Effects of Subsurface Drainage on Pavement Performance: Analysis of the SPS-1 and SPS-2 Field Sections (2007)

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.

85 Design of the SPS-1 and SPS-2 Experiments The LTPP Program’s SPS-1 and SPS-2 experiments were designed to assess the effects of several factors on the per- formance of flexible and rigid pavements. These factors include layer thickness, base type, subdrainage, climate, sub- grade, and truck traffic level. The SPS-2 experimental factors also include lane width and concrete flexural strength. There is a great deal that can be learned from the SPS-1 and SPS-2 experiments about the factors that influence AC and PCC pavement performance. Drainage is the experimental factor for which it is most dif- ficult to draw conclusions from the SPS-1 and SPS-2 experi- ments. This is because two experimental factors, base type and subdrainage, are confounded in both of the experiments. Some base types have drains and other base types do not; the experimental design does not include bases of the same type but with and without drains. This makes it difficult to discern how much any differences in performance between drained and undrained test sections are due to the presence or ab- sence of a functioning drainage system, versus how much any such differences are due to variations in base stiffness. It is somewhat easier to draw inferences about the role of drainage versus the role of base stiffness for the SPS-1 exper- iment than for the SPS-2 experiment, given the greater number of base type/drainage combinations in the SPS-1 experiment and the expectation that increased base stiffness in a flexible pavement will yield improved pavement per- formance. If SPS-1 sections with less stiff bases perform worse and sections with stiffer bases perform better than pavement sections with drained permeable asphalt-treated base layers, then it is reasonable to infer that base stiffness is the more important component of the base type/drainage experimental design factor. If, on the other hand, sections with permeable asphalt-treated layers perform better than both sections with less stiff bases and sections with stiffer bases, then it is reasonable to infer that drainage is the more important component. For rigid pavements, however, performance is optimized with a base that is neither too weak nor too stiff, the former being associated with high load-related stresses and the latter being associated with high curling stresses. Of the three base types in the SPS-2 experiment, the base of middle stiffness is also the one drained base type—permeable asphalt-treated aggregate—while the less stiff base (untreated aggregate) and the most stiff base (lean concrete) are both undrained. This makes it more difficult to draw inferences about whether base type or drainage is the more important component of the base type/drainage factor in the SPS-2 experiment. Supplemental sections at some SPS-2 sites provide additional insight, as do the results of regression analysis indicating the relative signif- icance of these as well as other experimental factors. Characteristics of the Sites Climate The SPS-1 and SPS-2 sites are comparable in terms of geo- graphic distribution throughout the United States and with respect to rainfall and temperature. Climatic data from the SPS-1 and SPS-2 sites were used to show that design storm parameters typically used to design subsurface drainage sys- tems can be estimated from correlations with more readily known average annual precipitation data. The Thornthwaite moisture index, which describes a loca- tion’s climate in a way that reflects the monthly variations in both precipitation and temperature, was suggested as a better climatic parameter for use in analysis of pavement perfor- mance than average annual precipitation or average annual temperature. In the regressions for roughness and distress in the SPS-1 and SPS-2 pavement sections, the Thornthwaite moisture index was almost always more significant than average precipitation or average annual temperature. C H A P T E R 6 Conclusions and Recommendations

86 Soils The natural drainage characteristics of the soils at the SPS-1 and SPS-2 sites were examined by studying county soil reports and the taxonomic classifications of the predominant soil series at the sites. The soils of the SPS-1 and SPS-2 sites range from very well drained (or, in natural drainage class ter- minology, “somewhat excessively drained” for agricultural purposes) to very poorly drained. Clearly some of the SPS-1 and SPS-2 sites have subgrade soils with such good natural drainage characteristics that any water that entered the pave- ment structure would be able to flow downward quickly into the subgrade without accumulating in the vicinity of the constructed base. Some other SPS-1 and SPS-2 sites have subgrade soils with such poor natural drainage characteristics that water that entered the pavement structure would proba- bly flow downward very slowly, if at all, and could accumu- late in and around the base unless drained laterally. Whether such accumulation occurs to a degree sufficient to detrimen- tally affect pavement performance is a separate question. Speaking very generally, the types of soils in the continen- tal United States that appear most likely to be classified as somewhat poorly to very poorly drained are the aquic (wet) suborders of the alfisol, ultisol, mollisol, and histosol soil or- ders. Alfisols are fertile but poorly draining soils that are com- monly found in a broad swath from the Great Lakes region down through the lower Mississippi Valley. Ultisols are low- nutrient clays found throughout much of the southeastern United States. Mollisols are the soft, dark, grassland soils that cover much of the Midwest and Great Plains; the wet subor- der of mollisols are found in parts of the Great Lakes region and northeastern Great Plains, as well as in some parts of Florida. Histosols are organic soils found in wetlands such as those along the Gulf of Mexico, the Atlantic coast, and the upper Midwest. Each of these soil orders has dry suborders as well, describing soils with much better drainage characteris- tics than those in the wet suborders. Thus, it is best not to assume too much about the drainage characteristics of a soil based on its general geographic location. Rather, the county soil report and taxonomic classification should be studied to determine the natural drainage class and other drainage- related properties of the soil at the specific place where a pave- ment is located or is to be located. Age and Traffic The SPS-1 sites ranged in age from 9 to 12 years at the time of this analysis. The SPS-1 sites for which traffic data were available had accumulated flexible pavement ESALs that ranged from less than 1 million to slightly more than 7 mil- lion. All but one of the sites, though, had less than 5 million accumulated flexible pavement ESALs. The SPS-2 sites ranged in age from 9 to 13 years. Of the SPS-2 sites for which traffic data were available, one had less than 1 million accumulated rigid pavement ESALs, and the rest had between 4 million and 24 million accumulated rigid pavement ESALs. (Traffic data are not available for six of the 18 SPS-1 sites or for three of the 14 SPS-2 sites.) An unfortu- nate consequence of so many of the SPS-1 sites having been located on lower volume roads is that the findings related to the performance of the SPS-1 flexible pavements are less reliably applicable to high traffic volume conditions than are findings related to the performance of the SPS-2 rigid pavements. Permeable Base Drainage System Flow Testing A procedure for field testing the rate of flow of water through a permeable base with edgedrains and outlets was developed for this study and used to test the flow times of the permeable base sections at all of the SPS-1 and SPS-2 sites. At many of the sites, the outlet headwalls were unmarked and were obscured by tall vegetation. At some sites, the outlet headwalls were also completely covered by dirt, gravel, and other vegetation that had to be dug out with hand tools. Even outlets that were fairly easy to locate usually had to be cleared out with hand tools before the drainage flow time testing could be conducted. It should be noted that, for the purposes of the flow testing, the in situ conditions were modified by the clearing of the outlets at many of the test sections. The degree to which the drainage outlets at the various SPS-1 and SPS-2 sites are or are not maintained is presumed to be typical of the level of maintenance conducted for drainage outlets on other, nonexperimental pavements. For drained test sections with measurable outflow, the cal- culated drainage system permeabilities ranged from 6,000 to more than 32,000 ft/day. It should be remembered, however, that (a) this reflects not only the permeability of the perme- able asphalt-treated base layer, but also the movement of water through the edgedrain and outlets, and (b) the actual magnitude of the drainage system permeability calculated is a function of an assumed width of the flow plume away from the core hole. The calculated drainage system permeabilities are perhaps better considered as indicators of the relative functioning of the drainage systems at the different sites. Furthermore, for the purpose of evaluating the permeability of the permeable asphalt-treated base material, independent of the flow rate through the edgedrain and outlet, the better measure obtained from the field testing might be the steady- state infiltration rate in gallons per minute. At all but two sites, water outflow from the drainage outlets occurred in at least one drained test section, even at some sites with subgrade soils classified as well drained to somewhat

excessively drained. This suggests that once a pavement struc- ture is constructed, water in a drainable base layer may find that lateral outflow is less restricted than downward flow even though the natural drainage characteristics of the subgrade soil, if exposed, would be conducive to downward flow. On the other hand, the lack of outflow for some drained test sections, along with observations of the fairly pristine condi- tion of the outlets, suggests that in such test sections, water has never moved laterally from the base into the edgedrains and to the outlets, but rather has flowed downward into the subgrade soil. Results from Deflection Analysis All of the deflection data collected on the SPS-1 and SPS-2 test sections, from the time of their construction to the upload of the data in Release 19 of the LTPP database, was retrieved from the LTPP database and analyzed for this study. For the sake of comparisons among the different test sections, the analysis results for the first set of deflection data are presented in this report. The results from the first set of deflection data and the most recent set of deflection data are also compared. While differences between the backcalculation results from the first year of testing and the most recent year of testing were observed, the differences were sufficiently variable that no statistically significant increase or decrease was detectable, for either the drained or the undrained test sections. For the flexible pavement sections in the SPS-1 experi- ment, a two-layer analysis procedure was used to determine the in-place elastic modulus of the subgrade and an elastic modulus of the pavement structure (all layers combined) above the subgrade, using deflections measured at load levels closest to 9,000 lb and normalized to 9,000 lb, and also normalized with respect to temperature. For the purpose of comparing the relative structural capacities of the pavement sections at each site, the actual total pavement thickness and backcalculated pavement modulus were used to calculate an equivalent thickness for a fixed asphalt concrete modulus of 500,000 psi. Similarly, for the rigid pavement sections in the SPS-2 experiment, a two-layer analysis procedure was used to determine the in-place dynamic k value of the subgrade and an elastic modulus of the pavement structure, above the sub- grade, using deflections measured at load levels closest to 9,000 lb and normalized to 9,000 lb. As part of this deflection analysis, slab size corrections were applied to the radius of rel- ative stiffness and the deflection under the load plate. Data on the temperature gradient through the concrete slab at the time of deflection testing was used to identify and correct for possible loss of contact between the concrete slab and under- lying base. For the purpose of comparing the relative struc- tural capacities of the pavement sections at each site, the actual total pavement thickness and backcalculated pavement modulus were used to calculate an equivalent thickness for a fixed concrete modulus of 5,000,000 psi. The weakest pavement sections in the SPS-1 experiment were found to be those with undrained, untreated aggregate bases, and the strongest pavements were found to be those with undrained, dense-graded asphalt-treated bases. The undrained sections with asphalt-treated base over aggregate and the drained sections with permeable asphalt-treated base over aggregate or asphalt-treated base over permeable asphalt-treated base fell between those two in terms of pavement strength (as indicated by backcalculated effective pavement thickness). The pavement sections in the SPS-2 experiment that were expected to be the weakest—those with untreated dense aggregate base—did not, in fact, have backcalculated effective pavement thicknesses much different than the pavement sec- tions with permeable asphalt-treated base. This may be because the concrete slab modulus is so much greater than the modulus of either of these two base types that it domi- nates the calculation of the effective pavement thickness. On the other hand, the effective thickness of the pavement sec- tions with lean concrete base was, in most cases, notably greater than the effective thickness of otherwise comparable pavement sections with the other two types of base. In short, the drained sections with permeable asphalt-treated base were not much more rigid than the sections with undrained aggregate base, but the undrained sections with lean concrete base were notably more rigid than the drained sections with permeable asphalt-treated base. SPS-2 deflection data were also analyzed to assess joint load transfer measurements from deflections measured with the load plate on the approach and leave sides of the joint. Con- trary to the conventional wisdom, but consistent with the finding of another recent analysis of LTPP deflection load transfer data (31), leave-side load transfer values were found to be greater than approach-side load transfer values, and the average difference between the two was found to be statisti- cally significant. The difference between approach and leave load transfer was found to be insensitive to slab temperature, which suggests that it is unrelated to the magnitude of joint opening. Leave-side load transfer values were selected for use in this study for comparisons among the three different base types in the SPS-2 experiment. Cumulative frequency distributions of load transfer values in the first year of deflection testing and in the most recent year of testing showed some decrease in load transfer associ- ated with all three base types. In fact, the greatest increase in the percentage of joints with load transfer at levels below 80% occurred in the sections with drained permeable asphalt- treated base. The percentage of joints with poor load transfer, however, remains low for all three base types, even after 10 or 87

88 more years of service and more than 15 million accumulated ESALs at some sites. It is not too surprising that only a small percentage of joints associated with any of the treatments are exhibiting poor load transfer, considering that the joints in nearly all of the SPS-2 sections are dowelled. Overall, after some 10 years in service and considerable truck traffic at many of the SPS-2 sites, load transfer values in the pavement sections with undrained aggregate base and undrained lean concrete base are no worse than in the pavement sections with drained permeable asphalt-treated base. Results from Performance Analysis Factors Affecting Distress and Roughness in SPS-1 Flexible Pavements Long-term IRI values for the SPS-1 flexible pavements were found to be more strongly correlated to initial IRI val- ues than to all of the SPS-1 experimental factors combined. The next most influential factors were found to be age, back- calculated equivalent thickness of the pavement structure, Thornthwaite moisture index, and average annual precipita- tion. The base type/drainage factors showed very little corre- lation to long-term IRI values. The cumulative frequency distributions for long-term IRI and change in IRI (latest minus initial) were all fairly similar for the five different base type/drainage combinations in the SPS-1 experiment. The SPS-1 sections without a dense-graded asphalt-treated base layer (undrained aggregate and drained permeable asphalt-treated base over aggregate) were found to have both higher initial IRI values and higher long-term IRI values than the SPS-1 sections with a dense-graded asphalt- treated base layer (undrained asphalt-treated base, undrained asphalt-treated base over aggregate, and undrained asphalt- treated base over permeable asphalt-treated base). The largest changes in IRI tended to occur in the undrained aggregate base sections, followed by the drained permeable asphalt- treated base sections, followed by the three groups of undrained sections with an asphalt-treated base layer. Whatever fairly minor effect the base type/drainage factor has had on the development of roughness in the SPS-1 pave- ment sections is concluded to be due to differences in base stiffness and not differences in drainage. Furthermore, the differences observed in IRI by base type are not entirely due to different rates of change in IRI over the time that the pave- ment sections have been in service, because there is evidence that the pavements with weaker bases, both drained and undrained, tended to be rougher initially than the pavements with stiffer bases. Although some pavement sections developed unusually high rutting at a young age (less than 6 years), the vast majority of the SPS-1 pavement sections do not appear to have started to develop increased rutting with increasing age and traffic. This is probably related to the low levels of truck traffic at most of the SPS-1 sites. Rutting levels were similar for all of the base type/drainage combinations, except that the undrained aggregate base group had a higher percentage of sections with unusually high rutting at an early age. Base stiff- ness, rather than drainage, is concluded to be the aspect of the base type/drainage experimental factor that is responsible for whatever role this factor has played in the development of rutting in the SPS-1 test sections. About half of the SPS-1 test sections in each base type/ drainage group have not yet developed any cracking. Among those that have, the weaker pavement sections (undrained ag- gregate base and drained permeable asphalt-treated base) have more cracking than the stronger pavement sections (undrained asphalt-treated base, undrained asphalt-treated base over aggregate, and drained asphalt-treated base over permeable asphalt-treated base). Whatever minor effect the base type/drainage factor has had on cracking in the SPS-1 pavement sections to date is concluded to be due to differ- ences in base stiffness, rather than differences in drainage. Factors Affecting Distress and Roughness in SPS-2 Rigid Pavements Long-term IRI values for the SPS-2 rigid pavements were more strongly correlated to initial IRI values than to any of the other factors considered, although the correlation between long-term IRI and initial IRI was not as strong for the SPS-2 pavements as it was for the SPS-1 pavements. Long- term IRI values for the pavement sections with the two undrained base types in the SPS-2 experiment (dense-graded aggregate and lean concrete) were similar and both higher than the long-term IRI values for the pavement sections with drained permeable asphalt-treated base. Changes in IRI over time were also greatest in the undrained dense-graded aggre- gate and lean concrete sections, followed by the drained permeable asphalt-treated base group. The SPS-2 experiment includes a small number of test sec- tions with undrained hot-mix asphalt concrete base and undrained cement-aggregate mixture base. The sections with undrained hot-mix asphalt concrete base had the highest median initial IRI values, but exhibited smaller changes in IRI than sections in the drained permeable asphalt-treated base group, and as a result had long-term IRI values similar to the sections in the drained permeable asphalt-treated base group. The sections with undrained cement-aggregate mixture base had the smallest changes in IRI and the lowest long-term IRI values of any of the base type/drainage combinations in the SPS-2 experiment. Whatever effect the base type/drainage factor has had on the SPS-2 pavement sections’ latest observed IRI values and

rates of change in IRI over time is concluded to be due pre- dominantly to differences in base stiffness. The potential ef- fect of drainage is not necessarily ruled out, but no particular evidence was detected for the role of drainage, independent of the role of base stiffness, in the development of roughness in the SPS-2 pavements. Furthermore, the differences in IRI by base type are not entirely attributable to different rates of change in IRI over the time that the SPS-2 pavements have been in service, because there is evidence of some significant differences in initial IRI values by base type. Of the three base types in the main SPS-2 experiment, the lean concrete base was associated with the highest initial IRI values, while the dense aggregate base was associated with the highest long- term IRI values. The SPS-2 sections with undrained lean concrete base and drained permeable asphalt-treated base have developed very similar levels of joint faulting, while the sections with undrained aggregate base have developed more faulting. This is true whether undowelled SPS-2 sections are included in or excluded from the comparisons. Of the eight undowelled pavement sections in the SPS-2 experiment, two have an undrained aggregate base, four have a drained permeable asphalt-treated base, one has an undrained lean concrete base, and one has an undrained hot- mix asphalt concrete base. Among these eight sections, the two with the highest faulting levels (2.9 mm and 3.0 mm), were the undowelled sections with aggregate base at the Arizona SPS-2 site (which has a Thornthwaite moisture index of –51 and subgrade soils that are naturally well drained to somewhat excessively drained). Faulting in the remaining six undowelled sections (at the Arizona, North Dakota, and Washington sites) ranged from 0 to 1.3 mm. These findings, particularly the similarity of results for undrained lean concrete base and drained permeable asphalt- treated base, suggest that whatever effect the base type/ drainage factor has had on the development of faulting in pavements in the SPS-2 experiment has been due to the stiffness of these bases compared with the lesser stiffness of the undrained dense-graded aggregate base. This conclusion is reinforced by the observation that the undowelled pave- ments with aggregate base developed more than twice as much faulting as undowelled pavements with drained or undrained stabilized bases, even those at the same sites. More than 60% of the SPS-2 sections with undrained lean concrete base have developed some cracking, while only about 30% of both the undrained aggregate base sections and the drained permeable asphalt-treated base sections have de- veloped some cracking. At one of the SPS-2 sites (the Nevada site), excessive drying shrinkage and premature slab cracking during or shortly after construction occurred in many of the test sections, particularly the sections with lean concrete base. However, more cracking occurred in the lean concrete base sections than in the aggregate and permeable asphalt-treated base sections at several other SPS-2 sites and did not appear to be construction related. The stiffest base type in the SPS-2 experiment, lean concrete base, may have been good for performance in terms of rough- ness and faulting, but it had a pronounced detrimental effect on cracking performance, particularly in the thinner concrete slabs in the experiment. Sections with the weakest base type, undrained aggregate base, also had more cracking than sec- tions with drained permeable asphalt-treated base. On the other hand, sections with undrained hot-mix asphalt concrete and cement-aggregate mixture bases had even less cracking than sections with drained permeable asphalt-treated base. As with the other SPS-2 performance measures, while the design of the experiment makes it difficult to rule out a potential effect of drainage on the development of cracking in the SPS-2 pavement sections, the above findings suggest that the differ- ences in cracking observed to date are due not to drainage dif- ferences but to differences in base stiffness. Final Comments and Recommendations This report began with the observation that pavement engineers have for many decades observed that an excess of water in pavement structures can accelerate certain types of distress in both AC and PCC pavements. It is undoubtedly true that poor subsurface drainage was detrimental to the performance of many pavements built in the United States in the decades following World War II, and that many of these pavements would have benefited from subsurface drainage systems. In many ways, the pavements built in the United States today, particularly those on Interstate highways and U.S. routes, are less vulnerable to the detrimental effects of exces- sive moisture than pavements built in the past. Flexible pave- ments are now built with thicker AC surface layers and thicker, usually stabilized, base layers. Rigid pavements are built with thicker PCC slabs, usually with dowelled joints and with stabilized base layers, and with better-quality aggregates. So while subsurface drainage systems may still be needed to achieve good performance in some pavements in some places, it appears to be far less true than it was 20 or more years ago that subsurface drainage systems are needed to achieve good performance in most pavements in most places. Consider, as an analogy, tire chains, which were routinely used to improve tire traction in snow throughout the north- ern United States, up through the late 1960s and early 1970s. Tire chains function as well today as they ever did, and they are still used in some locations in heavy snow conditions. But they are no longer needed for most wintertime driving because of improvements in tire tread, vehicle, and winter 89

90 maintenance technologies. Similarly, it is not that pavement subsurface drainage systems do not work (although some- times that is the case), but rather that many of the pavements being built today do not need them as much as many of the pavements built decades ago needed them. The analyses conducted for this study did not identify any aspect of the behavior or performance of the AC and PCC pavement structures in the SPS-1 and SPS-2 experiments that could be shown to have been improved by the presence of subsurface pavement drainage. What does appear to have influenced every aspect of pavement behavior and perfor- mance analyzed for these pavements—namely, deflection response, roughness, rutting, faulting, and cracking—is not the drainability of the base layers used but rather their stiff- ness. Overall, the best-performing pavements in the SPS-1 experiment were those with the stiffest bases (incorporating a dense-graded asphalt-treated base layer), whether drained or undrained. The best-performing pavements in the SPS-2 experiments were those with bases that were neither too weak (untreated aggregate) nor too stiff (lean concrete). These include not only the sections with drained permeable asphalt- treated concrete base, but also the sections with undrained hot-mix asphalt base and cement-aggregate mixture base. It is still important, however, for pavement engineers to be able to identify situations in which a pavement subsurface drainage system is likely to be necessary and cost-effective. Based on the findings from this study, the following recom- mendations are made for investigating the need for subsur- face drainage. • The Thornthwaite moisture index and monthly precipita- tion data are recommended for use in identifying sites with year-round or seasonal excesses of available moisture. • County soil reports and soil taxonomy information are recommended for use in identifying subgrade soils with poor natural drainage characteristics. • At sites with wet climates and poorly draining soils, the need for a subsurface drainage system should be consid- ered; this is particularly true for pavement designs likely to be vulnerable to moisture-related distress, including thin asphalt and thin concrete pavements on untreated aggre- gate base layers, especially when, in the case of jointed concrete pavements, the joints are to be undowelled. Evi- dence of adverse effects of poor drainage in nearby pave- ments (for example, pumping, faulting, potholes) should also be considered, as should local experience. Even in such situations, however, it is recommended that the possibility be considered that a stiffer base layer (for PCC pavements, not too stiff a base layer) may be a more cost-effective de- sign improvement than a subsurface drainage system.