Optimality of Highway Pavement Strategies in Canada

Transcription

1 TRANSPORTATON RESEARCH RECORD Optimality of Highway Pavement Strategies in Canada BRUCE HUTCHNSON, FRED P. NX, AND RALPH HAAS The flexible pavement performance predictions of a deterioration model proposed by Small, Winston, and Evans are compared with those estimated from a deterioration model developed for Ontario conditions. The proposed model by Small et al., incorporating even a very modest rate of annal environmentally indced degradation, severely nderestimates the initial service lives of Ontario pavements. The life-cycle cost characteristics of Ontario pavements are illstrated, and it is conclded that the optimal pavement strategy is insensitive to changes in the initial pavement thickness. This conclsion is in contrast to the "nderbilt" conclsions by Small et al. Typical average and marginal cost fnctions that mst form the basis of any rational axle weight-based road ser charge system are also presented. Several years ago, Small and Winston (1) and Small et al. (2) presented an elegant analysis of the optimality of highway pavement strctral designs on the U.S. primary highway system. They conclded with respect to optimal pavement drability that Or analysis of pavement drability, sing standard economic techniqes and a new statistical analysis of road test data, sggests that a sbstantial increase in drability cold be achieved at modest cost and wold lower the total costs of bilding and maintaining pavements over their life cycle. Their conclsion abot the optimality of pavement designs was based heavily on their reanalysis of the AASHO Road Test data. They arged that "AASHO's fnctional specifications and statistical estimation of the coefficients... were seriosly flawed." Hdson et al. (3) reviewed the analyses condcted by Small et al. (2) and performed additional analyses of the performance of the RoadTest rigid pavement sections. These new analyses by Hdson et al. inclded data obtained from the sbseqent long-term monitoring of the srviving Road Test sections that had been incorporated into the llinois nterstate highway network. They conclded that the Small/Winston analysis significantly nderestimated the life of thick rigid pavements. The original AASHO method overestimated the life of thick rigid pavements; bt not as significantly as the Small and Winston srvival analysis nderestimated the life. Becase of the lack of distress in the Road Test for thicker rigid pavements, the Small and Winston srvival analysis of AASHO Road Test rigid sections is not valid. (3) Many more of the flexible pavement sections failed at the AASHO Road Test, and the Small and Winston reanalysis of the flexible pavement performance data sggested only relatively small increases in pavement thickness to achieve optimal drability. Small and Winston noted that B. Htchinson and R. Haas, Department of Civil Engineering, University of Waterloo, Waterloo, Ontario, Canada N2L 3Gl. F.P Nix, Orangeville, Ontario, Canada L9W 2Y8.. se of the AASHQ eqations acconts for only abot one-third of the difference between optimal and crrent design (thicknesses); the rest mst be de to failre to incorporate economic optimization into design procedre. (J) The prpose of this paper is to highlight some of the findings of a stdy on road costs condcted for the recently completed Royal Commission on National Passenger Transportation in Canada on road costs ( 4). One of the major qestions addressed in this stdy of Canadian road costs was the "nderbilt thesis" advanced by Small et al. (2). The paper first addresses the isse of pavement deterioration models and compares the behaviors of flexible pavements estimated by a model developed for Ontario conditions with that developed by Small et al. from the AASHO Road Test, which contained some modifications for environmentally introdced pavement deterioration. The second part of the paper analyzes the life-cycle cost characteristics of pavements and identifies optimal pavement strategies. t also partitions the life cycle costs into those cased by vehicle damage and those cased by environmental degradation. The final part of the paper presents representative average and marginal cost fnctions along with their implications for heavy vehicle pricing. TYPCAL FLEXBLE PAVEMENT DESGNS Most of the pavements on the major highway system in Canada con- sist of flexible pavements. Smmarized in Table 1 are the characteristics of representative flexible pavement designs for primary highways in three provinces of Canada. The thicker pavements reqired in New Brnswick reflect the poorer-qality sbgrades in that province. Alberta has the smallest range of thicknesses becase most of the pavement deterioration is cased by the harsh winter climate rather than by traffic. The thinner base corse thickness in Alberta reflects the se of asphalt-stabilized base corses becase of the navailability of good granlar base corse material. The last row of Table 1 shows the range of eqivalent granlar thicknesses of the pavement sing granlar thickness eqivalencies for the srface of 2 and for the sbbase of.67. PAVEMENT DETERORATON n Canada pavement srface qality is sally expressed in terms of the riding comfort index (RC) rated on a 1-point scale instead of the 5-point scale of the U.S. present serviceability index (PS). New pavements typically have an initial RC of 8.5, and first-class highway pavements are normally considered to have deteriorated to an nacceptable condition when the RC has decreased to 4.5.

2 112 TRANSPORTATON RESEARCH RECORD 1455 TABLE 1 Typical Flexible Pavement Strctres Layer New Brnswick Ontario Alberta asphaltic concrete base corse (granlar or asphalt stabilized*) so granlar sb-base eqivalent granlar thickness entries are in mm Deterioration Model of Small and Colleages Small et al. (2) proposed the following pavement deterioration model on the basis of their reanalysis of the AAS HO Road Test data and the incorporation of a term to accont for environmental degradation: RC(t) = RCl(O) - [RC(O) - RC(f)] ( ) em' (1) where RCl(t) = RC at time t, RCl(O) = initial RC magnitde, - RC (f) = the RC magnitde at failre, Q =:the annal nmber of ESAL coverages; N = the total eqivalent single axle load (ESAL) coverages expected from a pavement before it deteriorates to the terminal serviceability magnitde RC (f), and m =a parameter to accont for the annal decrease in serviceability cased by climatic factors. The Small et al. (2) reanalysis of the AASHO Road Test data for flexible pavements reslted in the following eqation for predicting N, the cmlative ESAL coverages to failre: N = e12.62 (D + 1)1.161 (L 1 + Li) (Li)3.23s where D = the strctral nmber of a pavement, L 1 = the axle load (in thosands of lbs), and Li 1 for single axles and 2 for tandem axles. Setting L 1 eqal to 18 (i.e., the 18,-lb standard axle load) and Li eqal to 1, the pavement deterioration model in Eqation 1 may be rewritten as (2) llstrated in Figre 1 are the RC profiles calclated from Eqation 3 for D eqal to 4.9 for a range of annal ESAL loadings. The diagram shows that the model does not allow for deterioration in the absence :of axle loads. The RC profiles show that the model of Small et al. (2) predicts failre in abot 21 years, when the pavement experiences 1, ESALs per year, and failre in abot 7 years, for an annal ESAL loading of 5,: f m magnitdes are sed in Eqation 3 sch as those sggested for harsh Canadian winter conditfons by Paterson, say m eqal to 7., then the terminal serviceability of 4.5 wold be reached in abot 13 years instead of 21 years for an annal ESAL loading of 1,. OPAC Deterioration Model The Ontario flexible pavement deterioration model (OPAC) provides one of the few models that separates load- from climateindced deterioration. t was developed from the behavior observed at the AASHO Road Test, the theoretical behavior of layered elastic systems, and the longer-rn Brampton Test Road in Ontario (6) Rilett et al. (7) have discssed some of the characteristics of OPAC with respect to cost allocation stdies. The RC profiles predicted for a flexible pavement with D eqal to 4.9 by sing the OPAC model are illstrated in Figre 2 for the same range of annal ESAL loadings sed in Figre 1. The diagram illstrates that OPAC predicts that pavements deteriorate to an nacceptable condition in abot 4 years in the absence of axle loads. With annal ESAL loadings of 1 million, an RC of 4.5 is reached in abot 18 years [verss 4 years for the model of Small et al. (2)]; "Ui: \ '... ll... ' ) \ \ 5 -il:itil - " k \ K YEARS N SERVCE FGURE 1 RC verss age profiles from the model of Small et al. (2). D 1 x RC() 8 5 [ 4 Qt ] mi. t = (D + 1)7 761 e The vale 4 in Eqation 3 is the difference between RC(O) eqal to 8.5 and RC(f) eqal to 4.5. Eqation 3 can be sed to develop the RC profile of a representative flexible pavement with a strctral nmber D of 4.9. The magnitde of m eqal to 2.3 appears to be representative of the climaticindced deterioration in Canada from the comments made by Small et al. (2). Small et al. (2) sed m eqal to 4. in most of their calclations, and Paterson (5) sggests that an min the range of 5. to 1. might be appropriate for severe climates. There is considerable ncertainty abot the appropriate m magnitde, bt a vale of m of 2.3 is adeqate to illstrate the important featres of Eqation 3. (3) :>< c.::> YEARS N SERVCE FGURE 2 RC verss age profiles from OPAC model. D 1 x D 4

3 Htchinson et al. 113 with an annal ESAL loading of 2 million the terminal serviceability of 4.5 is reached in 14 years [verss abot 2 years for the model of Small, et al. (2)]. Rilett et al. (7) have pointed ot that the relative importance of axle loads and environment as sorces of pavement deterioration change with different intensities of loads. For example, the environment may accont for as mch as 8 percent of the deterioration on low-volme roads (fewer than 25, ESALS per year) and for only abot one-half of the deterioration on heavily loaded roads. Differences in Failre Life Prediction Demonstrated in the Table 2 are the differences in pavement lives predicted by the OPAC model and that proposed by Small et al. (2). The entries in Table 2 show the initial years to failre for a variety of annal axle load intensities by sing both models. The entries confirm the differences between the two models illstrated previosly; the model of Small et al. (2) forecasts are very sensitive to ESAL loading changes. Clearly, the economically optimm pavement strategy wold be mch more sensitive to the key design variable, the initial pavement life, if the model of Small et al. (2) rather than the OPAC model were sed. Small increases in pavement thickness wold have sbstantial payoffs if the model of Small et al. (2) were sed to predict performance. ECONOMC CHARACTERSTCS The economic characteristics of flexible pavements discssed in this section are based on pavement performance behavior forecast by the OPAC model. The pavement costs are based on nit costs for the srface corse of $Can95/mm/two-lane km (US$1 = abot $Can 1 :35) and for the base and sbbase corses of $Can2/mm/ two-lane-km. llstrated in Figre 3 are the well-known scale economies that exist for highway pavements, in which the present worths of total costs reqired to achieve a 15-year and a 2-year initial service life to failre are shown as a fnction of the magnitde of the annal ESAL loading. The pavement performance estimates that form the basis of Figre 3 were obtained from OPAC. Small increases in.pavement constrction costs can accommodate the sbstantial increases in ESAL loadings. llstrated in the Figre 4 is present vale of the constrction, resrfacing, and rotine maintenance costs for two magnitdes of annal ESAL loadings by sing a discont rate of 5 percent.. The life-cycle crves are derived from the initial pavement costs, the resrfacing costs reqired to achieve a 4-year service life (sally TABLE 2 Years in Service Estimated by Two Performance Models Small, Winston & Evans Model Annal ESAL OPAC Load Model m =. m =.23 m =.7 1, , ,, 2 2,, ,, :Z::CZ> i::z::cz> oo 25 2ooo '1lb 15,...ir i---- TC2YEARS E TC S YEARS ESALs N FRST YEAR FGURE 3 nitial constrction cost verss annal ESAL loadings. 3 -r ,---r--r----ir---.-"""-t---, 25 t:j:;;i..,;:iiii..-n=--;- ;;; e i:i... D CAPTAL MC 1 TOTAL 5_,_--1-=o..;ic:::'"+---f ooo oi_j J [:!Jo.o.J NTAL LFE (YEARS) NTAL LFE (YEARS) FGURE 4 Present worth (PW) life-cycle costs verss initial pavement life. two cycles are reqired), and the rotine pavement maintenance costs. The life-cycle cost crves for the high annal axle loadings are qite flat, sggesting that the choice of an initial pavement life is not critical in establishing minimm life-cycle cost thicknesses. For example, in the case of an annal ESAL loading of 2 million, there is only a 2.2 percent difference between the highest and the lowest costs. The life-cycle cost crve for annal ESALs of25, does rise continally, sggesting that the optimal strategy may be to bild pavements with short lives. However it shold be noted that the costs sed in Figre 4 do not inclde the delay costs to traffic indced by pavement maintenance and reconstrction operations. Nix et al. ( 4) provide additional analyses of the life-cycle costs of flexible pavements in Canada. They conclded that for the prpose of developing costing procedres for Canadian roads there is no evidence that pavements are being bilt with less than optimm

4 114 '- TRANSPORTATON RESEARCH RECORD 1455 drability. Any initial pavement life of abot 15 years seems to be optimal in minimizing total life-cycle costs. Use of the deterioration model of Small et al. (2) as the basis for the economic analysis wold yield very different conclsions abot the optimality f flexible pavement designs in Ontario. Total lifecycle costs wold be higher, and the rate of change of costs with initial pavement costs with changes in initial pavement life for each level of ESAL loading wold be higher. llstrated in Figre 5 is the present worth of the life-cycle pavement costs for an initial pavement life of 15 years for different annal ESAL loading magnitdes. The following fnction has been fitted to the data: C = 89, ,214 log (ESALs) (4) where C is the present worth of the life cycle costs. llstrated in Figre 6 is the marginal pavement cost (MC) fnction derived from Eqation 4, which has the following eqation: MC = 23,214 x _1_ ESALs lnlo The average cost fnction is also shown in Figre 6, and becase of the scale economies it falls above the marginal cost fnction. The implications of these marginal cost fnctions for three representative Ontario trck types and for three ESAL volmes by sing two methods for converting axle loads to ESALs are illstrated in Table 3. The pper nmber in each cell ses the AASHTO ESAL eqivalencies, whereas the lower nmber ses the fnctions v; en 22 en / v -- 5 loooooo _i FGURE 5 Present worth of life-cycle costs verss annal ESAL loadings. :;! wen!) < FGURE 6 Marginal and average pavement damage costs per ESAL verss annal ESAL loadings. (5) TABLE 3 Marginal Pavement Costs for Several Ontario Trck Types Annal ESAL Load ntensity Trck Type Low Medim High (5,) (5,) (2,,) 3 axle (25 t) axle tractor-semi (39 t) axle B-train (62 t) entries are cents per kilometre reported by Rilett and Htchinson ( 8). The marginal costs listed in Table 3 are for the traffic volmes listed at the head of each traffic volme colmn and are calclated from Eqation 5. The Table 3 entries show the large variations in the marginal costs per kilometer between trck types and across road types, for example, $.186/km for a 3-S3-S2 operating on a road with low annal ESAL loadings verss $.23/km for the trck when it operates on a road with high annal ESAL loadings. Charging trcks for load-associated pavement damage at the long-rn marginal cost wold not recover total pavement damage costs, and a second charge wold be reqired to recover flly the remainder of the damage costs, the nonload-associated costs, and the other costs associated with the provision of highway infrastrctres. CONCLUSONS The initial service lives of flexible pavements forecast by the model of Small et al. (2) containing a modest annal environment-indced degradation are mch shorter than those observed in Ontario. The more rapid decrease in RC estimated by the model of Small et al. (2) is probably cased by its se of the AASHO Road Test data and the accelerated loading of the pavements in this Road Test. Analyses of the life-cycle costs of flexible pavements by the OPAC degradation model for a range of annal ESAL loadings showed that weak minima existed. This means that optimal pavement strategies in Ontario are not sensitive to the isse of initial pavement drability (thickness) as sggested by Small et al. (2). The strong economies of scale present in flexible pavements prodce long-rn marginal pavement costs per ESAL that are mch lower than the long-rn average costs. This means that ESAL-kilometer charges for loaded trc\cs based on long-rn marginal costs wold not flly recover life-cycle costs. Other mechanisms sch as Ramsey pricing, higher vehicle registration fees, charges for externalities, and charges for capacity consmption wold be reqired to recover the additional costs of providing highways. REFERENCES 1. Small, K. A., and C. Winston. Optimal Highway Drability. The American Economic Review, Vol. 78, No. 3, 1988, pp Small, K. A., C. Winston, and C. A. Evans. Road Work: A New Highway Pr.icing and nvestment Policy. The Brookings nstittion, Washington, D.C., Hdson, W.R., M. T. McNerney, and T. Dossey. A Comparison and Re-Analysis of the AASHO Road Test Rigid Pavement Data. Presented at 7th Annal Meeting of the Transportation Research Board, Washington, D.C

5 Htchinson et al Nix, F. P., M. Bocher, and B. G. Htchinson. Road Costs. Directions, Vol. 3. Final Report. Royal Commission on National Passenger Transportation, Ottawa, Ontario, Canada, 1992, pp Paterson, W. D.. Road Deterioration and Maintenance Effects: Models for Planning and Management. The. World Bank, The Johns Hopkins University Press, Jng, F. W., R. Kher, and W. A. Phang. A Performance Prediction Sbsystem-Flexible Pavements. Research Report 2. Ontario Ministry of Transportation and Commnications, Downs view, Ontario, Canada, Rilett, L. R., B. G. Htchinson, and R. C. G. Haas. Cost Allocation mplications of Flexible Pavement Deterioration Models. n Transportation Research Record 1215, TRB, National Research Concil, Washington, D.C. 1989, pp Rilett, L. R., and B. G. Htchinson. LEF Fnctions from Canroad Pavement Load-Deflection Data. n Transportation Research Record 1196, TRB, National Research Concil, Washington, D.C. 1988,