Traffic Data Collection, Analysis, and Forecasting for Mechanistic Pavement Design (2004) / Chapter Skim
Currently Skimming:
Pages 1-59
The Chapter Skim interface presents what we've algorithmically identified as the most significant single chunk of text within every page in the chapter.
Select key terms on the right to highlight them within pages of the chapter.
From page 1... ...
Part 1 Traffic Data Collection, Analysis, and Forecasting for Mechanistic Pavement Design
|
From page 2... ...
by dividing by "traffic ratios" rather than by multiplying by "traffic factors." In Section 4.2, it is recommended that, if use is made of partialday classification counts, "truck traffic distribution factors" (TTDFs) not be used for converting these counts to estimates of 24-hour traffic volume by vehicle class; instead, it is recommended that "hourly traffic ratios" (or "hourly fractions")
|
From page 3... ...
for estimates of AADT by vehicle class.3 1-2 3 The original design for the Design Guide software would have required users to provide these CVs. However, the final version of this software makes no use of these CVs.
|
From page 4... ...
Successful communication between these groups hinges, in part, upon better knowledge between both groups as to what traffic data are needed for pavement design, how variations in traffic loads affect pavement design, and how to account for those variations. Consequently, the implementation requirements at both the state and the national level involve the following: • Development and execution of training programs to improve the knowledge of both groups, including both specific training in the TrafLoad software and more general instruction as to how traffic loads affect pavement design; and • Removal of institutional barriers that limit the interaction between pavement design and traffic data-collection and analysis staff.
|
From page 5... ...
The FHWA is the appropriate agency for providing this direction and for encouraging top-down direction within the highway agency itself. 2.2 State-Level Implementation Actions The implementation tasks needed at the state highway agency level relate to the following: • Training personnel, • Changing internal work processes to institutionalize the communication needed to ensure the collection and use of accurate traffic load data, and • Refining the current traffic data-collection and summarization process to improve the quality of the load estimates available for use in the Pavement Design software.
|
From page 6... ...
Members of both groups need to understand how traffic load and variations in that load affect pavement design and how traffic load varies with time and location. Only when both groups understand this interaction can cost-effective decisions be made as to how much traffic data to collect, where and when to collect the data, and how the data should be summarized and processed by the TrafLoad software for use in the Pavement Design Guide software.
|
From page 7... ...
Improvements in communication should also include other groups within the state government. For example, forecasting of future truck volumes and/or loads can be improved if the groups charged with support of statewide economic development are consulted about expected changes in statewide economic activity that might affect truck volumes.
|
From page 8... ...
3.1 Methodology All analyses were conducted using WIM data collected in 2000 by the state of California's traffic monitoring program and stored by the University of California's Pavement Research Center. The analyses used data for 55 sites for which there were at least 8 months of available data that met the checks on consistency of calibration.2 Estimates of AAEPV were developed for each of the 55 sites using a procedure that attempts to minimize the effects of missing days of the week and missing months.3 For each site, separate estimates of AAEPV were developed for each of three groups of FHWA vehicle classes (VCs)
|
From page 9... ...
In addition, several sets of AAEPV estimates were produced using ESAL ratios to "factor" the data. For 48-hour data, this factoring process involved three major steps performed separately for each VC group distinguished: 1.
|
From page 10... ...
This difference is due to differences in the extent of DOW variation in AEPV for SUTs and combinations. The daily ESAL ratios shown in Figure 3.1 reflect statewide average values of AEPV for each day of the week that have been normalized by dividing by annual AEPV.
|
From page 11... ...
1-10 Table 3.1 Errors Produced by Using Short-Duration WIM Data to Estimate Average ESALs per Vehicle Figure 3.1 Daily ESAL Ratios 0.70 0.80 0.90 1.00 1.10 1 2 3 4 5 6 7 Sunday Monday Tuesday Wednesday Thursday Friday Saturday Day of Week ESAL Ratio VC 5 VC 6-7 VC 8-13 Source: Cambridge Systematics, Inc., Accuracy of Traffic Load Monitoring and Projections, Volume II: The Accuracy of ESALs Estimates, prepared for FHWA, February 2003, Figure 5.1. Mean Absolute Percentage Error Vehicle Classes 5 6 and 7 8–13 Unfactored Data 48 hours 8.1% 13.0% 7.3% 7 days 5.7% 10.1% 6.6% Factored Data 48 hours 7.0% 12.7% 6.4% 7 days 5.5% 9.9% 5.7%
|
From page 12... ...
This result implies that, for VCs 8–13, there is a substantial degree of similarity in the month-of-year patterns in AEPV at different sites but relatively little similarity in the DOW patterns. For VCs 6 and 7, factoring produces relatively small (0.2- to 0.3-percent)
|
From page 13... ...
Section 4.3 summarizes the use of simple and weighted averages by TrafLoad, discusses the two cases in which weighted averages are used, and explains why simple averages (rather than weighted averages) are used for developing load spectra for "Truck Weight Road Groups." 4.1 Traffic Ratios versus Traffic Factors A 24-hour traffic count can be converted to an estimate of annual average daily traffic (AADT)
|
From page 14... ...
.2 Factoring procedures that use traffic ratios produce AADT estimates that generally differ only slightly from those produced by the corresponding procedures that use traffic factors. However, to the extent that the estimates differ, those produced using traffic ratios are likely to be the better ones.
|
From page 15... ...
Hence, when using short-duration counts for this month for sites in this group, the AADT estimates produced using MFs will be about 1 percent higher than those produced using MTRs. The information presented also suggests that a "neutral value" of 1.0 for the traffic ratio or factor is likely to be slightly preferable to the actual value of 1.01 for the MF.
|
From page 16... ...
that are analogous to the monthly and day-ofweek traffic ratios used by other TrafLoad factoring procedures.3 This factoring procedure was chosen over the more commonly used truck traffic distribution factor (TTDF) procedure because the latter procedure produces upward biases in truck volume estimates for sites at which most truck traffic is "business-day" truck traffic -- a characteristic of most urban sites.
|
From page 17... ...
79-80. 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Hour of Day 0 Percent 2 4 6 8 10 12 14 Percent Trucks Daily Average
|
From page 18... ...
and 2:00 p.m. 4.3 Simple versus Weighted Averages In several situations, traffic data analysis software, such as TrafLoad, is required to average data obtained from several sites in a group.
|
From page 19... ...
But the highest traffic volumes and truck loads exist only at a few sites, leading to estimates of average pavement load that overestimate loads at most sites.7 For this reason, TrafLoad was designed to use simple averages when developing a set of load spectra for a TWRG. 1-18 4 Although the derivation of weighted averages usually is more complex than the derivation of simple averages, the difference in complexity is sometimes quite small.
|
From page 20... ...
The procedure used for developing these factors uses truck volumes by day of week as weights in the averaging process.9 The process used is designed to develop average monthly load spectra that are not unduly influenced by load characteristics observed on days (usually weekend days) when truck volumes are low.
|
From page 21... ...
In particular, both the Pavement Design Guide procedure and TrafLoad's load spectra factoring procedure place substantial reliance on maintaining WIM calibration over time. In the case of the Design Guide procedure, temporal variations in WIM calibration may result in monthly load distribution factors that reflect the effects of variations in calibration rather than variations in axle load, thus compromising the use of these distribution factors in pavement design.
|
From page 22... ...
Similarly, TrafLoad develops sets of load spectra adjustment factors that are used to modify load spectra collected on certain days of the week and/or certain months so that they better reflect the overall pavement-damaging effects of loads traversing a site on an annual basis or during a particular month. This load spectra factoring procedure is a generalized version of the ESAL factoring procedure described in Chapter 2.0 that was developed and tested as part of a recent study for FHWA.1 These two procedures are the first procedures developed for adjusting WIM data for the effects of seasonal and DOW variations in axle loads.
|
From page 23... ...
These analyses could also provide the basis for modifying procedures for collecting traffic data so as to improve the resulting estimates of the input variables to which the Pavement Design Guide procedures are most sensitive. Averaging Procedures In several situations, systems for analyzing traffic data create averages of data obtained from a specified group of sites, such as a factor group.
|
From page 24... ...
This is the case of TWRGs, in which ESALs or axle-load data from WIM sites in a TWRG are used to provide default values for other sites in the TWRG.4 For most TWRGs, site-specific values of ESALs and axle loads per vehicle for combination trucks tend to be positively correlated with truck volumes, so that weighted averages will produce higher values for ESALs and load spectra than will unweighted averages. However, there is also a tendency to install WIM equipment at sites with relatively high truck volumes.
|
From page 25... ...
The final subsection lists potential improvements whose value is less clear or whose precise design may be better determined after some experience is gained in using TrafLoad. It is assumed that additional useful improvements will be identified by TrafLoad users in the course of using the system.
|
From page 26... ...
It would be desirable for TrafLoad to allow two design lanes at a site to be assigned to TWRGs and seasonal load spectra factor groups independently of each other.7 • Quality Control Checks. The software assumes that the vehicle classification and weight records have been through a quality-control process.
|
From page 27... ...
uses a set of DOW load spectra ESAL ratios.9 For any site, these ratios are developed for any month for which WIM data exist for each day of the week. There is at least one relatively unusual situation (when data are available for only 1 week and that week contains a holiday)
|
From page 28... ...
spectra using weights that vary with the number of observations for the individual months. Also, rules could be developed that would enable TrafLoad to develop DOW ESAL ratios using data for months for which data exist for some, but not all, of the five weekdays.
|
From page 29... ...
TrafLoad The traffic-data analysis software developed under this project. TTDFs Truck traffic distribution factors.
|
From page 30... ...
2A Site for which an AVC count is available for a period of at least 48 hours. 2B Site for which a manual classification count for a minimum of 6 weekday hours is available.
|
From page 31... ...
3 All other WIM sites.
|
From page 32... ...
For this reason, it usually will be better to develop separate growth rates for SUTs and CTs. However, for conciseness, the forecasting procedures presented in this appendix frequently refer simply to truck volumes and truck AADT.
|
From page 33... ...
Trend analysis formulas are very simple to develop and apply and can be contained in a single spreadsheet. A.1.1 The Procedures The simplest form of trend analysis is a linear regression method that forecasts future truck volumes based solely on historical truck volumes, developing a trend line of volumes into the future.
|
From page 34... ...
In particular, data on truck volumes on a given road usually are not collected annually, and even if they are, the factoring process used to convert short-duration truck counts to estimates of truck AADT introduces artifact into the resulting time series that adversely affects the regression results. When an exponential trend analysis is performed using truck VMT, the growth rate that is estimated for VMT on an entire set of roads usually is assumed to be valid for any road in the set.
|
From page 35... ...
. Table A.1 Historical Truck Volumes Year of Observation Number of Years Since First Observation Truck VMT (Billions)
|
From page 36... ...
Using this result, forecast VMT in 2022 is estimated as The same data can also be used to estimate an exponential trend or growth rate. Here the regression is performed using the natural log of truck VMT as the dependent variable and the number of years from 1992 as the independent variable.
|
From page 37... ...
When using a linear growth rate with TrafLoad, the software requires only this growth rate, 14.17 CTs per year. A.2 More Sophisticated Approaches This section describes three more sophisticated approaches for forecasting truck volumes: • Multivariate linear regression, • Growth-factors methods, and • Travel demand models.
|
From page 38... ...
. 1 Cambridge Systematics, Inc., et al., NCHRP Report 388: A Guidebook for Forecasting Freight Transportation Demand, Transportation Research Board, National Research Council, 1997.
|
From page 39... ...
. .+ × + × = billion Table A.2 Historical Truck Volumes and Grain Production Year of Observation Number of Years Since First Observation Truck VMT Billions Grain Production (Billion Bushels)
|
From page 40... ...
The conduct of such surveys is discussed briefly in Section A.2.4 (below) and discussed further in the Quick Response Freight Manual.6 3.
|
From page 41... ...
The annual growth factors for each industry sector are assumed to have been obtained from employment forecasts produced by a local economic development agency. The growth factors are applied to the base-year truck volumes using Equation A.6, with n, the number of years in the forecast period, set to 23 (2022 − 1999)
|
From page 42... ...
West Virginia West Virginia employs a growth-factor approach to forecast truck traffic. Growth factors are applied to baseline truck traffic data in order to project future truck activities.
|
From page 43... ...
Statewide forecasts of employment by industry group were used as the basis for predicting county-level growth factors. A.2.3 Travel Demand Models A well-established and relatively sophisticated technique for forecasting traffic volumes for an entire region is through the use of travel demand models (TDMs)
|
From page 44... ...
New Jersey New Jersey includes a vehicle-based truck model as part of the statewide travel demand model.9 The truck model develops truck trips for individual zones using trip rates for zonal 9 URS Greiner Woodward Clyde, Statewide Model Truck Trip Table Update Project, prepared for the New Jersey Department of Transportation, January 1999.
|
From page 45... ...
Forecasts for truck trips at major intermodal facilities were developed using forecasts from terminal operators, and external station forecasts were developed using trend analysis. Indiana The Indiana statewide freight model10 was developed by Professor William Black at Indiana University in the 1990s.
|
From page 46... ...
Until recently, the Bureau of Labor Statistics and the Bureau of Economic Analysis produced forecasts of several economic variables by state and industry at 2.5- and 5-year intervals.18 However, these forecasts are no longer available. Allocation of Truck Volumes to Commodities and Industries Vehicle Inventory and Use Survey The Vehicle Inventory and Use Survey (VIUS)
|
From page 47... ...
Other Sources Other sources of data that may be used for allocating truck volumes to commodity/industry groups are the following:
|
From page 48... ...
in each TAZ. Extensive information about truck-trip generation is contained in the Quick Response Freight Manual, which includes a compilation of truck-trip generation rates from a number of studies.
|
From page 49... ...
Exponential Growth Rate The formula for deriving an exponential annual growth rate using estimates of AADT of SUTs or CTs for any 2 years is (A.7) where: Yo is the base year, Yf is the future year, To and Tf are the corresponding estimates of AADT of SUTs or CTs (or any other variable of interest)
|
From page 50... ...
Changes in truck size and weight limits can have significant effects, in either direction, on truck configurations used and the load spectra of these configurations. Each of these three types of change in vehicle use will affect the load spectra of affected vehicle classes, and changes in size and weight limits also can have significant effects on the number of trucks in affected vehicle classes.
|
From page 51... ...
Except in the special case of changes in size and weight limits that have already been enacted but that were not in effect during the most recent year for which historical data are available, it is not possible to predict with confidence whether such changes will increase or decrease pavement stresses. Accordingly, the inability of the current version of the Pavement Design Guide software to use forecast changes in load spectra is of little consequence, and addressing this limitation should be a low priority.
|
From page 52... ...
By construction, these mean values of AADT are identical to the AADT estimates produced by a combined seasonal/DOW factoring procedure. 1 ERES Consultants and FUGRO-BRE, Draft Report, prepared for NCHRP Project 1-37A, 2000, pp.
|
From page 53... ...
The CV resulting from the use of factors that are derived using data from an entire factor group can be estimated as follows: (B.2) Similarly, assume that separate seasonal and DOW factors are being used, with the seasonal factors derived using data from n Level 1A sites and the DOW factors derived using data from m Level 1A sites (which may or may not overlap the first n sites)
|
From page 54... ...
For this reason, factoring generally is performed using data that come only from sites with relatively high truck volumes. Random variation can also be a problem for Level 1A sites at which a significant portion of total truck traffic is influenced by a small number of decision-makers.
|
From page 55... ...
Procedures for using a partial-day classification count to estimate full-day truck volumes 3 The alternative of creating a separate factor group for the lower functional systems may be considered. However, this alternative would increase data collection costs and, because of the random variation in truck volumes at sites with relatively low truck volumes, could produce even poorer estimates of AADT.
|
From page 56... ...
In addition to the errors discussed in the preceding subsection, AADT estimates developed for Level 2B sites incorporate errors that result from the conversion of partial-day counts to estimates of full-day truck volumes. There are two types of site: 1.
|
From page 57... ...
AADT at Level 1B sites is estimated by obtaining a set of short-duration classification counts at the site and applying factors obtained from the associated Level 1A site. Errors in the AADT estimates at Level 1B sites can be caused by the following: equipment malfunction at either the site or the associated Level 1A site, some of the inherent limitations of the factoring process (such as the use of an average "November/Tuesday" factor for any Tuesday in November)
|
From page 58... ...
The second kind of Level 3B site is on a planned new road. For these sites, estimates of total traffic and truck volumes are, at best, developed from travel demand models.
|
From page 59... ...
The goal of producing conservative pavement designs suggests that it would be preferable for the Pavement Design Guide software to adopt the latter option, i.e., to assume that the errors for all CT classes are perfectly correlated. Though the discussion here focused on Level 2 sites, this assumption appears to be appropriate for Level 1 and 3 sites as well.
|
Key Terms
This material may be derived from roughly machine-read images, and so is provided only to facilitate research.
More
information on Chapter Skim is available.