AP-G84/04 Best practice in road use data collection, analysis ... - WIM
AP-G84/04 Best practice in road use data collection, analysis ... - WIM
AP-G84/04 Best practice in road use data collection, analysis ... - WIM
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
<strong>AP</strong>-<strong>G84</strong>/<strong>04</strong><br />
<strong>Best</strong> <strong>practice</strong> <strong>in</strong> <strong>road</strong> <strong>use</strong> <strong>data</strong> <strong>collection</strong>,<br />
<strong>analysis</strong> and report<strong>in</strong>g
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
BEST PRACTICES IN ROAD USE DATA COLLECTION,<br />
ANALYSIS AND REPORTING
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
First Published 20<strong>04</strong><br />
© Aust<strong>road</strong>s Inc. 20<strong>04</strong><br />
This work is copyright. Apart from any <strong>use</strong> as permitted under the Copyright Act 1968,<br />
no part may be reproduced by any process without the prior written permission of Aust<strong>road</strong>s.<br />
National Library of Australia<br />
Catalogu<strong>in</strong>g-<strong>in</strong>-Publication <strong>data</strong>:<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
ISBN 0 85588 721 4<br />
Aust<strong>road</strong>s Project No.BS.A.N.520<br />
Aust<strong>road</strong>s Publication No. <strong>AP</strong>-<strong>G84</strong>/<strong>04</strong><br />
Project Manager<br />
Flori Mihai, Ma<strong>in</strong> Roads Western Australia<br />
Prepared by<br />
James Y K Luk, Charles Karl, Tim Mart<strong>in</strong><br />
ARRB Transport Research Ltd<br />
Peer Review Group<br />
Anthony Brown, Roads & Traffic Authority, NSW<br />
Raoul Casagrande, VicRoads<br />
Ray Daltrey, Roads & Traffic Authority, NSW<br />
Geoff Clarke, Department of Transport and Regional Services<br />
Jean<strong>in</strong>e Diederich, Ma<strong>in</strong> Roads Western Australia<br />
Ray Dyer, Department of Infrastructure, Energy & Resources, Tasmania<br />
Geoff Horni, Department of Infrastructure, Plann<strong>in</strong>g & Environment, NT<br />
Andrew Kra<strong>use</strong>, Department for Transport, Urban Plann<strong>in</strong>g and the Arts, SA<br />
Roger McLeay, Transit New Zealand<br />
David Pratt, Roads & Traffic Authority, NSW<br />
Geoff Smith, Queensland Department of Ma<strong>in</strong> Roads<br />
Tim Strickland, VicRoads<br />
Published by Aust<strong>road</strong>s Incorporated<br />
Level 9, Robell Ho<strong>use</strong><br />
287 Elizabeth Street<br />
Sydney NSW 2000 Australia<br />
Phone: +61 2 9264 7088<br />
Fax: +61 2 9264 1657<br />
Email: aust<strong>road</strong>s@aust<strong>road</strong>s.com.au<br />
www.aust<strong>road</strong>s.com.au<br />
This guide is produced by Aust<strong>road</strong>s as a general guide. Its application is discretionary. Road<br />
authorities may vary their <strong>practice</strong> accord<strong>in</strong>g to local circumstances and policies.<br />
Aust<strong>road</strong>s believes this publication to be correct at the time of pr<strong>in</strong>t<strong>in</strong>g and does not accept<br />
responsibility for any consequences aris<strong>in</strong>g from the <strong>use</strong> of <strong>in</strong>formation here<strong>in</strong>. Readers should<br />
rely on their own skill and judgement to apply <strong>in</strong>formation to particular issues.
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
BEST PRACTICES IN ROAD USE DATA COLLECTION,<br />
ANALYSIS AND REPORTING<br />
Sydney 20<strong>04</strong>
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
Aust<strong>road</strong>s profile<br />
Aust<strong>road</strong>s is the association of Australian and New Zealand <strong>road</strong> transport and traffic<br />
authorities whose purpose is to contribute to the achievement of improved Australian and<br />
New Zealand <strong>road</strong> transport outcomes by:<br />
♦ undertak<strong>in</strong>g nationally strategic research on behalf of Australasian <strong>road</strong> agencies and<br />
communicat<strong>in</strong>g outcomes<br />
♦ promot<strong>in</strong>g improved <strong>practice</strong> by Australasian <strong>road</strong> agencies<br />
♦ facilitat<strong>in</strong>g collaboration between <strong>road</strong> agencies to avoid duplication<br />
♦ promot<strong>in</strong>g harmonisation, consistency and uniformity <strong>in</strong> <strong>road</strong> and related operations<br />
♦ provid<strong>in</strong>g expert advice to the Australian Transport Council (ATC) and the Stand<strong>in</strong>g<br />
Committee on Transport (SCOT).<br />
Aust<strong>road</strong>s membership<br />
Aust<strong>road</strong>s membership comprises the six state and two territory <strong>road</strong> transport and traffic<br />
authorities and the Commonwealth Department of Transport and Regional Services <strong>in</strong> Australia,<br />
the Australian Local Government Association and Transit New Zealand. It is governed by a council<br />
consist<strong>in</strong>g of the chief executive officer (or an alternative senior executive officer) of each of its<br />
eleven member organisations:<br />
♦ Roads and Traffic Authority New South Wales<br />
♦ Roads Corporation Victoria<br />
♦ Department of Ma<strong>in</strong> Roads Queensland<br />
♦ Ma<strong>in</strong> Roads Western Australia<br />
♦ Department of Transport and Urban Plann<strong>in</strong>g South Australia<br />
♦ Department of Infrastructure, Energy and Resources Tasmania<br />
♦ Department of Infrastructure, Plann<strong>in</strong>g and Environment Northern Territory<br />
♦ Department of Urban Services Australian Capital Territory<br />
♦ Commonwealth Department of Transport and Regional Services<br />
♦ Australian Local Government Association<br />
♦ Transit New Zealand<br />
The success of Aust<strong>road</strong>s is derived from the collaboration of member organisations and others <strong>in</strong><br />
the <strong>road</strong> <strong>in</strong>dustry. It aims to be the Australasian leader <strong>in</strong> provid<strong>in</strong>g high quality <strong>in</strong>formation, advice<br />
and foster<strong>in</strong>g research <strong>in</strong> the <strong>road</strong> sector.
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
EXECUTIVE SUMMARY<br />
The objective of Aust<strong>road</strong>s Project BS.A.N.520 is to <strong>in</strong>vestigate the consistency of <strong>data</strong> on traffic<br />
load<strong>in</strong>g and traffic mix. Previous work delivered from the project <strong>in</strong>cludes a review of local and<br />
overseas <strong>practice</strong>s <strong>in</strong> <strong>road</strong> <strong>use</strong> <strong>data</strong> <strong>collection</strong>, and identification of the <strong>use</strong> of current and future<br />
<strong>road</strong> <strong>use</strong> <strong>data</strong>. The f<strong>in</strong>al task, and the aim of this report, is to develop a best <strong>practice</strong> report on the<br />
<strong>collection</strong>, <strong>analysis</strong> and report<strong>in</strong>g of <strong>road</strong> <strong>use</strong> <strong>data</strong>.<br />
An Aust<strong>road</strong>s Road Use Data Workshop was held <strong>in</strong> October 2003 to share <strong>in</strong>formation amongst<br />
Road Authorities (RAs), address issues of <strong>data</strong> <strong>in</strong>consistencies, and reach agreement on the<br />
contents of this report. The Workshop confirmed that RAs carry out their <strong>road</strong> <strong>use</strong> <strong>data</strong> tasks with<br />
similar methodologies and are quite consistent <strong>in</strong> key aspects <strong>in</strong>clud<strong>in</strong>g the calculation of Annual<br />
Average Daily Traffic (AADT) and Vehicle-Kilometre Travelled (VKT). Each <strong>road</strong> authority has<br />
developed its own traffic <strong>data</strong> count<strong>in</strong>g program over many years and these programs would have<br />
achieved good <strong>practice</strong>s <strong>in</strong> some ways.<br />
It is <strong>in</strong>tended that the materials <strong>in</strong> this report would further improve consistency <strong>in</strong> <strong>road</strong> <strong>use</strong> <strong>data</strong><br />
<strong>collection</strong>, <strong>analysis</strong> and report<strong>in</strong>g amongst RAs. The contents of this report are as follows:<br />
General pr<strong>in</strong>ciples <strong>in</strong> identify<strong>in</strong>g best <strong>practice</strong>s (Section 2),<br />
Technologies and <strong>practice</strong>s for traffic detection and classification (Section 3),<br />
Key elements of a <strong>road</strong> <strong>use</strong> count<strong>in</strong>g program (Section 4),<br />
Data <strong>in</strong>tegration <strong>in</strong>clud<strong>in</strong>g the issue of public-transport partnership (Section 5),<br />
Data <strong>in</strong>tegrity and deal<strong>in</strong>g with miss<strong>in</strong>g <strong>data</strong> (Section 6), and<br />
Stakeholder consultation (Section 7).<br />
The pr<strong>in</strong>ciples of best <strong>practice</strong> <strong>in</strong>clude accuracy, effectiveness, efficiency, reliability, accessibility,<br />
transparency, timel<strong>in</strong>ess and relevance. They also relate to the issue of <strong>data</strong> <strong>in</strong>tegrity. The issue of<br />
transparency deserves special attention beca<strong>use</strong> it is a key means at enhanc<strong>in</strong>g consistency <strong>in</strong><br />
<strong>data</strong> <strong>collection</strong>, <strong>analysis</strong> and report<strong>in</strong>g amongst RAs and facilitate accessibility of current and<br />
potential <strong>use</strong>rs of <strong>road</strong> <strong>use</strong> <strong>data</strong>.<br />
Another key <strong>data</strong> <strong>in</strong>tegrity issue is the method of manag<strong>in</strong>g miss<strong>in</strong>g <strong>data</strong>. It is recognised that<br />
each RA has its own procedure <strong>in</strong> handl<strong>in</strong>g miss<strong>in</strong>g <strong>data</strong>, and that there is no standard way of<br />
address<strong>in</strong>g the issue. This report provides the current procedures adopted <strong>in</strong> RTA NSW and Ma<strong>in</strong><br />
Roads WA as examples of good <strong>practice</strong>s. In particular, the software developed by RTA-CSIRO<br />
us<strong>in</strong>g the method of Hidden Markov Model could prove <strong>use</strong>ful amongst RAs after the necessary<br />
tests are completed.<br />
The topics of vehicle detection and classification are also discussed. Most issues related to these<br />
topics have been addressed over many years. It is generally accepted that the Aust<strong>road</strong>s 12-b<strong>in</strong><br />
classification by axle configuration is stable and well received s<strong>in</strong>ce its <strong>in</strong>ception. M<strong>in</strong>or variations<br />
are recommended as follows:<br />
Introduction of an error b<strong>in</strong> (b<strong>in</strong> 13) to account for the quality of classified counts;<br />
Distribution of vehicles <strong>in</strong> the error b<strong>in</strong> <strong>in</strong>to either b<strong>in</strong> 1 (the passenger car b<strong>in</strong>) or across all<br />
twelve b<strong>in</strong>s depend<strong>in</strong>g on the error b<strong>in</strong> size, or us<strong>in</strong>g both methods mak<strong>in</strong>g <strong>use</strong> of knowledge<br />
of on-site traffic conditions;<br />
Standardisation of classify<strong>in</strong>g vehicles <strong>in</strong>to four or five b<strong>in</strong>s by vehicle lengths;<br />
Recognition of m<strong>in</strong>or variations amongst RAs but sub-classified counts by either axle<br />
configurations or vehicle lengths must be capable of aggregat<strong>in</strong>g <strong>in</strong>to the Aust<strong>road</strong>s system<br />
(13 b<strong>in</strong>s by axles or 4 or 5 b<strong>in</strong>s by lengths).<br />
The issue of vehicle classification by vehicle lengths <strong>in</strong>to three, four or five b<strong>in</strong>s as an <strong>in</strong>tegral part<br />
of the Aust<strong>road</strong>s classification system is discussed. The current <strong>practice</strong>s amongst RAs are<br />
presented but have not been harmonised <strong>in</strong> this study.
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
The <strong>use</strong> of a s<strong>in</strong>gle axle sensor for traffic count <strong>collection</strong> is quite common amongst RAs, even<br />
though most have recognised the need for classified counts. It is recommended that traffic counts,<br />
as a default, be reported <strong>in</strong> vehicle numbers. If the raw <strong>data</strong> is <strong>in</strong> the form of axles or axle-pairs,<br />
then the count should be converted to vehicles with the conversion factor also reported if possible,<br />
adopt<strong>in</strong>g the pr<strong>in</strong>ciple of transparency.<br />
The issue of utilis<strong>in</strong>g weigh-<strong>in</strong>-motion (<strong>WIM</strong>) equipment as part of a traffic count<strong>in</strong>g program is also<br />
discussed <strong>in</strong> some detail. As long as all lanes of traffic are monitored at a <strong>WIM</strong> site, the traffic<br />
count <strong>data</strong> (and <strong>WIM</strong> <strong>data</strong>) should add extra value to a count<strong>in</strong>g program. The related issue of<br />
correlat<strong>in</strong>g classified count <strong>data</strong> with vehicle mass <strong>data</strong> is more complex and there is not much<br />
<strong>in</strong>formation published on this topic. In <strong>practice</strong>, if the ratio of loaded and unloaded vehicles at a<br />
count<strong>in</strong>g station is consistent with a nearby <strong>WIM</strong> site, then an accurate estimate of pavement<br />
load<strong>in</strong>g can be determ<strong>in</strong>ed.<br />
Road <strong>use</strong> <strong>data</strong> can come from many sources meet<strong>in</strong>g a variety of needs. This leads to the issues<br />
of <strong>data</strong> <strong>in</strong>tegration and accessibility. Many stakeholders are also <strong>in</strong>volved. This report provides a<br />
stakeholder consultation model from Ma<strong>in</strong> Roads WA. These issues are rather <strong>in</strong>volved, and RAs<br />
at present have quite varied policies on <strong>data</strong> availability and its pric<strong>in</strong>g. As the trend to outsource<br />
<strong>data</strong> cont<strong>in</strong>ues amongst RAs, third parties may become both supplier and purchaser of <strong>road</strong> <strong>use</strong><br />
<strong>data</strong>.<br />
In summary, this report has provided materials on best <strong>practice</strong>s amongst RAs not previously<br />
covered <strong>in</strong> NAASRA (1982) and the recently updated Guide to Traffic Eng<strong>in</strong>eer<strong>in</strong>g Practice Part 3<br />
(Aust<strong>road</strong>s 1988a and 20<strong>04</strong>). In particular, the materials cover topics such as: error b<strong>in</strong> (b<strong>in</strong> 13) <strong>in</strong><br />
the Aust<strong>road</strong>s vehicle classification system by axle configurations, vehicle classification by lengths,<br />
def<strong>in</strong><strong>in</strong>g a homogeneous traffic section, correlation of classified counts with <strong>WIM</strong> <strong>data</strong>, publicprivate<br />
partnership, quality checks, deal<strong>in</strong>g with miss<strong>in</strong>g <strong>data</strong>, calibration of a <strong>WIM</strong> system us<strong>in</strong>g<br />
CULWAY as an example, and a stakeholder consultation model. A list of thirteen<br />
recommendations is summarised <strong>in</strong> the report with future research identified <strong>in</strong> the follow<strong>in</strong>g areas:<br />
Determ<strong>in</strong>e the threshold values for automatic vehicle classification system by lengths;<br />
Develop a model for the correlation of <strong>WIM</strong> <strong>data</strong> with classified counts;<br />
Develop a policy or bus<strong>in</strong>ess model on <strong>data</strong> <strong>in</strong>tegration, accessibility and pric<strong>in</strong>g.<br />
It takes time to achieve consistency amongst RAs <strong>in</strong> this important area of <strong>road</strong> <strong>use</strong> <strong>data</strong>. This<br />
report represents a step <strong>in</strong> this on-go<strong>in</strong>g process.
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
TABLE OF CONTENTS<br />
1 INTRODUCTION ...................................................................................................................1<br />
2 BEST PRACTICE PRINCIPLES............................................................................................2<br />
3 VEHICLE DETECTION AND CLASSIFICATION...................................................................3<br />
3.1 Axle Sensor and Axle Counts .......................................................................................3<br />
3.2 Inductive Loop Sensor and Vehicle Counts..................................................................5<br />
3.3 Vehicle Classification System.......................................................................................6<br />
3.4 Other Considerations..................................................................................................10<br />
4 TRAFFIC COUNTING PROGRAM ......................................................................................13<br />
4.1 Background.................................................................................................................13<br />
4.2 AADT and VKT Estimation .........................................................................................14<br />
4.3 Network Coverage ......................................................................................................23<br />
4.4 Report<strong>in</strong>g ....................................................................................................................29<br />
5 DATA INTEGRATION..........................................................................................................33<br />
5.1 Road Use Data Types for Integration .........................................................................33<br />
5.2 Correlation of Traffic Composition and <strong>WIM</strong> Data ......................................................35<br />
5.3 Public-Private Partnership ..........................................................................................38<br />
6 DATA INTEGRITY ...............................................................................................................39<br />
6.1 Issues .........................................................................................................................39<br />
6.2 Specifications for Data Collection Activities................................................................40<br />
6.3 Quality Check and Deal<strong>in</strong>g with Miss<strong>in</strong>g Data ............................................................41<br />
6.4 Calibration of a <strong>WIM</strong> System ......................................................................................44<br />
6.5 Accuracy Specifications..............................................................................................48<br />
7 STAKEHOLDER CONSULTATION .....................................................................................49<br />
8 CONCLUSIONS ..................................................................................................................51<br />
REFERENCES ...........................................................................................................................53<br />
<strong>AP</strong>PENDIX A – NATIONAL GLOSSARY OF TERMS ................................................................53<br />
<strong>AP</strong>PENDIX B – NUMBER OF AADT OBSERVATIONS .............................................................53<br />
<strong>AP</strong>PENDIX C – EXECUTIVE SUMMARY OF LITERATURE REVIEW REPORT (RC2050-2) ..53<br />
<strong>AP</strong>PENDIX D – EXECUTIVE SUMMARY OF CURRENT AND FUTURE ROAD USE REPORT<br />
(RC2050-3) .................................................................................................................................53<br />
Page
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
TABLES<br />
Table 1 Use of axle sensors under different AADT values......................................................4<br />
Table 2 Current Aust<strong>road</strong>s vehicle classification systems (updated <strong>in</strong> 1994) .........................7<br />
Table 3 Vehicle classification systems currently <strong>in</strong> <strong>use</strong>...........................................................8<br />
Table 4 Vehicle classification systems by lengths...................................................................9<br />
Table 5 Accuracy requirements for <strong>in</strong>dividual sites ...............................................................17<br />
Table 6 Accuracy requirements for groups of <strong>road</strong>s .............................................................17<br />
Table 7 Examples of round<strong>in</strong>g AADT ....................................................................................18<br />
Table 8 Adjustment factors by jurisdictions ...........................................................................19<br />
Table 9 Some adjustment factors from short-term counts to AADT estimates......................20<br />
Table 10 Densities for traffic count<strong>in</strong>g stations........................................................................24<br />
Table 11 National highway traffic count<strong>in</strong>g station density by location and status..................24<br />
Table 12 Groups by functional classes: ..................................................................................26<br />
Table 13 Suggested volume stratification of <strong>road</strong> segments...................................................27<br />
Table 14 Suggested volume stratification based on % change <strong>in</strong> AADT ................................28<br />
Table 15 Complete list of <strong>data</strong> classes collected/computed and stored <strong>in</strong> Queensland .........31<br />
Table 16 Mean axle group and gross vehicle mass ................................................................35<br />
Table 17 RTA NSW rejection rules on classified counts .........................................................42<br />
Table18 High-speed Weigh-<strong>in</strong>-Motion Systems Used and available <strong>in</strong> Australia<br />
(source: Koniditsiotis 2000) ......................................................................................45<br />
Table 19 Accuracy requirements.............................................................................................48<br />
Table B1 Calculation of the Number of VKT Stations..............................................................53<br />
FIGURES<br />
Figure 1 A pair of pneumatic tubes and a vehicle classifier on a local <strong>road</strong> monitor<strong>in</strong>g<br />
two lanes of vehicle traffic and a bicycle path on the nearside of the photograph .....3<br />
Figure 2 Four positions of a two-axle vehicle pass<strong>in</strong>g over two axle sensors ..........................5<br />
Figure 3 Typical response of a loop detector crossed by a passenger car ..............................6<br />
Figure 4 Loop configurations for bicycle detection .................................................................10<br />
Figure 5 A sensor array for separate lane axle counts on a three-lane carriageway .............11<br />
Figure 6 Survey of <strong>road</strong> <strong>use</strong> <strong>data</strong> needs amongst Road Authorities ......................................15<br />
Figure 7 Expected percentage error for AADT from ADT and count duration<br />
(n-Day counts) for a 75% confidence level...............................................................18<br />
Figure 8 A design for traffic count<strong>in</strong>g program .......................................................................22<br />
Figure 9 An <strong>in</strong>tegrated <strong>road</strong> <strong>use</strong> <strong>data</strong>base from Queensland Ma<strong>in</strong> Roads.............................34<br />
Figure 10 Total mass of three-axle prime movers and three-axle semi-trailers.........................36<br />
Figure 11 Total mass of n<strong>in</strong>e-axle B-doubles ............................................................................37<br />
Figure 12 Laden and unladen vehicles <strong>in</strong> each vehicle class at Port Fremantle, WA ...............37<br />
Figure 13 Flow diagram of the CSIRO method for detection of corrupt traffic <strong>data</strong> ...................44<br />
Figure 14 Three axle masses of an articulated six-axle vehicle (Class 9) .................................47<br />
Figure 15 Adjustment factors for diurnal variation of CULWAY axle mass <strong>data</strong><br />
(Grundy et al. 2002) .................................................................................................47<br />
Figure 16 Adjustment factors for seasonal variation of CULWAY axle mass <strong>data</strong><br />
(Grundy et al 2002) ..................................................................................................47
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
ABBREVIATIONS<br />
(see also Appendix A)<br />
AADT Annual Average Daily Traffic<br />
AAWT Annual Average Weekday Traffic<br />
AASHTO American Association of Highway and Transportation Officials<br />
ADT Average Daily Traffic<br />
AWDT Average Weekday Daily Traffic<br />
AWT Average Weekday Traffic<br />
DAM Drive Axle Mass<br />
FOI Freedom of Information<br />
GIS Geographic Information System<br />
GPS Global Position<strong>in</strong>g System<br />
GTEP Guide to Traffic Eng<strong>in</strong>eer<strong>in</strong>g Practice<br />
HMM Hidden Markov Model<br />
I<strong>AP</strong> Intelligent Access Program<br />
ITS Intelligent Transport System<br />
LoS Level of Service<br />
MAWDT Monthly Average Weekday Traffic<br />
MRWA Ma<strong>in</strong> Roads Western Australia<br />
OH&S Occupational Health and Safety<br />
PCS Permanent Count<strong>in</strong>g Station (or Site)<br />
RA, RAs Road Authority, Road Authorities<br />
RTA NSW Roads and Traffic Authority of New South Wales<br />
SAM Steer<strong>in</strong>g Axle Mass<br />
SCATS Sydney Coord<strong>in</strong>ated Adaptive Traffic System<br />
SCS Short-term Count<strong>in</strong>g Station<br />
TAM Tandem Axle Mass<br />
TRB Transportation Research Board<br />
UTS Uniform Traffic Section<br />
VicRoads Roads Corporation Victoria<br />
VKT Vehicle-Kilometre Travelled<br />
<strong>WIM</strong> Weigh-In-Motion<br />
WRS Weighted Road Segments
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
1 INTRODUCTION<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 1 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
The objective of Aust<strong>road</strong>s Project BS.A.N.520 is to <strong>in</strong>vestigate the consistency of <strong>data</strong> on traffic<br />
load<strong>in</strong>g and traffic mix. Previous work delivered from the project <strong>in</strong>cludes a review of local and<br />
overseas <strong>practice</strong>s <strong>in</strong> <strong>road</strong> <strong>use</strong> <strong>data</strong> <strong>collection</strong> (Mart<strong>in</strong> and Chiang 2002), and identification of the<br />
current and future <strong>use</strong> of <strong>road</strong> <strong>use</strong> <strong>data</strong> (Roper et al. 2002). The Executive Summaries of those<br />
two Reports are <strong>in</strong>cluded <strong>in</strong> Appendix C and D respectively. The f<strong>in</strong>al task, and the aim of this<br />
report, is to develop a best <strong>practice</strong> report on the <strong>collection</strong>, <strong>analysis</strong> and report<strong>in</strong>g of <strong>road</strong> <strong>use</strong><br />
<strong>data</strong>.<br />
An Aust<strong>road</strong>s Road Use Data Workshop was held <strong>in</strong> October 2003 to share <strong>in</strong>formation amongst<br />
Road Authorities (RAs), address issues of <strong>data</strong> <strong>in</strong>consistencies, and reach agreement on the<br />
content of this report (Luk and Karl 2003).<br />
The Workshop confirmed that RAs carry out their <strong>road</strong> <strong>use</strong> <strong>data</strong> tasks with similar methodologies<br />
and are quite consistent <strong>in</strong> key aspects <strong>in</strong>clud<strong>in</strong>g the calculation of Annual Average Daily Traffic<br />
(AADT) and Vehicle-Kilometre Travelled (VKT). The methodologies have largely followed the<br />
NAASRA (1982) pr<strong>in</strong>ciples. It is recognised that there are differences <strong>in</strong> apply<strong>in</strong>g these pr<strong>in</strong>ciples<br />
and there are jurisdictional requirements and constra<strong>in</strong>ts. Further, each <strong>road</strong> authority has<br />
developed its own traffic <strong>data</strong> count<strong>in</strong>g program over many years and the program would have<br />
been ref<strong>in</strong>ed, i.e. already achiev<strong>in</strong>g good <strong>practice</strong>s <strong>in</strong> some ways.<br />
It is <strong>in</strong>tended that the materials <strong>in</strong> this report would further improve consistencies <strong>in</strong> <strong>road</strong> <strong>use</strong> <strong>data</strong><br />
<strong>collection</strong>, <strong>analysis</strong> and report<strong>in</strong>g amongst RAs. The contents of this report are as follows:<br />
General pr<strong>in</strong>ciples <strong>in</strong> identify<strong>in</strong>g best <strong>practice</strong>s (Section 2),<br />
Technologies and <strong>practice</strong>s for traffic detection and classification (Section 3),<br />
Key elements of a <strong>road</strong> <strong>use</strong> count<strong>in</strong>g program (Section 4),<br />
Data <strong>in</strong>tegration <strong>in</strong>clud<strong>in</strong>g the issue of public-private partnership (Section 5),<br />
Data <strong>in</strong>tegrity and deal<strong>in</strong>g with miss<strong>in</strong>g <strong>data</strong> (Section 6), and<br />
Stakeholder consultation (Section 7).
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
2 BEST PRACTICE PRINCIPLES<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 2 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
This section provides a brief discussion on the pr<strong>in</strong>ciples of best <strong>practice</strong>. The aim is to provide a<br />
b<strong>road</strong> context <strong>in</strong> mak<strong>in</strong>g recommendations <strong>in</strong> this report. These pr<strong>in</strong>ciples also relate to <strong>data</strong><br />
<strong>in</strong>tegrity issues, which will be addressed <strong>in</strong> more detail <strong>in</strong> Section 6. They are as follows:<br />
Accuracy – This is a measure of how well the <strong>data</strong> collected represent true values. In traffic<br />
<strong>data</strong>, perfect accuracy is unnecessary and it is actually impossible to achieve.<br />
Consequently, it is good <strong>practice</strong> to quote a percentage error and the associated confidence<br />
level. A trade-off between accuracy with other quality criteria is also necessary – higher<br />
accuracy often implies higher cost.<br />
Effectiveness – This is the likelihood that a work program achieves desired objectives. It<br />
measures how closely outcomes match predef<strong>in</strong>ed targets.<br />
Efficiency – This is the ratio of output to <strong>in</strong>put. A lot of output <strong>data</strong> with m<strong>in</strong>imal <strong>in</strong>put<br />
resources is an efficient <strong>data</strong> <strong>collection</strong> program. However, collect<strong>in</strong>g the wrong <strong>data</strong> is an<br />
<strong>in</strong>effective program irrespective of the amount of <strong>in</strong>put resources <strong>in</strong>volved.<br />
Reliability – This is related to accuracy. A count<strong>in</strong>g program needs to produce consistent and<br />
repeatable results. Reliability is therefore an outcome of how well a <strong>road</strong> network is sampled<br />
or covered to produce statistically consistent <strong>data</strong>.<br />
Accessibility – Road <strong>use</strong> <strong>data</strong> should be easily accessible with<strong>in</strong> a jurisdiction and to the<br />
public due to the Freedom of Information (FOI) Act, possibly not <strong>in</strong> the raw <strong>data</strong> form but <strong>in</strong><br />
aggregated form. The <strong>data</strong> structure and formats should be amenable to <strong>data</strong> accessibility.<br />
Transparency – A clear enunciation of the procedures for <strong>collection</strong>, <strong>analysis</strong> and report<strong>in</strong>g<br />
should help reduc<strong>in</strong>g differences amongst RAs and other <strong>use</strong>rs <strong>in</strong> the <strong>use</strong> and <strong>in</strong>terpretation<br />
of <strong>road</strong> <strong>use</strong> <strong>data</strong>.<br />
Timel<strong>in</strong>ess – This relates to the frequency of <strong>data</strong> <strong>collection</strong>, its report<strong>in</strong>g and updates. A<br />
yearly update and report<strong>in</strong>g should be the m<strong>in</strong>imal update period. With more and better<br />
<strong>data</strong>base technologies, it is expected that <strong>road</strong> <strong>use</strong> <strong>data</strong> would become more timely <strong>in</strong> the<br />
near future.<br />
Relevance – The <strong>data</strong> collected should meet the needs of a <strong>use</strong>r, i.e. should fulfil the<br />
pr<strong>in</strong>ciple of fitness of purpose. There is always a great need for traffic counts and weigh-<strong>in</strong>motion<br />
(<strong>WIM</strong>) <strong>data</strong>. A current issue is what other types of <strong>data</strong> should be <strong>in</strong>cluded <strong>in</strong> a<br />
comprehensive <strong>road</strong> <strong>use</strong> <strong>data</strong>base.<br />
The above list is certa<strong>in</strong>ly not exhaustive but does cover most of the important elements <strong>in</strong><br />
identify<strong>in</strong>g best <strong>practice</strong>s. These pr<strong>in</strong>ciples will be applied <strong>in</strong> the follow<strong>in</strong>g sections to identify and<br />
recommend <strong>practice</strong>s <strong>in</strong> the f<strong>in</strong>al <strong>road</strong> <strong>use</strong> <strong>data</strong> report.<br />
Further, a glossary of terms is <strong>in</strong>cluded <strong>in</strong> Appendix A. The glossary is <strong>in</strong>tended to be a ‘national<br />
glossary’ to promote consistency. It is recognised that <strong>in</strong>dividual jurisdictions would adopt slight<br />
variations <strong>in</strong> def<strong>in</strong>itions and therefore <strong>practice</strong>s due to constra<strong>in</strong>ts specific to a jurisdiction. Over<br />
time, the national glossary also requires updates and is a reference towards which jurisdictional<br />
<strong>practice</strong>s would migrate.
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
3 VEHICLE DETECTION AND CLASSIFICATION<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 3 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
The Aust<strong>road</strong>s (1988a and 20<strong>04</strong>) Guide to Traffic Eng<strong>in</strong>eer<strong>in</strong>g Practice (GTEP) Part 3 on Traffic<br />
Studies provides a b<strong>road</strong> background to vehicle detection and classification and other traffic<br />
studies. Discussions on technologies such as image process<strong>in</strong>g and Global Position<strong>in</strong>g System<br />
(GPS) will only be mentioned <strong>in</strong> pass<strong>in</strong>g <strong>in</strong> this report. The focus of this section is on gett<strong>in</strong>g the<br />
most out of currently <strong>use</strong>d technologies, which are axle and <strong>in</strong>ductive loop sensors.<br />
3.1 Axle Sensor and Axle Counts<br />
The most common type of axle sensor is the pneumatic tube with an air switch. It is suitable <strong>in</strong><br />
both urban and rural environments for temporary axle counts. Two tubes are usually needed for<br />
speed measurement and vehicle classification. A two-tube <strong>in</strong>stallation for vehicle classification on a<br />
local <strong>road</strong> is shown <strong>in</strong> Figure 1. The photograph shows the set-up for the monitor<strong>in</strong>g of two lanes of<br />
vehicular traffic and a bicycle path on the near side of the photograph.<br />
Figure 1 – A pair of pneumatic tubes and a vehicle classifier on a local <strong>road</strong> monitor<strong>in</strong>g<br />
two lanes of vehicle traffic and a bicycle path on the nearside of the photograph<br />
A s<strong>in</strong>gle axle sensor has been found <strong>use</strong>ful <strong>in</strong> a wide range of <strong>road</strong> traffic conditions. For s<strong>in</strong>gle<br />
lane traffic, one axle sensor usually presents no problems <strong>in</strong> obta<strong>in</strong><strong>in</strong>g axle counts. At remote sites,<br />
s<strong>in</strong>gle axle sensors are sometimes <strong>use</strong>d <strong>in</strong> permanent count stations beca<strong>use</strong> they are reasonably<br />
reliable and repair costs at remote sites are expensive. For multi-lane traffic (<strong>in</strong> one or two<br />
directions), axle counts are missed when the wheels of two different vehicles arrive at an axle<br />
sensor at the same time. Install<strong>in</strong>g two axle sensors could improve accuracy but the <strong>in</strong>stallation of<br />
two axle sensors also <strong>in</strong>crease the Occupational Health and Safety (OH&S) risk.
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 4 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
With the prevalent <strong>use</strong> of s<strong>in</strong>gle axle sensor for traffic counts, it is important that the number of axle<br />
pairs obta<strong>in</strong>ed be corrected for vehicles with more than two axles. The correction factor is<br />
dependent on traffic composition and site conditions and should be calibrated on-site. Some<br />
guidel<strong>in</strong>es on the correction factor are given <strong>in</strong> GTEP Part 3. For example, if heavy vehicles<br />
account for 12% of the total traffic on an urban highway, the number of axle-pairs obta<strong>in</strong>ed should<br />
be reduced by about 6%, i.e. the correction factor is 0.94 ±0.15 at the 95 % confidence limits.<br />
There are no standard <strong>practice</strong>s amongst RAs regard<strong>in</strong>g when axle pair counts are collected rather<br />
than classified counts. At the Aust<strong>road</strong>s Road Use Data Workshop, there was general agreement<br />
that classified counts be collected wherever possible. As mentioned earlier, a pair of axle sensors<br />
is essential for classified counts <strong>in</strong> spite of higher <strong>in</strong>stallation and ma<strong>in</strong>tenance costs. An axle<br />
sensor pair is also more appropriate than a s<strong>in</strong>gle axle sensor <strong>in</strong> obta<strong>in</strong><strong>in</strong>g bi-directional counts as<br />
traffic volume <strong>in</strong>creases.<br />
Table 1 recommends what could be <strong>use</strong>d under different axle arrangements and AADT values.<br />
In Table 1, the threshold AADT values correspond to different Levels of Service (LoS) for two-lane<br />
rural <strong>road</strong>s and arterial <strong>road</strong>s (GTEP Part 2, Aust<strong>road</strong>s 1988b; TRB 2000). At a LoS of E, a <strong>road</strong><br />
facility is operat<strong>in</strong>g at capacity. Usually a facility is designed for a LoS of C. Delay is frequently<br />
experienced at LoS of D and E. The values <strong>in</strong> Table 1 are for a peak-hour to daily factor of 0.15.<br />
Road environment<br />
Comb<strong>in</strong>ed direction (e.g. two-lane<br />
two-way rural <strong>road</strong>s)<br />
Separate direction (e.g. multi-lane<br />
carriageway at 80 km/h speed limit)<br />
Table 1 - Use of axle sensors under different AADT values<br />
Indicative maximum AADT (veh)<br />
One axle Two axles<br />
9,000 (two lanes)<br />
(Level of Service = D)<br />
8,500 per lane<br />
(Level of Service = C)<br />
15,000 (two lanes)<br />
(Level of Service = E<br />
or at capacity)<br />
11,000 per lane<br />
(Level of Service = D)<br />
On multi-lane highways, temporary surface mounted pneumatic tube sensors may not be suitable<br />
under heavy traffic conditions, poor pavement surface or OH&S problems for field staff. The count<br />
accuracy also decreases with <strong>in</strong>crease <strong>in</strong> traffic flow and the number of lanes that an axle sensor<br />
(or an axle pair) has to cover. In these situations, it could be better to <strong>use</strong> <strong>in</strong>-ground loop sensors<br />
described <strong>in</strong> Section 3.2.<br />
The development of piezo-cables for axle detection cont<strong>in</strong>ues to progress (Luk and Brown 1987;<br />
ARRB TR 2003). The cost is higher than pneumatic tubes but should be able to offer a work<strong>in</strong>g life<br />
similar to that of an <strong>in</strong>-ground <strong>in</strong>ductive loop - four to five years depend<strong>in</strong>g on traffic volume, its<br />
composition and pavement type and condition. Piezo-cables can be <strong>in</strong>stalled lane-by-lane so lanebased<br />
counts are possible. The piezo-sensor needs to cover only half to three-quarters of a lane<br />
width to reduce cost and possibly improve accuracy with ‘cleaner’ axle actuations.<br />
Axle sensors give more dist<strong>in</strong>ct axle actuation times and therefore better accuracy than loops <strong>in</strong><br />
speed measurement and vehicle classification. Figure 2 shows the four tim<strong>in</strong>g pulses for speed<br />
calculation and hence vehicle classification. For very slow or stationary traffic, it is difficult to<br />
separate vehicles us<strong>in</strong>g the axle actuation times T1 to T4 and determ<strong>in</strong>e speeds and classify<br />
vehicles, especially if a vehicle straddles over both axle sensors for a long period of time. Inductive<br />
loops and the related electronic circuitry are designed to monitor loop occupancy (i.e. the time that<br />
a vehicle is on top of a loop) and are more suitable for slow traffic, e.g. urban traffic <strong>in</strong> peak hours.
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 5 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
The ability to detect gaps between adjacent vehicles and obta<strong>in</strong> a valid vehicle count is dependent<br />
on the length of the <strong>in</strong>ductive loop <strong>in</strong> the direction of travel along a lane. Undercount<strong>in</strong>g will occur<br />
when the loop detection zone is too long.<br />
Mendigor<strong>in</strong> et al. (2003) described studies <strong>in</strong> Sydney on us<strong>in</strong>g axle sensors for classified counts <strong>in</strong><br />
urban traffic. Their results suggested that a separation of 4 m between two axle sensors provide<br />
good accuracy and these results are consistent with those obta<strong>in</strong>ed <strong>in</strong> Luk and Besley (1985) us<strong>in</strong>g<br />
different count<strong>in</strong>g equipment. The optimal separation is a balance between equipment accuracy<br />
and fluctuations of vehicle speeds between the two axle sensors. It is dependent on the type of<br />
equipment and the associated software <strong>use</strong>d <strong>in</strong> a study.<br />
T 1<br />
T 3<br />
T 2<br />
T 4<br />
Pair of axle<br />
sensors<br />
Vehicle actuation<br />
pulses<br />
T 1<br />
Figure 2 – Four positions of a two-axle vehicle pass<strong>in</strong>g over two axle sensors<br />
In summary, a pair of axle sensors should be considered where possible to give better accuracy,<br />
better dist<strong>in</strong>ction by direction, and the ability to produce classified counts through axle<br />
configurations. If a s<strong>in</strong>gle axle is <strong>use</strong>d, it is a recommended <strong>practice</strong> to correct for traffic<br />
composition and convert axle-pairs <strong>in</strong>to vehicle counts for report<strong>in</strong>g purposes.<br />
3.2 Inductive Loop Sensor and Vehicle Counts<br />
The operation of a loop sensor requires an alternat<strong>in</strong>g current pass<strong>in</strong>g through the <strong>in</strong>ductive loop.<br />
When a mass of metal such as a vehicle chassis and an eng<strong>in</strong>e passes through the<br />
electromagnetic field of the loop, the <strong>in</strong>ductance of the loop reduces. These reductions are <strong>use</strong>d to<br />
<strong>in</strong>dicate the passage or presence of a vehicle.<br />
Figure 3 shows the typical response of a loop detector when crossed by a passenger car. The<br />
reduction <strong>in</strong> loop <strong>in</strong>ductance depends on the size and metallic content of the vehicle. For example,<br />
a passenger car generates a change of about five per cent for the loops typically <strong>use</strong>d <strong>in</strong> traffic<br />
count<strong>in</strong>g. The <strong>in</strong>ductive loop detector is <strong>use</strong>d extensively for automatic traffic count<strong>in</strong>g, traffic<br />
surveillance and traffic signal control. The <strong>analysis</strong> of <strong>in</strong>ductive loop profiles can also be <strong>use</strong>d for<br />
vehicle classification but is seldom <strong>use</strong>d amongst RAs.<br />
−<br />
T 13<br />
T 3<br />
−<br />
T 2<br />
−<br />
T 34<br />
T 4<br />
−<br />
Time
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 6 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
Figure 3 - Typical response of a loop detector crossed by a passenger car<br />
Two types of loop designs are commonly <strong>use</strong>d by <strong>road</strong> authorities for vehicle count<strong>in</strong>g on a lane of<br />
traffic flow. A loop for count<strong>in</strong>g at mid-block is often a square loop of 2 m × 2 m, and a pair of<br />
these loops at a known distance (4 - 5 m) apart is usually <strong>use</strong>d for speed measurement and<br />
vehicle classification. A second loop design is the SCATS loop for signal operation and traffic<br />
count<strong>in</strong>g. The SCATS loop has a width of 2 m and an overall length of about 4.5 m and is located<br />
near the stopl<strong>in</strong>e of an approach. It provides reasonably accurate counts over a measurement<br />
period of, say, 15 m<strong>in</strong>utes but the count<strong>in</strong>g accuracy is less over the length of a signal cycle of one<br />
to two m<strong>in</strong>utes due to the loop location and length.<br />
For accurate count<strong>in</strong>g, care needs to be taken <strong>in</strong> select<strong>in</strong>g the size, shape and position<strong>in</strong>g of the<br />
loop with<strong>in</strong> the traffic lanes. Over-count<strong>in</strong>g can occur when loops <strong>in</strong> adjacent lanes are too close<br />
and the same vehicle is detected <strong>in</strong> both lanes. On the other hand, motor cycles or small vehicles,<br />
straddl<strong>in</strong>g both lanes may be missed by loops placed too far apart.<br />
The ability of the loop detection circuit to hold the change <strong>in</strong> <strong>in</strong>ductance (or electric charges) allows<br />
a loop sensor to be a presence detector. This property allows a loop sensor to measure volume<br />
and speed at very low speed and is therefore suitable for congested urban traffic conditions.<br />
However, speed measurement us<strong>in</strong>g loops is not as accurate as axle sensors due to more<br />
uncerta<strong>in</strong>ty <strong>in</strong> the actuation tim<strong>in</strong>gs. This also affects the accuracy of vehicle classification us<strong>in</strong>g a<br />
pair of loops. The classification has to be by vehicle lengths and <strong>in</strong> a smaller number of classes or<br />
b<strong>in</strong>s (see Section 3.3). Also, <strong>in</strong>ductive loop detectors are not suitable for unsealed <strong>road</strong>s <strong>in</strong><br />
general.<br />
3.3 Vehicle Classification System<br />
Vehicle classification <strong>data</strong> is important for all transport eng<strong>in</strong>eer<strong>in</strong>g applications. Comprehensive<br />
classification <strong>data</strong> is now possible us<strong>in</strong>g new technologies. Vehicle classification surveys are<br />
concerned with the distribution of traffic <strong>data</strong> <strong>in</strong>to different types of vehicles. Classified <strong>data</strong> can be<br />
<strong>use</strong>d for design and performance of <strong>road</strong> pavements and bridges, traffic capacity and operations<br />
<strong>analysis</strong>, vehicle categories for traffic legislation and regulation, <strong>road</strong> safety studies, and economic<br />
studies.
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 7 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
A number of classification systems are currently <strong>in</strong> <strong>use</strong>. These systems <strong>in</strong>clude the 1994 Aust<strong>road</strong>s<br />
system (Table 2) and a few other systems ma<strong>in</strong>ly for urban traffic, which are also shown <strong>in</strong> Table 3.<br />
Table 2 - Current Aust<strong>road</strong>s vehicle classification systems<br />
(updated <strong>in</strong> 1994)<br />
Level 1 Level 2 Level 3<br />
Length<br />
(<strong>in</strong>dicative)<br />
Axles and<br />
Axle Groups<br />
Vehicle Type<br />
Aust<strong>road</strong>s<br />
Classification<br />
Type Axle Groups Description<br />
LIGHT VEHICLES<br />
Class Parameters<br />
Short<br />
Up to 5.5m 2 1 or 2<br />
Medium<br />
5.5m to 14.5m<br />
Long<br />
11.5m to 19.0m<br />
3, 4 or 5 3<br />
Short<br />
Sedan, Wagon, 4WD,<br />
Utility, Light Van, Bicycle,<br />
Motorcycle, etc.<br />
Short – Tow<strong>in</strong>g<br />
Trailer, Caravan, Boat, etc.<br />
1<br />
2<br />
d1 ≤ 3.2m<br />
and Axles = 2<br />
Groups = 3,<br />
2.1 m ≤ d1 ≤ 3.2m<br />
d2 ≥ 2.1m,<br />
and Axles = 3, 4 or 5<br />
2 2<br />
HEAVY VEHICLES<br />
Two Axle Truck or Bus 3 d1 > 3.2m and Axles = 2<br />
3 2 Three Axle Truck or Bus 4 Axles = 3 and Groups = 2<br />
> 3 2 Four Axle Truck 5 Axles > 3 and Groups = 2<br />
3 3<br />
4 > 2<br />
5 > 2<br />
6<br />
> 6<br />
> 2<br />
3<br />
Medium<br />
Comb<strong>in</strong>ation<br />
> 6 4<br />
17.5m to 36.5m > 6 5 or 6<br />
Long<br />
Comb<strong>in</strong>ation<br />
Over 33 m<br />
> 6 > 6<br />
Three Axle Articulated or<br />
Rigid vehicle & trailer<br />
Four Axle Articulated or<br />
Rigid vehicle & trailer<br />
Five Axle Articulated or<br />
Rigid vehicle & trailer<br />
Six Axle (or more) Articulated<br />
or Rigid vehicle & trailer<br />
'B' Double or<br />
Heavy truck trailer<br />
Double Road Tra<strong>in</strong> or<br />
Heavy truck and trailers<br />
Triple Road Tra<strong>in</strong> or<br />
Heavy truck and three trailers<br />
Def<strong>in</strong>itions: Group: (axle group) - where adjacent axles are less than 2.1m apart<br />
Groups: number of axle groups<br />
Axles: number of axles (maximum axle spac<strong>in</strong>g of 10 m)<br />
d1: distance between first and second axle<br />
d2: distance between second and third axle<br />
6<br />
7<br />
8<br />
9<br />
d1 > 3.2m, Axles = 3<br />
and Groups = 3<br />
d2 ≤ 2.1m,<br />
2.1 m ≤ d1 ≤ 3.2 m<br />
Axles = 4 and Groups > 2<br />
d2 ≤ 2.1m,<br />
2.1 m ≤ d1 ≤ 3.2 m<br />
Axles = 5 and Groups > 2<br />
Axle = 6 and Groups > 2<br />
or Axles > 6 and Groups = 3<br />
10 Groups = 4 and Axles > 6<br />
11<br />
Groups = 5 or 6<br />
and Axles > 6<br />
12 Groups > 6 and Axles > 6
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
Table 3 – Vehicle classification systems currently <strong>in</strong> <strong>use</strong><br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 8 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
Methodology Urban traffic Rural traffic<br />
By axle configuration us<strong>in</strong>g two axle<br />
sensors<br />
By vehicle length us<strong>in</strong>g two <strong>in</strong>ductive<br />
loops<br />
By manual observation, loop<br />
<strong>in</strong>ductance profil<strong>in</strong>g or video imag<strong>in</strong>g<br />
Not suitable for congested <strong>road</strong>s, turn<strong>in</strong>g<br />
traffic, vehicles chang<strong>in</strong>g lanes<br />
3- or 4-b<strong>in</strong> systems<br />
Aust<strong>road</strong>s 12-b<strong>in</strong> system (Table 2)<br />
Suitable for sites where loops can be<br />
embedded <strong>in</strong>to a sealed carriageway;<br />
requires good pavement surface<br />
2-b<strong>in</strong> system (light and heavy vehicles; Table 2) and various vehicle classification<br />
schemes<br />
3.3.1 Current Aust<strong>road</strong>s Vehicle Classification System<br />
As already mentioned, the Aust<strong>road</strong>s vehicle classification system currently <strong>in</strong> <strong>use</strong> is based on<br />
classify<strong>in</strong>g axle configurations. B<strong>in</strong>s 1 and 2 are classified as light vehicles (two axle vehicles with<br />
or without tow<strong>in</strong>g a caravan, trailer, etc.) and b<strong>in</strong>s 3 to 12 are called heavy vehicles. This two-b<strong>in</strong><br />
system is suitable for manual classified counts but now <strong>use</strong>d ma<strong>in</strong>ly for <strong>in</strong>tersection movement<br />
counts.<br />
The 12-b<strong>in</strong> system has been found valuable s<strong>in</strong>ce its <strong>in</strong>ception and has provided <strong>use</strong>ful <strong>data</strong> for all<br />
RAs. The system has provided consistency across jurisdictions and the <strong>data</strong> can also be analysed<br />
over time. There have been requests for changes that <strong>in</strong>clude:<br />
Different jurisdictions require variations to cater for their own needs; sub-classifications with<strong>in</strong><br />
a b<strong>in</strong> would be a solution. Examples are motor cycles <strong>in</strong> b<strong>in</strong> 1 and sand trucks <strong>in</strong> b<strong>in</strong> 9.<br />
These sub-classes can be easily aggregated to provide the standard 12-b<strong>in</strong> system for<br />
consistent report<strong>in</strong>g and trend <strong>analysis</strong>.<br />
New vehicle-axle configurations such as the performance-based standard (PBS) vehicles are<br />
be<strong>in</strong>g developed. These vehicles aim to carry heavier loads with better safety capability and<br />
more <strong>use</strong>r-friendly suspension. It is expected that PBS vehicles would still fit <strong>in</strong>to the 12-b<strong>in</strong><br />
system but their development needs to be monitored. It is anticipated that the migration of<br />
the current heavy vehicle fleet to a PBS regime will take some time.<br />
There are requests to create additional classification b<strong>in</strong>s with<strong>in</strong> Aust<strong>road</strong>s b<strong>in</strong>s 11 and 12<br />
(Type 1 and Type 2 <strong>road</strong> tra<strong>in</strong>s respectively; Roper et al. 2002).<br />
It is therefore recommended that the current Aust<strong>road</strong>s vehicle classification be ma<strong>in</strong>ta<strong>in</strong>ed but<br />
allow<strong>in</strong>g sub-classes as variations of the basic system. This was also endorsed at the Aust<strong>road</strong>s<br />
Road Use Data Workshop held <strong>in</strong> Melbourne. It may be appropriate to <strong>in</strong>troduce a new notation to<br />
designate different jurisdictional versions, e.g. the Aust<strong>road</strong>s (Vic) for the Victoria version.<br />
The issue of <strong>in</strong>troduc<strong>in</strong>g a separate b<strong>in</strong> to collect errors <strong>in</strong> classified counts is discussed below.<br />
3.3.2 The Error B<strong>in</strong><br />
It is good <strong>practice</strong> to always check where the errors come from, e.g. poor <strong>in</strong>stallation,<br />
malfunction<strong>in</strong>g of sensors, etc. Different ways are currently <strong>use</strong>d to treat an error or ‘nonclassifiable’<br />
vehicle count. Error counts can be lumped <strong>in</strong>to a separate error b<strong>in</strong> (say, b<strong>in</strong> 13) that<br />
also acts as a quality check of the classified <strong>data</strong>. Methods for manag<strong>in</strong>g the error counts <strong>in</strong>clude<br />
the follow<strong>in</strong>g possible actions:
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
Put them <strong>in</strong>to b<strong>in</strong> 1 (passenger-car b<strong>in</strong>);<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 9 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
Distribute across the 12 b<strong>in</strong>s <strong>in</strong> proportion to the measured vehicle numbers <strong>in</strong> each b<strong>in</strong> (i.e.<br />
new b<strong>in</strong> j vehicles = old b<strong>in</strong> j vehicles × sum of b<strong>in</strong>s 1 to 13 / sum of b<strong>in</strong>s 1 to 12); and<br />
Discard them.<br />
Each of these methods has its own <strong>in</strong>adequacy. The last method <strong>in</strong> discard<strong>in</strong>g the error counts<br />
would underestimate the true traffic demand <strong>in</strong> a <strong>road</strong> network. The second method of proportional<br />
distribution across all b<strong>in</strong>s may significantly distort the true distribution if the sum of vehicles <strong>in</strong> b<strong>in</strong>s<br />
1 to 12 is small relative to the vehicles <strong>in</strong> the error b<strong>in</strong>. The first method of load<strong>in</strong>g error counts <strong>in</strong>to<br />
the passenger-car b<strong>in</strong> (b<strong>in</strong> 1) would not significantly affect the true count <strong>in</strong> b<strong>in</strong> 1 only if there are<br />
relatively large numbers of vehicles <strong>in</strong> this b<strong>in</strong>.<br />
It is thus recommended that:<br />
(a) An error b<strong>in</strong> (no. 13) be <strong>in</strong>troduced <strong>in</strong>to the Aust<strong>road</strong>s system as good <strong>practice</strong> to <strong>in</strong>dicate the<br />
quality of the classified counts;<br />
(b) The <strong>in</strong>clusion of error counts <strong>in</strong> b<strong>in</strong> 13 <strong>in</strong> the total counts requires some judgement. If the<br />
number of error counts is high relative to the total counts, it is necessary to identify the<br />
reasons for such a situation before error counts are <strong>in</strong>cluded;<br />
(c) If it is deemed appropriate to distribute b<strong>in</strong> 13 vehicles, the two distribution methods<br />
mentioned previously can be considered, or both methods are <strong>use</strong>d and guided by relevant<br />
local knowledge.<br />
(d) The number of error counts should be monitored and, if these rema<strong>in</strong> high relative to the<br />
traffic stream, the reasons for these errors should be identified and the problem rectified.<br />
3.3.3 Vehicle Classification <strong>in</strong> Urban Traffic<br />
As <strong>in</strong>dicated <strong>in</strong> Table 3, loop sensors are more suitable for collect<strong>in</strong>g vehicle classified counts on<br />
congested multi-lane highways. These sensors are usually located mid-block on arterial <strong>road</strong>s and<br />
<strong>data</strong> retrieved remotely us<strong>in</strong>g modems. Another source of classified counts is the freeway<br />
management systems now <strong>in</strong> place <strong>in</strong> capital cities. Loop sensors or virtual loops us<strong>in</strong>g imag<strong>in</strong>g<br />
technologies are <strong>in</strong>stalled at regular <strong>in</strong>tervals (e.g. 500 m) to monitor traffic flow and identify<br />
<strong>in</strong>cidents.<br />
These loop pairs calculate the spot speed and classify vehicles by length. Both 3-b<strong>in</strong> and 4-b<strong>in</strong><br />
systems have been employed <strong>in</strong> various jurisdictions. VicRoads, Ma<strong>in</strong> Roads Queensland and<br />
Transit New Zealand have reported us<strong>in</strong>g a 4-b<strong>in</strong> system. Transit New Zealand and RTA NSW<br />
also <strong>use</strong> three b<strong>in</strong>s to classify vehicles by lengths. Table 4 shows the threshold values of these<br />
systems.<br />
Vehicle class Ma<strong>in</strong> Roads<br />
Queensland<br />
Table 4 – Vehicle classification systems by lengths<br />
Vehicle lengths (m)<br />
VicRoads Transit New Zealand RTA NSW<br />
B<strong>in</strong> 1 (short) < 5.8 < 6.0 < 5.5 11.5<br />
B<strong>in</strong> 4 (comb<strong>in</strong>ation) > 21.2 > 17.5 > 17.0 -
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 10 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
Transit New Zealand’s 3-b<strong>in</strong> system was developed for vehicle classification on motorways us<strong>in</strong>g<br />
video imag<strong>in</strong>g, and found to give adequate <strong>in</strong>formation to handle traffic flows and understand the<br />
makeup of fleet <strong>in</strong> each lane.<br />
A variety of threshold values are therefore be<strong>in</strong>g <strong>use</strong>d for vehicle classification by lengths. It is<br />
recommended that RAs come to agreement on a s<strong>in</strong>gle classification system by lengths. Some<br />
empirical work us<strong>in</strong>g a standard loop design would appear necessary. On reach<strong>in</strong>g agreement, the<br />
classification system <strong>in</strong> Table 2 should <strong>in</strong>corporate a def<strong>in</strong>itive multi-b<strong>in</strong> system by vehicle lengths<br />
as a component of the total Aust<strong>road</strong>s classification system. This harmonisation should facilitate<br />
equipment manufactures to design their equipment for a s<strong>in</strong>gle specification.<br />
3.4 Other Considerations<br />
The follow<strong>in</strong>g aspects of traffic count<strong>in</strong>g/classify<strong>in</strong>g require further discussion:<br />
(a) Bicycles – Cycl<strong>in</strong>g is now recognised as an important mode of <strong>road</strong> transport. Its count<strong>in</strong>g<br />
and therefore management <strong>in</strong> a mixed stream of traffic is a challeng<strong>in</strong>g task (see also GTEP<br />
Part 3). It can be detected by <strong>in</strong>ductive loops, pneumatic tubes or piezoelectric cables.<br />
Lesch<strong>in</strong>ski (1994) reported that the standard symmetripole loop (configuration 1 <strong>in</strong> Figure 4)<br />
for signal operation is not suitable for bicycles and that the best configuration is an<br />
asymmetric loop slant<strong>in</strong>g at a 45o angle to the direction of traffic (configuration 4 <strong>in</strong> Figure 4).<br />
Golden River (2001) also reported the feasibility of separat<strong>in</strong>g bicycles from other traffic by<br />
measur<strong>in</strong>g the actual tyre contact width.<br />
The detection of bicycles (and pedestrians) is an area of on-go<strong>in</strong>g research, and a recent<br />
study can be found <strong>in</strong> FHWA (2003). This study tested a range of technologies <strong>in</strong>clud<strong>in</strong>g<br />
loop, <strong>in</strong>frared, video, microwave and magnetic field <strong>in</strong> a controlled environment. The results<br />
show that most of these technologies are promis<strong>in</strong>g, but the challenge is really <strong>in</strong> undertak<strong>in</strong>g<br />
these tests <strong>in</strong> a real-world, mixed traffic environment.<br />
Figure 4 - Loop configurations for bicycle detection
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 11 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
(b) Vehicle concatenation – When the space between two consecutive vehicles is small, it<br />
becomes difficult to separate the two vehicles us<strong>in</strong>g, say, a standard two axle sensor<br />
classifier. Where very high accuracy is required <strong>in</strong> a traffic operation (e.g. toll <strong>collection</strong>), a<br />
loop sensor may be <strong>use</strong>ful <strong>in</strong> separat<strong>in</strong>g vehicles by monitor<strong>in</strong>g the loop occupancy time. The<br />
loop can be placed between the two axles. The software for classification must be capable of<br />
us<strong>in</strong>g the <strong>in</strong>formation from the loop sensor. An <strong>in</strong>frared beam can also perform the function of<br />
monitor<strong>in</strong>g occupancy as an alternative to the loop (see (d) below).<br />
(c) Axle sensor array for multi-lane highway – There are situations when short-term classified<br />
counts are needed for each lane on multi-lane highways. An array of axle sensors may be<br />
needed if the equipment is unable to dist<strong>in</strong>guish traffic travell<strong>in</strong>g <strong>in</strong> each lane, as shown <strong>in</strong><br />
Figure 5. By deduction, lane traffic can be determ<strong>in</strong>ed. For simplicity, only the configuration<br />
for axle or axle-pair counts is shown. An extra parallel set of sensors is needed for classified<br />
counts. An alternative arrangement is to locate an extra <strong>data</strong> logger <strong>in</strong> the median. This<br />
counter or logger can monitor one or more lanes of traffic closest to the median. Instructions<br />
from equipment manufacturers should be followed <strong>in</strong> a study.<br />
tube sensors<br />
and <strong>data</strong> logger<br />
median<br />
kerb<br />
Figure 5 – A sensor array for separate lane axle counts on a three-lane carriageway<br />
(d) Infrared axle sensor – Infrared beams are rout<strong>in</strong>ely <strong>use</strong>d <strong>in</strong> commerce and <strong>in</strong>dustry, e.g. as a<br />
‘doorbell’ <strong>in</strong> many shops. A beam is aimed from a transmitter to a receiv<strong>in</strong>g unit, and any<br />
break <strong>in</strong> the beam is processed to produce an electric signal, which can be <strong>use</strong>d to<br />
<strong>in</strong>crement a counter or sound a buzzer. In traffic applications a beam may be aimed across a<br />
<strong>road</strong>way or reflected from a raised pavement marker. The reflectors can be on the centrel<strong>in</strong>e<br />
(count<strong>in</strong>g one direction), or the far side of a two-lane <strong>road</strong> (count<strong>in</strong>g both directions).<br />
The beams should be close to the <strong>road</strong> surface as an alternative to a surface-mounted axle<br />
sensor. Potential problems <strong>in</strong>clude detector robustness and the ma<strong>in</strong>tenance of beam<br />
alignment. The concept does offer the advantage of m<strong>in</strong>imum <strong>in</strong>terference to the pavement<br />
and traffic stream, but it is still to be proven for multi-lane highway applications.
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 12 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
(e) Pr<strong>in</strong>ciple of uncerta<strong>in</strong>ty - Completely accurate counts do not exist due to human errors,<br />
mechanical failures, environmental factors such as ra<strong>in</strong> and magnetic fields, vehicles<br />
straddl<strong>in</strong>g two lanes, <strong>in</strong>terference from oppos<strong>in</strong>g traffic, multi-axled vehicles and other factors<br />
which affect the reliability of raw traffic counts. The best <strong>practice</strong> is often a compromise<br />
between accuracy and cost.<br />
(f) Known test <strong>data</strong> set – the uncerta<strong>in</strong>ty mentioned <strong>in</strong> (e) can be m<strong>in</strong>imised with a known test<br />
<strong>data</strong> set. A file with a know distribution of classified counts can test out the accuracy of a<br />
vehicle classifier. This approach was adopted <strong>in</strong> the development of the ARRB Vehicle<br />
Detection Data Acquisition System <strong>in</strong> the late 1970s, and should aga<strong>in</strong> be deployed <strong>in</strong> test<strong>in</strong>g<br />
a new generation of software and hardware.<br />
As described <strong>in</strong> GTEP Part 3, the count<strong>in</strong>g personnel should be well prepared <strong>in</strong> terms of count<strong>in</strong>g<br />
equipment, ancillary equipment such as tools and seats if necessary, appropriate cloth<strong>in</strong>g <strong>in</strong>clud<strong>in</strong>g<br />
cont<strong>in</strong>gency provisions for adverse weather and amenities (e.g. food and toilets), and carry<br />
authority certificates. Conspicuous signage is not recommended as it may affect count results.<br />
Other issues of quality control are addressed <strong>in</strong> Section 6.
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
4 TRAFFIC COUNTING PROGRAM<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 13 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
This section addresses best <strong>practice</strong>s <strong>in</strong> a traffic count<strong>in</strong>g program. It beg<strong>in</strong>s with a brief<br />
description of the current situation <strong>in</strong> traffic count<strong>in</strong>g programs amongst <strong>road</strong> authorities (RAs).<br />
The description is from the perspective of traffic count<strong>in</strong>g techniques and methods as well as from<br />
the perspective of the RAs hav<strong>in</strong>g to balance rigour <strong>in</strong> their traffic count<strong>in</strong>g programs aga<strong>in</strong>st an<br />
<strong>in</strong>creas<strong>in</strong>g rate of change and variation <strong>in</strong> traffic flows ca<strong>use</strong>d by economic activity, <strong>in</strong>creased<br />
scrut<strong>in</strong>y and <strong>use</strong>r demands as well as budgetary cost pressures (Section 4.1). This section covers<br />
best <strong>practice</strong>s <strong>in</strong> the three ma<strong>in</strong> areas of traffic count<strong>in</strong>g programs: obta<strong>in</strong><strong>in</strong>g the raw <strong>data</strong> (Section<br />
4.2), transform<strong>in</strong>g the raw <strong>data</strong> <strong>in</strong>to AADT and VKT (Section 4.3) and report<strong>in</strong>g of the raw and<br />
transformed <strong>data</strong> (Section 4.4).<br />
4.1 Background<br />
Traffic counts and classifications are the basic raw materials for the study and <strong>analysis</strong> of traffic,<br />
form<strong>in</strong>g a key <strong>in</strong>put and <strong>in</strong>fluenc<strong>in</strong>g the design, ma<strong>in</strong>tenance and management decision mak<strong>in</strong>g of<br />
every RA. In fact, the importance of traffic <strong>data</strong> extends beyond the <strong>in</strong>terests of each RA to <strong>in</strong>clude<br />
various commercial <strong>use</strong>rs (for plann<strong>in</strong>g and commercial studies) and other government agencies at<br />
a national level on policies lead<strong>in</strong>g to an <strong>in</strong>tegrated national land transport <strong>in</strong>frastructure network.<br />
In general, there are three ma<strong>in</strong> themes of traffic <strong>data</strong>:<br />
Variation of traffic flow by location,<br />
Time-variation of traffic flow,<br />
Composition of the traffic flow.<br />
The above <strong>data</strong> are collected by count<strong>in</strong>g stations. Count<strong>in</strong>g stations are located at various po<strong>in</strong>ts<br />
<strong>in</strong> the <strong>road</strong> network to ensure the <strong>collection</strong> of a representative sample of <strong>data</strong> that will, <strong>in</strong> isolation,<br />
report on particular <strong>road</strong> segments or corridors and, <strong>in</strong> aggregate, report on the entire <strong>road</strong> network<br />
itself. Two types of count<strong>in</strong>g stations are employed. They are the Permanent Count<strong>in</strong>g Stations<br />
(PCS, also known as ‘Pattern Stations’) and Short-term Count<strong>in</strong>g Stations (SCS, also known as<br />
‘Sample’ or ‘Coverage’ Stations). Pattern Stations cont<strong>in</strong>uously monitor or frequently sample traffic<br />
to determ<strong>in</strong>e patterns and seasonal fluctuations. Such stations are permanent or semi-permanent<br />
for count<strong>in</strong>g over a def<strong>in</strong>ed ‘season’ (usually a year). Short-term stations are count<strong>in</strong>g stations that<br />
are utilised for brief traffic surveys rang<strong>in</strong>g from periods of a few weeks to one hour, with a typical<br />
sample site count<strong>in</strong>g traffic for seven consecutive days.<br />
Traffic count<strong>in</strong>g faces pressure from <strong>use</strong>rs who want additional <strong>data</strong> for reports and projects, as<br />
well as from owners who want improved cost efficiencies. Out-sourc<strong>in</strong>g many aspects of <strong>data</strong><br />
<strong>collection</strong> and ma<strong>in</strong>tenance is more commonplace today and one particular traffic <strong>data</strong> unit <strong>in</strong> a RA<br />
has become totally off-budget, requir<strong>in</strong>g it to secure customers amongst the various divisions<br />
with<strong>in</strong> the parent RA.<br />
At the same time, rapid changes <strong>in</strong> the economy, the development of toll <strong>road</strong>s, <strong>in</strong>creased freight<br />
flows, chang<strong>in</strong>g vehicle regulation and technology have resulted <strong>in</strong> potentially large changes <strong>in</strong><br />
traffic patterns as well as the potential to <strong>use</strong> new technology to track and count vehicles.<br />
In this environment, it is timely to revisit the fundamentals of traffic count<strong>in</strong>g programs. The<br />
contribution of the practitioners <strong>in</strong> this field through the earlier fieldwork and the recent Aust<strong>road</strong>s<br />
Road Use Data Workshop has been <strong>in</strong>valuable towards a common sense approach <strong>in</strong> address<strong>in</strong>g<br />
and develop<strong>in</strong>g best <strong>practice</strong> for traffic count<strong>in</strong>g programs.
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 14 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
Traffic count<strong>in</strong>g programs are well established and documented through a series of guides<br />
beg<strong>in</strong>n<strong>in</strong>g with NAASRA (1982) and the two versions of GTEP Part 3 (Aust<strong>road</strong>s 1998a and 20<strong>04</strong>).<br />
A national review of traffic count<strong>in</strong>g procedures (Roper 2001) and subsequent recommendations<br />
(Bennett 2002), addressed the issues of additional <strong>data</strong> <strong>collection</strong> facilities and consistency across<br />
the report<strong>in</strong>g RAs. The studies of Roper (2001), Bennett (2002), a parallel research activity<br />
address<strong>in</strong>g an <strong>in</strong>ternational literature review on consistency of <strong>data</strong> (Mart<strong>in</strong> and Chiang 2002), and<br />
a review of the current traffic <strong>data</strong> of the RAs (Roper et al. 2002), led to the Road Use Data<br />
Workshop <strong>in</strong> October 2003 <strong>in</strong>volv<strong>in</strong>g representatives from the RAs specifically address<strong>in</strong>g the key<br />
issues <strong>in</strong> traffic <strong>data</strong> (Luk and Karl 2003).<br />
In addition to the studies and Workshop already mentioned, a range of other sources (such as the<br />
US FHWA, the European Commission and other reports) were accessed <strong>in</strong> the preparation of this<br />
section.<br />
The NAASRA (1982) Guide provided a solid ground<strong>in</strong>g for RAs on which to base their traffic<br />
count<strong>in</strong>g operations and, <strong>in</strong> general, this has resulted <strong>in</strong> a consistent and diverse set of traffic <strong>data</strong><br />
available from RAs, as can be seen <strong>in</strong><br />
Figure 6 (Roper et al. 2002). A cursory exam<strong>in</strong>ation of the tables will show that most of the<br />
tabulated <strong>data</strong> items of <strong>in</strong>terest are already reported by most of the RAs.<br />
4.2 AADT and VKT Estimation<br />
Annual Average Daily Traffic (AADT) and Vehicle-Kilometres Travelled (VKT) are two of the most<br />
commonly <strong>use</strong>d descriptors of traffic derived from raw traffic counts. Data expressed <strong>in</strong> AADT and<br />
VKT are universally understood. This section beg<strong>in</strong>s with brief <strong>in</strong>troductions to AADT and VKT,<br />
followed by discussions on accuracy and adjustment factors.<br />
4.2.1 AADT<br />
AADT is the Average Annual Daily Traffic pass<strong>in</strong>g a <strong>road</strong>side observation po<strong>in</strong>t over the period of a<br />
calendar year. It can be obta<strong>in</strong>ed by a number of ways. For example, it is obta<strong>in</strong>ed by measur<strong>in</strong>g<br />
the total volume of traffic pass<strong>in</strong>g the observation po<strong>in</strong>t over a year and then divided by the number<br />
of days <strong>in</strong> that year, i.e.<br />
AADT<br />
=<br />
n<br />
∑Vi<br />
/ n<br />
i=<br />
1<br />
where Vi = the total traffic counts on day i and n is the number of days <strong>in</strong> that year (n can be 365 and 366 days).<br />
Some jurisdictions adopt n = 365.25 days but the difference should be small.<br />
The production of AADT <strong>data</strong> would be straightforward if all count<strong>in</strong>g stations were to operate<br />
without fail for every day of the year, record<strong>in</strong>g daily traffic <strong>data</strong>. However, <strong>in</strong> reality, reliability is<br />
not 100% and not every count<strong>in</strong>g station is a permanent station that operates all year round.<br />
AADTs derived from permanent or pattern sites may require ‘plugg<strong>in</strong>g the gaps’ due to miss<strong>in</strong>g<br />
<strong>data</strong>. Such adjustments are preferably m<strong>in</strong>or and <strong>in</strong>volve a few days of lost or corrupt <strong>data</strong> over a<br />
twelve month period. Permanent sites, however, comprise only a small amount of the traffic<br />
count<strong>in</strong>g network. The issue of handl<strong>in</strong>g miss<strong>in</strong>g <strong>data</strong> will be addressed <strong>in</strong> Section 6.
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
(a) Traffic volume <strong>data</strong> items by jurisdiction<br />
Data Item ACT NSW NT QLD SA TAS VIC WA NZ<br />
ADT <br />
AADT <br />
AWT <br />
AAWT <br />
Midweek Volume <br />
VKT <br />
DHV <br />
Growth & Trends <strong>in</strong><br />
AADT<br />
Growth & Trends <strong>in</strong><br />
VKT<br />
<br />
<br />
(b) Traffic flow <strong>data</strong> items by jurisdiction<br />
Data Item ACT NSW NT QLD SA TAS VIC WA NZ<br />
Rank<strong>in</strong>g <strong>in</strong> highest<br />
hourly flows<br />
Distribution of Hourly<br />
Flows through the<br />
year<br />
<br />
Directional Patterns <br />
Directional/Lane<br />
Distribution of traffic<br />
<br />
Seasonal Patterns <br />
Use of Aggregated<br />
Data by Time or<br />
Location (e.g.: 15,<br />
30, 60-m<strong>in</strong>ute<br />
<strong>in</strong>tervals etc.)<br />
<br />
(c ) Traffic composition <strong>data</strong> items by jurisdiction<br />
Data Item ACT NSW NT QLD SA TAS VIC WA NZ<br />
Use of Aust<strong>road</strong>s<br />
Vehicle Classification<br />
schemes<br />
<br />
<br />
Other schemes <br />
Traffic Composition<br />
(e.g. % HV, % Artic.<br />
% Rigid, %<br />
Comb<strong>in</strong>ation, etc.)<br />
<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 15 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
(d) Weigh-<strong>in</strong>-motion <strong>data</strong> items by jurisdiction<br />
Data Item ACT NSW NT QLD SA TAS VIC WA NZ<br />
Weigh-<strong>in</strong>-Motion (<strong>WIM</strong>) <br />
Truck Weight Data <br />
Freight Mass <br />
(e) Other <strong>data</strong> items by jurisdiction<br />
Data Item ACT NSW NT QLD SA TAS VIC WA NZ<br />
Functional Classification of<br />
Road<br />
<br />
Geographical Regions <br />
Types of Counts Collected<br />
(e.g., Turn<strong>in</strong>g Counts, Axle<br />
Counts, Vehicle Counts,<br />
Bicycle Counts)<br />
Other Types of Traffic<br />
Data Collected (e.g.,<br />
Travel Time Surveys)<br />
<br />
<br />
(f) Adjustment factors by jurisdiction<br />
Data Item ACT NSW NT QLD SA TAS VIC WA NZ<br />
Seasonal Adjustment<br />
Factors<br />
Expansion Factors (e.g..<br />
12 hour count to 24 hour<br />
count)<br />
Other Adjustment Factors<br />
(e.g.. public holiday,<br />
weekends, etc )<br />
Daily Flow Factors<br />
Figure 6 – Survey of <strong>road</strong> <strong>use</strong> <strong>data</strong> needs amongst Road Authorities
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 16 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
All AADTs are estimates. A large amount of AADT <strong>data</strong> is derived from sample or short-term sites<br />
and estimated by utilis<strong>in</strong>g short-term counts to calculate an Average Daily Traffic (ADT) and<br />
apply<strong>in</strong>g an adjustment factor. This factor is derived from a seasonal (12-month) count on the <strong>road</strong><br />
<strong>in</strong> question, or on a <strong>road</strong> exhibit<strong>in</strong>g similar seasonal traffic variations.<br />
Other related traffic parameters are also processed by RAs. These <strong>in</strong>clude (see Appendix A):<br />
Annual Average Weekday Traffic (AAWT), Average Weekday Daily Traffic (AWDT) or Average<br />
Weekday Traffic (AWT). Different jurisdictions at present adopt slightly different def<strong>in</strong>itions and<br />
hence procedures <strong>in</strong> calculat<strong>in</strong>g these quantities. The actual values should not be significantly<br />
different. However, it would be good <strong>practice</strong> for RAs to state the methods <strong>use</strong>d to generate AADT<br />
and other values. Further, the def<strong>in</strong>ition of AADT or other quantities should not imply a particular<br />
method of calculation or estimation. It is, however, important that the def<strong>in</strong>itions <strong>in</strong> Appendix A be<br />
accepted as a national glossary.<br />
4.2.2 VKT<br />
VKT on a network is estimated by measur<strong>in</strong>g or estimat<strong>in</strong>g AADT on selected <strong>road</strong> sections, and<br />
mak<strong>in</strong>g an estimate of the total network travel on the basis of travel on each <strong>road</strong> section. Simply<br />
expressed,<br />
VKT<br />
=<br />
m<br />
∑ AADT ×<br />
j=<br />
1<br />
j l j<br />
where<br />
VKT = the vehicle-kilometre travelled on a network of <strong>road</strong> sections j = 1, … m<br />
AADTj = the AADT for <strong>road</strong> section j, and<br />
= the length of <strong>road</strong> section j <strong>in</strong> km.<br />
lj<br />
The total daily travel (the daily VKT) on all segments <strong>in</strong> a <strong>road</strong> network is the sum of the product of<br />
AADT on each <strong>road</strong> section and the section length. The yearly VKT or total yearly travel is<br />
therefore the total daily travel multiplied by the number of days <strong>in</strong> that year (365 or 366 days). It is<br />
good <strong>practice</strong> to specify units to dist<strong>in</strong>guish daily VKT compared to annual VKT.<br />
The accuracy of the estimation of VKT depends primarily on how close the sample of <strong>road</strong> sections<br />
represent the traffic flow for the entire <strong>road</strong> (or <strong>road</strong> network) and the accuracy of the estimates of<br />
AADT. It also depends on the dispersion or ‘scatter’ of the estimates. Accuracy can be improved<br />
by group<strong>in</strong>g together <strong>road</strong> segments with similar traffic characteristics. These groups may be<br />
geographic regions, functional categories, or volume strata (see also Section 4.3.2). For example,<br />
seasonal variations have the greatest impact on AADT estimates and regional group<strong>in</strong>gs are<br />
therefore preferable. On the other hand, volume stratification is <strong>use</strong>d to improve the accuracy <strong>in</strong><br />
estimat<strong>in</strong>g a VKT and will be described further below.<br />
A number of different forms of VKT are commonly <strong>use</strong>d:<br />
Total VKT,<br />
VKT by vehicle class,<br />
VKT by <strong>road</strong> or route,<br />
VKT by <strong>road</strong> type,<br />
VKT by region.
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 17 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
These breakdowns provide further mean<strong>in</strong>gful <strong>in</strong>formation for <strong>use</strong>rs. VKT by vehicle class is<br />
particularly <strong>use</strong>ful <strong>in</strong> differentiat<strong>in</strong>g freight flows. The classification of traffic counts, a necessary<br />
prerequisite <strong>in</strong> order to produce this <strong>data</strong>, has been discussed earlier <strong>in</strong> Section 3. The correlation<br />
of classified counts and VKT by vehicle classes with weigh-<strong>in</strong>-motion <strong>data</strong> (e.g. from CULWAY)<br />
facilitates the estimation of pavement and bridge load<strong>in</strong>g and is discussed <strong>in</strong> Section 5.2.<br />
4.2.3 Accuracy and Round<strong>in</strong>g<br />
Traffic counts taken at any one location are affected by occurrences elsewhere <strong>in</strong> the <strong>road</strong> network<br />
(RTA 2003). Some effects may be known, such as a new <strong>road</strong> open<strong>in</strong>g or major disruption to traffic<br />
due to a recent major <strong>in</strong>cident. Others may be m<strong>in</strong>or and unknown, such as m<strong>in</strong>or <strong>road</strong> crashes.<br />
Other effects <strong>in</strong>clude weather conditions, equipment malfunction<strong>in</strong>g and miscod<strong>in</strong>g, and community<br />
effects such as day-of-week, school and public holidays. Each traffic count taken, say, daily should<br />
be viewed as a sample from an unknowable distribution. AADT, whether estimated from short-term<br />
or permanent count stations, should not be viewed as a uniquely def<strong>in</strong>ed quantity, but rather as a<br />
mean value with a distribution of errors.<br />
Accuracy requirements for both AADT and VKT have been previously reported <strong>in</strong> NAASRA (1982)<br />
and are given <strong>in</strong> Table 5 and Table 6. As can be seen, larger error bands are acceptable at low<br />
levels of AADT and 68% confidence limits represent ± one standard deviation from the mean.<br />
Traffic characteristic<br />
Traffic characteristic<br />
AADT < 100<br />
101-300<br />
301-1100<br />
>1100<br />
Trend <strong>in</strong> AADT<br />
Traffic composition<br />
VKT - Annual total<br />
VKT - Annual change<br />
VKT - Annual composition<br />
Table 5 - Accuracy requirements for <strong>in</strong>dividual sites<br />
Table 6 - Accuracy requirements for groups of <strong>road</strong>s<br />
Maximum error requirements<br />
(at 68 % confidence limits)<br />
50%<br />
35%<br />
25%<br />
15%<br />
10%<br />
20%<br />
Errors at 95 % confidence limits<br />
Australia State City*<br />
3%<br />
1%<br />
3%<br />
5-7%<br />
3%<br />
5%<br />
* If these accuracies are achieved, those for <strong>in</strong>dividual State and Australia will also be obta<strong>in</strong>ed.<br />
10%<br />
5%<br />
10%<br />
Rural<br />
area*<br />
10%<br />
5%<br />
10%<br />
AADT range<br />
(Table 13)<br />
Errors are reduced if the number of days that traffic is counted is <strong>in</strong>creased, as can be seen <strong>in</strong><br />
Figure 7 (NAASRA 1982). For all AADTs, percentage error decreases as the number of days<br />
counted <strong>in</strong>creases and the six-day count will produce less than 12.5% error for AADTs over 1000.<br />
While error bands are naturally higher for smaller AADTs, this has to be realistically acceptable<br />
and is tolerable <strong>in</strong> terms of the limited impact on the overall traffic volume.<br />
25%<br />
10%<br />
15%
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 18 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
Figure 7 - Expected percentage error for AADT from ADT<br />
and count duration (n-Day counts) for a 75% confidence level.<br />
An appreciation of the accuracy of AADT <strong>data</strong> permits the consideration of the round<strong>in</strong>g of<br />
numbers.<br />
Table 7 provides some examples of round<strong>in</strong>g and compares the percentage change <strong>in</strong> AADT<br />
aga<strong>in</strong>st the maximum error requirements stipulated <strong>in</strong> NAASRA (1982). It shows that, <strong>in</strong> all cases,<br />
round<strong>in</strong>g is with<strong>in</strong> the permitted error range. The maximum error permitted is based on random<br />
sampl<strong>in</strong>g and round<strong>in</strong>g should be carried out only if necessary, e.g. for report<strong>in</strong>g purposes.<br />
Table 7 - Examples of round<strong>in</strong>g AADT<br />
AADT Example Round<strong>in</strong>g Percentage change Maximum error permitted<br />
0-100 46 50 8.7% 50%<br />
101-300 157 200 27.4% 35%<br />
301-1100 678 700 3.2% 25%<br />
>1100 2345 2300 1.9% 15%<br />
4.2.4 Adjustment Factors<br />
Traffic counts from both permanent and short-term sites often are <strong>in</strong>complete. Seasonal<br />
adjustment factors, conversion and expansion factors are typically applied to traffic counts <strong>in</strong> order<br />
to statistically convert shorter counts to annual averages. FHWA (2001) describes it as follows:<br />
Assum<strong>in</strong>g that temporal characteristics affect all <strong>road</strong>s and s<strong>in</strong>ce cont<strong>in</strong>uous<br />
temporal <strong>data</strong> exist at several po<strong>in</strong>ts (the permanent/pattern sites) to describe<br />
temporal variation, it is possible to transfer knowledge by develop<strong>in</strong>g factor<strong>in</strong>g<br />
mechanisms.
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 19 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
Most RAs have utilised a range of adjustment factors, as can be seen <strong>in</strong> Table 8 (Roper et al.<br />
2002).<br />
Table 8 - Adjustment factors by jurisdictions<br />
Data Item ACT NSW NT QLD SA TAS VIC WA NZ<br />
Seasonal Adjustment Factors <br />
Expansion factors (e.g. 12 hour count to<br />
24 hour count)<br />
Other adjustment factors (e.g. public<br />
holiday, weekends, or daily flow factors<br />
from control sites)<br />
<br />
<br />
A number of adjustment factors have been applied, <strong>in</strong> various jurisdictions, which <strong>in</strong>clude:<br />
Seasonal adjustment factors,<br />
Hour-of the day expansion factors,<br />
Day-of-week adjustment factor,<br />
Lane distribution factors,<br />
Growth trends at that location,<br />
Factors for short-term counts,<br />
Factors for total volume and estimates for trucks,<br />
Visual 8, 12 hour expansion factors,<br />
Visual truck expansion factors.<br />
GTEP Part 3 provides a general guidel<strong>in</strong>e, not<strong>in</strong>g that where a <strong>road</strong> count<strong>in</strong>g program has been<br />
established and seasonal patterns identified, the AADT at a particular location may be estimated<br />
by multiply<strong>in</strong>g a sample count (say, two to six days duration) by the Seasonal Adjustment Factor<br />
derived from a Pattern Station representative of the required location by:<br />
(AADT)j = ADT ij × (SAF) i,k<br />
where<br />
(AADT)j = the AADT at the required location j;<br />
(ADT)ij = the sample count <strong>in</strong> the season (or month, week, day, etc.) i at the location j;<br />
(SAF)ik = the Seasonal Adjustment Factor for the season (or month, week, day, etc.) i at a Pattern Station, k,<br />
representative of the required location j.<br />
As conditions will vary <strong>in</strong> each jurisdiction, the specific adjustment factors utilised by RAs will each<br />
be unique to their jurisdiction. It is good <strong>practice</strong> for each RA to derive its own adjustment factors.<br />
Also, adjustment factors should be clearly detailed, say, <strong>in</strong> appendices to provide clarity and<br />
transparency to the process of adjust<strong>in</strong>g the raw <strong>data</strong>.
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 20 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
Transfund New Zealand (2001) provides the adjustment factors to convert short-term counts to<br />
estimates of AADT and some of these factors are shown <strong>in</strong> Table 9. These factors are categorised<br />
by <strong>road</strong> types, day-of-week, time periods of day (part-days) and vehicle types. The day and partday<br />
conversion factors are shown <strong>in</strong> Table 2.4 for Auckland and non-Auckland <strong>road</strong>s <strong>in</strong> New<br />
Zealand. The validity of these factors for other places has yet to be established, and the error for<br />
AADT estimates us<strong>in</strong>g part-day factors can be greater than 30% and exceed Aust<strong>road</strong>s (20<strong>04</strong>)<br />
guidel<strong>in</strong>es.<br />
Table 9 - Some adjustment factors from short-term counts to AADT estimates<br />
Road types<br />
Day factors<br />
Mon Tues Wed Thu Fri Sat Sun<br />
Urban arterial 1 (Auckland) 0.98 0.95 0.93 0.92 0.88 1.17 1.37<br />
Urban arterial 1 (non-Auckland) 1.01 0.98 0.95 0.94 0.89 1.09 1.24<br />
Urban arterial 2 (Auckland) 0.99 0.96 0.95 0.93 0.89 1.13 1.31<br />
Urban arterial 2 (non-Auckland) 1.00 0.97 0.94 0.93 0.88 1.11 1.30<br />
Urban CBD (Auckland) 1.02 1.00 0.97 0.94 0.86 1.07 1.34<br />
Urban Industrial (Auckland) 0.85 0.85 0.84 0.84 0.84 1.84 2.66<br />
Rural urban fr<strong>in</strong>ge 1.14 1.14 1.10 1.07 1.07 0.94 0.86<br />
Rural strategic 1 1.05 1.01 0.99 0.97 0.97 1.09 1.10<br />
Rural strategic 2 1.11 1.14 1.10 1.05 1.05 1.03 0.91<br />
Rural recreation summer 1.07 1.17 1.13 1.05 1.05 1.11 0.88<br />
Rural recreation w<strong>in</strong>ter 1.15 1.25 1.21 1.12 1.12 1.03 0.82<br />
Road types<br />
Part-day factors for typical Monday - Thursday<br />
7-9 a.m. 9a.m.-12p.m. 1-4 p.m. 4-6 p.m.<br />
Urban arterial (Auckland) 5.63 5.66 5.10 5.78<br />
Urban arterial (non-Auckland) 7.31 5.17 4.60 6.00<br />
Urban CBD (Auckland) 7.86 5.77 5.22 6.36<br />
Urban Industrial (Auckland) 5.57 4.13 3.72 5.97<br />
Rural urban fr<strong>in</strong>ge 8.61 5.59 5.49 6.64<br />
Rural strategic 1 8.40 4.74 4.41 5.99<br />
Rural strategic 2 10.2 5.17 4.89 6.88<br />
Rural recreation summer 16.9 4.84 4.08 7.77<br />
Rural recreation w<strong>in</strong>ter 13.8 5.24 4.76 7.91<br />
(Source: Transfund New Zealand 2001)
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 21 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
Some examples of us<strong>in</strong>g Table 9 are as follows:<br />
(a) If the short-term day count on a Wednesday on a rural urban fr<strong>in</strong>ge site is 10,000 vehicles<br />
per 24 hour, then the estimated AADT is 1.10 × 10,000 = 11,000 vehicles.<br />
(b) If the short-term part-day (7 a.m. – 9 a.m.) count on a Wednesday at the same rural urban<br />
fr<strong>in</strong>ge site is 1,300 vehicle per 2 hour, then the estimated AADT is 8.61 × 1300 = 11,193<br />
vehicles.<br />
In general, counts of very short durations of a few hours or even 12 hours are not reliable for traffic<br />
forecast<strong>in</strong>g. With automatic count<strong>in</strong>g equipment now commonly deployed, the extra cost of<br />
collect<strong>in</strong>g <strong>data</strong> on more days should be m<strong>in</strong>imal. The best way to ensure consistency amongst RAs<br />
is for each RA to collect robust <strong>data</strong> sets – collect<strong>in</strong>g <strong>data</strong> over a consecutive period of, say, seven<br />
days.<br />
Based on some NSW count <strong>data</strong>, Botterill and Luk (1998) reported that regression <strong>analysis</strong> is<br />
suitable for determ<strong>in</strong><strong>in</strong>g an adjustment factor to convert a sample count to an estimated AADT and<br />
estimat<strong>in</strong>g the relevant accuracy. Regression <strong>analysis</strong> could be preferable to cluster <strong>analysis</strong> for<br />
some <strong>use</strong>rs beca<strong>use</strong> the statistics from regression are well-known. The NSW <strong>data</strong> were separated<br />
<strong>in</strong>to monthly counts and regression equations were derived relat<strong>in</strong>g AADT to <strong>road</strong> type, count type<br />
(axle counts or classified counts), and AADT strata. It was found that AADT strata had m<strong>in</strong>imal<br />
effect and a recommended regression equation is as follows (without us<strong>in</strong>g the constant term):<br />
AADT = β × Vol + ε<br />
where AADT is the count for each Permanent Station,<br />
β is the seasonal adjustment factor to be estimated,<br />
Vol is the daily volume for each nom<strong>in</strong>al period (either a three-day or a six-day week) <strong>in</strong> the four weeks<br />
<strong>in</strong> any month for each permanent stations,<br />
ε is the error term represent<strong>in</strong>g the deviation from the regression l<strong>in</strong>e.<br />
Botterill and Luk (1998) further provided a systematic view of the tasks <strong>in</strong>volved <strong>in</strong> a traffic count<strong>in</strong>g<br />
program and the calculation of VKTs (Figure 8). The framework highlights the follow<strong>in</strong>g issues:<br />
Need for a strategic view <strong>in</strong> implement<strong>in</strong>g a count<strong>in</strong>g program (fund<strong>in</strong>g issues and on-go<strong>in</strong>g<br />
ma<strong>in</strong>tenance);<br />
Appropriate <strong>analysis</strong> techniques (cluster or regression for pattern recognition, seasonal<br />
adjustments, etc.);<br />
Reliability check over many years and between years;<br />
Clear operational producers.<br />
Some of the issues have been mentioned earlier <strong>in</strong> Section 2 and are addressed further below.
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
Stages<br />
Set up network<br />
Set up stations<br />
Data<br />
Relate counters<br />
Relate <strong>data</strong><br />
Network VKT<br />
Figure 8 - A design for traffic count<strong>in</strong>g program<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 22 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
Concepts Feedback<br />
Operations<br />
Road<br />
network<br />
Group<strong>in</strong>g<br />
variables<br />
Strategy:<br />
funds, ma<strong>in</strong>t’e,<br />
cont<strong>in</strong>uity<br />
Count<br />
duration,<br />
tim<strong>in</strong>g<br />
Cluster,<br />
regression<br />
method<br />
Conversion<br />
(seasonalis<strong>in</strong>g)<br />
method<br />
Reliability<br />
over many<br />
years<br />
Reliability<br />
between<br />
years<br />
Improve<br />
accuracy<br />
PCS site<br />
choice<br />
- reliability<br />
Yearly<br />
count <strong>data</strong><br />
Determ<strong>in</strong>e<br />
<strong>road</strong> sections<br />
SCS site<br />
choice<br />
- coverage<br />
Sample<br />
count <strong>data</strong><br />
Group PCS Allocate to<br />
PCS group<br />
Conversion<br />
factors<br />
(Seasonal, etc)<br />
- by group<br />
Total VKT<br />
on network<br />
- accuracy<br />
Adjusted<br />
estimates of<br />
VKT, AADT<br />
- by group
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 23 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
4.3 Network Coverage<br />
Reports of AADT and VKT relate to the network where traffic is counted. As traffic counts are<br />
almost always a sample of the <strong>road</strong> network, the issue of network coverage or l<strong>in</strong>k-node<br />
development is important. The key issues are as follows:<br />
Difficulty <strong>in</strong> prescrib<strong>in</strong>g rules like number of permanent sites per km as each RA network is<br />
unique. Good judgement is required to def<strong>in</strong>e <strong>road</strong> sections or l<strong>in</strong>ks;<br />
Use of performance measures such as an <strong>in</strong>dication of the number of permanent sites for a<br />
given number of homogeneous sections;<br />
Frequency of review<strong>in</strong>g uniform traffic sections (UTS) or <strong>road</strong> sections; and<br />
Def<strong>in</strong>ition of homogeneous or uniform <strong>road</strong> sections.<br />
The accuracy of AADTs at sample stations depends critically on the selection of Pattern Stations,<br />
which alone provide the full year of traffic counts and from which the related short-term sites<br />
depend for their adjustment factors. In order to <strong>in</strong>crease the confidence <strong>in</strong> the pattern <strong>in</strong>formation<br />
ga<strong>in</strong>ed, the <strong>collection</strong> of <strong>data</strong> at all Short-term Stations <strong>in</strong> a pattern group should be made at the<br />
same time as the related Pattern Station (if it is seasonal) is be<strong>in</strong>g sampled.<br />
The accuracy of an AADT estimate depends partly on the reliability of group<strong>in</strong>g Short-term Stations<br />
with Pattern Stations, and the method of group<strong>in</strong>g is important. Areas with similar economy,<br />
culture and development tend to show similar traffic patterns, and proximity on its own may not be<br />
a good <strong>in</strong>dicator.<br />
This section deals with those issues and discusses good <strong>practice</strong>s found <strong>in</strong> each area. The<br />
follow<strong>in</strong>g sub-sections cover sampl<strong>in</strong>g (Section 4.3.1), frequency (Section 4.3.2), def<strong>in</strong><strong>in</strong>g<br />
homogeneous sections (Section 4.3.3) and directionality (Section 4.3.4).<br />
4.3.1 Sampl<strong>in</strong>g<br />
Sampl<strong>in</strong>g <strong>in</strong> a traffic count program relates to two separate yet <strong>in</strong>ter-related issues; the sampl<strong>in</strong>g<br />
unit and the frequency of sampl<strong>in</strong>g. Both issues are discussed <strong>in</strong> this section.<br />
Sampl<strong>in</strong>g Unit<br />
Pattern stations are best located by randomly select<strong>in</strong>g the number of sites on <strong>road</strong>s from each of<br />
a number of AADT strata, with<strong>in</strong> each geographical region. As a general rule, the density of<br />
stations <strong>in</strong> each stratum should be proportional to the product of the total length of <strong>road</strong> and the<br />
square root of the mean AADT <strong>in</strong> that stratum (Aust<strong>road</strong>s 20<strong>04</strong>).<br />
In <strong>practice</strong>, the number of Short-term Stations is determ<strong>in</strong>ed by the extent of the <strong>road</strong> system and<br />
the availability of funds. The number of Pattern Stations, and the ratio of Pattern Stations to Shortterm<br />
Stations, cannot be determ<strong>in</strong>ed <strong>in</strong> advance, beca<strong>use</strong> the assessment will depend on the<br />
regional group<strong>in</strong>g and the reliability of matched patterns. Thus, the achievement of consistent and<br />
specified accuracy levels <strong>in</strong> AADT estimation is essentially an iterative process that develops over<br />
time and with experience.<br />
The guidel<strong>in</strong>es for the densities of traffic count<strong>in</strong>g stations are shown <strong>in</strong> Table 10.
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
Rural<br />
Type of <strong>road</strong><br />
Table 10 - Densities for traffic count<strong>in</strong>g stations<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 24 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
Count<strong>in</strong>g station density (km/station)<br />
Pattern Stations<br />
Permanent* Seasonal*<br />
Freeways, arterials 100-200 200-300<br />
Local <strong>road</strong>s 5,000-9,000 1,000-2,000<br />
Urban<br />
Freeways, arterials 20-50 30-50<br />
Collectors, distributors 100-150 50-100<br />
Local <strong>road</strong>s >1,000 4,000-6,000<br />
* lower figures (or higher densities) applicable if relatively few Seasonal Stations <strong>use</strong>d<br />
Short-term<br />
Stations<br />
Up to 25<br />
(usually 13-17)<br />
Short-term Stations may be<br />
characterised by one Pattern<br />
Station<br />
In a recent review of traffic count<strong>in</strong>g stations, Roper (2001) found that RAs generally exceed the<br />
guidel<strong>in</strong>es <strong>in</strong> Table 11 for traffic count<strong>in</strong>g station densities. In most cases, densities were higher<br />
than the guide of 100 – 200 km for rural <strong>road</strong> and 20 - 50 km for urban count stations.<br />
Rural<br />
Urban<br />
Table 11 – National highway traffic count<strong>in</strong>g station density by location and status<br />
Count<strong>in</strong>g station type Station density (km/site)<br />
(location and status) NSW VIC QLD WA SA TAS NT ACT NZ<br />
Permanent 68.3 68.1 19.3 217.6 221.9 51.0 175.4 6.3 111<br />
Temporary 10.3 6.4 74.1 25.8 35.0 2.7 202.4 9.5 9.8<br />
Overall 9.0 5.9 15.3 23.1 30.3 2.6 94.0 3.8 9.1<br />
Permanent 2.9 2.6 3.2 -* 43.5 7.0 3.3 - 22.5<br />
Temporary 3.0 4.2 20.6 2.8 7.3 0.5 3.9 - 4.5<br />
Overall 1.5 1.6 2.8 2.8 6.2 0.5 1.8 - 3.1<br />
* there are 40 sites <strong>in</strong> metropolitan Perth although none on National Highways due to the urban form of Perth<br />
Beca<strong>use</strong> judgemental sampl<strong>in</strong>g is often employed to locate Permanent and Short-term Stations,<br />
the AADT values obta<strong>in</strong>ed cannot be regarded as random samples. The estimation of VKT from<br />
these AADT values is therefore not as representative as those obta<strong>in</strong>ed by, say, the sampl<strong>in</strong>g<br />
frame of the Survey of Motor Vehicle Usage (SMVU) of the Australian Bureau of Statistics (ABS<br />
1963 and other years). The VKT from a <strong>road</strong> authority count<strong>in</strong>g program (e.g. the NSW program)<br />
is comparable to SMVU estimates if VKT from all <strong>road</strong> types <strong>in</strong> a State are accounted for,<br />
especially the VKT from local <strong>road</strong>s.<br />
The NAASRA (1982) method of calculat<strong>in</strong>g the number of count<strong>in</strong>g stations to determ<strong>in</strong>e the VKT<br />
of a region is shown <strong>in</strong> Appendix B. The method has been developed further <strong>in</strong> Botterill and Luk<br />
(1998) by <strong>in</strong>clud<strong>in</strong>g accuracy and statistical significance requirements and relax<strong>in</strong>g the requirement<br />
on preset AADT strata. Us<strong>in</strong>g the orig<strong>in</strong>al method, the number of stations required could be<br />
underestimated. The derivation of the new formulae is quite <strong>in</strong>volved and will not be <strong>in</strong>cluded <strong>in</strong><br />
Appendix B. In any case, as mentioned previously, the provision of count<strong>in</strong>g stations is more than<br />
generally recommended <strong>in</strong> NAASRA (1982).
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 25 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
Frequency of Sampl<strong>in</strong>g<br />
The issue of frequency of traffic counts relates <strong>in</strong> particular to the frequency at which short-term<br />
counts are conducted at specific sites. In determ<strong>in</strong><strong>in</strong>g the frequency of <strong>data</strong> <strong>collection</strong>, a <strong>road</strong><br />
authority will need to consider the rate of change of values of particular <strong>data</strong> types. The accuracy,<br />
quality and currency of the <strong>data</strong> should be determ<strong>in</strong>ed with reference to the cost of collect<strong>in</strong>g those<br />
<strong>data</strong> and the value and benefit of that <strong>data</strong> (Western European Road Directorate 2003). The issues<br />
are summarised as follows:<br />
With<strong>in</strong> a year – number of sampl<strong>in</strong>g periods, e.g. once, twice or more times per year;<br />
duration of each sampl<strong>in</strong>g period, e.g. 1 h, 1 day, 1 week or more; when to survey, e.g.<br />
randomly, outside school holidays or at each season; consistency with previous surveys;<br />
Between years – predictability and consistency of traffic growth; predictability and<br />
consistency of factors lead<strong>in</strong>g to traffic growth; existence or otherwise of commodity <strong>in</strong> traffic<br />
flows between areas or <strong>road</strong> types, e.g. rural versus urban or National Highways versus ma<strong>in</strong><br />
<strong>road</strong>s;<br />
External demands – <strong>in</strong>ternal <strong>road</strong> authority requirement; external government agency<br />
requirements; political or public relations demands.<br />
The frequency of collect<strong>in</strong>g short-term traffic counts varies amongst RAs. The current <strong>practice</strong>s<br />
range from a frequency of a sample once every year to as <strong>in</strong>frequent as ten years apart. Most RAs<br />
sample at a frequency of once every two to three years. With<strong>in</strong> this sample program, one half or<br />
one third of the sites are sampled <strong>in</strong> any year, result<strong>in</strong>g <strong>in</strong> all sites covered over a two to three year<br />
cycle (described as a ‘roll<strong>in</strong>g’ program). Increas<strong>in</strong>g budgetary pressures are lead<strong>in</strong>g RAs to lower<br />
<strong>collection</strong> frequencies, e.g. from three to four years to five to six years <strong>in</strong> South Australia.<br />
Further ‘f<strong>in</strong>e-tun<strong>in</strong>g’ is applied with<strong>in</strong> a sampl<strong>in</strong>g frequency regime, by differentiat<strong>in</strong>g between<br />
classes of <strong>road</strong> (urban and rural) as well as with<strong>in</strong> classes of <strong>road</strong> (ma<strong>in</strong>ly rural). Thus, Western<br />
Australia reported a two year metropolitan cycle and a five year rural cycle while Tasmania<br />
reported sampl<strong>in</strong>g 100 sites every two years, 300 sites every five years and 450 sites every ten<br />
years.<br />
It is good <strong>practice</strong> that the sampl<strong>in</strong>g <strong>in</strong>terval should not extend beyond three years, under what can<br />
generally be described as ‘normal’ conditions. In <strong>practice</strong>, sampl<strong>in</strong>g frequency is a balance of<br />
costs and accuracy, and recognition of chang<strong>in</strong>g traffic conditions, best determ<strong>in</strong>ed by the <strong>use</strong>rs of<br />
the <strong>data</strong> themselves. As such, for rural <strong>road</strong>s where traffic counts hardly change, a ten year<br />
sampl<strong>in</strong>g frequency may be justifiable, as with some <strong>road</strong>s <strong>in</strong> Tasmania. This, of course, would not<br />
apply to rural <strong>road</strong>s closer to urban areas, where it is likely that conditions will change more rapidly.<br />
The local knowledge of RAs enables f<strong>in</strong>e tun<strong>in</strong>g with<strong>in</strong> portfolios of count stations, to achieve cost<br />
effectiveness <strong>in</strong> the count<strong>in</strong>g program.<br />
4.3.2 Def<strong>in</strong><strong>in</strong>g a Homogeneous Section<br />
Homogeneous sections are utilised to transpose AADTs calculated for specific survey locations, as<br />
part of the sample traffic count<strong>in</strong>g program, across to <strong>road</strong> sections <strong>in</strong> the total <strong>road</strong> network. By<br />
assign<strong>in</strong>g an AADT for each <strong>road</strong> section <strong>in</strong> the network, a calculation for VKT can be made,<br />
multiply<strong>in</strong>g the <strong>road</strong> section length by the AADT and add<strong>in</strong>g up across all <strong>road</strong> sections.<br />
Homogeneous sections are groups of <strong>road</strong>s or ‘strata’ or ‘uniform traffic segments’ (UTS), each of<br />
which <strong>in</strong>cludes a relatively large number of segments that can be regarded as a ‘population’ for<br />
statistical sampl<strong>in</strong>g. Homogeneous sections also relate to groups of <strong>road</strong>s to which adjustment<br />
factors, described earlier, can be applied. FHWA (2001) describes the creation of ‘factor groups’<br />
(similar to homogeneous sections) as follows:
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 26 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
The factor<strong>in</strong>g process def<strong>in</strong>es a set of <strong>road</strong>s as a ‘group’. All <strong>road</strong>s with<strong>in</strong> that<br />
group are assumed to behave similarly. Then a sample of locations on <strong>road</strong>s<br />
from with<strong>in</strong> that group is taken and <strong>data</strong> are collected. The mean condition for<br />
that sample is computed and that mean value is <strong>use</strong>d as the ‘best’ measure of<br />
how all <strong>road</strong>s <strong>in</strong> the group behave. If the sample of <strong>data</strong> <strong>collection</strong> sites is<br />
randomly selected and moderately large, the distribution of that measure about<br />
the mean is a good measure of how well that mean applies to <strong>road</strong> sections <strong>in</strong><br />
the group (FHWA 2001).<br />
The def<strong>in</strong>ition of a homogeneous <strong>road</strong> section is complicated by differences between rural and<br />
urban sections across a State and region. Homogeneity can be derived from functional classes as<br />
well as from stratification by volume. Various homogeneous group<strong>in</strong>gs <strong>in</strong>clude those <strong>in</strong> Table 12<br />
by functional classes and Table 13 by AADT strata.<br />
US<br />
FHWA (2001):<br />
• <strong>in</strong>terstate rural<br />
• other rural<br />
• <strong>in</strong>terstate urban<br />
• other urban<br />
• recreational<br />
Table 12 - Groups by functional classes:<br />
Canada<br />
(Mart<strong>in</strong> and Chiang 2002)<br />
• regional commuter<br />
• average rural<br />
• partially recreational<br />
• recreational<br />
New Zealand<br />
(Transfund NZ 2001)<br />
• urban arterial – major, m<strong>in</strong>or<br />
• urban – commercial, <strong>in</strong>dustrial,<br />
other<br />
• rural urban fr<strong>in</strong>ge<br />
• rural strategic<br />
• rural recreational – w<strong>in</strong>ter, summer<br />
VicRoads is implement<strong>in</strong>g a classification system for its declared <strong>road</strong> types based on uniform <strong>road</strong><br />
geometry and performance specifications to <strong>in</strong>voke consistent <strong>use</strong>r expectations. The <strong>road</strong> network<br />
then becomes a set of clearly stratified sub-networks each of reasonably similar AADT.<br />
Table 13 provides the current Aust<strong>road</strong>s guide on deriv<strong>in</strong>g homogeneous sections based on AADT<br />
volume <strong>data</strong>, categorised by urban and rural <strong>road</strong>s. In the study on traffic counts for RTA, Botterill<br />
and Luk (1998) found that the rural AADT strata of (0, 100) and 7,000-plus could be expanded to<br />
give better estimation.
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
Table 13 - Suggested volume stratification of <strong>road</strong> segments<br />
Stratum Suggested range<br />
for urban <strong>road</strong>s<br />
(AADT)<br />
1<br />
2<br />
3<br />
4<br />
5<br />
6<br />
7<br />
8<br />
9<br />
10<br />
0-300<br />
301-1,100<br />
1,101-2,000<br />
2,001-4,000<br />
4,001-7,000<br />
7,001-16,000<br />
16,001-20,000<br />
20,001-35,000<br />
35,001-50,000<br />
50,000 plus<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 27 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
Suggested range<br />
for rural <strong>road</strong>s<br />
(AADT)<br />
0-100<br />
101-300<br />
301-700<br />
701-1,100<br />
1,101-2,000<br />
2,001-4,000<br />
4,001-7,000<br />
7,000 plus<br />
Homogeneous or uniform traffic sections should be determ<strong>in</strong>ed <strong>in</strong> a consistent way based on a<br />
predef<strong>in</strong>ed set of rules. Road authorities would have their own <strong>in</strong>dividual procedures <strong>in</strong> def<strong>in</strong><strong>in</strong>g<br />
and process<strong>in</strong>g homogeneous sections. An example of rules for def<strong>in</strong><strong>in</strong>g homogeneous traffic<br />
section, sourced from Ma<strong>in</strong> Roads WA and Transit NZ and reported below. those rules can be<br />
<strong>use</strong>d as a start<strong>in</strong>g po<strong>in</strong>t to generate traffic sections that can be reviewed by stakeholders with local<br />
knowledge for consistency and accuracy. Each homogeneous section to be selected up to its<br />
maximum length, <strong>in</strong>clud<strong>in</strong>g ramps;<br />
Homogeneous sections to be delimited by <strong>in</strong>tersections with arterial and sub-arterial <strong>road</strong>s;<br />
A homogeneous section should not extend over two <strong>road</strong>s;<br />
A homogeneous section should not conta<strong>in</strong> a s<strong>in</strong>gle and dual carriage way or a ‘one-way’<br />
<strong>road</strong>;<br />
A homogeneous section cannot span across different speed zones<br />
The start and end po<strong>in</strong>ts of homogeneous sections should not co<strong>in</strong>cide with start and end<br />
po<strong>in</strong>ts of strategic l<strong>in</strong>ks (cannot have a homogeneous section overlapp<strong>in</strong>g the lengths of two<br />
consecutive strategic l<strong>in</strong>ks); and<br />
A difference <strong>in</strong> AADT volumes between two consecutive count sites of 15 - 100% generates<br />
a new traffic section, as shown <strong>in</strong> Table 14. Similar rules can be applied <strong>in</strong> regard to traffic<br />
composition.
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 28 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
Table 14 - Suggested volume stratification based on % change <strong>in</strong> AADT<br />
From AADT or AAWT To AADT or AAWT<br />
(Source: Ma<strong>in</strong> Roads WA)<br />
% change above which a new traffic<br />
section is required<br />
40,000 50,000 15%<br />
30,000 40,000 20%<br />
20,000 30,000 25%<br />
10,000 20,000 30%<br />
7500 10,000 35%<br />
5,000 7,500 40%<br />
4,000 5,000 45%<br />
200 400 50%<br />
10 20 100%<br />
Road Authorities deal with a total number of homogeneous <strong>road</strong> sections vary<strong>in</strong>g from 500<br />
segments to 5,000 segments. These segments will change from time to time and various recent<br />
techniques have offered potentially new solutions <strong>in</strong> the updat<strong>in</strong>g of <strong>road</strong> segments. Thoresen and<br />
Michel (2002) reported on the <strong>use</strong> of hourly flow histograms from Pattern Stations to develop<br />
default hourly histograms for other sites and ultimately across the entire network. Such a<br />
technique provides, <strong>in</strong>directly, another means of deriv<strong>in</strong>g homogeneous sections across the<br />
network.<br />
RTA NSW employs an alternative technique <strong>in</strong> def<strong>in</strong><strong>in</strong>g a homogeneous section and a network is<br />
divided at po<strong>in</strong>ts of major flow changes. RTA (2003) proposed a computational solution based on a<br />
statistical set of rules to update UTSs (Uniform Traffic Sections or Segments) automatically with<br />
m<strong>in</strong>imal manual ma<strong>in</strong>tenance. The techniques are based on work<strong>in</strong>g with an alternative set of<br />
<strong>road</strong> sections, called ‘Weighted Road Segments’ (WRS) that encompass a number of UTSs and<br />
def<strong>in</strong><strong>in</strong>g UTSs <strong>in</strong> terms of changes <strong>in</strong> AADTs with<strong>in</strong> each WRS by a 95% statistically significant<br />
measure of twice the square root of AADT, based on the assumption that AADT is distributed as a<br />
Poisson variable.<br />
While it is easy to def<strong>in</strong>e a <strong>road</strong> section on AADT volume stratification, the same <strong>road</strong> section may<br />
fall under a number of <strong>road</strong> classifications (urban, rural, etc). Furthermore, the def<strong>in</strong>ition of the<br />
group changes depend<strong>in</strong>g on the vehicle classification. This raises another implication for good<br />
<strong>practice</strong> - the need to ma<strong>in</strong>ta<strong>in</strong> a historical record of changes to homogeneous sections and<br />
classifications over time.<br />
In summary, a good <strong>practice</strong> among RAs <strong>in</strong> def<strong>in</strong><strong>in</strong>g homogeneous sections, and hence the<br />
calculation of AADT and VKT <strong>in</strong> a statewide count<strong>in</strong>g program, would <strong>in</strong>volve the follow<strong>in</strong>g iterative<br />
steps:<br />
(a) Locate permanent and short-term count stations based on judgement and <strong>road</strong> functional<br />
classes (e.g. Table 12);<br />
(b) Collect counts;<br />
(c) Divide <strong>road</strong> network us<strong>in</strong>g volume stratification and traffic pattern behaviour;<br />
(d) Determ<strong>in</strong>e the number of stations required by apply<strong>in</strong>g a volume stratification technique (e.g.<br />
Tables 13 and 14);
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 29 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
(e) Review count station density and location;<br />
(f) Calculate adjustment factors to convert short-term counts to AADT estimates;<br />
(g) Identify homogenous traffic sections and calculate VKT; and<br />
(h) Check, review and prepare a roll<strong>in</strong>g program with stakeholders who have good local<br />
knowledge.<br />
4.3.3 Directionality<br />
The count<strong>in</strong>g of traffic on <strong>road</strong> sections can be either directional or bi-directional (two-way).<br />
Typically, most RAs report directional AADTs for urban <strong>road</strong> sections and two-way AADTs for rural<br />
<strong>road</strong> sections. Directional VKT for rural environment has some value and should cont<strong>in</strong>ue to be<br />
measured where possible and most permanent count<strong>in</strong>g stations can provide the <strong>in</strong>formation on<br />
directionality.<br />
More specifically, directionality report<strong>in</strong>g is usually based on carriageway or the sum of all lanes <strong>in</strong><br />
the reported direction. In greater detail, VicRoads reported AADT by direction with breakdowns for<br />
a.m., p.m. and other time periods, while RTA NSW reported only by direction.<br />
In general, directionality should be specified as reported either <strong>in</strong> one direction or bi-directionally,<br />
and that traffic counts are the sum of counts from all lanes <strong>in</strong> the direction reported. When<br />
directionality is unspecified, then an AADT value always refers to a bi-directional (two-way) count.<br />
4.4 Report<strong>in</strong>g<br />
Report<strong>in</strong>g of traffic counts reaches a wide audience. While <strong>in</strong>ternal <strong>use</strong>rs with<strong>in</strong> a <strong>road</strong> authority<br />
may be considered to be familiar <strong>use</strong>rs of such <strong>data</strong>, an <strong>in</strong>creas<strong>in</strong>g number of external <strong>use</strong>rs have<br />
vary<strong>in</strong>g levels of familiarity from novice to experienced. Furthermore, <strong>use</strong>r familiarity with traffic<br />
count <strong>data</strong> and reports may be based on experience with reports and <strong>data</strong> from other jurisdictions,<br />
locally and <strong>in</strong>ternationally, where procedures and methods may be different. In some cases, <strong>use</strong>rs<br />
may be <strong>in</strong>terested <strong>in</strong> further aggregat<strong>in</strong>g <strong>data</strong>, across a number of States, or compar<strong>in</strong>g such <strong>data</strong><br />
with similar <strong>data</strong> from other States.<br />
Road Authorities have on-go<strong>in</strong>g demand from external <strong>use</strong>rs for <strong>road</strong> <strong>use</strong> <strong>data</strong>, on an ad-hoc feefor-service<br />
basis or under Freedom of Information (FOI) legislation. At the same time, RAs are<br />
themselves provid<strong>in</strong>g <strong>use</strong>rs with more access through the Internet, by CDs and publications (both<br />
hard and soft copies).<br />
The October 2003 Road Use Data Workshop addressed issues <strong>in</strong> consistency of reports and this<br />
section covers examples of good <strong>practice</strong> <strong>in</strong> report<strong>in</strong>g (Section 4.4.1), mak<strong>in</strong>g <strong>data</strong> available to<br />
<strong>use</strong>rs (Section 4.4.2). Data <strong>in</strong>tegration and public-private partnership <strong>in</strong> mak<strong>in</strong>g <strong>data</strong> available are<br />
mentioned <strong>in</strong> this section and will be addressed more fully <strong>in</strong> a separate progress report.<br />
4.4.1 Examples of Good Practices<br />
Estimates of AADT and VKT, together with their trend projections, are of key <strong>in</strong>terest to <strong>road</strong><br />
agencies. These <strong>data</strong> are kept <strong>in</strong> numerical tabulation and graphical forms. Primary <strong>in</strong>formation<br />
such as location of the count<strong>in</strong>g site, location plan, weather features and public holidays should be<br />
documented with the respective <strong>data</strong>. The common presentation of the follow<strong>in</strong>g <strong>data</strong> parameters<br />
is discussed below.<br />
AADT<br />
Published maps with historical and current AADT shown at each pattern and short-term<br />
count<strong>in</strong>g station.
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 30 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
Plott<strong>in</strong>g of AADT as ord<strong>in</strong>ates on map of urban <strong>road</strong> network, aga<strong>in</strong>st each segment as an<br />
abscissa.<br />
Axle pairs or Vehicle Counts<br />
Provide raw <strong>data</strong> on axle pairs and adjusted vehicle counts if a s<strong>in</strong>gle axle sensor is <strong>use</strong>d;<br />
Record the method of adjustment;<br />
Mention the level of accuracy.<br />
Seasonal Patterns and Trends of Traffic<br />
Monthly, weekly and daily variation patterns <strong>in</strong> traffic can be illustrated <strong>in</strong> graphical and<br />
tabular forms;<br />
Information on the variation about the average level of traffic can be drawn from the graphical<br />
<strong>data</strong> arranged <strong>in</strong> decreas<strong>in</strong>g order.<br />
Turn<strong>in</strong>g Movements at Intersections<br />
A tabular method is <strong>use</strong>d when classification by vehicle type is required to be present;<br />
Peak and daily hour turn<strong>in</strong>g movements can be illustrated graphically.<br />
Other issues that should be considered are:<br />
Report<strong>in</strong>g <strong>in</strong>terval,<br />
What <strong>data</strong> are reported,<br />
Geographic referenc<strong>in</strong>g of the <strong>in</strong>formation, and<br />
Report presentation format.<br />
As a m<strong>in</strong>imum, it is recommended that report<strong>in</strong>g should be at yearly <strong>in</strong>tervals, provid<strong>in</strong>g AADT for<br />
each section of the <strong>road</strong> network, <strong>in</strong> terms of vehicles per section per time period. Each traffic<br />
count<strong>in</strong>g station should be identified by geographic referenc<strong>in</strong>g us<strong>in</strong>g maps or coord<strong>in</strong>ates.<br />
M<strong>in</strong>imum report<strong>in</strong>g format should <strong>in</strong>clude location, AADT by <strong>road</strong> section and vehicle class, details<br />
of homogeneous <strong>road</strong> sections, VKT (total, by vehicle class, by <strong>road</strong> type, by region), details of<br />
duration of count<strong>in</strong>g and traffic growth over the previous year.<br />
An example of the <strong>data</strong> classes collected and stored for report<strong>in</strong>g <strong>in</strong> Queensland is detailed <strong>in</strong><br />
Table 15.<br />
It is also good <strong>practice</strong> that statistical results be reported jo<strong>in</strong>tly <strong>in</strong> terms of percentage error and<br />
confidence <strong>in</strong>terval. If measurements could be made over the totality of a target population rather<br />
than over just a sample then there would be no need to specify a confidence <strong>in</strong>terval. Beca<strong>use</strong><br />
usually only a sample is taken, there rema<strong>in</strong>s the possibility that this sample is unrepresentative of<br />
the population, so that the error is likewise unrepresentative. It is also good <strong>practice</strong> to specify a<br />
risk factor that the sample could be unrepresentative.<br />
For example, a risk of 1 <strong>in</strong> 20 chance (sample) is taken, which translates to the 95% confidence<br />
<strong>in</strong>terval and this happens with the confidence limits at ±1.96 standard deviations from the mean.<br />
Similarly, with confidence limits at ±1 standard deviation from the mean, the confidence level is at<br />
68 %. Risk should be specified as a percentage error for determ<strong>in</strong><strong>in</strong>g confidence limits.
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 31 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
Table 15 - Complete list of <strong>data</strong> classes collected/computed and stored <strong>in</strong> Queensland<br />
Class Description Class Description<br />
0 Volume 33 Free Speed – Distribution – Low Range<br />
1 Annual Average Daily Traffic 34 Free Speed – Distribution – Mid Range<br />
2 Highest Hourly Volumes 35 Free Speed – Distribution – High Range<br />
3 AM and PM Peaks 40 Gross Vehicle Mass – Mean<br />
5 Exponential Growth 41 Gross Vehicle Mass – Standard Deviation<br />
6 Set Factors 42 Gross Vehicle Mass - 95%<<br />
10 Speed – Mean 43 Gross Vehicle Mass – Distribution<br />
11 Speed – Standard Deviation 50 Axle Group Mass – Mean<br />
12 Speed – 85%< 51 Axle Group Mass – Standard Deviation<br />
13 Speed Distribution – Low Range 52 Axle Group Mass – 95%<<br />
14 Speed Distribution – Mid Range 53 Axle Group Mass – Distribution<br />
15 Speed Distribution – High Range 60 Unloaded<br />
20 Percentage Follow<strong>in</strong>g 61 Loaded<br />
21 Gap Acceptance Overtak<strong>in</strong>g 62 Overloaded Axle Group<br />
22 Headway – Mean 63 Overloaded Vehicle – by Axle Group<br />
23 Headway – Standard Deviation 64 Overloaded Vehicle – by Gross Vehicle Mass<br />
24 Headway – 85%< 69 Mean Net Load – (Total Freight tonne)<br />
25 Headway – Distribution 70 ESA – Unloaded<br />
30 Free Speed – Mean 71 ESA – Loaded<br />
31 Free Speed – Standard Deviation 72 ESA – Overloaded<br />
32 Free Speed – 85%< 73 ESA – Total<br />
80 Vehicle-Kilometre Travelled<br />
4.4.2 Data Availability<br />
Report<strong>in</strong>g also entails mak<strong>in</strong>g <strong>data</strong> available to a range of <strong>use</strong>rs at different levels of access. Such<br />
considerations <strong>in</strong>clude:<br />
Storage formats which should be clearly stated for easy unpack<strong>in</strong>g and process<strong>in</strong>g;<br />
Download formats which should <strong>in</strong>cluded the most commonly <strong>use</strong>d applications (pdf,<br />
spreadsheets, zipped, etc);<br />
Methods of access which should <strong>in</strong>clude traditional hard copies and more common electronic<br />
forms of access (Internet, dial up, help desk, etc);<br />
Limitations on sensitive and commercial <strong>in</strong>formation (if any); and<br />
Copyright and legal issues, regard<strong>in</strong>g the on-sale and distribution of such <strong>in</strong>formation to third<br />
parties.<br />
The existence of Freedom of Information Acts <strong>in</strong> each RAs’ jurisdiction implies that <strong>road</strong> <strong>data</strong><br />
should be available to the public. Indeed, this is already the experience of most RAs <strong>in</strong> deal<strong>in</strong>g<br />
with an <strong>in</strong>creas<strong>in</strong>g amount of requests for such <strong>in</strong>formation. As requests can vary from simple<br />
queries to detailed <strong>in</strong>vestigations requir<strong>in</strong>g further effort, process<strong>in</strong>g or <strong>analysis</strong> of the <strong>data</strong>, it is<br />
reasonable that a <strong>use</strong>r-pay policy applies.
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 32 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
It is good <strong>practice</strong> to develop a schedule of standard <strong>data</strong> and <strong>in</strong>formation that are available (such<br />
as audited and processed <strong>data</strong> required for normal RA report<strong>in</strong>g purposes) and the cost of<br />
provision of those <strong>data</strong> to the public <strong>in</strong> the required formats (hard copy or electronic copy).<br />
A schedule of charges for labour and overheads should also be prepared for non-standard<br />
requests that <strong>in</strong>volve utilis<strong>in</strong>g RA resources <strong>in</strong> fulfill<strong>in</strong>g a private or commercial <strong>in</strong>quiry.<br />
Further sources of <strong>road</strong> <strong>use</strong> <strong>data</strong> are available through <strong>in</strong>frastructure deployed for other purposes.<br />
These <strong>in</strong>clude <strong>WIM</strong> <strong>data</strong> and other <strong>data</strong> collected under the Aust<strong>road</strong>s program on national<br />
performance <strong>in</strong>dicators (Aust<strong>road</strong>s 2001). More recently, SCATS, freeway <strong>in</strong>cident management<br />
systems and other Intelligent Transport Systems now also provide ‘cheap’ traffic <strong>data</strong> efficiently.<br />
Commercial organisations also, <strong>in</strong>creas<strong>in</strong>gly, have access to traffic <strong>data</strong> through the <strong>use</strong> of track<strong>in</strong>g<br />
systems such as GPS and mobile phone position<strong>in</strong>g technology.<br />
New sources of ITS <strong>data</strong> will orig<strong>in</strong>ate, not only from RAs, but also from commercial organisations.<br />
The commercial structures of such arrangements (licens<strong>in</strong>g, jo<strong>in</strong>t venture) will need to be<br />
exam<strong>in</strong>ed. This may lead further down the path of disaggregat<strong>in</strong>g traffic count<strong>in</strong>g <strong>in</strong>to its<br />
component activities, i.e. <strong>collection</strong>, aggregation and dissem<strong>in</strong>ation. Already, out-sourc<strong>in</strong>g occurs<br />
at the <strong>collection</strong> level <strong>in</strong> many RAs.<br />
Furthermore, the delivery or dissem<strong>in</strong>ation channels for this expanded <strong>data</strong> and <strong>in</strong>formation will<br />
<strong>in</strong>clude specialist third parties such as the media, Internet and Telematics service providers as well<br />
as mobile phone operators.<br />
In this new environment, report<strong>in</strong>g of traffic <strong>data</strong> would become highly customised and<br />
personalised, with service providers creat<strong>in</strong>g value by differentiat<strong>in</strong>g and customis<strong>in</strong>g traffic <strong>data</strong><br />
for their <strong>use</strong>r base. There could be complementary roles where the RAs provide the basic raw or<br />
simply processed <strong>data</strong> for <strong>in</strong>ternal <strong>use</strong>rs. The general public and the service providers undertake<br />
further <strong>analysis</strong> and transformation of traffic <strong>data</strong> for their customers.<br />
Data accessibility and its pric<strong>in</strong>g will cont<strong>in</strong>ue to be an issue fac<strong>in</strong>g RAs <strong>in</strong> the years to come.<br />
Policies need to be established. RTA NSW has <strong>in</strong> place manuals for the sale of <strong>data</strong> and f<strong>in</strong>ance<br />
manual on pric<strong>in</strong>g. The development of a pric<strong>in</strong>g policy especially for external <strong>use</strong>rs is complex and<br />
common amongst RAs. It is an area recommended for further research.<br />
The issue of <strong>data</strong> <strong>in</strong>tegration and public-private partnership are further addressed <strong>in</strong> Sections 5<br />
and 6 respectively.
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
5 DATA INTEGRATION<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 33 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
Road authorities are <strong>in</strong>creas<strong>in</strong>gly <strong>in</strong>volved <strong>in</strong> collect<strong>in</strong>g a wide range of <strong>road</strong> <strong>use</strong> <strong>data</strong>. These <strong>data</strong><br />
types can come from a number of sources. There is also an <strong>in</strong>creas<strong>in</strong>g need to efficiently access<br />
the <strong>data</strong> with<strong>in</strong> a RA, or from external parties <strong>in</strong>clud<strong>in</strong>g the public. This raises the issue that <strong>road</strong><br />
<strong>use</strong> <strong>data</strong> should be better <strong>in</strong>tegrated for easy access. This section describes the issues <strong>in</strong>volved,<br />
current <strong>practice</strong>s with<strong>in</strong> RAs and future directions.<br />
Data <strong>in</strong>tegration can be b<strong>road</strong>ly def<strong>in</strong>ed as the br<strong>in</strong>g<strong>in</strong>g together of different types of <strong>road</strong> <strong>use</strong> <strong>data</strong><br />
and other <strong>road</strong> <strong>data</strong> such as the <strong>in</strong>ventory of <strong>road</strong> conditions <strong>in</strong> a central <strong>data</strong>base. The <strong>in</strong>tegration<br />
can utilise several <strong>data</strong>bases, with the capability of referenc<strong>in</strong>g the <strong>data</strong> us<strong>in</strong>g a unique referenc<strong>in</strong>g<br />
system such as <strong>road</strong> numbers and Geographical Information System (GIS) coord<strong>in</strong>ates.<br />
By comb<strong>in</strong><strong>in</strong>g <strong>data</strong>, it is possible to generate further <strong>in</strong>formation than <strong>in</strong> its parts, e.g. the Aust<strong>road</strong>s<br />
(2001) National Performance Indicators Program. However, <strong>in</strong>tegration should not be <strong>in</strong>terpreted<br />
as a simple comb<strong>in</strong>ation of <strong>data</strong> such as averag<strong>in</strong>g <strong>data</strong> sets, merg<strong>in</strong>g <strong>data</strong> from different time<br />
periods of a day or aggregat<strong>in</strong>g <strong>data</strong> across lanes. Also, raw <strong>data</strong> should be ma<strong>in</strong>ta<strong>in</strong>ed <strong>in</strong> a<br />
<strong>data</strong>base as far as possible.<br />
5.1 Road Use Data Types for Integration<br />
The range of <strong>road</strong> <strong>use</strong> and related <strong>data</strong> generally suitable for <strong>in</strong>tegration, and reflect<strong>in</strong>g the<br />
diversity of traffic <strong>in</strong>formation needs, <strong>in</strong>cludes the follow<strong>in</strong>g:<br />
Traffic volumes, traffic composition (i.e. classified counts) and turn<strong>in</strong>g movement counts;<br />
Pavement load<strong>in</strong>g <strong>data</strong> from weigh-<strong>in</strong>-motion (<strong>WIM</strong>) systems and weigh bridges;<br />
Pavement condition rat<strong>in</strong>g;<br />
Transport demand <strong>data</strong> such as orig<strong>in</strong>-dest<strong>in</strong>ation <strong>data</strong>;<br />
Route performance <strong>data</strong>, e.g. travel time, level of service or congestion <strong>in</strong>dex;<br />
Traffic management <strong>data</strong> from ITS, signal systems or freeway management systems,<br />
Road crash <strong>data</strong>, e.g. exposure measure, risk or black spots; and<br />
Asset management <strong>data</strong> such as bridges, tunnels, signs and mark<strong>in</strong>gs and guard rails.<br />
The above list is not exhaustive and other <strong>data</strong> types can be <strong>in</strong>cluded <strong>in</strong> an <strong>in</strong>tegrated <strong>data</strong>base<br />
now and <strong>in</strong> the future. Various RAs have already started the <strong>in</strong>tegration process, although there<br />
are no standard procedures, software or types of <strong>data</strong> for <strong>in</strong>tegration. Efficient access us<strong>in</strong>g<br />
<strong>data</strong>base technologies will benefit a wide range of <strong>use</strong>rs with<strong>in</strong> and external to a jurisdiction for<br />
many different needs. An example is the need for a <strong>road</strong> crash exposure <strong>in</strong>dicator and one that<br />
can correlate with <strong>road</strong> crashes already stored <strong>in</strong>, say, a GIS framework. For urban <strong>road</strong>s, that<br />
exposure <strong>in</strong>dicator can be a low-cost <strong>in</strong>dicator available from <strong>road</strong> traffic counts <strong>in</strong> a signal control<br />
system, assum<strong>in</strong>g that signal counts are stored rout<strong>in</strong>ely.<br />
Figure 9 shows the framework of the ARMIS <strong>data</strong>base of the Queensland Department of Ma<strong>in</strong><br />
Roads. The ARMIS framework covers most of the <strong>data</strong> types mentioned above, supplemented with<br />
<strong>road</strong> and bridge <strong>in</strong>ventory <strong>data</strong>. These <strong>data</strong> types are entered <strong>in</strong>to a <strong>data</strong>base, which has an audit<br />
utility. At the core of the <strong>data</strong>base is a <strong>road</strong> reference and <strong>road</strong> <strong>in</strong>ventory that specifies locations for<br />
traffic <strong>analysis</strong>, <strong>road</strong> crash <strong>analysis</strong>, pavement condition, <strong>road</strong> ma<strong>in</strong>tenance performance contracts<br />
and construction management. After an audit<strong>in</strong>g process, the <strong>data</strong> is sent to the Road Information<br />
Data Centre to support a wide range of <strong>road</strong> <strong>data</strong> <strong>use</strong>rs.
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
ARMIS<br />
Remote Bridge<br />
Logg<strong>in</strong>g Systems<br />
Road Crash<br />
Traffic<br />
Count<strong>in</strong>g Tools<br />
Condition Data<br />
Acquisition<br />
Remote Feature<br />
Logg<strong>in</strong>g Systems<br />
Traffic Analysis<br />
& Report<strong>in</strong>g<br />
Pavement<br />
Condition<br />
External Ma<strong>in</strong>tenance<br />
Management Systems<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 34 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
Figure 9 – An <strong>in</strong>tegrated <strong>road</strong> <strong>use</strong> <strong>data</strong>base from Queensland Ma<strong>in</strong> Roads<br />
It is therefore recommended as a good <strong>practice</strong> that traffic <strong>data</strong> from count<strong>in</strong>g stations, manual<br />
counts, <strong>WIM</strong> sites and signal systems be <strong>in</strong>tegrated as a first step. Other <strong>data</strong> to follow should<br />
<strong>in</strong>clude <strong>data</strong> previously mentioned - Aust<strong>road</strong>s National Performance Indicators, <strong>road</strong> management<br />
<strong>data</strong>, crash rates and others.<br />
Ma<strong>in</strong> Roads WA also reported the <strong>use</strong> of a relational <strong>data</strong>base management system to receive and<br />
validate raw traffic count <strong>data</strong> (Mihai 20<strong>04</strong>). The <strong>data</strong> is then uploaded to the Integrated Road<br />
Information System (IRIS) which has the functions of:<br />
Ma<strong>in</strong>ta<strong>in</strong><strong>in</strong>g traffic section def<strong>in</strong>itions on a <strong>road</strong> network;<br />
Manag<strong>in</strong>g the edit<strong>in</strong>g, storage and access to <strong>data</strong> on each traffic section, and at each count<br />
site;<br />
Ma<strong>in</strong>ta<strong>in</strong><strong>in</strong>g count site locations;<br />
Road Crash 2<br />
Road Reference<br />
and<br />
Road Inventory<br />
Road Ma<strong>in</strong>tenance<br />
Performance Contracts<br />
Bridge<br />
Information<br />
Construction<br />
Management<br />
System<br />
Ma<strong>in</strong>ta<strong>in</strong><strong>in</strong>g relationships between count sites; and<br />
Roads<br />
Information<br />
Data<br />
Centre<br />
Roads Info<br />
Directory<br />
Service<br />
Chartview<br />
Ma<strong>in</strong>ta<strong>in</strong><strong>in</strong>g and publish<strong>in</strong>g high level annual traffic statistics for count sites and <strong>road</strong><br />
sections.<br />
Some <strong>road</strong> authorities further reported the <strong>use</strong> of proprietary weigh-<strong>in</strong>-motion <strong>data</strong>bases (e.g. <strong>WIM</strong><br />
Net) to facilitate the process<strong>in</strong>g, storage and access of weigh-<strong>in</strong>-motion <strong>data</strong>. These <strong>data</strong>bases<br />
perform additional functions such as calibrat<strong>in</strong>g for weight drift and time clock <strong>in</strong>accuracies (see<br />
Section 6 on quality assurance). They can assist with identify<strong>in</strong>g overload<strong>in</strong>g trend and associated<br />
axle configuration patterns.<br />
RIPA<br />
F<strong>in</strong>ancial<br />
Data<br />
Presentation<br />
& Analysis<br />
RIO<br />
ARMIS GIS<br />
Mapview<br />
Road Crash<br />
GIS<br />
Databrowser<br />
Query Tool<br />
Spreadsheet<br />
Scenario
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 35 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
Other RAs have also reported the <strong>use</strong> of <strong>data</strong>base technologies to facilitate <strong>data</strong> <strong>in</strong>tegration. The<br />
frameworks currently adopted (or be<strong>in</strong>g developed) <strong>in</strong> Queensland Ma<strong>in</strong> Roads and Ma<strong>in</strong> Roads<br />
WA represent examples of good <strong>practice</strong> <strong>in</strong> <strong>data</strong> <strong>in</strong>tegration.<br />
Many other opportunities are possible and can be exploited perhaps with private sector fund<strong>in</strong>g <strong>in</strong><br />
a public-private partnership arrangement (Section 5.3). Also, the many applications of <strong>road</strong> <strong>use</strong><br />
<strong>data</strong> necessitate consultation with various stakeholders, and this issue is addressed <strong>in</strong> Section 7.<br />
5.2 Correlation of Traffic Composition and <strong>WIM</strong> Data<br />
<strong>WIM</strong> sites can provide classified counts, gross vehicle masses and axle group masses, with some<br />
<strong>WIM</strong> systems also provid<strong>in</strong>g <strong>in</strong>dividual axle load <strong>in</strong>formation. <strong>WIM</strong> sites are ma<strong>in</strong>ly located on<br />
major freight corridors, and many of them are permanent sites, e.g. CULWAY sites. These<br />
count<strong>in</strong>g sites can be <strong>use</strong>d to calculate AADTs if all lanes of traffic are monitored. They can be<br />
treated as Pattern Stations, although they may not be at the best statistically representative<br />
locations. Short-term <strong>WIM</strong> sites can certa<strong>in</strong>ly provide an extra sample of counts to supplement<br />
current Short-term Stations. A good <strong>practice</strong> is therefore treat<strong>in</strong>g them as part of an <strong>in</strong>tegrated<br />
count<strong>in</strong>g program.<br />
An important issue is whether traffic composition at a count site can correlate with axle/vehicle load<br />
<strong>data</strong> at a nearby or related <strong>WIM</strong> site. In other words, the issue is whether it is possible to relate the<br />
tonnage of a heavy vehicle from its axle configuration us<strong>in</strong>g historical <strong>WIM</strong> <strong>data</strong>. Pearson and<br />
Foley (2001) identified a relationship shown <strong>in</strong> Table 16 from <strong>data</strong> collected at 16 CULWAY sites <strong>in</strong><br />
Victoria from 1995-1998 for two vehicle classes.<br />
Class 9 Articulated vehicle<br />
[0 00 000]<br />
Table 16 - Mean axle group and gross vehicle mass<br />
Mean mass (tonne)<br />
1995 1996 1997 1998<br />
Axle Group Mass 2 8.9 9.4 9.5 9.2<br />
Axle Group Mass 3 10.1 10.7 10.7 10.5<br />
Gross Vehicle Mass (GVM) 23.5 25.3 25.3 24.9<br />
Class 10 B-double<br />
[0 00 000 000]<br />
Axle Group Mass 2 8.6 9.5 9.5 9.3<br />
Axle Group Mass 3 9.7 10.8 10.9 10.7<br />
Axle Group Mass 4 9.2 10.4 10.4 10.2<br />
Gross Vehicle Mass (GVM) 32.0 35.9 36.0 35.6<br />
Table 16 shows that the average GVMs of the two vehicle types were 25 tonnes and 36 tonnes<br />
respectively. It is therefore possible to ga<strong>in</strong> an approximate <strong>in</strong>dication of the axle and gross vehicle<br />
load from particular axle configurations. However, further <strong>analysis</strong> of the heavy vehicle <strong>data</strong><br />
reported for the National Transport Commission (formerly the National Road Transport<br />
Commission, Ramsay et al. 2001) showed that the mass distribution for a particular vehicle class is<br />
not a normal distribution. Ramsay et al. reported that the <strong>data</strong> were compiled from the <strong>WIM</strong><br />
<strong>data</strong>bases of each State, totall<strong>in</strong>g some 29 million records of vehicle spac<strong>in</strong>gs and weights. The<br />
<strong>data</strong> were collected <strong>in</strong> the period 1993 to 2000, with predom<strong>in</strong>antly pre-1999 <strong>data</strong>.
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 36 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
Figure 10 shows the distribution for six-axle Class 9 (‘123’) vehicles <strong>in</strong> all <strong>WIM</strong> sites surveyed. It<br />
has two peaks, one at around 16 tonnes most likely for an unladen vehicle. Another peak is at 42<br />
tonnes most likely for a fully laden vehicle (the maximum mass limit is 42.5 tonnes).<br />
Figure 11 shows a similar bi-polar distribution of GVM distribution for B-doubles (‘1233’ or n<strong>in</strong>eaxle<br />
Class 10 vehicle). The first peak is at 26 tonnes and the second at 58 tonnes (the maximum<br />
mass limit of a B-double is 62.5 tonnes).<br />
Note that the y-axis show<strong>in</strong>g the distributions were not scaled <strong>in</strong> the orig<strong>in</strong>al document (Ramsay et<br />
al. 2001). This omission is due to the sensitive nature of the <strong>data</strong>.<br />
Peters (2002) showed similar results for all heavy vehicles enter<strong>in</strong>g and exit<strong>in</strong>g the Port of<br />
Fremantle <strong>in</strong> WA (Figure 12). He also reported that 8% of six-axle vehicles, 13% of B-doubles and<br />
4% of double <strong>road</strong>-tra<strong>in</strong>s were overloaded. The majority of overload<strong>in</strong>g occurred on the <strong>in</strong>-bound<br />
route <strong>in</strong>to the Port. The degree of overload<strong>in</strong>g varied and static weighbridges are needed for<br />
enforc<strong>in</strong>g compliance.<br />
Distribution of ’123’ vehicles<br />
0 10 20 30 40 50 60 70 80<br />
Gross Vehicle Mass (t)<br />
Figure 10 – Total mass of three-axle prime movers and three-axle semi-trailers<br />
(Source: Ramsay et al. 2001)
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
Vehicles<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 37 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
Distribution of ’1233’ vehicles<br />
0 10 20 30 40 50 60 70 80 90 100<br />
Gross Vehicle Mass (t)<br />
Figure 11 – Total mass of n<strong>in</strong>e-axle B-doubles<br />
(Source: Ramsay et al. 2001)<br />
Figure 12 – Laden and unladen vehicles <strong>in</strong> each vehicle class at Port Fremantle, WA<br />
It is thus concluded that mean values such as those <strong>in</strong> Table 16 can only give an approximate<br />
<strong>in</strong>dication of bridge and pavement load<strong>in</strong>g from axle configurations. At a particular count<strong>in</strong>g station,<br />
an <strong>in</strong>dication of whether a heavy vehicle is laden or unladen can substantially improve the<br />
accuracy of load estimation from axle configurations. This can be obta<strong>in</strong>ed from freight surveys,<br />
proximity to an exist<strong>in</strong>g <strong>WIM</strong> site, on-site calibration us<strong>in</strong>g portable <strong>WIM</strong> equipment, and subjective<br />
observations. For example, if the ratio of loaded and unloaded vehicles at a count<strong>in</strong>g station is<br />
consistent with a nearby <strong>WIM</strong> site, then an accurate estimate of pavement load<strong>in</strong>g could be<br />
determ<strong>in</strong>ed.
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 38 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
Road authorities have undertaken correlation work on <strong>WIM</strong> <strong>data</strong> with classified counts to determ<strong>in</strong>e<br />
equivalent standard axle factors and other <strong>data</strong> for pavement deterioration modell<strong>in</strong>g. Accord<strong>in</strong>g to<br />
Transit New Zealand, a key issue is that the storage, retrieval and process<strong>in</strong>g of the large amount<br />
of <strong>WIM</strong> <strong>data</strong> consume a lot of time and stretches the capacity of the computer system manag<strong>in</strong>g<br />
the <strong>data</strong>.<br />
5.3 Public-Private Partnership<br />
The future development of <strong>road</strong> <strong>use</strong> <strong>data</strong> as a corporate asset is dependent on how much <strong>data</strong><br />
<strong>collection</strong> and management is carried out by the government sector. Budgetary constra<strong>in</strong>ts with<strong>in</strong><br />
RAs may limit a <strong>data</strong> <strong>collection</strong> program. Many <strong>data</strong> <strong>collection</strong> activities have also been<br />
outsourced. At the same time, the private sector may see opportunities to participate <strong>in</strong> the <strong>data</strong><br />
<strong>collection</strong> program, add value and market the <strong>data</strong> as a commercially viable bus<strong>in</strong>ess.<br />
The current Aust<strong>road</strong>s Intelligent Access Program (I<strong>AP</strong>) is an example of exploit<strong>in</strong>g this publicprivate<br />
partnership (Koniditsiotis 2003). This project <strong>in</strong>vites third party service providers (the<br />
private sector) to monitor <strong>in</strong> real time the compliance of heavy vehicles <strong>in</strong> follow<strong>in</strong>g jurisdictional<br />
regulations such as speed, load and route access. In exchange for be<strong>in</strong>g monitored with the <strong>use</strong><br />
of ITS technologies such as on-board GPS units, the transport <strong>in</strong>dustry expects to ga<strong>in</strong> a<br />
concession <strong>in</strong> access<strong>in</strong>g a <strong>road</strong> network. The service provider charges the transport <strong>in</strong>dustry for<br />
track<strong>in</strong>g vehicles that subscribe to the service, and also have the freight/heavy vehicle <strong>data</strong> that<br />
can potentially be another source of <strong>in</strong>come. The I<strong>AP</strong> is currently <strong>in</strong> the early stages of<br />
implementation after a two-year feasibility study.<br />
Integration and <strong>analysis</strong> of these new sources of <strong>data</strong> will be costly and clear bus<strong>in</strong>ess cases will<br />
be required to support <strong>in</strong>vestment <strong>in</strong> this area. Progress <strong>in</strong> <strong>data</strong> fusion will most likely occur <strong>in</strong><br />
close consultation and cooperation with <strong>use</strong>rs to ensure that specific customer-driven traffic <strong>data</strong><br />
needs are be<strong>in</strong>g met <strong>in</strong> aggregat<strong>in</strong>g such <strong>data</strong>. In most cases, there will not be one s<strong>in</strong>gle <strong>use</strong>r,<br />
but potentially multi-<strong>use</strong>r scenarios, br<strong>in</strong>g<strong>in</strong>g benefits, for example, to traffic management<br />
operations as well as to <strong>road</strong> <strong>use</strong>rs who want to know the impact of <strong>in</strong>cidents along their routes.<br />
Other similar public-private bus<strong>in</strong>ess models have been promoted for the future development of<br />
various ITS services. These services <strong>in</strong>clude, <strong>in</strong> particular, the provision of travel time <strong>in</strong>formation<br />
(see, e.g. Karl et al. 2003). Others could be speed camera operation and adm<strong>in</strong>istration, toll <strong>road</strong>s,<br />
operations of traffic <strong>in</strong>formation centre.<br />
These trends require policy and pric<strong>in</strong>g decisions for RAs <strong>in</strong> terms of the extent of private sector<br />
<strong>in</strong>volvement and price or cost of services or <strong>data</strong> to be provided to the private sector. Road<br />
authorities at present adopt a variety of policies on this issue. Some have gone further down the<br />
path of out-sourc<strong>in</strong>g and are charg<strong>in</strong>g fees for provid<strong>in</strong>g <strong>data</strong> to other parties. A best <strong>practice</strong> on<br />
this issue has yet to be developed, but the emerg<strong>in</strong>g trends suggest that the public-private<br />
partnership could be a w<strong>in</strong>-w<strong>in</strong> bus<strong>in</strong>ess model.
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
6 DATA INTEGRITY<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 39 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
6.1 Issues<br />
Data <strong>in</strong>tegrity pr<strong>in</strong>cipally means the provision of accurate <strong>data</strong> <strong>in</strong> a <strong>use</strong>r-friendly format on demand,<br />
but it can often mean different th<strong>in</strong>gs to different people (Daltrey 2002).<br />
There is little doubt that <strong>in</strong>tegrity or quality is of major concern <strong>in</strong> <strong>road</strong> <strong>use</strong> <strong>data</strong> <strong>collection</strong> and<br />
process<strong>in</strong>g. Consider the issues faced <strong>in</strong> RTA NSW (and similarly issues <strong>in</strong> other RAs). Traffic<br />
<strong>data</strong> <strong>in</strong> NSW are collected at 1,300 Short-term Stations every year and from 500 Pattern Stations<br />
on a cont<strong>in</strong>uous basis. Both types of <strong>data</strong> must then be edited to remove corrupt <strong>data</strong> and adjust<br />
for miss<strong>in</strong>g <strong>data</strong> prior to release and preparation of summary statistics. Edit<strong>in</strong>g is a timeconsum<strong>in</strong>g<br />
task that is partly computer-assisted. There could be different levels of consistency <strong>in</strong><br />
edit<strong>in</strong>g, which would impact upon the quality of the processed <strong>data</strong>. The task can be onerous,<br />
repetitive, on-go<strong>in</strong>g and time-consum<strong>in</strong>g. The manual edit<strong>in</strong>g needs to be replaced with an<br />
automated process. RTA NSW has been test<strong>in</strong>g some edit<strong>in</strong>g software jo<strong>in</strong>tly developed by RTA<br />
NSW and CSIRO (Donnelly et al. 2003) that could be of <strong>use</strong> to other RAs when thoroughly tested.<br />
Data quality can be affected by a variety of factors. These factors could be environmental factors<br />
such as ra<strong>in</strong> and magnetic fields that may <strong>in</strong>fluence equipment accuracy or abnormal traffic events<br />
<strong>in</strong> the connect<strong>in</strong>g <strong>road</strong> network. In consequence, it is necessary to treat events <strong>in</strong> traffic systems as<br />
random variables and <strong>use</strong> statistical theory to <strong>in</strong>vestigate the impact of <strong>in</strong>teract<strong>in</strong>g factors. Unusual<br />
variability with<strong>in</strong> traffic statistics derived from one survey station should be verified if possible from<br />
surround<strong>in</strong>g survey stations or previous <strong>data</strong> at the same station. The success of a statistical<br />
<strong>in</strong>vestigation depends largely on the size and validity of the traffic survey <strong>use</strong>d to collect the <strong>data</strong>.<br />
Daltrey (2002) further suggested that quality is also def<strong>in</strong>ed <strong>in</strong> terms of fitness of purpose.<br />
Consider the def<strong>in</strong>ition of AADT as the total yearly count divided by the number of days. If there is<br />
a <strong>data</strong> loss beca<strong>use</strong> of equipment failure, then the miss<strong>in</strong>g <strong>data</strong> can be estimated by established<br />
procedures. However, if a <strong>road</strong> is closed for some proportions of a year, the AADT statistics based<br />
on sample counts (estimated by factor<strong>in</strong>g from associated permanent traffic count stations) would<br />
differ from those statistics obta<strong>in</strong>ed from a permanent <strong>in</strong>-situ site. As mentioned previously,<br />
weather, <strong>road</strong> works and other environmental factors can ca<strong>use</strong> transient reductions <strong>in</strong> traffic<br />
levels. These factors could lead to AADT values that differ from trend and expectation <strong>in</strong> the<br />
context of AADT values from surround<strong>in</strong>g traffic survey sites. The issue is whether the trend-based<br />
AADT value should be <strong>use</strong>d (e.g. for plann<strong>in</strong>g purposes), or whether the <strong>in</strong>cident-affected values<br />
be <strong>use</strong>d (e.g. as an exposure measure <strong>in</strong> <strong>road</strong> crash research). Users of <strong>data</strong> should therefore pay<br />
attention to this issue of fitness of purpose.<br />
This section discusses the follow<strong>in</strong>g key quality issues:<br />
Specify<strong>in</strong>g <strong>data</strong> <strong>collection</strong> Quality Assurance (QA) procedures for field survey activities,<br />
Deal<strong>in</strong>g with miss<strong>in</strong>g <strong>data</strong>,<br />
Calibration of a <strong>WIM</strong> site us<strong>in</strong>g CULWAY as an example,<br />
Specification of accuracies.<br />
It is worth repeat<strong>in</strong>g that <strong>road</strong> <strong>use</strong> <strong>data</strong> <strong>collection</strong> and process<strong>in</strong>g is not a new discipl<strong>in</strong>e but an<br />
established operation of a RA over many years. Each RA has established its own <strong>data</strong> quality<br />
procedures ref<strong>in</strong>ed over many years, and these procedures are therefore good <strong>practice</strong>s. This<br />
section highlights some of these <strong>practice</strong>s as examples of best <strong>practice</strong>s.
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 40 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
6.2 Specifications for Data Collection Activities<br />
It is recommended that specifications for <strong>data</strong> <strong>collection</strong> activities be developed irrespective of<br />
whether the survey is undertaken by <strong>in</strong>-ho<strong>use</strong> staff or contractors. These specifications can be<br />
<strong>use</strong>d to ensure consistency <strong>in</strong> <strong>data</strong> <strong>collection</strong> and provide a basis for assess<strong>in</strong>g <strong>data</strong> quality.<br />
Surveys undertaken by third parties def<strong>in</strong>e the work by means of a brief. It is recommended that a<br />
modified version of the brief be <strong>use</strong>d for surveys undertaken by <strong>in</strong>-ho<strong>use</strong> staff. GTEP Part 3 states<br />
that the follow<strong>in</strong>g po<strong>in</strong>ts must be <strong>in</strong>cluded <strong>in</strong> a brief:<br />
General outl<strong>in</strong>e of type of survey to be undertaken and purpose;<br />
Details of field survey and survey design (if <strong>in</strong>cluded) required;<br />
Competencies required and field <strong>in</strong>struction detailed;<br />
Criteria for <strong>data</strong> acceptance;<br />
Structure of <strong>data</strong> outcomes; and<br />
Provision of method for conduct<strong>in</strong>g survey, <strong>data</strong> quality plan and tra<strong>in</strong><strong>in</strong>g proposed by<br />
consultants.<br />
Tenders should then be assessed on the basis of management capability (schedul<strong>in</strong>g, cont<strong>in</strong>gency<br />
plann<strong>in</strong>g, staff recruitment and tra<strong>in</strong><strong>in</strong>g, quality plan, resource programm<strong>in</strong>g), technical capabilities<br />
(project management skills of key personnel and level of technical support), price and documented<br />
prior experience.<br />
The key to successful <strong>data</strong> <strong>collection</strong> lies <strong>in</strong> the attention paid by the study manager to the details<br />
of survey adm<strong>in</strong>istration. The follow<strong>in</strong>g checklist may help to reduce the severity of problems<br />
aris<strong>in</strong>g dur<strong>in</strong>g a traffic study. The checklist is particularly suitable for the manual <strong>collection</strong> of <strong>data</strong>,<br />
but the pr<strong>in</strong>ciples are suitable for the <strong>collection</strong> of most <strong>road</strong> <strong>use</strong> <strong>data</strong>.<br />
(a) Personnel must receive tra<strong>in</strong><strong>in</strong>g <strong>in</strong> the purpose of the survey and <strong>in</strong> the methods of<br />
measurements to be employed.<br />
(b) Replacement field staff should be available, especially on the first day of a multi-day survey.<br />
Dur<strong>in</strong>g the survey, rosters of rest periods for observers are essential, and replacement staff<br />
will be needed to cover these periods.<br />
(c) Survey forms should be prepared and distributed, as far as possible, on the eve of the<br />
survey. The pre-study brief<strong>in</strong>g is ideal for this distribution.<br />
(d) Procedures should be <strong>in</strong> place for early check<strong>in</strong>g of <strong>data</strong> quality and for consequential<br />
modifications to the survey operations to be made. Plans and procedures should be <strong>in</strong> place<br />
to accommodate failure to obta<strong>in</strong> <strong>in</strong>formation.<br />
(e) Occupational health and safety policies must be followed. Rest and meal breaks must<br />
comply with relevant <strong>in</strong>dustrial awards. These are essential <strong>in</strong> ma<strong>in</strong>ta<strong>in</strong><strong>in</strong>g the alertness and<br />
concentration of field personnel and reduc<strong>in</strong>g the errors of observation and record<strong>in</strong>g.<br />
(f) Privacy legislation must be followed. Data of confidential nature should not be disclosed<br />
without proper clearance.<br />
Transit New Zealand has out-sourced the <strong>collection</strong> of traffic <strong>data</strong> for a number of years. It has 89<br />
Pattern Stations and 1,400 Short-term Stations. A s<strong>in</strong>gle contractor provides the service of<br />
collect<strong>in</strong>g <strong>data</strong> from the Pattern Stations which are all telemetry sites throughout New Zealand.<br />
Several <strong>in</strong>dependent contractors collect <strong>data</strong> from the Short-term Stations. The specifications<br />
provided to these contractors follow the pr<strong>in</strong>ciples outl<strong>in</strong>ed above and represent a set of best<br />
<strong>practice</strong> specifications (Transit New Zealand 2002). The follow<strong>in</strong>g two procedures adopted by<br />
Transit New Zealand are of <strong>in</strong>terest <strong>in</strong> <strong>data</strong> quality control:<br />
(a) Data validation aga<strong>in</strong>st <strong>in</strong>dependent means – the contractor specifications <strong>in</strong>clude a traffic<br />
counter operational check. The contractor is to fax the form to Transit New Zealand with<strong>in</strong> 48
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 41 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
hours of the counter be<strong>in</strong>g <strong>in</strong>stalled <strong>in</strong> the field. This allows Transit to undertake <strong>in</strong>dependent<br />
visual surveys at random to ensure the contractor is perform<strong>in</strong>g as expected and provid<strong>in</strong>g<br />
quality <strong>data</strong>. Any variation between recorded <strong>data</strong> and the visual surveys of ± 5% will result <strong>in</strong><br />
the contractor hav<strong>in</strong>g to repeat the survey at their cost.<br />
(b) Cross-checks - The contractors responsible for the Short-term Stations are asked to<br />
undertake random (once very three years) classified counts at each of the permanent<br />
telemetry sites as a way of <strong>in</strong>dependently validat<strong>in</strong>g the accuracy of each other’s equipment<br />
and <strong>data</strong> collected. Any variance of ± 5% is then the source of <strong>in</strong>vestigation to assess why<br />
the results are different. As both parties are <strong>in</strong>dependent, the contractors have a strong<br />
<strong>in</strong>centive to ensure their methodology and equipment accurately record traffic at these sites.<br />
In summary, the count<strong>in</strong>g personnel whether contractors or <strong>in</strong>-ho<strong>use</strong> staff should be well prepared<br />
<strong>in</strong> terms of count<strong>in</strong>g equipment, suitable seats, appropriate cloth<strong>in</strong>g <strong>in</strong>clud<strong>in</strong>g cont<strong>in</strong>gency<br />
provisions for adverse weather, amenities (food, toilet), and carry authority certificates.<br />
Conspicuous signage is not recommended as it may affect count results.<br />
6.3 Quality Check and Deal<strong>in</strong>g with Miss<strong>in</strong>g Data<br />
The pr<strong>in</strong>ciple of edit<strong>in</strong>g <strong>data</strong> and identify<strong>in</strong>g miss<strong>in</strong>g <strong>data</strong> is endorsed as a good <strong>practice</strong> at the<br />
October 2003 Road Use Data Workshop. The Department of Infrastructure, Energy and<br />
Resources <strong>in</strong> Tasmania recommended the need to describe clearly how the edit<strong>in</strong>g is carried out.<br />
This may <strong>in</strong>volve the <strong>use</strong> of meta<strong>data</strong> to provide traceability (meta<strong>data</strong> is ‘<strong>data</strong> on <strong>data</strong>’ <strong>use</strong>d to<br />
describe the content, condition and other characteristics <strong>in</strong> the <strong>data</strong>). In US, Smith et al. (2003)<br />
endorsed a similar pr<strong>in</strong>ciple and suggested the need to reconsider the AASHTO (1992) policy of<br />
not estimat<strong>in</strong>g miss<strong>in</strong>g (or ‘imput<strong>in</strong>g’) <strong>data</strong>.<br />
Each RA has its own <strong>data</strong> quality check<strong>in</strong>g procedures, which are also different for different <strong>data</strong><br />
types. It is beyond the scope of this report to record all procedures; however, few examples of<br />
current <strong>practice</strong>s from Queensland Ma<strong>in</strong> Roads, RTA NSW, Ma<strong>in</strong> Roads WA and the CSIRO<br />
software developed for RTA, are presented below for illustration.<br />
The current Queensland Ma<strong>in</strong> Roads <strong>practice</strong> is to ensure complete <strong>data</strong> sets at Pattern Stations<br />
dur<strong>in</strong>g annual process<strong>in</strong>g. Miss<strong>in</strong>g <strong>data</strong> are computed at the hourly level us<strong>in</strong>g weekly and hourly<br />
adjustment factors, i.e.<br />
Miss<strong>in</strong>g hour volume = Anticipated AADT × Week factor × Hour factor<br />
RTA NSW pays special attention to classified counts and the <strong>data</strong> rejection rules for classified<br />
counts at RTA NSW are shown <strong>in</strong> Table 17.
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
Table 17 – RTA NSW rejection rules on classified counts<br />
Daily reject criterion ≤ 5 veh / day or<br />
Hourly reject criterion 0 veh for more than 12 hours<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 42 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
> 14% Class 3 (two-axle truck or bus with wheelbase more than 3.2 m) or<br />
> 20% Class 13 (error b<strong>in</strong>, with less than 1,000 error vehicles per day)<br />
> 20% Class 13 (error b<strong>in</strong> with less than 100 error vehicles per hour)<br />
Ma<strong>in</strong> Roads WA adopts a def<strong>in</strong>ition of AADT slightly different from GTEP Part 3. AADT is<br />
determ<strong>in</strong>ed us<strong>in</strong>g the Monthly Average Day of the Week Traffic (MADW) and Annual Average Days<br />
of the Week (AADW) as follows:<br />
MADW<br />
ij<br />
=<br />
V<br />
n w ij<br />
ij<br />
∑<br />
w= 1 nij<br />
12<br />
∑<br />
j=<br />
1<br />
(84 values <strong>in</strong> a year)<br />
1<br />
AADW i = MADWij<br />
(7 values <strong>in</strong> a year)<br />
12<br />
1<br />
AADT =<br />
7<br />
7<br />
∑ AADWi<br />
i=<br />
1<br />
where i is the weekday <strong>in</strong>dex with i = 1 to 7 (Monday to Sunday),<br />
j is the monthly <strong>in</strong>dex with j = 1 to 12 (January to December),<br />
w is the week <strong>in</strong>dex with w = 1 to nij = 1 to 5,<br />
nij is the number of a weekday <strong>in</strong> a month; (e.g. one to five Mondays <strong>in</strong> a month),<br />
w<br />
V ij is the traffic volume on ith weekday <strong>in</strong> the wth week <strong>in</strong> the jth month,<br />
MADW is the monthly average traffic for each day of the week, over the period of one month, and<br />
AADW is the annual average traffic for each day of the week, over the period of one year.<br />
Ma<strong>in</strong> Roads WA has a well-def<strong>in</strong>ed set of quality check rules. Us<strong>in</strong>g the above three traffic volume<br />
statistics for illustration, the m<strong>in</strong>imum number of statistics for a site to be treated as active is as<br />
follows:<br />
At least one weekday must exist for MADW to be calculated for that weekday <strong>in</strong> the month,<br />
e.g. at least one Monday out of four or five Mondays <strong>in</strong> a month;<br />
At least n<strong>in</strong>e MADWs for a weekday must exist before AADW is calculated for that weekday<br />
(i.e. n<strong>in</strong>e out of twelve);<br />
All seven AADWs must exist before AADT is calculated for a Pattern Count Site (PCS);<br />
In all other circumstances, if a site has <strong>in</strong>sufficient <strong>data</strong>, it will be treated as <strong>in</strong>active unless<br />
miss<strong>in</strong>g <strong>data</strong> is estimated.<br />
Other similar rules apply to average weekday traffic and other related quantities when Saturdays,<br />
Sundays and public holidays are not considered.
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 43 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
A PCS can exceed the m<strong>in</strong>imum <strong>data</strong> requirement and still be unsuitable to calculate AADT or be<br />
<strong>use</strong>d <strong>in</strong> seasonal adjustment. Sites which have unusual or significant seasonal variations fall <strong>in</strong>to<br />
this category. Therefore all <strong>data</strong> for all PCS must be reviewed prior to acceptance of the <strong>data</strong> and<br />
production of f<strong>in</strong>al AADT and seasonal factors. The <strong>in</strong>itial procedures are as follows:<br />
Identify and remove all anomalous <strong>data</strong>.<br />
Re-run monthly/ annual reports and identify which sites have AADT affected by miss<strong>in</strong>g <strong>data</strong>.<br />
Identify which of these sites and months are required for seasonal adjustment. These<br />
sites/months will have estimates submitted.<br />
Estimates will be submitted to meet m<strong>in</strong>imum <strong>data</strong> requirements and to produce seasonal<br />
factors for months where associated short-term <strong>data</strong> requires adjustment.<br />
Estimate MADW <strong>data</strong> for these sites.<br />
It is not necessary to submit estimates for miss<strong>in</strong>g <strong>data</strong> where the m<strong>in</strong>imum requirement is already<br />
achieved, seasonal factors are not required and the miss<strong>in</strong>g <strong>data</strong> has the same result as the<br />
MADW. For example, a PCS has n<strong>in</strong>e MADW values and July’s MADW for Monday is miss<strong>in</strong>g. If<br />
the AADW for Monday is 6225 and the estimated MADW for July is 6228 then it is not necessary to<br />
submit this MADW. If there is no <strong>data</strong> for a specific weekday <strong>in</strong> a month where estimates are<br />
required then an MADW must be submitted.<br />
If an actual MADW exists for a specific weekday for a month but this has been derived from less<br />
than the m<strong>in</strong>imum number of those weekdays required <strong>in</strong> that month, then the MADW must be<br />
reviewed and where required, an estimate will replace the actual value.<br />
Public holidays have a significant effect on MADW. Any months with miss<strong>in</strong>g daily flows for<br />
weekdays on which public holidays occur (particularly Mondays) will usually require an estimate to<br />
replace the actual value.<br />
For the estimation of miss<strong>in</strong>g <strong>data</strong>, it is necessary to refer to historical <strong>in</strong>formation, current<br />
<strong>in</strong>formation for previous and subsequent weeks and available <strong>data</strong> for the pattern relat<strong>in</strong>g to<br />
weekday flows. Estimation of the value is either by computer modell<strong>in</strong>g or manual means us<strong>in</strong>g<br />
graphical <strong>analysis</strong>.<br />
The framework of the RTA-CSIRO software for detect<strong>in</strong>g and correct<strong>in</strong>g corrupt <strong>data</strong> is shown <strong>in</strong><br />
Figure 13. It fits a l<strong>in</strong>ear model on all available <strong>data</strong> as follows (Donnelly 2003):<br />
Predicted value = grand median + trend + season + day + hour + school + public<br />
The model then <strong>use</strong>s a Hidden Markov Model (HMM) to detect periods of erratic counts. If any<br />
detected period exceeds seven days then that period is removed from the <strong>data</strong>set <strong>in</strong> the HMM<br />
modell<strong>in</strong>g process. A second l<strong>in</strong>ear model is fitted to the rema<strong>in</strong><strong>in</strong>g <strong>data</strong> and outlier detection<br />
method is <strong>use</strong>d to detect and correct <strong>in</strong>dividual extreme counts. If no periods of erratic counts are<br />
detected then the outlier method can be run us<strong>in</strong>g the orig<strong>in</strong>al l<strong>in</strong>ear model. The second l<strong>in</strong>ear<br />
model is <strong>use</strong>d to predict counts that are miss<strong>in</strong>g from the <strong>data</strong>set, either beca<strong>use</strong> they were<br />
removed dur<strong>in</strong>g this process (before l<strong>in</strong>ear modell<strong>in</strong>g or dur<strong>in</strong>g the HMM process) or beca<strong>use</strong> they<br />
were miss<strong>in</strong>g from the orig<strong>in</strong>al <strong>data</strong>.<br />
In summary, the management of miss<strong>in</strong>g <strong>data</strong> still requires substantially more research. A vital<br />
po<strong>in</strong>t is to flag amended <strong>data</strong> and ensure that the orig<strong>in</strong>al raw <strong>data</strong> sets can be recovered should<br />
better estimation techniques become available. There is much expectation that the CSIRO<br />
software developed for RTA NSW would soon be <strong>in</strong> production phase and become <strong>use</strong>ful also to<br />
other RAs.
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
Input raw<br />
<strong>data</strong><br />
Add calendar<br />
<strong>in</strong>formation<br />
Fit l<strong>in</strong>ear<br />
model<br />
Run Hidden<br />
Markov<br />
Model<br />
7+ days<br />
block of<br />
bad <strong>data</strong><br />
found?<br />
Remove these<br />
bad periods<br />
from <strong>data</strong><br />
Refit l<strong>in</strong>ear<br />
model<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 44 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
Run outlier<br />
method<br />
Predict<br />
miss<strong>in</strong>g<br />
counts<br />
If hourly<br />
aggregate to<br />
daily<br />
Output<br />
Figure 13 – Flow diagram of the CSIRO method for detection of corrupt traffic <strong>data</strong><br />
6.4 Calibration of a <strong>WIM</strong> System<br />
Table 18 shows the range of (high-speed) <strong>WIM</strong> technologies employed <strong>in</strong> Australia (Koniditsiotis<br />
2000). These technologies <strong>in</strong>clude plate-<strong>in</strong>-ground, load cell, piezo-cable and stra<strong>in</strong> gauge. The<br />
supplier/manufacturer of a <strong>WIM</strong> system will provide detailed <strong>in</strong>structions for calibration. In view of<br />
the fact that most of the <strong>WIM</strong> sites <strong>in</strong> Australia employ CULWAY, with 142 <strong>in</strong> operation <strong>in</strong> year<br />
2000, this section therefore <strong>use</strong>s CULWAY as an example to illustrate some aspects of good<br />
<strong>practice</strong>s <strong>in</strong> <strong>WIM</strong> calibration. Aga<strong>in</strong>, each RA has its own operational guidel<strong>in</strong>es on CULWAY<br />
<strong>in</strong>stallation and operation and it is unnecessary to repeat the full procedures <strong>in</strong> this report.<br />
CULWAY weighs and classifies traffic by us<strong>in</strong>g mechanical stra<strong>in</strong> gauges deployed <strong>in</strong> a box<br />
culvert. Vehicles pass<strong>in</strong>g over the culvert ca<strong>use</strong> the roof to deflect produc<strong>in</strong>g stra<strong>in</strong> forces that can<br />
be converted to axle loads. The CULWAY system employs two piezoelectric sensors, placed 10 m<br />
apart <strong>in</strong> each <strong>in</strong>strumented lane to detect vehicle passage and synchronise the stra<strong>in</strong><br />
measurement at each axle. The accuracy of CULWAY relative to other <strong>WIM</strong> systems currently <strong>in</strong><br />
<strong>use</strong> is also shown <strong>in</strong> Table 18. CULWAY and all other <strong>WIM</strong> sites should be carefully calibrated to<br />
achieve optimal <strong>use</strong> of the system.
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 45 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
This section highlights one general aspect and one specific aspect <strong>in</strong> the <strong>use</strong> of CULWAY:<br />
General guidel<strong>in</strong>es on site selection; and<br />
Corrections for the time-dependent drift between calibrations.<br />
Table18 – High-speed Weigh-<strong>in</strong>-Motion Systems Used and available <strong>in</strong> Australia<br />
(source: Koniditsiotis 2000)<br />
Mass sensor type Bend<strong>in</strong>g plate Load cell Piezoelectric<br />
cable<br />
<strong>WIM</strong> system name PAT DAW100 ARRB TR HSEMU ARRB TR Expressweigh<br />
Installation mode Semi-permanent<br />
flush mounted<br />
Semi-permanent<br />
flush mounted<br />
Permanent flush<br />
mounted<br />
Stra<strong>in</strong> gauge<br />
ARRB TR CULWAY<br />
Semi-permanent<br />
with<strong>in</strong> pavement<br />
culvert <strong>in</strong>stalled<br />
Max. no. of lanes 4 User-def<strong>in</strong>ed 2 1 to 4<br />
Axles<br />
weighed<br />
Weigh<strong>in</strong>g capacity for<br />
axle load (tonnes)<br />
All All All All<br />
20 80 - 50<br />
Temperature (°C) -40 to +75 -20 to +70 -50 to +80 -10 to +70<br />
Sensor life span<br />
(years)<br />
Accuracy<br />
Vendor<br />
3 rd party<br />
Installations <strong>in</strong><br />
Australia (as reported<br />
<strong>in</strong> year 2000)<br />
− 15 − 10<br />
*GVM<br />
(± 10% at 95%)<br />
−<br />
10<br />
(another 6 us<strong>in</strong>g<br />
other bend<strong>in</strong>g plates)<br />
GVM<br />
(± 5% at 95%)<br />
GVM<br />
(± 3% at 66%)<br />
GVM<br />
(± 10% at 95%)<br />
−<br />
GVM<br />
(± 10% at 95%)<br />
GVM<br />
(± 7% at 95%)<br />
2 1 140<br />
(another 2 for multilane<br />
operation)<br />
* GVM (±10% at 95%) means that the accuracy of measur<strong>in</strong>g gross vehicle mass is ±10% for 95% of vehicles weighed.<br />
6.4.1 Site Location<br />
The selection of a suitable location for a CULWAY <strong>in</strong>stallation is critical to the accuracy of the<br />
<strong>in</strong>formation collected. The follow<strong>in</strong>g criteria (from Ma<strong>in</strong> Roads WA) are recommended when<br />
assess<strong>in</strong>g the suitability of a site for CULWAY:<br />
(a) Alignment and overtak<strong>in</strong>g opportunities - a m<strong>in</strong>imum of 120 m of straight and level <strong>road</strong>way<br />
should be on each side of the CULWAY. Sight distance should not be so good as to make it<br />
an attractive overtak<strong>in</strong>g opportunity for motorists. This encourages poor lane discipl<strong>in</strong>e<br />
result<strong>in</strong>g <strong>in</strong> poor detection rates and low levels of <strong>data</strong> confidence. Avoid plac<strong>in</strong>g CULWAY<br />
too close to curves s<strong>in</strong>ce vehicles travell<strong>in</strong>g at speed may sw<strong>in</strong>g wide tak<strong>in</strong>g the curve and<br />
consequently affect the CULWAY results by <strong>in</strong>valid vehicle actuations.<br />
(b) Longitud<strong>in</strong>al grade - ideally the longitud<strong>in</strong>al gradient shall be less than 1%. However, the<br />
gradient may exceed this where considered appropriate but gradients of more than 2% are<br />
not suitable.
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 46 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
(c) Crossfall - ideally the crossfall shall be between 2% and 3% with a m<strong>in</strong>imum of 2%. The<br />
designer needs to balance the needs of CULWAY, i.e. as flat and smooth as possible aga<strong>in</strong>st<br />
the eng<strong>in</strong>eer<strong>in</strong>g necessity of adequately dra<strong>in</strong><strong>in</strong>g the <strong>road</strong>.<br />
(d) Open speed section - the speed of the traffic may also impact CULWAY results. Normal<br />
highway speeds are desirable for optimal operation. Slow speed, <strong>in</strong> particular speeds of less<br />
than 70 km/h may affect accuracy. Vehicles travell<strong>in</strong>g below 30 km/h will probably be<br />
overweighed. The presence of nearby <strong>in</strong>tersect<strong>in</strong>g <strong>road</strong>s or any other <strong>road</strong> geometry that<br />
ca<strong>use</strong>s traffic acceleration or deceleration over the CULWAY is undesirable.<br />
(e) Culvert cover - a section of fill provid<strong>in</strong>g adequate cover for the culvert after ensur<strong>in</strong>g a<br />
culvert of m<strong>in</strong>imum height 600 mm can be placed above natural surface.<br />
(f) Culvert to be dry - the CULWAY should be located <strong>in</strong> a section where the fall is sufficient to<br />
achieve essentially what will be a dry culvert or at least will ensure that any water will dra<strong>in</strong><br />
away quickly from the CULWAY.<br />
6.4.2 Time-Dependent Drift between Calibrations<br />
It has been known for some time that the CULWAY <strong>data</strong> is affected by time-of-day and week-ofyear<br />
(Lim 1992). The variation or drift is dependent on site and environment conditions. The<br />
strength of the bitum<strong>in</strong>ous surface on top of a culvert box is affected by temperature, and how well<br />
air and water are dra<strong>in</strong>ed through the surface.<br />
Grundy (2002) reported the study of two CULWAY sites <strong>in</strong> Victoria, mak<strong>in</strong>g <strong>use</strong> of the property that<br />
the steer<strong>in</strong>g axles mass (SAM) of an articulated six-axle is the least affected by temperature and<br />
dra<strong>in</strong>age conditions. This is beca<strong>use</strong> there is no direct load on the steer<strong>in</strong>g axle (<br />
Figure 14). It is therefore possible to <strong>use</strong> this constancy of average SAM to correct for diurnal and<br />
seasonal drift <strong>in</strong> <strong>WIM</strong> sensitivity.<br />
The results are shown <strong>in</strong> Figure 15 for the diurnal effect of hour for the sites – Western R<strong>in</strong>g Road<br />
<strong>in</strong> Melbourne and Yarra Glen just outside Melbourne. Figure 16 shows the seasonal variation over<br />
six months (July to December). The variation can be substantial <strong>in</strong> the case of the Western R<strong>in</strong>g<br />
Road. The correction factor varied from 0.93 (w<strong>in</strong>ter) to 1.1 (summer). The Yarra Glen site<br />
showed less variation, possibly due to the <strong>use</strong> of a more traditional seal whereas the more porous<br />
surface of the Western R<strong>in</strong>g Road exhibited more variation with temperature and dra<strong>in</strong>age.<br />
A CULWAY system has its own temperature-compensat<strong>in</strong>g mechanism <strong>in</strong> its electronic hardware<br />
and provides reasonable results with acceptable accuracy levels but does not consider<br />
temperature/climatic variations of the pavement on top of the culvert box. It is therefore good<br />
<strong>practice</strong> to pay attention to site conditions and how seasonal changes <strong>in</strong> climate could affect a <strong>WIM</strong><br />
<strong>in</strong>stallation. Some RAs already allow for seasonal and diurnal drifts of CULWAY.
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
Steer<strong>in</strong>g<br />
Axle<br />
Mass<br />
SAM<br />
Drive Axle<br />
Mass DAM<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 47 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
Tandem<br />
Axle Mass<br />
TAM<br />
Figure 14 – Three axle masses of an articulated six-axle vehicle (Class 9)<br />
Figure 15 – Adjustment factors for diurnal variation of CULWAY axle mass <strong>data</strong><br />
(Grundy et al. 2002)<br />
Figure 16 – Adjustment factors for seasonal variation of CULWAY axle mass <strong>data</strong><br />
(Grundy et al 2002)
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 48 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
6.5 Accuracy Specifications<br />
Section 6.1 mentions the complexity of traffic phenomena, which often arises by chance from a<br />
comb<strong>in</strong>ation of many factors. These factors have high degrees of variability. Completely accurate<br />
counts do not exist and all traffic quantities are statistics with distributions of errors.<br />
Table 19 provides b<strong>road</strong> guidel<strong>in</strong>es on accuracy requirements for different traffic statistics. The<br />
values were adapted ma<strong>in</strong>ly from GTEP Part 3. Note that the quality of the <strong>data</strong> obta<strong>in</strong>ed depends<br />
on the survey method selected and the amount of quality control performed. More precise survey<br />
techniques with higher quality control generally consume more survey resources. A trade-off must<br />
be made on the basis of time, money and people so that a feasible survey method and sample size<br />
can be found for the successful completion of a survey. Also, the specification of accuracy should<br />
also follow the pr<strong>in</strong>ciple of fitness of purpose discussed <strong>in</strong> Section 6.1.<br />
Table 19 – Accuracy requirements<br />
Traffic and other quantities Maximum error<br />
AADT < 100<br />
101-300<br />
301-1100<br />
> 1100<br />
±50%<br />
±35%<br />
±25%<br />
±15%<br />
Confidence level and<br />
comments<br />
90% confidence limits<br />
Trends <strong>in</strong> AADT ±10% 90% confidence limits<br />
Classified counts<br />
(% error <strong>in</strong> vehicle class proportions)<br />
VKT (annual total for National and State<br />
<strong>road</strong>s)<br />
<strong>WIM</strong> applications<br />
Economic <strong>analysis</strong>, transport studies,<br />
vehicle classification<br />
Road and bridge management,<br />
overload warn<strong>in</strong>g, <strong>road</strong> safety<br />
Enforcement compliance<br />
±20%<br />
±3 (Australia)<br />
±7% (State)<br />
±10% (City)<br />
±20%<br />
±15%<br />
±5%<br />
90% confidence limits<br />
95% confidence limits<br />
Gross Vehicle Mass for 95% of<br />
vehicles
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
7 STAKEHOLDER CONSULTATION<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 49 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
Many technical issues have been addressed <strong>in</strong> previous sections. This section is a summary<br />
section that serves to illustrate key components of a best <strong>practice</strong> <strong>road</strong> <strong>use</strong> <strong>data</strong> program. It must<br />
be emphasised aga<strong>in</strong> that many activities require <strong>road</strong> <strong>use</strong> <strong>data</strong> and a diverse group of <strong>road</strong> <strong>use</strong><br />
<strong>data</strong> stakeholders are <strong>in</strong>volved. These stakeholders can be <strong>in</strong>ternal or external to a RA. Examples<br />
of external stakeholders are <strong>road</strong> safety agencies, Local Governments and plann<strong>in</strong>g authorities.<br />
As such, <strong>road</strong> <strong>use</strong> <strong>data</strong> is a corporate asset that requires its share of corporate plann<strong>in</strong>g and<br />
management. The follow<strong>in</strong>g model developed by Mihai (20<strong>04</strong>) for Ma<strong>in</strong> Roads WA is <strong>use</strong>d for<br />
illustration and is shown <strong>in</strong> Figure 17.<br />
The governance structure <strong>in</strong> Figure 17 aims to br<strong>in</strong>g together all parties <strong>in</strong>terested <strong>in</strong> <strong>road</strong> <strong>use</strong> <strong>data</strong><br />
<strong>collection</strong>, and ensures their needs are considered. District or regional <strong>in</strong>terests with<strong>in</strong> a RA must<br />
be represented. A good governance structure should facilitate secur<strong>in</strong>g corporate support. Such a<br />
structure would <strong>in</strong>clude the follow<strong>in</strong>g three groups:<br />
Bus<strong>in</strong>ess reference groups that <strong>in</strong>cludes expert practitioners, strategic planners external and<br />
<strong>in</strong>ternal to a RA;<br />
A steer<strong>in</strong>g committee that <strong>in</strong>cludes an RA’s corporate executives; and<br />
Cross-regional team that <strong>in</strong>cludes representatives from regions and districts <strong>in</strong> a RA.<br />
With a governance structure <strong>in</strong> place, a <strong>road</strong> <strong>use</strong> <strong>data</strong> plann<strong>in</strong>g and management process can be<br />
put <strong>in</strong>to operation. Such a process consists of six key elements as shown <strong>in</strong> Figure 17. The<br />
process is an on-go<strong>in</strong>g process with regular reviews and updates. Some of the elements have<br />
been mentioned <strong>in</strong> previous sections and brief descriptions are given below.<br />
(a) Stakeholder consultation – a register of stakeholders should be kept and their needs <strong>in</strong> terms<br />
of <strong>data</strong> types and formats periodically updated. Ma<strong>in</strong> Roads WA conducts comprehensive<br />
stakeholder consultation every five years. The needs can have priorities set at various levels<br />
to determ<strong>in</strong>e fund<strong>in</strong>g priorities.<br />
(b) Data plann<strong>in</strong>g – necessary activities <strong>in</strong>clude the review of needs, review of exist<strong>in</strong>g <strong>data</strong><br />
<strong>collection</strong> sites, sett<strong>in</strong>g (five-year) goals, estimation of costs and benefits, sett<strong>in</strong>g and<br />
updat<strong>in</strong>g of policies on issues such as the level of coverage (Section 4.3), def<strong>in</strong><strong>in</strong>g a uniform<br />
traffic section (Section 4.3) and <strong>data</strong> quality (Section 6).<br />
(c) Budget and programm<strong>in</strong>g – this task <strong>in</strong>cludes prepar<strong>in</strong>g estimates of capital equipment costs<br />
and on-go<strong>in</strong>g ma<strong>in</strong>tenance costs, roll<strong>in</strong>g programs over, say, five years accord<strong>in</strong>g to different<br />
scenarios (outsourced, <strong>in</strong>-ho<strong>use</strong> or a comb<strong>in</strong>ation), <strong>data</strong> process<strong>in</strong>g costs, and costs of<br />
monitor<strong>in</strong>g count<strong>in</strong>g sites periodically.<br />
(d) Data management – this process is as described previously on the <strong>collection</strong>, validation,<br />
process<strong>in</strong>g, <strong>in</strong>tegration and storage.<br />
(e) Data delivery – after <strong>data</strong> management, the key issue is facilitat<strong>in</strong>g the efficient and effective<br />
delivery of <strong>data</strong> to <strong>use</strong>rs. The tasks of <strong>data</strong> report<strong>in</strong>g and accessibility have been discussed<br />
<strong>in</strong> earlier sections. In particular, a public-private partnership could be an important means to<br />
make <strong>data</strong> more accessible. Tra<strong>in</strong><strong>in</strong>g sessions on <strong>data</strong> usage should be provided.<br />
(f) Review and audit – reviews and audits of the <strong>data</strong> under management are part of a corporate<br />
process. Feedback <strong>in</strong> the form of stakeholder surveys should also be considered.
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
STAKEHOLDER<br />
CONSULTATION<br />
- Stakeholder<br />
Register<br />
- Stakeholder<br />
Needs Matrix<br />
DATA PLANNING<br />
- Needs Review<br />
- Exist<strong>in</strong>g Sites<br />
- Policies<br />
- 5 Years Goals<br />
- Costs and Benefits<br />
BUDGET AND<br />
PROGRAMMING<br />
- Equipment Costs<br />
- Ma<strong>in</strong>tenance<br />
Scenarios<br />
- Ma<strong>in</strong>tenance Costs<br />
- Other Costs<br />
Figure 17 – Road <strong>use</strong> <strong>data</strong> plann<strong>in</strong>g and management framework (Source: Mihai 20<strong>04</strong>)<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 50 —<br />
DATA<br />
MANAGEMENT<br />
- Data Collection<br />
- Data Validation<br />
- Data Process<strong>in</strong>g<br />
- Data Storage<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
DATA DELIVERY<br />
- Statistics<br />
- Data Release<br />
- Data Access<br />
- Tra<strong>in</strong><strong>in</strong>g Sessions<br />
REVIEW AND<br />
AUDIT<br />
- Data Audit<br />
- Process Review<br />
- Stakeholder<br />
Satisfaction Surveys<br />
Steer<strong>in</strong>g Committee Bus<strong>in</strong>ess Reference Group Cross Regional Team<br />
Procedures and Guidel<strong>in</strong>es
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
8 CONCLUSIONS<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 51 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
This report has successfully compiled and highlighted best <strong>practice</strong>s <strong>in</strong> <strong>road</strong> <strong>use</strong> <strong>data</strong> <strong>collection</strong>,<br />
<strong>analysis</strong> and report<strong>in</strong>g. Road <strong>use</strong> <strong>data</strong> is a complex topic and the materials <strong>in</strong> this report are the<br />
f<strong>in</strong>d<strong>in</strong>gs of an Aust<strong>road</strong>s Road Use Data Workshop held <strong>in</strong> October 2003 <strong>in</strong> Melbourne, and the<br />
many feedback and suggestions from the Project Work<strong>in</strong>g Group members. A key f<strong>in</strong>d<strong>in</strong>g of this<br />
project is that RAs have been follow<strong>in</strong>g similar pr<strong>in</strong>ciples <strong>in</strong> traffic count <strong>data</strong> <strong>collection</strong> and<br />
<strong>analysis</strong>. Road authorities have ref<strong>in</strong>ed their count<strong>in</strong>g programs and <strong>in</strong> many ways have achieved<br />
good <strong>practice</strong>s.<br />
The pr<strong>in</strong>ciples of best <strong>practice</strong>s are enunciated <strong>in</strong> the report (Section 2). These pr<strong>in</strong>ciples <strong>in</strong>clude<br />
accuracy, effectiveness, efficiency, reliability, accessibility, transparency, timel<strong>in</strong>ess and relevance.<br />
They also relate to the issue of <strong>data</strong> <strong>in</strong>tegrity (Section 6). The issue of transparency deserves<br />
special attention beca<strong>use</strong> it is a key means at enhanc<strong>in</strong>g consistency <strong>in</strong> <strong>data</strong> <strong>collection</strong>, <strong>analysis</strong><br />
and report<strong>in</strong>g amongst RAs and facilitate accessibility of current and potential <strong>use</strong>rs of <strong>road</strong> <strong>use</strong><br />
<strong>data</strong>.<br />
Another key <strong>data</strong> <strong>in</strong>tegrity issue is the method of manag<strong>in</strong>g miss<strong>in</strong>g <strong>data</strong>. It is recognised that<br />
each RA has its own procedure <strong>in</strong> handl<strong>in</strong>g miss<strong>in</strong>g <strong>data</strong>, and that there is no standard way of<br />
address<strong>in</strong>g the issue. This report provides the current procedures adopted <strong>in</strong> RTA NSW and Ma<strong>in</strong><br />
Roads WA as examples of good <strong>practice</strong>s. In particular, the software developed by RTA-CSIRO<br />
us<strong>in</strong>g the method of Hidden Markov Model could prove <strong>use</strong>ful amongst RAs after the necessary<br />
tests are completed.<br />
The topics of vehicle detection and classification are also discussed. Most issues related to these<br />
topics have been addressed over many years. It is generally accepted that the Aust<strong>road</strong>s 12-b<strong>in</strong><br />
classification by axle configuration is stable and well-received s<strong>in</strong>ce its <strong>in</strong>ception <strong>in</strong> 1994. M<strong>in</strong>or<br />
variations are recommended as follows:<br />
Introduction of an error b<strong>in</strong> (b<strong>in</strong> 13) to account for the quality of classified counts;<br />
Distribution of vehicles from the error b<strong>in</strong> <strong>in</strong>to either b<strong>in</strong> 1 (the passenger-car b<strong>in</strong>) or across<br />
all twelve b<strong>in</strong>s depend<strong>in</strong>g on the error b<strong>in</strong> size, or us<strong>in</strong>g both methods mak<strong>in</strong>g <strong>use</strong> of<br />
knowledge of on-site traffic conditions;<br />
Standardisation of classify<strong>in</strong>g vehicles <strong>in</strong>to four or five b<strong>in</strong>s by vehicle lengths; and<br />
Recognition of m<strong>in</strong>or variations amongst RAs, but sub-classified counts by either axle<br />
configurations or vehicle lengths must be capable of aggregat<strong>in</strong>g <strong>in</strong>to the Aust<strong>road</strong>s system<br />
(13 b<strong>in</strong>s by axles or 4 or 5 b<strong>in</strong>s by lengths).<br />
In the case of vehicle classification by lengths, there is no standard established system yet.<br />
Indeed, it may be appropriate to adopt a 5-b<strong>in</strong> system to sub-classify comb<strong>in</strong>ation vehicles so<br />
manufacturers can design their equipment accord<strong>in</strong>gly. The five b<strong>in</strong>s can be aggregated <strong>in</strong>to four<br />
b<strong>in</strong>s where necessary. Some field tests and calibration to determ<strong>in</strong>e the appropriate threshold<br />
values of vehicle lengths for each b<strong>in</strong> are recommended.<br />
The <strong>use</strong> of a s<strong>in</strong>gle axle sensor for traffic count <strong>collection</strong> is quite common amongst RAs, even<br />
though most have recognised the need for classified counts. It is recommended that traffic counts,<br />
as a default, be reported <strong>in</strong> vehicle numbers. If the raw <strong>data</strong> is <strong>in</strong> the form of axles or axle-pairs,<br />
then the count should be converted to vehicles with the conversion factor also reported if possible,<br />
adopt<strong>in</strong>g the pr<strong>in</strong>ciple of transparency.
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 52 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
The issue of utilis<strong>in</strong>g <strong>WIM</strong> equipment as part of a traffic count<strong>in</strong>g program is also discussed <strong>in</strong><br />
some detail. As long as all lanes of traffic are monitored at a <strong>WIM</strong> site, the traffic count <strong>data</strong> (and<br />
<strong>WIM</strong> <strong>data</strong>) should add extra value to a count<strong>in</strong>g program. The related issue of correlat<strong>in</strong>g classified<br />
count <strong>data</strong> with vehicle mass <strong>data</strong> is more complex and there is not much <strong>in</strong>formation published on<br />
this topic. In <strong>practice</strong>, if the ratio of loaded and unloaded vehicles at a count<strong>in</strong>g station is consistent<br />
with a nearby <strong>WIM</strong> site, then an accurate estimate of pavement load<strong>in</strong>g can be determ<strong>in</strong>ed.<br />
Road <strong>use</strong> <strong>data</strong> can come from many sources meet<strong>in</strong>g a variety of needs. This leads to the issues<br />
of <strong>data</strong> <strong>in</strong>tegration and accessibility. Many stakeholders are consequently <strong>in</strong>volved. This report<br />
provides a stakeholder consultation model from Ma<strong>in</strong> Roads WA. These issues are rather <strong>in</strong>volved,<br />
and RAs at present have quite varied policies on <strong>data</strong> availability and its pric<strong>in</strong>g. As the trend to<br />
outsource <strong>data</strong> cont<strong>in</strong>ues amongst RAs, third parties may become both supplier and purchaser of<br />
<strong>road</strong> <strong>use</strong> <strong>data</strong>. While these issues were discussed at the October 2003 Workshop and described <strong>in</strong><br />
some detail <strong>in</strong> this report, they will cont<strong>in</strong>ue to confront RAs <strong>in</strong> the future. It is appropriate to<br />
develop, as a future research task, an appropriate policy or bus<strong>in</strong>ess model <strong>in</strong> this area of <strong>data</strong><br />
<strong>in</strong>tegration, accessibility and pric<strong>in</strong>g.<br />
In summary, a list of recommendations is as follows:<br />
1. Vehicle detection - a pair of axle sensors should be considered where possible to give better<br />
accuracy, better dist<strong>in</strong>ction by direction, and the ability to produce classified counts through<br />
axle configurations. If a s<strong>in</strong>gle axle is <strong>use</strong>d, it is a recommended <strong>practice</strong> to correct for traffic<br />
composition and convert axle-pairs <strong>in</strong>to vehicle counts for report<strong>in</strong>g purposes.<br />
2. Vehicle classification - the current Aust<strong>road</strong>s vehicle classification be ma<strong>in</strong>ta<strong>in</strong>ed but allow<strong>in</strong>g<br />
sub-classes as variations of the basic system. It is appropriate to <strong>in</strong>troduce a new notation to<br />
designate different jurisdictional versions, e.g. the Aust<strong>road</strong>s (Vic) for the Victoria version.<br />
3. Error b<strong>in</strong> - an error b<strong>in</strong> (no. 13) be <strong>in</strong>troduced <strong>in</strong>to the Aust<strong>road</strong>s system as good <strong>practice</strong> to<br />
<strong>in</strong>dicate the quality of the classified counts. The number of error counts should be monitored<br />
and, if these rema<strong>in</strong> high relative to the traffic stream, the reasons for these errors should be<br />
identified and the problem rectified. The <strong>in</strong>clusion of error counts <strong>in</strong> b<strong>in</strong> 13 <strong>in</strong> the total counts<br />
requires some judgement. If the number of error counts is high relative to the total counts, it<br />
is necessary to identify the reasons for such a situation before error counts are <strong>in</strong>cluded. If it<br />
is deemed appropriate to distribute b<strong>in</strong> 13 vehicles, the method of distribut<strong>in</strong>g across all b<strong>in</strong>s<br />
from 1 to 12 or only to b<strong>in</strong> 1 can be considered, or both methods are <strong>use</strong>d and guided by<br />
relevant local knowledge.<br />
4. Vehicle classification by length - a variety of threshold values are be<strong>in</strong>g <strong>use</strong>d for vehicle<br />
classification by lengths. It is recommended that RAs come to agreement on a s<strong>in</strong>gle<br />
classification system by lengths. Some empirical work us<strong>in</strong>g a standard loop design would<br />
appear necessary. On reach<strong>in</strong>g agreement, the Aust<strong>road</strong>s system should <strong>in</strong>corporate a<br />
def<strong>in</strong>itive multi-b<strong>in</strong> system by vehicle lengths as a component of the total Aust<strong>road</strong>s<br />
classification system. This harmonisation should facilitate equipment manufactures to design<br />
their equipment for a s<strong>in</strong>gle specification.
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 53 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
5. Transparency - Different jurisdictions at present adopt slightly different def<strong>in</strong>itions and hence<br />
procedures <strong>in</strong> calculat<strong>in</strong>g traffic statistics. The actual values should not be significantly<br />
different. However, it is good <strong>practice</strong> for RAs to state the methods <strong>use</strong>d to generate AADT<br />
and other values. Further, the def<strong>in</strong>ition of AADT or other quantities should not imply a<br />
particular method of calculation or estimation. It is, however, important that the national<br />
glossary <strong>in</strong> Appendix A be adopted.<br />
6. Count<strong>in</strong>g program - a statewide count<strong>in</strong>g program should <strong>in</strong>volve the follow<strong>in</strong>g iterative steps:<br />
− Locate permanent and short-term count stations based on judgement and <strong>road</strong><br />
functional classes;<br />
− Collect counts;<br />
− Divide <strong>road</strong> network us<strong>in</strong>g volume stratification and traffic pattern behaviour;<br />
− Determ<strong>in</strong>e the number of stations required by apply<strong>in</strong>g a volume stratification<br />
technique;<br />
− Review count station density and location;<br />
− Calculate adjustment factors to convert short-term counts to AADT estimates;<br />
− Identify homogenous traffic sections and calculate VKT; and<br />
− Check, review and prepare a roll<strong>in</strong>g program with stakeholders who have good local<br />
knowledge.<br />
7. Report<strong>in</strong>g of statistics - statistical results be reported jo<strong>in</strong>tly <strong>in</strong> terms of percentage error and<br />
confidence <strong>in</strong>terval. If measurements could be made over the totality of a target population<br />
rather than over just a sample then there would be no need to specify a confidence <strong>in</strong>terval.<br />
Beca<strong>use</strong> usually only a sample is taken, there rema<strong>in</strong>s the possibility that this sample is<br />
unrepresentative of the population, so that the error is likewise unrepresentative.<br />
8. Data quality control – the follow<strong>in</strong>g two procedures be considered when employ<strong>in</strong>g a<br />
contractor for <strong>data</strong> <strong>collection</strong> or undertak<strong>in</strong>g the task <strong>in</strong>-ho<strong>use</strong>:<br />
− Data should be validated aga<strong>in</strong>st <strong>in</strong>dependent means. The contractor specifications<br />
<strong>in</strong>clude a traffic counter operational check. The contractor is to fax the form to a RA<br />
with<strong>in</strong> 48 hours of the counter be<strong>in</strong>g <strong>in</strong>stalled <strong>in</strong> the field. This allows the RA to<br />
undertake <strong>in</strong>dependent visual surveys at random to ensure the contractor is perform<strong>in</strong>g<br />
as expected and provid<strong>in</strong>g quality <strong>data</strong>. Any variation between recorded <strong>data</strong> and the<br />
visual surveys of ± 5% will result <strong>in</strong> the contractor hav<strong>in</strong>g to repeat the survey at their<br />
cost.<br />
− There should be cross-checks. The contractors responsible for the Short-term Stations<br />
are asked to undertake random (once every three years) classified counts at each of<br />
the permanent telemetry sites as a way of <strong>in</strong>dependently validat<strong>in</strong>g the accuracy of<br />
each other’s equipment and <strong>data</strong> collected. Any variance of ± 5% is then the source of<br />
<strong>in</strong>vestigation to assess why the results are different. As both parties are <strong>in</strong>dependent,<br />
the contractors have a strong <strong>in</strong>centive to ensure their methodology and equipment<br />
accurately record traffic at these sites.
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 54 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
9. Miss<strong>in</strong>g <strong>data</strong> – it is good <strong>practice</strong> to edit and provide miss<strong>in</strong>g <strong>data</strong>. For estimat<strong>in</strong>g miss<strong>in</strong>g<br />
<strong>data</strong>, it is necessary to refer to historical and current <strong>in</strong>formation from previous and<br />
subsequent weeks and available <strong>data</strong> for the pattern relat<strong>in</strong>g to weekdays flows. The<br />
management of miss<strong>in</strong>g <strong>data</strong> still requires substantially more research. A vital po<strong>in</strong>t is to flag<br />
amended <strong>data</strong> and ensure that the orig<strong>in</strong>al raw <strong>data</strong> sets can be recovered should better<br />
estimation techniques become available.<br />
10. Short duration counts - Counts of very short durations of a few hours or even 12 hours are<br />
not reliable for traffic forecast<strong>in</strong>g. With automatic count<strong>in</strong>g equipment now commonly<br />
deployed, the extra cost of collect<strong>in</strong>g <strong>data</strong> on more days should be m<strong>in</strong>imal. The best way to<br />
ensure consistency amongst RAs is for each RA to collect robust <strong>data</strong> sets, i.e. collect<strong>in</strong>g<br />
<strong>data</strong> over a consecutive period of, say, seven days at Short-term Stations.<br />
11. Weigh-<strong>in</strong>-motion <strong>data</strong> - <strong>WIM</strong> site selection and equipment calibration ensure <strong>data</strong> accuracy<br />
and are requirements of good <strong>practice</strong>. The potential impact of weather conditions on some<br />
pavement surfaces should be noted. As long as all lanes of traffic are monitored at a <strong>WIM</strong><br />
site, the traffic count and <strong>WIM</strong> <strong>data</strong> should add extra value to a count<strong>in</strong>g program.<br />
12. Stakeholder consultation - good <strong>practice</strong> is achieved by adopt<strong>in</strong>g a corporate <strong>road</strong> <strong>use</strong> <strong>data</strong><br />
<strong>collection</strong> framework and engag<strong>in</strong>g stakeholders. Many activities require <strong>road</strong> <strong>use</strong> <strong>data</strong> and a<br />
diverse group of stakeholders are <strong>in</strong>volved. As such, <strong>road</strong> <strong>use</strong> <strong>data</strong> is a corporate asset that<br />
requires its share of corporate plann<strong>in</strong>g and management.<br />
13. Future research - the follow<strong>in</strong>g areas are identified:<br />
− Determ<strong>in</strong>e the threshold values for automatic vehicle classification system by lengths;<br />
− Develop a model for the correlation of <strong>WIM</strong> <strong>data</strong> with classified counts; and<br />
− Develop a policy or bus<strong>in</strong>ess model on <strong>data</strong> <strong>in</strong>tegration, accessibility and pric<strong>in</strong>g.
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
REFERENCES<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 55 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
ARRB Transport Research Ltd (2003). Express Weigh<br />
<br />
AASHTO (1992). Guidel<strong>in</strong>es for traffic <strong>data</strong> programs. American Association of Highway and<br />
Transport Officials, Wash<strong>in</strong>gton, DC.<br />
Australian Bureau of Statistics (1963). Survey of Motor Vehicle Usage. Catalogue No. 9202.0.<br />
ABS, Sydney.<br />
Aust<strong>road</strong>s (1988a). Guide to Traffic Eng<strong>in</strong>eer<strong>in</strong>g Practice Part 3 – Traffic Studies. Report No.<br />
<strong>AP</strong>-11.3/88. Aust<strong>road</strong>s, Sydney.<br />
Aust<strong>road</strong>s (1988b). Guide to Traffic Eng<strong>in</strong>eer<strong>in</strong>g Practice Part 2 – Roadway Capacity. Report No.<br />
<strong>AP</strong>-11.2/88. Aust<strong>road</strong>s, Sydney.<br />
Aust<strong>road</strong>s (2001). The Australian and New Zealand <strong>road</strong> system and <strong>road</strong> authorities - national<br />
performance <strong>in</strong>dicators 2000. Report No. <strong>AP</strong>-43/01. Aust<strong>road</strong>s, Sydney (and earlier annual<br />
editions: 43/96, 43A/96, 43/98, 43/99, 43/00). <br />
Aust<strong>road</strong>s (20<strong>04</strong>). Guide to Traffic Eng<strong>in</strong>eer<strong>in</strong>g Practice Part 3 – Traffic Studies. Report No.<br />
<strong>AP</strong>-G11.3/<strong>04</strong>. Aust<strong>road</strong>s, Sydney.<br />
Bennett, D. (2002). Assessment of national highway traffic count<strong>in</strong>g. Contract Report RC2219-1.<br />
ARRB Transport Research, Vermont South.<br />
Botterill, R and Luk, J.Y.K. (1998). Investigation of a new method of estimat<strong>in</strong>g vehicle kilometres<br />
of travel. Contract Report RC6028. ARRB Transport Research, Vermont South.<br />
Daltrey, R. (2002). Data <strong>in</strong>tegrity. Discussion Paper, Roads and Traffic Authority, Sydney.<br />
Donnelly, J., Chan, D., Duncan, J., Keighley, T. and Angelovski, K. (2003). SSOFTRAC – a<br />
statistical software package for detection and correction of corrupt traffic counts. User’s Manual<br />
Version 2.2. CMIS Report No. 03/157. CSIRO Division of Mathematical and Information Sciences,<br />
North Ryde.<br />
Federal Highway Adm<strong>in</strong>istration (2001). Traffic monitor<strong>in</strong>g guide. US Department of<br />
Transportation, FHWA, Office of Highway Policy Information, Wash<strong>in</strong>gton DC<br />
<br />
Federal Highway Adm<strong>in</strong>istration (2003). Bicycle and pedestrian detection. F<strong>in</strong>al Report prepared<br />
by SRF consult<strong>in</strong>g Group, M<strong>in</strong>neapolis.<br />
Golden River Traffic Ltd (2001). M410 Bicycle Counter Mixed Traffic count<strong>in</strong>g.<br />
<br />
Karl, C.A., Powell, T. and Luk, J.Y.K. (2003). Development of a predictive travel time model for<br />
S<strong>in</strong>gapore based on GLIDES. Proc. 21 st ARRB and 11 th REAAA Conf. 18-23 May 2003, Cairns.<br />
Koniditsiotis, C. (2000). Weigh-<strong>in</strong>-motion technology. Publication No. <strong>AP</strong>-R168/00. Aust<strong>road</strong>s,<br />
Sydney.<br />
Lesch<strong>in</strong>ski, R. (1994). Bicycle detection at signalised <strong>in</strong>tersections. Research Report ARR No. 245,<br />
Australian Road Research Board, Vermont South.
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 56 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
Lim, E. (1992). Secondary <strong>use</strong>s of CULWAY <strong>data</strong>: comparison of three Victorian sites. Work<strong>in</strong>g<br />
document WD TE 92/009. Australian Road Research Board Ltd., Vermont South.<br />
Luk, J.Y.K. and Besley, M.A. (1985). Automatic vehicle identification as a traffic management<br />
measure, Forum Papers, 10th Australian Transport Research Forum, 10(2), 13-15 May,<br />
Melbourne, pp.19 - 44.<br />
Luk, J.Y.K. and Brown, J.I. (1987). The <strong>use</strong> of piezoelectric cable as an axle sensor. Proc. New<br />
Zealand Road<strong>in</strong>g Symposium, Well<strong>in</strong>gton, pp297 - 3<strong>04</strong>.<br />
Luk, J.Y.K. and Karl, C.A. (2003). Aust<strong>road</strong>s Road Use Data Workshop. Progress Report RC3370-<br />
1. ARRB Transport Research Ltd, Vermont South.<br />
Ma<strong>in</strong> Roads WA (2003). Notes on procedures to determ<strong>in</strong>e homogeneous traffic sections. Data<br />
Plann<strong>in</strong>g and Standards Section, Ma<strong>in</strong> Roads, East Perth.<br />
Mart<strong>in</strong>, T. and Chiang, K. (2002). Consistency of <strong>data</strong> on traffic load<strong>in</strong>g and traffic mix: literature<br />
review. Contract Report RC2050-2. ARRB Transport Research Ltd, Vermont South.<br />
Mendgor<strong>in</strong> L., Peachman, J. and White, R. (2003). The <strong>collection</strong> of classified counts <strong>in</strong> an urban<br />
area – accuracy issues and results. Forum Paper, 26 th Australasian Transport Research Forum, 1-<br />
3 October, Well<strong>in</strong>gton. Paper No. 27 (CD ROM).<br />
Mihai, F. (20<strong>04</strong>). Road <strong>use</strong> <strong>data</strong>, vital <strong>in</strong>formation for decision mak<strong>in</strong>g, MRWA framework for <strong>data</strong><br />
management. Discussion Paper, Ma<strong>in</strong> Roads WA, East Perth.<br />
National Association of Australian State Road Authorities (1982). Guide to Traffic Count<strong>in</strong>g <strong>in</strong> State<br />
Road Authorities. <strong>AP</strong> TEC-13. NAASRA, Sydney.<br />
Peters, B. (2002). Heavy vehicle mass management at the Port of Fremantle. Proc. AITPM<br />
National Conference, 8-9 August 2002, Perth. pp. 269-285.<br />
Ramsay, E., Prem, H. and Pearson, B. (2001). Dimensions and mass characterisation of the<br />
Australian heavy vehicle fleet. NRTC Work<strong>in</strong>g Paper. National Road Transport Commission,<br />
Melbourne.<br />
Roads and Traffic Authority (2003). Notes on <strong>data</strong> consistency and the determ<strong>in</strong>ation of uniform<br />
traffic sections, Traffic Management Branch, RTA, Sydney.<br />
Roper, R. (2001). National highway traffic count procedures study: f<strong>in</strong>al report. Contract Report<br />
RC1632-3. ARRB Transport Research, Vermont South.<br />
Roper, R., Chiang, K. and Mart<strong>in</strong>, T. (2002). Consistency of <strong>data</strong> on traffic load<strong>in</strong>g and traffic mix:<br />
current and future <strong>road</strong> <strong>use</strong> <strong>data</strong> report. Contract Report RC2050-3. ARRB Transport Research,<br />
Vermont South.<br />
Smith, B.L., Scherer, W.T. and Conk<strong>in</strong> J.H. (2003). Explor<strong>in</strong>g imputation techniques for miss<strong>in</strong>g<br />
<strong>data</strong> <strong>in</strong> transportation management systems. Transportation Research Record TRR 1836. pp. 132-<br />
142.<br />
Thoresen, T. and Michel, N. (2002). Improved hourly traffic volume measurement. Contract Report<br />
RC 1346. ARRB Transport Research, Vermont South.<br />
Transfund New Zealand (2001). Guide to estimation and monitor<strong>in</strong>g of traffic count<strong>in</strong>g and traffic<br />
growth. Research Report No. 205. Transfund, Well<strong>in</strong>gton, N.Z.
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 57 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
Transit New Zealand (2002). Standard Professional Services Specifications – State Highway<br />
Traffic Monitor<strong>in</strong>g. Version 1. TNZ, Well<strong>in</strong>gton.<br />
Transportation Research Board (2000). Highway Capacity Manual HCM 2000. TRB, Wash<strong>in</strong>gton,<br />
DC.<br />
Western European Road Directorate (2003). Data management for <strong>road</strong> adm<strong>in</strong>istrations: A best<br />
<strong>practice</strong> guide. Version 2. WERD, Sub-group Road Data, Brussels.
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
<strong>AP</strong>PENDIX A – NATIONAL GLOSSARY OF TERMS<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 58 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
Annual Average Daily Traffic AADT is a measure of the average daily traffic pass<strong>in</strong>g a<br />
<strong>road</strong>side observation po<strong>in</strong>t over the period of a calendar<br />
year. One method of calculat<strong>in</strong>g AADT is by count<strong>in</strong>g the<br />
total volume of traffic pass<strong>in</strong>g an observation po<strong>in</strong>t and<br />
divided by the number of days <strong>in</strong> that year (365 or 366 days).<br />
Annual Average Weekday Traffic AAWT is the average 24 hour traffic volume on weekdays<br />
(Mondays to Fridays exclud<strong>in</strong>g public holidays) throughout a<br />
12 month period, at a specific observation po<strong>in</strong>t.<br />
Average Daily Traffic ADT is a sample of the AADT and is the traffic count<br />
averaged over a particular month, a week or a few days.<br />
Average Weekday Daily Traffic AWDT is taken as the average 24-hour count over a<br />
consecutive seven-day period from Monday to Sunday. It is<br />
often considered beca<strong>use</strong> of the empirical observation that<br />
the longer the count<strong>in</strong>g period <strong>use</strong>d to observe a traffic<br />
stream, the better the result<strong>in</strong>g estimates of the design<br />
parameters AADT or HHV.<br />
Average Weekday Traffic AWT is taken as the average 24-hour count over the period<br />
Monday to Friday. In some situations the average seven-day<br />
weekly traffic may be required and referred to as Average<br />
Weekday Daily Traffic (AWDT).<br />
Axle counts This is the number of actuations on an axle sensor such as a<br />
pneumatic tube as the wheels of vehicles cross over the<br />
sensor.<br />
Axle pair counts This is one half of the number of axle counts and is <strong>use</strong>d as<br />
an <strong>in</strong>dication of the number of vehicle counts. Axle pair<br />
counts are always less than the number of vehicles and can<br />
only be an approximation beca<strong>use</strong> a traffic stream will have<br />
vehicles with more than two axles. A correction factor is<br />
usually applied after calibration.<br />
Design Hour Volume DHV is the traffic flow rate chosen as the design traffic load<br />
for a facility over its design life. Common <strong>practice</strong> is to<br />
choose an ‘nth’ HHV as the design volume, with the 30 th<br />
highest hourly volume often <strong>use</strong>d <strong>in</strong> a rural environment and<br />
the 80 th HHV <strong>in</strong> an urban area. The 100 th HV (or 100 HV) is<br />
<strong>use</strong>d on National Highways. This probabilistic concept is<br />
chosen beca<strong>use</strong> it is uneconomic, if not impossible, to<br />
design a facility realistically to meet the highest traffic flow<br />
rate.
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 59 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
Highest Hourly Volume HHV is the highest hourly volume of any cont<strong>in</strong>u<strong>in</strong>g 60<br />
m<strong>in</strong>ute period over a whole year. There are a number of<br />
these volumes <strong>use</strong>d <strong>in</strong> <strong>road</strong> design, and HHV is usually rated<br />
<strong>in</strong> terms of an ‘nth’ highest hour volume, mean<strong>in</strong>g the hourly<br />
traffic volume (veh/h) exceeded <strong>in</strong> only n hours of a year.<br />
The 30th and 80th highest hourly volumes (denoted as 30<br />
HV and 80 HV respectively) are commonly <strong>use</strong>d parameters<br />
<strong>in</strong> assess<strong>in</strong>g the design volume for sett<strong>in</strong>g the capacity of a<br />
traffic facility.<br />
L<strong>in</strong>k count A l<strong>in</strong>k count is the number of vehicles pass<strong>in</strong>g an<br />
observation po<strong>in</strong>t along a <strong>road</strong> l<strong>in</strong>k over a given period. The<br />
count may be bi-directional (i.e. a two-way total), or may be<br />
split <strong>in</strong>to separate counts for the two directions of flow.<br />
n th Highest Hourly Volume This is the nth largest hourly volume recorded for a year and<br />
denoted as n HV. The 30 th highest hourly volume <strong>in</strong> the year<br />
(30 HV) is often <strong>use</strong>d to as a representative traffic volume for<br />
<strong>road</strong> design.<br />
Pattern Station A Pattern Station is a count<strong>in</strong>g station that has at least 12<br />
months’ cont<strong>in</strong>uous or frequently sampled count<strong>in</strong>g. There<br />
are two types of Pattern Stations – Seasonal and Permanent<br />
Stations. They are <strong>use</strong>d for calculat<strong>in</strong>g seasonal adjustment<br />
factors for traffic <strong>data</strong> collected at Short-term count<strong>in</strong>g<br />
stations. These factors are <strong>use</strong>d for generat<strong>in</strong>g the annual<br />
seasonally adjusted statistics for those same count<strong>in</strong>g<br />
stations, e.g. AADT.<br />
Peak Hour Factor A Peak Hour Factor (PHF) is the ratio of total hourly volume<br />
to the maximum flow rate over a specific time period with<strong>in</strong><br />
that hour. If the time period is 15 m<strong>in</strong>utes, then PHF = hourly<br />
volume / (4 × maximum 15 m<strong>in</strong>ute counts). The ratio<br />
represents the variation of traffic volume with<strong>in</strong> an hour and<br />
is less than or equal to 1. PHF = 1 represents an even<br />
distribution over an hour.<br />
Peak Hour Volume A Peak Hour Volume is the maximum traffic count observed<br />
<strong>in</strong> any 60-m<strong>in</strong>ute <strong>in</strong>terval dur<strong>in</strong>g a day. In rural areas it is<br />
usually sufficient to quote a s<strong>in</strong>gle peak hour volume. In<br />
urban areas two peak hour volumes are often considered:<br />
one for the morn<strong>in</strong>g and one for the even<strong>in</strong>g. This <strong>practice</strong> is<br />
adopted beca<strong>use</strong> of the likelihood of significant differences <strong>in</strong><br />
the directional flows on urban <strong>road</strong>s at different times of day.<br />
Permanent Station A permanent station is a count<strong>in</strong>g station <strong>in</strong>stalled for long<br />
cont<strong>in</strong>uous periods, usually many years.<br />
Seasonal Station This is a count<strong>in</strong>g station <strong>in</strong>stalled for a specific period<br />
(usually one year) to obta<strong>in</strong> <strong>in</strong>formation for factor<strong>in</strong>g shortterm<br />
counts to establish AADT.<br />
Short-term Station This is a count<strong>in</strong>g station set up for a short time. The counter<br />
is then moved to the next site to provide the widest possible
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 60 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
traffic <strong>data</strong> coverage with<strong>in</strong> a region. Seasonal factors are<br />
<strong>use</strong>d to estimate AADT from this <strong>data</strong>.<br />
Traffic patterns These are the variation of traffic volumes over a period of<br />
time.<br />
Turn<strong>in</strong>g movement A turn<strong>in</strong>g movement count is the number of vehicles<br />
observed to make a particular turn<strong>in</strong>g movement (left or right<br />
turn, or through movement) at an <strong>in</strong>tersection over a<br />
specified period. All traffic counts fall <strong>in</strong>to one or other<br />
category of l<strong>in</strong>k counts or turn<strong>in</strong>g movement counts.<br />
Twelve hour volume The 12-Hour Volume on a <strong>road</strong> is the number of vehicles<br />
pass<strong>in</strong>g an observation po<strong>in</strong>t over a given 12-hour <strong>in</strong>terval<br />
dur<strong>in</strong>g a day.<br />
Vehicle-Kilometres Travelled Vehicle-Kilometres Travelled (VKT) is the amount of travel<br />
given by the product of the total number of vehicles and their<br />
distance travelled. VKT is usually expressed as an annual<br />
statistic, and sometimes expressed as a daily statistic due to<br />
the convention of express<strong>in</strong>g traffic volumes as an average<br />
daily value. The yearly VKT is the daily VKT multiplied by the<br />
number of days <strong>in</strong> that year (365 or 366 days).
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
<strong>AP</strong>PENDIX B – NUMBER OF AADT OBSERVATIONS<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 61 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
The recommended method of estimat<strong>in</strong>g the number of AADT observations required to calculate<br />
VKTij <strong>in</strong> any region i and <strong>in</strong> stratum j is as follows:<br />
Calculate the follow<strong>in</strong>g estimates from the most recent available <strong>data</strong>:<br />
VKT i<br />
SD<br />
ij<br />
L<br />
ij<br />
ni<br />
- amount of travel <strong>in</strong> region i (=∑ VKTij<br />
),<br />
j=<br />
1<br />
- standard deviation of the AADT observations <strong>in</strong> region i and stratum j,<br />
- total length of <strong>road</strong> <strong>in</strong> region i and stratum j,<br />
ni - the number of strata <strong>in</strong> region i.<br />
Calculate the required number of AADT observations <strong>in</strong> region i and stratum j us<strong>in</strong>g the follow<strong>in</strong>g<br />
expression (NAASRA 1982):<br />
N<br />
ij<br />
Lij<br />
SDij<br />
=<br />
( 0.<br />
05VKT<br />
)<br />
i<br />
ni<br />
2 ∑<br />
j=<br />
1<br />
L<br />
ij<br />
SD<br />
ij<br />
where:<br />
N<br />
ij<br />
is the required number of AADT observation <strong>in</strong> region i and stratum j.<br />
(Equation B1)<br />
An example calculation to illustrate the procedure is as shown <strong>in</strong> Table B1 with eight strata (ni = 8).<br />
The site is on a rural <strong>road</strong>. The estimates of VKTij, Lij and SDij are from exist<strong>in</strong>g <strong>in</strong>ventory <strong>data</strong> <strong>in</strong><br />
region i.<br />
AADT stratum<br />
j =1 to 8<br />
0-100<br />
101-300<br />
301-700<br />
701-1100<br />
1101-2000<br />
2001-4000<br />
4001-7000<br />
7000 plus<br />
VKT ij<br />
(×10 3 veh-km)<br />
2586<br />
2750<br />
3277<br />
2409<br />
3496<br />
4013<br />
2574<br />
2703<br />
Table B1 - Calculation of the Number of VKT Stations<br />
Inventory Data <strong>in</strong> Region i<br />
L<br />
ij<br />
(km)<br />
94821<br />
15275<br />
6931<br />
2699<br />
2310<br />
1428<br />
496<br />
275<br />
SD<br />
ij<br />
(veh)<br />
29<br />
60<br />
111<br />
110<br />
279<br />
543<br />
848<br />
2459<br />
L<br />
ij<br />
× SD<br />
ij<br />
(veh-km)<br />
2,749,809<br />
916,500<br />
769,341<br />
296,890<br />
644,490<br />
775,4<strong>04</strong><br />
420,608<br />
676,225<br />
No. of stations<br />
N ij<br />
14.07 (14)<br />
4.7 (5)<br />
3.94 (4)<br />
1.52 (2)<br />
3.30 (4)<br />
3.97 (4)<br />
2.15 (3)<br />
3.46 (4)<br />
Sum 23808 - - 7,249,267 40
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 62 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
The total VKTi for this region i us<strong>in</strong>g eight strata = ∑ VKT ij = 23,808 × 10<br />
j=<br />
1<br />
3 veh-km<br />
8<br />
∑<br />
j=<br />
1<br />
The summation ij ij SD L = 7,249,267 veh-km<br />
The number of count<strong>in</strong>g stations for each stratum ( N ij ) is calculated us<strong>in</strong>g Equation B1. For<br />
example, Li1 = 94,821 km and SDi1 = 29 veh for j=1 and<br />
N i 1<br />
94,<br />
821×<br />
29<br />
= × 7,<br />
249,<br />
267 = 14.07 or 14 stations<br />
3 2<br />
( 0.<br />
05 × 23,<br />
808 × 10 )<br />
This method has been developed further <strong>in</strong> Botterill and Luk (1998) by <strong>in</strong>clud<strong>in</strong>g accuracy and<br />
statistical significance requirements and relax<strong>in</strong>g the requirement on preset AADT strata. Us<strong>in</strong>g the<br />
orig<strong>in</strong>al method, the number of stations required could be underestimated.<br />
8
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 63 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
<strong>AP</strong>PENDIX C – EXECUTIVE SUMMARY OF LITERATURE REVIEW<br />
REPORT (RC2050-2)<br />
Scope<br />
This Report documents a literature review of typical traffic <strong>data</strong> <strong>collection</strong>, process<strong>in</strong>g, <strong>analysis</strong> and<br />
report<strong>in</strong>g <strong>practice</strong>s by major <strong>road</strong> agencies <strong>in</strong> Australia, Canada, United K<strong>in</strong>gdom (UK), New<br />
Zealand and the United States (US). The emphasis of the review was on highlight<strong>in</strong>g <strong>practice</strong>s<br />
that differ or are <strong>in</strong> addition to the NAASRA 1982 Guide. The features of the non-Australian<br />
<strong>practice</strong>s were also summarised and compared with the NAASRA Guide.<br />
F<strong>in</strong>d<strong>in</strong>gs<br />
In some <strong>in</strong>stances, the review was not able to f<strong>in</strong>d some specific <strong>in</strong>formation from the <strong>practice</strong>s<br />
reviewed. For example, <strong>use</strong>ful <strong>in</strong>formation about the frequency of traffic <strong>data</strong> <strong>collection</strong> was not<br />
available for New Zealand and the US and <strong>in</strong>formation about traffic <strong>data</strong> presentation was not<br />
found for New Zealand and the UK.<br />
Inconsistencies between <strong>practice</strong>s<br />
B<strong>road</strong> <strong>in</strong>consistencies between the reviewed <strong>road</strong> agency <strong>practice</strong>s were identified <strong>in</strong> regard to:<br />
the def<strong>in</strong>ition of the traffic <strong>in</strong>formation to be collected and reported;<br />
the traffic guidel<strong>in</strong>es and standards on the appropriate amount of traffic count<strong>in</strong>g (cont<strong>in</strong>uous<br />
vs. temporary count<strong>in</strong>g), the duration of temporary traffic count<strong>in</strong>g and the methodology of<br />
arrang<strong>in</strong>g traffic counter sites;<br />
vehicle classification systems, the type of equipment <strong>use</strong>d and the record<strong>in</strong>g of vehicle types;<br />
and,<br />
recommendations for achiev<strong>in</strong>g adequate levels of quality control and the required accuracy<br />
of traffic measurement.<br />
Future guidance<br />
The follow<strong>in</strong>g f<strong>in</strong>d<strong>in</strong>gs identified as a result of this review may form a basis to resolve the current<br />
<strong>in</strong>consistencies <strong>in</strong> traffic monitor<strong>in</strong>g procedures:<br />
the development of a consistent generic framework for traffic measurement, process<strong>in</strong>g and<br />
report<strong>in</strong>g that can accommodate variations <strong>in</strong> equipment technology and differ<strong>in</strong>g <strong>road</strong><br />
network characteristics;<br />
the development of a systematic approach to traffic monitor<strong>in</strong>g and report<strong>in</strong>g that captures<br />
changes from previous monitor<strong>in</strong>g and report<strong>in</strong>g;<br />
a consistent approach to the estimation of annual average daily traffic (AADT) and other<br />
fundamental traffic measurement parameters;<br />
the development of a sound and defensible basis for traffic adjustment factors <strong>use</strong>d dur<strong>in</strong>g<br />
<strong>data</strong> process<strong>in</strong>g; and,<br />
the development of pr<strong>in</strong>ciples for clear quality control procedures cover<strong>in</strong>g equipment<br />
calibration, quantification of equipment accuracy and the accuracy required for the estimated<br />
traffic parameters.
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 64 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
<strong>AP</strong>PENDIX D – EXECUTIVE SUMMARY OF CURRENT AND FUTURE<br />
ROAD USE REPORT (RC2050-3)<br />
Scope:<br />
This Report documents a review of current and future needs for <strong>road</strong> <strong>use</strong> <strong>data</strong> cover<strong>in</strong>g the<br />
follow<strong>in</strong>g:<br />
identification of current <strong>road</strong> <strong>use</strong> <strong>data</strong> related to traffic volume and flow, traffic composition,<br />
weigh-<strong>in</strong>-motion and other <strong>road</strong> <strong>use</strong> characteristics;<br />
a discussion of the limitations of the current <strong>road</strong> <strong>use</strong> <strong>data</strong>;<br />
identification of future needs for <strong>road</strong> <strong>use</strong> <strong>data</strong> related to traffic volume and flow, traffic<br />
composition and other derived traffic parameters; and,<br />
a discussion of the issues aris<strong>in</strong>g from the capture of future <strong>road</strong> <strong>use</strong> <strong>data</strong>.<br />
This Report forms the next logical step <strong>in</strong> the process of develop<strong>in</strong>g the best <strong>practice</strong> <strong>in</strong> <strong>road</strong> <strong>use</strong><br />
<strong>data</strong> <strong>collection</strong>, <strong>analysis</strong> and its report<strong>in</strong>g.<br />
Outcomes:<br />
Current Road Use Data<br />
Current <strong>road</strong> <strong>use</strong> <strong>data</strong> <strong>collection</strong> <strong>practice</strong>s relat<strong>in</strong>g to traffic volume, flow, composition, weigh-<strong>in</strong>motion<br />
and other <strong>road</strong> characteristics across the States are <strong>in</strong>consistent beca<strong>use</strong> of differences <strong>in</strong><br />
the standards adopted, the budget allocations available, level of network coverage, specific State<br />
requirements and the technology <strong>use</strong>d.<br />
Future Road Use Data<br />
Traffic volume and flow are likely to rema<strong>in</strong> the basic <strong>in</strong>dicators of <strong>road</strong> <strong>use</strong> <strong>in</strong> the foreseeable<br />
future and it is expected that further advances <strong>in</strong> the technology available to measure volume and<br />
flow will cont<strong>in</strong>ue to evolve. Emphasis will be on improv<strong>in</strong>g measurement accuracy and vehicle<br />
classification across the network and ga<strong>in</strong><strong>in</strong>g the benefits from new technology for <strong>road</strong> <strong>use</strong> <strong>data</strong><br />
<strong>collection</strong> applications.<br />
The <strong>use</strong> of new technologies is likely to be <strong>in</strong>itially targeted towards heavily trafficked routes, major<br />
freight routes and other strategic routes of commercial or community value where the greatest<br />
potential exists to ga<strong>in</strong> additional value from the <strong>in</strong>creased knowledge of traffic flows.<br />
A <strong>road</strong> <strong>use</strong> <strong>data</strong> <strong>collection</strong> framework needs to be developed for the long-term, <strong>in</strong>clud<strong>in</strong>g storage<br />
and report<strong>in</strong>g standards and systems, implementation plans, audit and review mechanisms and<br />
long-term strategies.<br />
Improved def<strong>in</strong>ition of traffic composition through new technology will allow for greater detailed<br />
<strong>in</strong>formation about the characteristics of <strong>in</strong>dividual vehicles travell<strong>in</strong>g on the <strong>road</strong> network. This<br />
additional <strong>in</strong>formation could allow an expanded vehicle classification system to be developed. This<br />
detailed <strong>in</strong>formation could be <strong>use</strong>d for enforcement and regulatory purposes, if it is acceptable.<br />
Issues of privacy and access to this form of detailed <strong>data</strong> are likely to restrict this <strong>use</strong> of the <strong>data</strong>.
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 65 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
A number of potentially <strong>use</strong>ful traffic parameters can be derived from the detailed traffic volume,<br />
flow, composition, weigh-<strong>in</strong>-motion and other <strong>road</strong> characteristics <strong>data</strong> if it is made accessible and<br />
an adequate level of resources are <strong>use</strong>d to extract these parameters. Detailed access to this <strong>data</strong><br />
is not likely to occur unless it is <strong>in</strong>sulated from enforcement.<br />
Other Issues<br />
A portion of the future <strong>road</strong> <strong>use</strong> <strong>data</strong> is likely to be <strong>in</strong> private hands plac<strong>in</strong>g access arrangements<br />
on a contractual basis. Different levels of access to the <strong>data</strong> also need to be developed to suit the<br />
<strong>in</strong>dividual requirements of <strong>data</strong> <strong>use</strong>rs. Internet based systems are expected to play a major role <strong>in</strong><br />
transferr<strong>in</strong>g <strong>data</strong> to the <strong>use</strong>rs.<br />
Data dissem<strong>in</strong>ation systems need to be made <strong>in</strong> a form compatible and consistent with other<br />
collect<strong>in</strong>g agencies and <strong>data</strong> <strong>use</strong>rs. Some attention to develop<strong>in</strong>g such systems need to be made<br />
<strong>in</strong> cooperation with all Aust<strong>road</strong>s Member Authorities (MAs).<br />
The development of the best <strong>road</strong> <strong>use</strong> <strong>data</strong> <strong>collection</strong>, <strong>analysis</strong> and report<strong>in</strong>g <strong>practice</strong> will need the<br />
<strong>in</strong>put of all MAs. The development of this <strong>practice</strong> should allow the flexibility to <strong>in</strong>troduce new<br />
technologies and ideas as they emerge to ensure that the f<strong>in</strong>al system rema<strong>in</strong>s as relevant and<br />
adaptable as possible. A b<strong>road</strong> non-prescriptive approach needs to be <strong>use</strong>d that can<br />
accommodate all these differences and future changes <strong>in</strong> technology.
Accessed by AR - ARRB TRANSPORT RESEARCH on <strong>04</strong> Feb 2005<br />
INFORMATION RETRIEVAL<br />
Aust<strong>road</strong>s 20<strong>04</strong><br />
— 66 —<br />
<strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis and Report<strong>in</strong>g<br />
Aust<strong>road</strong>s (20<strong>04</strong>), <strong>Best</strong> Practices <strong>in</strong> Road Use Data Collection, Analysis<br />
and Report<strong>in</strong>g, Sydney, A4, 76pp, <strong>AP</strong>-<strong>G84</strong>/<strong>04</strong><br />
KEYWORDS:<br />
Bicycle, Classification (vehicle), Data Integration, Delay, Dimensions (vehicle),<br />
Guidel<strong>in</strong>es, Quality Assurance, Traffic, Traffic Survey, Travel Time, Weigh-<strong>in</strong>-<br />
Motion.<br />
ABSTRACT:<br />
This report describes the current <strong>practice</strong>s of <strong>road</strong> authorities (RAs) <strong>in</strong> <strong>road</strong><br />
<strong>use</strong> <strong>data</strong> <strong>collection</strong>, <strong>analysis</strong> and report<strong>in</strong>g, and <strong>use</strong>s examples to demonstrate<br />
best <strong>practice</strong>s. It cover topics such as: error b<strong>in</strong> (b<strong>in</strong> 13) <strong>in</strong> the Aust<strong>road</strong>s<br />
vehicle classification system, vehicle classification by lengths, def<strong>in</strong><strong>in</strong>g a<br />
homogeneous traffic section, correlation of classified counts with <strong>WIM</strong> <strong>data</strong>,<br />
public-private partnership, quality checks, deal<strong>in</strong>g with miss<strong>in</strong>g <strong>data</strong>,<br />
calibration of a <strong>WIM</strong> system us<strong>in</strong>g CULWAY as an example, and a stakeholder<br />
consultation model. A key f<strong>in</strong>d<strong>in</strong>g is that RAs have been follow<strong>in</strong>g ref<strong>in</strong><strong>in</strong>g<br />
their count<strong>in</strong>g programs and <strong>in</strong> many ways have achieved good <strong>practice</strong>s, and<br />
that the best way to ensure consistency amongst RAs is for each <strong>road</strong><br />
authority to collect robust <strong>data</strong> sets.