS FOR RIGID PAVEMENTS

Mean transverse joint faulting predictions as a function of age are shown in figure 58 for each climatic scenario. The models used to develop the predictions were based on average faulting over the entire test section; therefore, the model does not predict faulting for one joint but rather for the average for a pavement segment.

Figure 59 provides confidence interval data at a pavement age of 20 years. Faulting in climates with considerable amounts of annual precipitation is not significantly different between any of the frost settings.

In drier climates, deep-frost penetration contributes to an increased accumulation of faulting as compared with the multiple FTC and no-freeze wet regions.

The same observations can be made for the majority of the base/subgrade combinations. The variability inherent in all combinations with ROCK/STONE subgrade is relatively high, which confounds the differences in predicted performance. For structures with ATB and NONBIT, deep frost (in the dry climate) contributes to significantly larger accumulations of faulting than the multiple freeze-thaw (dry) climate at a pavement age of 20 years; however, predictions in these climates are not statistically different than the no-freeze wet region.

Figure 58. Scatter chart. Mean predicted rigid pavement FLT values for each climatic region (BASE=DGAB/SG=FINE).

View Alternate Text
1 mm = 0.04 inch

Figure 59. Bar chart. Predicted rigid pavement FLT values at 20 years for each climatic region (BASE=DGAB/SG=FINE).

View Alternate Text
1 mm = 0.04 inch

Previous | Contents | Next

Federal Highway Administration | 1200 New Jersey Avenue, SE | Washington, DC 20590 | 202-366-4000
Turner-Fairbank Highway Research Center | 6300 Georgetown Pike | McLean, VA | 22101