Use of Main Roads WA Corporate Data for Road Safety Risk Data …€¦ · Corporate Data for Road Safety Risk Data Sets Exploring the use of Corporate Data for AusRAP/ANRAM Data Sets

Use of Main Roads WA

Corporate Data for Road Safety

Risk Data Sets

Exploring the use of Corporate Data

for AusRAP/ANRAM Data Sets

Commercial in confidence September 2018

Use of Main Roads WA Corporate Data for Road Safety Risk Data Sets 2017-011

ABN 68 004 620 651

Victoria

80A Turner St

Port Melbourne VIC 3207

Australia

P: +61 3 9881 1555

F: +61 3 9887 8104

[email protected]

Western Australia

191 Carr Place

Leedervil le WA 6007

Australia

P: +61 8 9227 3000

F: +61 8 9227 3030

[email protected]

New South Wales

2-14 Mountain St

Ultimo NSW 2007

Australia

P: +61 2 9282 4444

F: +61 2 9280 4430

[email protected]

Queensland

21 McLachlan St

Fortitude Valley QLD 4006

Australia

P: +61 7 3260 3500

F: +61 7 3862 4699

[email protected]

South Australia

Level 11,

101 Grenfell Street

Adelaide SA 5000

Australia

P: +61 8 7200 2659

F: +61 8 8223 7406

[email protected]

for Mark Parr (Main Roads WA)

Reviewed

Project Leader

Anna Brett

Quality Manager

Lisa Steinmetz PSS17081-1 September 2018

Commercial in confidence

- i - September 2018

SUMMARY

This project explored the suitability and applicability of using corporate data extracted from the Main Roads WA corporate database to supplement and ultimately replace the traditional (manually coded data) approach for developing AusRAP and ANRAM data sets.

Data collected during the WARRIP TSD trial was used in the comparison of Main Roads corporate data against the traditional approach. An analysis of the differences between the data sets found similarities in the coding of results of some attributes but substantial differences in many others. It is suggested that corporate data could be used for some, although not currently all, attributes for producing AusRAP / ANRAM data sets. With further investigation and fine tuning, additional attributes could potentially be extracted from the corporate database. It is noted that additional work would need to be undertaken to improve alignment and to ensure confidence in the corporate data. It is recommended that for attributes that require review and refinement, traditional (manual) rating continue in parallel to the (evolving) corporate extraction to facilitate ongoing comparison, analysis and refinement of these until there is confidence that the corporate data can be consistently and accurately extracted.

While use of corporate data is anticipated to deliver relatively small cost savings over the short-term, further cost saving opportunities would arise over time.

A number of key challenges, observations and lessons were noted during this project, including:

▪ There are challenges in aligning data from different sources.

▪ Differences in terminology and understanding / familiarity with different data within ARRB and Main Roads WA can lead to miscommunication and errors.

▪ For some attributes, corporate data may provide more accurate information than the AusRAP traditional (manual) rating approach.

▪ Assumptions relating to the accuracy and / or currency of information in the corporate database, and / or interpretation of these attributes may be erroneous.

▪ Use of corporate data will require consideration of new issues such as deterioration of assets over time.

▪ Differences in interpretation or categorisation of attributes (between corporate and traditional approach) need to be understood (particularly if these lead to differences in star rating outcomes).

▪ Recognising that some attributes are (currently) difficult to systematically or automatically extract from the corporate database. However, new and evolving techniques and technologies will help improve accuracy of data attributes and will ultimately lead to automatic classification of these.

Although the Report is believ ed to be

correct at the time of publication,

Australian Road Research Board, to the

extent lawf ul, excludes all liability f or loss

(whether arising under contract, tort,

statute or otherwise) arising f rom the

contents of the Report or f rom its use.

Where such liability cannot be excluded,

it is reduced to the f ull extent lawf ul.

Without limiting the f oregoing, people

should apply their own skill and

judgement when using the inf ormation

contained in the Report.


- ii - September 2018

ACKNOWLEDGEMENTS

The authors would like to acknowledge the large contributions to this project by key data management staff from both Main Roads WA and ARRB.


- iii - September 2018

CONTENTS

1 INTRODUCTION ........................................................................................................................ 1

1.1 Background................................................................................................................................. 1

1.2 Purpose for Project..................................................................................................................... 1

1.3 Method Overview........................................................................................................................ 2

1.4 Comment on Other Similar Projects .......................................................................................... 2

2 METHOD .................................................................................................................................... 3

2.1 AusRAP Data Sets ..................................................................................................................... 3 2.1.1 Traditional Coded Data Set (Based on TSD Data)...................................................... 3 2.1.2 Corporate Data Set (Extracted from Main Roads WA Corporate Asset Database) ... 3

2.2 ANRAM Data Sets ...................................................................................................................... 4

2.3 Analysis ...................................................................................................................................... 5 2.3.1 Data Set Comparison ................................................................................................... 5 2.3.2 AusRAP Results Comparison .................................................................................... 15 2.3.3 ANRAM Results Comparison..................................................................................... 21

3 FINDINGS AND DISCUSSION ................................................................................................ 23

3.1 Differences in Coding and Risk Assessment Results.............................................................. 23

3.2 Key Challenges, Observations and Lessons Learnt................................................................ 24

3.3 Comment on Efficiencies and Value for Money....................................................................... 26

3.4 Next Steps and Future Technology ......................................................................................... 27

REFERENCES................................................................................................................................... 29

APPENDIX A DERIVING IRAP AND ANRAM ATTRIBUTES FROM IRIS ...................... 30 APPENDIX B EXAMPLE ROAD SEGMENT LENGTH APPROACH ............................... 60 APPENDIX C CODING COMPARISON BETWEEN DATA SETS.................................... 63 APPENDIX D CODING COMPARISON BY J KARPINSKI............................................... 80 APPENDIX E GLOSSARY ................................................................................................. 81


- iv - September 2018

TABLES

Table 2.1: Calibration factors ...................................................................................................... 4 Table 2.2: Data set comparison (length & rows) ........................................................................ 6 Table 2.3: Data set comparison overview .................................................................................. 8 Table 2.4: Critical attributes in crash risk estimation of different crash types .......................... 14 Table 2.5: AusRAP star rating results....................................................................................... 16 Table 2.6: AusRAP star rating comparison .............................................................................. 18 Table 2.7: AusRAP star rating detailed comparison................................................................. 18 Table 2.8: ANRAM results summary (traditional vs corporate) ................................................ 22

FIGURES

Figure 2.1: Corporate data: frequency of segment lengths ......................................................... 6 Figure 2.2: AusRAP star rating map .......................................................................................... 17 Figure 2.3: SRS correlation results ............................................................................................ 20

Use of Main Roads WA Corporate Data for Road Safety Risk Data Sets PSS17081-1


- 1 - September 2018

1 INTRODUCTION

1.1 Background

The Safe System approach establishes an ethical position that no one should die or be seriously injured on the road. To fulfil this vision there is a requirement to identify parts or segments of the road network that require treatment.

The traditional approach to identify crash risk is through crash history (reactive approach). Given that human error may occur at any time on the road network, and that the human tolerance to impact forces may be exceeded on many sections of the network, a proactive approach is valuable. A proactive approach enables Main Roads to be more confident in determining locations that will maximise the number of lives and serious injuries saved for dollars invested; this directly links with Main Roads Safety Policy Statement, the Western Australia State Government’s Road Safety Strategy Towards Zero and the National Road Safety Strategy.

AusRAP and ANRAM can provide insights into identifying crash risk across the road network, and ultimately provide a network-based approach to guide severe crash risk reduction programs:

▪ AusRAP focusses on SRS outputs presented as star ratings. ANRAM has a greater focus on reporting future crash risk in terms of expected FSI crashes and also gives users greater control over treatment selection.

▪ AusRAP uses the total number of fatalities and serious injuries and crash types across the road network to calibrate the fatality estimation model. It does not rely on the spatial location of crashes.

▪ ANRAM uses crash history across the road network within the Crash Validation Module as part of calculations to estimate expected crashes. The spatial location of crashes is used to achieve a more accurate estimate of expected fatal and serious injury crashes for a given road network or route. ANRAM provides enhanced fatality and serious injury estimates for road sections, given infrastructure and operation information (e.g. sealed shoulder width, lane width, speed limit, traffic volume) and observed crash data. ANRAM also provides the ability to ‘test’ treatment options and estimate the reduction in fatal and serious injury crashes given the proposed countermeasures.

1.2 Purpose for Project

Main Roads has an extensive and sophisticated asset database. This project looked to explore the suitability and applicability of using corporate data extracted from this database for AusRAP and ANRAM assessments. More specifically, the project looked to:

▪ provide / explore the level of confidence on attributes that may be drawn from the Main Roads WA corporate data set

▪ opportunities to deliver increased efficiency and value for money through using corporate data for multiple purposes (in this case use of asset data to inform crash risk on the network).

The project also builds on Main Roads WA and ARRB’s experience and knowledge in relation to application of ANRAM and sources of data, and also fosters close collaboration to improve and share knowledge for Main Roads WA and Australian road agencies more widely.




1.3 Method Overview

Data collected during the WARRIP TSD1 trial was used in the comparison of Main Roads corporate data against the traditional approach (manually coded data). A comparison of the results was undertaken whereby challenges with particular variables were identified and addressed with the view to quantifying the accuracy, efficiency and cost of using corporate data for this type of analysis.

This project seeks to provide greater clarity on attributes that may be drawn from the Main Roads corporate data set for use in ANRAM or AusRAP assessments. To facilitate this, the reporting / collection for the corporate and manually coded data were from similar time periods (meaning the data should be comparable from a time perspective). As part of the investigation, there was strong collaboration with key Main Roads staff, particularly in preparation of the data sets in order to identify, manage and resolve issues as they arose, in order to leverage and maximise the technical expertise residing within ARRB and Main Roads.

1.4 Comment on Other Similar Projects

Previous studies have demonstrated the potential for reduced coding requirements with respect to manual risk assessment ratings (Karpinski 2014; ARRB internal communication) through the integration of corporate data. This project aims to apply lessons learnt from previous analyses against a high-level quality-controlled data set.

Appendix D shows the coding comparison undertaken by J Karpinski between traditionally coded AusRAP (manual coding) and corporate data sets.

Jan Karpinski (2014) noted:

▪ ‘Care needs to be taken when using inventory for the preparation of these data sets, as some inventory may be more up to date than the video used for rating purposes’.

▪ ‘Providing the inventory is current and accurate (which is difficult to quantify), the creation of data from inventory removes rater bias for’ attributes that may have a level of subjectivity. It was noted that manually rated data tended to regularly code road sections with patching as having ‘medium’ or ‘poor’ road condition, when the measured road condition data indicated that these segments were ‘good’ (in terms of roughness and rutting).

▪ ‘Depending on the methodology used to assemble the [corporate] data’ other biases may be evident or introduced ‘based on assumptions used to create the [corporate] data’, and such biases may result in systematically and consistently higher or lower measurements (for instance).

The above observations are true, although it is also noted that ARRB’s experience has also found that:

▪ on occasion, aspects of corporate databases may be outdated themselves, if that part of the corporate database has not been kept up to date

▪ other bias or subjectivity may be evident in any data set, particularly in relation to data that involves human interpretation or input. This may be present in the corporate data set or traditionally coded data set.

1 The TSD is now known as iPAVE in Australia.




2 METHOD

The project involved:

▪ preparation of AusRAP data sets

— creation of traditional data set – coding of TSD video data

— creation of data set based on Main Roads WA corporate asset data

▪ preparation of ANRAM data sets

▪ analysis

— comparison of ‘coding standard’ of the two data sets

— AusRAP and ANRAM analysis of each data set

— interpretation of the outcomes.

2.1 AusRAP Data Sets

As part of preparation of the AusRAP data sets:

▪ quality checks were undertaken

▪ extensive work was undertaken to ensure alignment of the data sets

▪ traffic volume data was incorporated and the data sets uploaded to ViDA to obtain the AusRAP results.

2.1.1 Traditional Coded Data Set (Based on TSD Data)

The previously collected TSD data was referenced to the road distance2 as used by Main Roads. Liaison with Main Roads was required to ensure the referencing lined up correctly with existing road distance and data to ensure the coded data matched the corporate data.

The video data with the corrected referencing was coded in accordance with the accepted AusRAP coding procedures (iRAP 2014).

Quality and sense checks were conducted on coding results as per the requirements of the AusRAP quality assurance (QA) guideline requirements.

2.1.2 Corporate Data Set (Extracted from Main Roads WA Corporate Asset Database)

A second data set was created that incorporates Main Roads WA corporate data for the agreed fields. The choice of categories incorporated and the mechanism for extracting the data was determined in consultation with Main Roads, outcomes from the previous analysis (Karpinski 2014) and ARRB knowledge.

Preparation of this data set involved close collaboration between Main Roads WA and ARRB to build the data set. This included:

▪ Discussion and agreement on attributes that would be drawn from the corporate database versus adoption of default values, and some instances where coded data values were to be used in the corporate data set.

2 For this project, ‘road distance’ was referenced to Main Roads WA ‘True Distance’.




▪ A large amount of work was required to test and ensure alignment of the data. Multiple iterations of producing versions of the corporate data set were required. In doing this, part of the data needed to be dropped from the comparison because alignment could not be achieved for those sections.

▪ Early in the project it was noted that differences in terminology and understanding / familiarity with different data within ARRB and Main Roads WA was leading to miscommunication and errors. These issues were identified and resolved.

AADT data (provided by Main Roads) was incorporated into both data sets prior to final completeness checks / QA being carried out, processing and data upload to ViDA. Where AADT was not available, Main Roads provided default vehicle flows to be applied.

Appendix A provides an outline of the attributes and the underlying approach for extracting the relevant information from the corporate data base by Main Roads WA. As the project progressed, the approach for corporate data extraction needed to be revised. The code for extracting the corporate data was produced using the corporate statistical program SAS.

2.2 ANRAM Data Sets

Data was exported from ViDA and final ANRAM sense checks applied to both data sets. Fatal and serious injury (FSI) data (provided by Main Roads) was incorporated and the data sets were translated to ANRAM format.

This project benefited from refinements over the last few years that have been incorporated into the ANRAM translation process developed by ARRB to improve the quality and consistency of AusRAP and ANRAM data sets.

FSI crash data (for a five-year period) were then assigned to each ANRAM road section. The crash data period was selected to reflect the infrastructure data (i.e. corresponding to the ANRAM input data set).

ANRAM calibration values were prepared for the road network, using the five-year crash data provided for this project.

Crash data was also used to calibrate ANRAM (2014, ANRAM v1.04) for the rural undivided road stereotype. The calibration factors used in this project are shown Table 2.1 .

Table 2.1: Calibration factors

Crash type Rural undivided

Traditionally coded data

Rural undivided

Corporate data

Run-off-road 0.75 1.53

Head-on 1.01 0.69

Intersection 0.95 2.17

Other 1.45 1.73




2.3 Analysis

An analysis was undertaken to compare results between the two approaches, using the traditional coded data set as the baseline to further refine the methods and identifying lessons learnt.

2.3.1 Data Set Comparison

A review of the differences between the data sets was undertaken. This included checking for differences between the number of rows, the length (km) and coding differences:

▪ Number of rows: the ‘traditionally coded’ data set and ‘Main Roads WA corporate’ data set files (in both AusRAP upload file and ANRAM data file format) have the exact same number of rows of data (6252), as well as the corresponding crash data file (see Table 2.2).

▪ Volumes: It was noted that the ‘traditionally coded’ data set had old AADT data. The traffic volumes in the ‘traditionally coded’ data set were updated to reflect the volumes in the ‘Main Roads WA corporate’ data set.

▪ Coding differences: The section below (‘Coding differences’) provides an overview of the coding differences between the two data sets, while Appendix C includes a more detailed breakdown between the coding of the two data sets. The ‘Coding differences’ section also provides further investigation of critical attributes.

▪ Length (km): It is noted that the ‘length’ field in the traditionally coded data is consistently 100 m. In comparison, the corporate data set length field varies from 0.003 (i.e. 3 m) to 0.203 (i.e. 203 m) (see Table 2.2). The Main Roads WA team was advised of this discrepancy at the time that the data sets were being collated. Because the road distance was used to extract data broken down by segment, a key learning is that this method does not result in producing 100 m segment lengths (which may affect results). Note, the Main Roads WA team advised that the short lengths were likely to be related to where the road goes from a single carriageway to dual or similar. A histogram showing the frequency of different segment lengths in the corporate data shows that the vast majority (99.6%) of the segment lengths were within the range of 90 to 110 m (Figure 2.1). Therefore, while segment lengths in the data do vary from 3 m to 203 m, this range will have a small effect on results, since these extremes form a very low proportion of segment lengths in the overall data set.

The varying segment length issue is understood, and Main Roads WA advised that the data can be easily corrected within the code for extracting the corporate data. Note, this issue does also occur in traditionally coded data sets. Initially, Main Roads WA adopted the approach of extracting the known true distance then trying to match to that provided by ARRB and correlate this to the ARRB GPS coordinate. This proved problematic with lots of variation. Several other methods were trialled producing similar issues. It was finally decided that, to maintain consistency, Main Roads WA would adopt the ARRB GPS coordinates and simply extract the corresponding true distance. The variations in road distance not matching 100 m segments resulted. However, it is important to note that when calculating the actual distance based on the GPS coordinates given using simple trigonometry (see Appendix B), the majority (99.5%) equates to 100 m rounded to the nearest two decimal places. This is also true for the ARRB data set where, based on the above approach, this data set produces the same results. In view of this, Main Roads WA could simply make each road distance segment equate to the 100 m value.




Table 2.2: Data set comparison (length & rows)

Data set Number of rows Sum of length (km)

TSD AusRAP upload file 6252 (+ a heading row) 625.2

Main Roads WA corporate

AusRAP upload file

6252 (+ a heading row) 625.566

TSD ANRAM data file 6252 (+ a heading row) 625.2


ANRAM data file

6252 (+ a heading row) 625.566

ANRAM crash data file 6252

Figure 2.1: Corporate data: frequency of segment lengths

When data was processed in iRAP’s ViDA platform, a couple of issues were identified:

▪ TSD AusRAP upload file: H005 distance 16.165 changed to carriageway type undivided (code value = 3) (this did not need to be corrected for Main Roads WA corporate data set).

▪ AusRAP download files may re-order data rows. There is a need to be mindful of this if comparisons between data sets are being undertaken (this issue was encountered during processing of the AusRAP data for this project).

Coding differences:

Table 2.3 summarises the correlation between corporate and manually coded attributes (% correct column). The legend below indicates recommendations in relation to the use of corporate data for building data sets, as well as other relevant comments relating to each attribute. As noted above, Appendix B includes a more detailed breakdown between the coding of the two data sets.

Segment length

Fre

quency




This project primarily focussed on rural routes, which meant that default values for some attributes (particularly relating to pedestrians) could be relatively confidently applied. For more urbanised assessments, the recommendation would be not to apply this philosophy.

Legend

Recommend Use of

Corporate Data

Recommend Use of

Corporate Data – Some

minor adjustment /

ongoing assessment

required

Recommend further

analysis of results,

application of Corporate

Data to be determined

Recommend data be

manually coded /

Corporate Data not

available

Main Roads WA elected

to apply default values

for this project. Main

Roads WA could

investigate drawing

these data from

corporate database in

the future

Table 2.3 also includes:

▪ Commentary on what may be the reason for the differences, and recommendations on where to focus efforts to potentially achieve greater alignment between manually coded and corporate data.

▪ Associated attributes (which are used as a cross-reference during the QA process). Note that this list does not capture every associated field but does indicate the critical ones.

▪ The iRAP quality benchmark as a useful reference. Note, that this is used during the final step of the QA process, where a sample of data is visually reviewed against the coding results for all attributes. The quality benchmark figure indicates the minimum required to satisfy iRAP requirements.

Table 2.3 lists AusRAP attributes that are rated during the traditional coding process. Other attributes (not included in this table) are the ‘post coding’ attributes, where default values may be applied, where appropriate:

▪ Attributes that may be provided by the client: AADT, Motorcycle %, pedestrian peak hour flow across and along the road, bicycle peak hour flow, and operating speed (85%ile and mean speed). Note that default values for some of these attributes may be applied (if not available in the corporate data)

▪ Attributes for which default values are applied include (as currently not used in Australia): roads that cars can read and star rating policy targets.

It is noted that the coding comparison project undertaken in 2014 (Karpinski 2014, see Appendix D) achieved greater alignment between manually coded and corporate for many attributes, although it is noted that some attributes achieved a similar or better correlation in this current project. Differences in these outcomes may be due to a variety of factors, such as the method used for extraction of corporate data (and other refinements applied) and the location that the data is drawn from (i.e. greater alignment may be more easily achieved for certain road types).




Table 2.3: Data set comparison overview

Field /

category

% match

between

Corp and

Traditional

iRAP quality

benchmark Comments / recommendations

Field linked to

for QA

Area Type 86.1% 90%

For the corporate data, speed has been used as the primary factor in

determining area type. Manual coding also considers residential density .

Recommend rev iew of corporate methodology based on the above.

Speed Limit

Speed Limit 95.7% 95%

Investigation may be undertaken into where the variations occurred to

determine cause (i.e. was this related to data alignment). Variation may also

have occurred where there is a change of speed mid segment. Where coded

manually , the higher speed would be applied for that segment. Recommend

rev iewing methodology applied by Main Roads WA where speed changes mid

segment.

Area Type

Carriageway

Label 99.8% 100%

Recommend investigating further to determine cause for minor misalignment

(would expect this attribute to be 100% match, although difference could be

due to differences in start and end points of segments between the two

approaches).

Median

Motorcycle

Flow Obs 99.9% 95%

Default value of 1 (none) applied.

Main Roads WA advised that unable to record due to how Austroads vehicle

classification applied. May be able to extract using raw data based on min ax le

spacing. Need to be mindful of micro cars e.g. Smart for 2 wheelbase 1.8m

Bicycle Flow

Obs 100% 95%


Pedestrian

Flow Obs –

LHS

99.8% 95%


Pedestrian

Flow Obs –

RHS

100% 95%


Pedestrian

Flow Obs –

Crossing Road

N/A 95%


No. of Lanes 26% 98%

Recommend rev iew results and methodology. Expect this field to have good

alignment. There is a high variation between manually coded data and

corporate data, especially relating to one and two lanes.

Main Roads WA can confirm there is a potential issue with the code producing

erroneous results for this attribute.

Lane Width 97.9% 95%

Manual coding approach records the narrowest lane width from within the

segment. Recommend rev iew of corporate methodology to further refine

approach.

Paved

Shoulder Width

– LHS

59.8% 95%

Main variation where manually coded considered narrow, whilst corporate data

indicates wide. Recommend rev iew results especially in relation to undiv ided

roads. Methodology for corporate data is basing width on RHS whereas for

manually coded data, consideration given to both sides with the highest risk

(i.e. least wide) result being recorded.


erroneous results for this attribute, as it is related to No. of Lanes code.

Sidewalk

Prov ision




Field /

category

% match

between

Corp and

Traditional

iRAP quality


Field linked to

for QA

Paved

Shoulder Width

– RHS

51% 95%

Refer above. Sidewalk

Prov ision

Shoulder

Rumble Strips 70.5% 98%

Recommend investigation to determine trends for variance.

Curvature 82.5% 95%

Most variations between the data sets are only one steps different (i.e. straight

vs moderate). Review curve radius applied to determine if adjustment improves

results. Small number of outliers, investigate to determine if there is a pattern

for the variance.

The traditional rating approach for classification of the curvature attribute

involves a hybrid of measured data inputs (from network survey vehicles) and

rater input (note the rater input is generally to ensure that road segments with

features such as a roundabout are not classified as a curve, etc)

Main Roads WA data is drawn from actual known curve radius information,

then allocates a category based on the iRAP coding manual values for each

group. However, Main Roads WA code doesn't account for the use of a speed

value in addition to the radius, which may have an impact category allocation if

incorporated.

Speed Limit

Quality of

Curve

Quality of

Curve 81.9% 90%

Likely mainly due to differences in coding for curvature attribute. Further

investigation may be required after curvature attribute achieves better

alignment.

Based on the iRAP coding manual content, Main Roads WA considers that this

attribute has a high degree of subjective interpretation and is reliant on how an

indiv idual perceives the safety of the approach to the curve. This will be largely

impacted on driver experience. Main Roads WA code is based on comparing

the calculated horizontal safe speed a vehicle can travel to the actual sign

posted speed limit. The posted speed limit used may not include any advisory

speed signs (e.g. yellow cautionary speed signs) which may impact on the

results if included.

The iRAP coding manual approach is based on signage prov ision, and ARRB

coders also take into consideration sight distance and delineation. ARRB notes

that the two approaches for determining curve quality are different, and

therefore may lead to variations in results. Where a jurisdiction elects to adopt

a different approach to determining attribute classification, iRAP requires this

variation to be clearly documented (with an understanding in differences in

coding outcomes).

Curvature

Sight Distance

Delineation

Grade 99.9% 90%

Sight Distance 100% 90% Default value of 1 (Adequate) applied. Curvature

Delineation 21.7% 90%

Review of corporate methodology required, consider broadening requirements

(i.e. inclusion of rumble strips).

Main Roads WA to rev iew code to establish if rumble strips and other line

types incorporated in delineation attribute.

Quality of

Curve




Field /

category

% match

between

Corp and

Traditional

iRAP quality


Field linked to

for QA

Road Condition 43.6% 90%

Corporate classification for road condition seems to be more stringent than the

iRAP model. Consideration required for how this will impact star rating results.

Main Roads WA advised that this attribute is calculated based on three tiers of

customer facing levels of serv ice (LOS) relating to ride quality and the

underly ing levels of roughness using the International Roughness Index (IRI).

There is likely to be significant differences between data extracted and a

subjective v isual interpretation.

Skid

Resistance 31.4% 90%

Corporate methodology for determining skid resistance to be rev iewed further

with ongoing analysis of results – results not well-aligned with iRAP (v isual

rating)

approach. Variations in results may be due to differences in equipment

measured vs v isual ratings, currency of SCRIM / BPM or v ideo data, etc.

Main Roads WA advised that as similar to road condition, the skid resistance

attribute is based on a technical approach using calculations relating to the

type of surface (e.g. asphalt, sprayed seal, unsealed) Again, there is likely to

be significant differences between data extracted and a subjective v isual

interpretation.

Land Use –

LHS 0% 90%

Default value of 5 (Not Recorded) applied.

Land Use –

RHS 0% 90%

Default value of 5 (Not Recorded) applied.

Vehicle Parking 98.1% 95% Default Value of 1 (None) applied.

Median Type 34.7% 95%

Recommend rev iew of methodology applied for corporate data extraction.

Further analysis comparison required.


erroneous results for this attribute.

Carriageway

Centreline

Rumble Strips 90.1% 98%

Recommend rev iewing results noting that manual coding assigns ‘present’

where rumble strip is in ev idence for the full (or almost full) length of the

segment (excluding where an intersection occurs).

It would appear Main Roads WA code is assigning values based on defined

road distance ranges, rather than looking up a data set to establish if centreline

rumble strips (CRS) are present for this attribute. It may be that it was

established in an earlier piece of work that the defined segments had CRS. It is

also noted that this piece of code also allocates a median as type 14 for the

same road segments. This may explain some of the differences both this and

the median type attributes.

Street Lighting 89.5% 98%

Default value of 1 (Not Present) applied.

Main Roads WA appears to only hold street lighting data for freeways.

Information needs to be sourced from the relevant utility owners to prov ide an

extract of assets preferably with GPS locations to enable integration into the

data set.




Field /

category

% match

between

Corp and

Traditional

iRAP quality


Field linked to

for QA

Property

Access Points 88.2% 95%

Recommend investigating how corporate data accounts for commercial access

points.

Main Roads WA suggested that this can do with additional refinement. No

inclusion is prov ided for residential driveways from a known residential

database. It appears the code has been written to look at unknown road

number ‘Z’ through median opening or u-turns. Although the % hit is fairly high,

it requires further investigation.

Land Use

Object

Distance – LHS 47.4% 95%

ARRB believes roadside object and distance is currently difficult to accurately

code due to complex ity of these attributes, i.e. the nearest object that is likely

to cause the greatest injury outcome is the item coded for the segment. This is

particularly challenging as a corporate database may not be able to capture

every roadside object, and in some cases the roadside object may be on

private property (e.g. trees).

It is important to note that Main Roads WA utilised the Roadside Hazard Rating

(RHR) data set for ‘Object Distance’ and ‘Roadside Object’ for both LHS and

RHS. The RHR information was formulated using similar principles to that of

iRAP in v isually identify ing the various types of roadside hazardous objects,

categorising them and measuring their distance from the roadside. The

information is nearly 10 years old (2009) and there may be issues in the

underly ing methodology used that results in differences being introduced when

compared to iRAP. Additionally , the application of the worst-case scenario may

require further analysis in rev iewing the coding and methodology. More

investigation is required in the development and application of these attributes.

Roadside

Object – LHS 30% 95%

See above Median

Object

Distance –

RHS

40.6% 95%

See above

Roadside

Object – RHS 33.9% 95%

See above Median

Facilities for

Bicycles 82.4% 98%

Default value of 4 (None) applied.

Pedestrian

Crossing

Facilities

99.2% 98%

Default value of 7 (No Facility ) applied.

Intersections

Pedestrian

Crossing

Quality

99.3% 90%

Default values of 3 (Not Applicable) applied.




Field /

category

% match

between

Corp and

Traditional

iRAP quality


Field linked to

for QA

Sidewalk

Prov ision –

LHS

5.9% 98%

Default value of 5 (None) applied. Note, much of the difference is due to

manual coding of ‘informal path’.

Main Roads WA believes that ARRB assessors are misinterpreting the majority

of these roads to include informal paths incorrectly . It appears that raters are

confusing the unsealed road edges as informal paths. There is generally no

ev idence of pedestrians along the sides of these roads. A similar result was

found when rev iewing older AusRAP data for Albany Hwy in a prev ious project.

Main Roads WA suggests this be raised as a training point for overseas raters.

ARRB has rev iewed the commentary in the iRAP coding manual and concurs

that a rev iew of methodology is required for this attribute moving forward.

Paved

Shoulder

Sidewalk

Prov ision –

RHS

28.4% 98%

Default value of 5 (None) applied. Note, much of the difference is due to

manual coding of ‘informal path’. May be worth rev iewing how this attribute is

interpreted.

Paved

Shoulder

Pedestrian

Crossing on

Side Road

99.5% 95%

Default value of 7 (No Facility ) applied. Intersection

Type

School Zone

Warning 100% 95%

For the corporate data set school zones where identified based on roadside

signs being present or not. However, on rev iewing the code, it does not

account for ‘flashing beacons’ or variable road signs. Code rev iew required.

Note this is coded blue as the data sets for this project had no school zones.

School Zone

Crossing

Superv isor

100% 95%

Default value of 3 applied.

Pedestrian

Fencing 100% 95%


Intersection

Type 92.6% 98%

Further rev iew / analysis of results recommended. Determine if alignment

between the data sets contributed to variations.

There is a lot of complex code surrounding how this attribute is determined and

populated in the corporate data set. An audit of the differences could be

undertaken with a v iew of establishing mistakes in either ARRB or Main Roads

WA allocations and refining the code if required.

Intersection

Quality /

Channelization

and Road

Volume

Intersection

Quality 92.5% 90%

Recommend further rev iew into alignment of intersections between data sets. Intersection

Type

Intersecting

Road Volume 92.2% 98%

Recommend further analysis of methodology and results. Corporate data

source likely to prov ide more accurate results. Variations in results expected

but will need to be quantified.

Intersection

Type

Intersection

Channelization 94.5% 98%

Some rev iew / further refinement to methodology recommended to ensure

fields align. Main Roads WA Methodology indicates measured between

adequate / poor and not applicable whilst the manual coding measure between

not present / present and not applicable.

This attribute relates to intersection type. There is a lot of complex code

surrounding how this attribute is determined and populated for the corporate

data set. An audit of the differences could be undertaken with a v iew of

establishing mistakes in either ARRB or Main Roads WA allocations and

refining the code if required.

Intersection

Type




Field /

category

% match

between

Corp and

Traditional

iRAP quality


Field linked to

for QA

Upgrade Costs 99.8% 95% Default Value of 2 (Medium) applied.

Note, not used in risk (including SRS) calculations.

Roadworks 94.1% 95% Default value of 1 (No Road Works) applied.

Serv ice Road 99.9% 95% Default value of 1 (Not Present) applied.

Traffic Calming

/ Speed

Management

100% 90%


Truck Speed

Limit 13.1% 98%

Expect this field to closely align with the results of speed limit. Review indicates

differences due to methodology applied for the corporate data matching posted

limits including in 110 km/hr zones (truck speed limits are 100 km/h). Adjusting

this would achieve 95.7% match.


erroneous results for this attribute. Main Roads WA has simply made this

attribute equal the ‘speed limit’ attribute. A simple adjustment to code would

rectify this.

Critical attributes

Current network-level crash risk assessment tools such as ANRAM or AusRAP require significant amount of data coding. In total, 72 attributes need to be coded for each 100 m in order to populate the input data file. Of the 72, 46 affect the risk scores with others being identifiers and non-critical data items.

This high number of attributes is driven by the tools’ treatment-focussed design, rather than crash risk estimation only. For example, the current approach requires that in order to recommend linemarking improvement as a safety treatment, information on its presence and quality needs to be collected first.

The number of attributes that has the greatest contribution to crash risk and crash estimation is fewer than the full set. Previous work undertaken by ARRB looked to identify attributes critical to estimation of risk, i.e. the key variables influencing risk scores and crash estimation in AusRAP and ANRAM. These factors are presented in Table 2.4.

A breakdown of the coding for each of the following attributes is also presented in Appendix C.1:

▪ AADT – same coding in both data sets.

▪ Mean speed – speed limit has been compared here instead. 95.7% of the coding is the same between the data sets. Where the coding is different, the corporate data set often has a higher speed limit. Main Roads WA can confirm there is a potential issue with the code producing erroneous results for this attribute. Main Roads WA has simply made this attribute equal to the ‘speed limit’ attribute. A simple adjustment to the code would rectify this. The same applies to ‘operating speed (85th percentile)’.




▪ Pavement (road) condition – 43% same coding. Traditional data set has much more coded as ‘1’ (good), while corporate data set has a large proportion coded ‘2’ or ‘3’ (medium or poor). This difference in coding is not necessarily wrong; however, it should be recognised that this will have an impact on results. For the corporate data set, this attribute is calculated based on three tiers of customer facing levels of service (LOS) relating to ride quality and the underlying levels of roughness using the International Roughness Index (IRI). There is likely to be significant differences between data extracted and a subjective visual interpretation.

▪ Grade – 99.9% same.

▪ Pavement width – a combination of lane width and shoulder width. Lane width is 97.9% the same; however, paved shoulder is 51% on drivers’ side and 59.8% on passengers’ side. For the shoulder measurements, where the coding is different, a large proportion is coded as narrower than the corporate data (large proportion of traditional (manual) data coded 3 (medium (≥ 1.0 m to < 2.4 m) when corporate data coded 2 (narrow (≥ 0 m to < 1.0 m)). The Main Roads WA team suspects the difference in coding results here may be partly due to the approach taken for extraction and assigning categories for the corporate data set. Main Roads WA can confirm there is a potential issue with the code producing erroneous results for this attribute.

▪ Curvature – 82.5% same.

▪ Roadside distance and severity (object) – quite a large difference in the coding of these attributes between the data sets.

▪ Intersection type – 92.6% same.

▪ Intersection volume – 92.2% same.

▪ Median type – 34.7% same. A large proportion of corporate data is coded 14 (wide centreline) when traditional (manual) data is coded 1 (centreline). It is also noted that a substantial proportion of corporate data is coded 11 (centreline) when traditional (manual) data is 1, 2, 3, 4, 5, 6,17 (barrier or physical median of different types / widths). The Main Roads WA team suspects the difference in coding results here may be due to the approach taken for extraction and assigning categories for the corporate data set.

Table 2.4: Critical attributes in crash risk estimation of different crash types

Attribute

Run-off-road

Passenger-side

Run-off-road

Driver-side Head-on Intersection

Property

Access

Mean speed* ✓ ✓ ✓ ✓ ✓

AADT ✓ ✓ ✓ ✓ ✓

Pavement condition** ✓ ✓ ✓ ✓

Grade ✓ ✓ ✓ ✓

Pavement width*** ✓ ✓ ✓

Curvature ✓ ✓ ✓

Roadside severity –

passenger-side distance

✓

Roadside severity –

passenger-side object

✓

Roadside severity – driver-side

distance

✓




Attribute

Run-off-road

Passenger-side

Run-off-road

Driver-side Head-on Intersection

Property

Access

Roadside severity – driver-side

object

✓

Intersection type

✓ ✓

Intersecting road volume

✓

Property access points

✓

Median type ✓ ✓

* Mean speed can be adopted from probe data, or assumed to be a derivative of the Speed limit AusRAP attribute.

** Pav ement condition is a combination of Road condition and Skid resistance attributes in AusRAP.

*** Pav ement width is a combination of Number of lanes, Lane width and Paved shoulder width attributes in AusRAP.

Source: ARRB internal document (2015).

2.3.2 AusRAP Results Comparison

Star rating results comparison

Table 2.5 and Figure 2.2 show the star rating results from iRAP’s ViDA platform for the two data sets. While the proportion of the network is similar for two-star roads, the proportions for most other star rating categories is quite different. Figure 2.2 shows that the AusRAP star rating results were sometimes different (between the traditional and corporate coding approaches) for the same sections of road (see also Table 2.6 and Table 2.7). This aspect was investigated further in the ‘SRS results comparison’ section.




Table 2.5: AusRAP star rating results

Traditional (TSD) coded data set Main Roads WA corporate data set

Smoothed

Star

Ratings

(smoothed

by length)

Raw Star

Ratings

Source: Tables created using ViDA software, ©iRAP (2018).




Figure 2.2: AusRAP star rating map

(a)

Traditional

(TSD) coded

data

(b) Main

Roads WA

corporate

data

Source: Maps created using ViDA software, ©iRAP (2018), map data © google.

A closer review of the AusRAP data output files (looking at the results for each 100 m segment of the data) for the traditionally coded vs corporate data sets found that 51% of the smoothed star rating results and 43% of the raw results were the same (Table 2.6).

Table 2.7 shows a further breakdown of the traditional and corporate data ratings (green shaded cells highlight same results). Where star ratings are different between the two methods, systematic differences in the star rating results do not show a particular trend for consistently higher or lower results.




Table 2.6: AusRAP star rating comparison

Smoothed star ratings Raw star ratings

Count % Count %

Same 3219 51% 2701 43%

±1 2706 43% 2585 41%

> 1 327 5% 966 15%

Total 6252

6252

Table 2.7: AusRAP star rating detailed comparison

Raw star results comparison


Traditional 1 2 3 4 5 Total

1 246 434 365 21 11 1077

2 280 903 944 46 38 2211

3 159 504 1456 117 126 2362

4 27 69 217 59 81 453

5 2 9 93 8 37 149

Total 714 1919 3075 251 293 6252

Smoothed star results comparison


Traditional 1 2 3 4 5 Grand Total

1

660 71 10

741

2 142 1878 1152 77

3249

3 80 529 1312 165 50 2136

4

39 57 27

123

5

1 2 3

Total 222 3106 2592 280 52 6252




SRS results comparison

Some star rating bands (see Table 2.5) do have similar proportions over the road network in this project (for example, the AusRAP Raw Star Ratings results show 35% of the network being 2-star based on the traditional approach compared to 31% for the corporate approach); however, these 2-star ratings have not necessarily been allocated to the same sections of road by the two approaches (see Figure 2.2, Table 2.6 and Table 2.7). In addition, different risk scores may be within the range covered by a single iRAP band.

In order to understand this better, the relationship between SRS results for each data set was reviewed (Figure 2.3). This helped to provide an insight into how well the risk scoring matched between the traditional and corporate approaches.

If the two approaches delivered identical results (i.e. same SRS – or very similar – for each segment), then:

▪ the blue line would be the same as the orange line (i.e. the equation of best fit (the blue line) through the data (the blue dots) would fit the 1:1 relationship (the orange line))

▪ the blue dots would tend to follow the blue line, rather than being dispersed away from the blue line (this is reflected by the R2 value).

The correlation for smoothed and raw SRS scores, as well as for each crash type reported in AusRAP was found to be low (see Figure 2.3):

▪ Most of the figures show that the equation of the line of best fit (blue line) is not close to a 1:1 relationship between the results produced by the two data set methods (i.e. for each segment, SRS corporate (plotted on the X-axis) should equal SRS traditional (plotted on the Y-axis) if the two data sets generate the same results.

▪ Most of the figures return a very low R2 value.

R2 is a statistical measure of how close the data are to the fitted regression line (in this case the blue dashed line in each graph). A high R2 value (> 0.85) indicates a strong relationship. A low R2 value (< 0.7) indicates a very low connection.

It was noted that:

▪ The equation of the line of best fit (blue line) was close to a 1:1 relationship for ‘vehicle SRS total smoothed’ (i.e. was close to the orange line). However, the low R2 value (i.e. –0.2) indicates a very poor fit between the data and the line of best fit (essentially, it is a random dispersion around the blue line).

▪ The R2 value for head-on overtaking SRS is high (0.88), which indicates that the scores (corporate and traditional) have little spread around their line of best fit. However, the relationship that this line of best fit indicates is a clear demonstration that the two methods produce different scores (if they were producing the same scores, the line of best fit would be a 1:1 relationship). The current corporate data extraction method produces scores that are related to, but appear to be a large multiple of, the traditional scores. Further investigation of this relationship could show a way that the corporate method may, with modification, produce results acceptably close to those produced using the AusRAP method for head-on overtaking SRS results only.




Figure 2.3: SRS correlation results




2.3.3 ANRAM Results Comparison

Table 2.8 provides a summary of the ANRAM results for the two data methods (traditional and corporate). The summary shows that the Predicted and ANRAM FSI results were quite different for the two data methods, and the proportion of risk attributed to each key crash type was also quite different. The traditional method attributed approximately 50% of the FSI crash risk to run-off-road crash types, compared to approximately 30% for the corporate data set.




Table 2.8: ANRAM results summary (traditional vs corporate)

Traditional (TSD) coded

data set

Main Roads WA

corporate data set

Observed FSI crashes Run-off-road 100 (48% ) 100 (48% )

Head-on 12 (6% ) 12 (6% )

Intersection 42 (20% ) 42 (20% )

Other 55 (26% ) 55 (26% )

Total (excl ped crashes) 209 (100% ) 209 (100% )

Predicted FSI crashes Run-off-road 72.91 (56% ) 51.69 (33% )

Head-on 7.99 (6% ) 16.58 (11% )

Intersection 19.98 (15% ) 22.85 (15% )

Other 30.15 (23% ) 65.64 (42% )

Total 131.03 (100% ) 156.76 (100% )

ANRAM FSI crashes Run-off-road 73.05 (53% ) 51.98 (34% )

Head-on 7.9 (6% ) 15.97 (10% )

Intersection 23.68 (17% ) 26.1 (17% )

Other 32.43 (24% ) 59.45 (39% )

Total 137.06 (100% ) 153.51 (100% )




3 FINDINGS AND DISCUSSION

This project sought to provide greater clarity on attributes that may be drawn from the Main Roads corporate data set. This section presents the project findings, including a discussion on the differences in coding and risk assessment results, key challenges as well as observations and lessons learnt, opportunities to deliver increased efficiency and value for money, and finally next steps the impact of technology developments.

3.1 Differences in Coding and Risk Assessment Results

An analysis of the differences between the data sets found similarities in the coding of results of some attributes, but substantial differences in many others. Section 2.3.1 shows that (during this project), the corporate data coding:

▪ was similar to the traditionally coded data set for five attributes (green), and that corporate data could be used

▪ should be suitable to use following some minor adjustment of the corporate extraction method for another five attributes (blue)

▪ may be suitable following a review, analysis and refinement of the corporate extraction method for another 12 attributes (yellow).

The Main Roads WA team noted that coding differences for attributes such as median type and, potentially, pavement (lane and shoulder width) may be due to the approach that was adopted for extraction and assigning corporate information to the attribute categories. Further investigation and refinement of corporate approach should provide closer alignment with the visual rating approach.

The Main Roads WA team noted that for the road condition attribute the traditional (visual rating) approach had no sections assigned as category 3 (poor) and only few category 2 (medium) results compared to the large proportion of category 2 and 3 for this attribute in the corporate data set. For the corporate data, Main Roads WA used a combination of road roughness and rutting in conjunction with standards to determine the classification for both road condition and skid resistance attributes. It would be useful to explore this issue further (for both the road condition and the skid resistance attribute). ARRB notes that for attributes such as these, corporate data may be more accurate than the traditional (visual) rating approach.

It is also noted that for the traditionally coded (manual) rating approach, a subjective visual interpretation is needed for some attributes which may have less reliability than measured data.

The comparison of AusRAP results found that while the proportion of the network is similar for two-star roads, the proportions for most other star rating categories is quite different. A closer review of the AusRAP data (results for each 100 m segment of the data) found that only 51% of the smoothed star rating results and 43% of the raw results were the same. Where star ratings were different, there did not seem to be an indication of a trend for consistently higher or lower results.

An investigation to understand the correlation of AusRAP risk scores for the same road segments found that the traditional and corporate approaches tended to have quite different results (i.e. for smoothed and raw SRS scores, as well as for each cash type reported in AusRAP, the correlation was found to be low). It was noted that for ‘head-on overtaking’ type crashes, the current corporate data extraction method produces scores that are related to, but appear to be a large multiple of, the traditional scores. Further investigation of this relationship could show a way that the corporate method may, with modification, produce results acceptably close to those produced using the AusRAP method for head-on overtaking results only.




The comparison of the ANRAM results found that the predicted and ANRAM FSI results, as well as the proportion of risk attributed to each key crash type, were quite different for the two data methods. The traditional method attributed approximately 50% of the FSI crash risk to run-off-road crash types, compared to approximately 30% for the corporate data set.

3.2 Key Challenges, Observations and Lessons Learnt

A number of key challenges, observations and lessons were noted during this project:

▪ Data alignment – Challenges exist in aligning data from different sources. For this project there were challenges in matching road distance versus using driven video data and the method as to how it was collected. There were some additional challenges for this project because of the recycling of trial TSD data. Because the TSD data was collected for trial purposes (focussed on demonstrating the TSD technology), distance was not aligned to the road distance to the same level of accuracy as would be for a normal project. This distance issue led to matching distances between corporate and TSD data for this project. This would not be an issue for a future data set which would employ ARRBs normal quality processes and procedures.

▪ Terminology – It was recognised that differences in terminology and understanding / familiarity with different data within ARRB and Main Roads WA led to miscommunication and errors early in the project (similar issues have been encountered with other jurisdictions). This may be resolved by ensuring liaison between staff who are intimately familiar with the data (from both ARRB and Main Roads WA) to circumvent communication issues.

▪ Data accuracy / currency

— Assumptions relating to the accuracy and / or currency of information in the corporate database, and / or interpretation of these attributes may trigger inaccurate error alerts in the manually coded data. Clear communications on how false positive alerts are managed through the quality assurance stage would be required. For example, if corporate speed limit data was used, the assumption would be that this data is accurate. However, if it were not consistently accurate, issues related to associated attributes may be falsely flagged (e.g. an incorrect 50 km/h speed limit in a rural area). For other attributes, the age (currency) of the data in the corporate database may need to be taken into consideration.

— For some attributes, corporate data may provide more accurate information than the AusRAP traditional (manual) rating approach. A key example is pavement quality / skid resistance, where the AusRAP visual rating approach may not provide accurate ratings, and where corporate data may be more accurate. Further investigation would be valuable to confirm whether the data extraction approach used for this project is appropriate, or whether further refinement may be required. This will also largely depend on how Main Roads WA considers the best way to measure road condition and skid resistance. At present a very simplistic approach is taken. However, it may be more beneficial to include other elements to provide a better gauge (more aligned to a visual assessment) of these attributes (especially road condition).

— For the traditionally coded (manual) rating approach, a subjective visual interpretation is needed for some attributes which result in less reliability than measured data. It should be remembered that bias or subjectivity may be evident in any data set, particularly in relation to data that involves human interpretation or input. This may be present in the corporate data set or traditionally coded data.




▪ Data interpretation

— For some attributes, during the interpretation and categorisation of attributes, Main Roads WA may have a higher (or potentially lower in some cases) stringency (i.e. when an attribute’s quality may be considered good or poor). This difference in interpretation may not be wrong (and in fact, the road agency’s interpretation may be considered better), but this would have an influence on coding results across the network (and may lead to differences in star ratings). This may introduce challenges (or at least issues to be aware of) if comparing risk score (AusRAP or ANRAM) results where data sets may utilise different approaches (particularly national comparisons).

— Future work should also take into account whether interpretation of the data is applicable in different road environments. This project focussed on primarily rural routes. The approach for classification of some attributes may be applicable across urban and rural road environments, while the approach may need to be refined for urban environments. Similarly, default values may be relatively confidently adopted for some attributes in rural environments (particularly relating to pedestrians – as pedestrians are generally not present on rural roads). For more urbanised assessments, this approach would need to be reviewed.

— For some attributes, Main Roads may elect to adopt a different approach to determining attribute classification to the approach outlined in the iRAP coding manual. A difference in approach may be appropriate (and preferred) for a jurisdiction; in such cases, iRAP requires this variation to be clearly documented (with an understanding in differences in coding outcomes).

▪ Asset condition – Use of corporate data will require consideration of new issues such as deterioration of assets over time (and whether this is currently (or will be) captured in corporate data, or whether a rule may need to be applied). For attributes that experience deterioration over time (e.g. pavement quality, linemarking etc.), consideration may need to be given to issues such as currency of corporate data, and whether that information is likely to reflect conditions. Drawing these attributes from the corporate database should take into consideration (1) the currency of the information in the corporate database for that attribute (e.g. linemarking installed on x date, or network survey undertaken on x date indicating quality level for that attribute), and (2) the likely deterioration of that asset over time (e.g. should the linemarking be assumed to be poor after x years). Main Roads WA may want to consider an approach for assuming conditions for these types of attributes (for example, if recent data on condition is not available in corporate database, perhaps an assumed level of deterioration over time may be adopted).

▪ Roadside hazards

— Some attributes would be extracted from the corporate database, for example roadside hazard type and distance. The iRAP manual requires the most dangerous hazard within each 100 m segment be identified, and then the distance noted. (The iRAP process nominates the order for the roadside hazard severity, for example cliff precedes tree ≥10 cm diameter) ARRB and Main Roads WA recognises that there is difficulty for assessing roadside hazard type and distance especially when taking into account roadside object distance. The Main Roads WA team noted the need to question and / or review the accuracy of their Road Side Hazard (RHR) rating data sets including the methodology used. The Main Roads WA team indicated that it may be valuable to undertake a detailed comparison of the AusRAP approach for roadside hazard object and distance (e.g. through visual audits on a sample of the network) and compare this with the approach applied to determine these attributes from the corporate data in this project. This might provide some insights as to improvements




that could be made to one or both approaches (traditional rating vs corporate database approach).

It is unknown exactly how close the Main Roads WA RHR is aligned to the worst-case principles how closely it corresponds to the iRAP attribute categorisations for the roadside object dataset.

Unless a rater is 100% diligent during the entire rating process, it is noted that there is potential for errors to be introduced at this point in the traditionally rated data set. A rater reviews each 100 m segment then decides which object is considered the worst case, based on guidance in the iRAP coding manual and relying on an intricate knowledge and memorisation of roadside object risk hierarchy order. Coding errors may occur (e.g. selection of the incorrect ‘worst’ roadside object’) due to repetitive nature of reviewing long stretches of road that have similar objects such as in WA with roads lined with trees. The QA process looks to identify and resolve such issues.

Newer technologies (e.g. use of LiDAR information) are being developed and research is continuing, which could assist in reducing the human element in identifying these related attributes.

— It is noted that, in a different project (ARRB internal communication 26 June 2018), ARRB undertook an analysis of 2600 crashes to ascertain the correlation between the object hit in a crash (as recorded in police reports) compared to the hazard identified in AusRAP coding. This analysis found that vehicles often did not hit the object identified in the AusRAP coding (e.g. for the particular data set reviewed, the percentage of crashes where object hit (in the crash data) matched the AusRAP coding was 52% for trees, 44% for ‘no object’, 26% for guardrail and 19% for non-frangible sign, post or pole). This raises the question whether roadside rating is an accurate predictor of hazards of concern.

3.3 Comment on Efficiencies and Value for Money

The project also looked to identify opportunities to deliver increased efficiency and value for money through using corporate data for multiple purposes (in this case, use of asset data to inform crash risk on the network):

▪ With a small amount of adjustment, there are 10 attributes (highlighted in green / blue, Table 2.2) that can easily be drawn from the Main Roads WA corporate database. ARRB anticipates that projected initial cost savings for the 10 attributes identified (at a network-level assessment) would be in the order of 5%.

▪ A further 12 attributes (highlighted yellow, Table 2.2), with some further analysis and refinement in the method of corporate data extraction, should be able to be drawn from the Main Roads WA corporate database. ARRB anticipates that cost savings associated with use of the 22 identified attributes (green, blue & yellow) would be in the order of 10%.

▪ These anticipated savings are relatively low due to (1) a large amount of QA effort still required, and (2) complexities in integrating and aligning the two data sets. However, further cost saving opportunities would arise over time as processes relating to the alignment and integration of two independent data sets matured (due to mature processes delivering efficiencies in alignment and integration). In addition, over time, as the number of attributes that may be drawn from corporate database increases (due to refinements in corporate extraction techniques, new information available in corporate databases and / or the influence of new technologies), savings may be compounded.




Main Roads WA noted that there were some attributes where default values were adopted for this project (due to time constraints and for ease of coding) that could be drawn from the corporate database in the future. These attributes include:

▪ land use – LHS

▪ land use - RHS

▪ street lighting

▪ facilities for bicycles

▪ pedestrian crossing facilities

▪ pedestrian crossing quality

▪ sidewalk provision – LHS

▪ sidewalk provision – RHS

▪ school zone warning

▪ pedestrian crossing on side road

▪ pedestrian fencing.

There are a number of key points to be aware of, which provide further insight into complexities relating to building a data set from two sources (manual and corporate):

▪ Potential cost reduction with a reduction in the number of fields being manually coded over time. Achieving cost reduction requires removing multiple associated fields. Consideration should be given to how categories are grouped (i.e. using corporate source for intersection type and intersecting volume but traditional (manual coding) for intersection channelization would still require coding in relation to intersections).

▪ It should be noted that some fields may require ongoing manual coding in the short-medium term to test and refine changes to corporate extraction methodology to achieve consistent results. This may need to be tested in different road environments, different road stereotypes, and different regions (to gain understanding in accuracy and repeatability in results in these different road configurations, and from different sections of the corporate database).

▪ It should be noted that some fields may require ongoing manual coding for the purpose of QA even where corporate data meets requirements (difficulties arise if associated fields are not drawn from the same data source. In such instances ARRB would propose parallel data coding (both corporate and manual coding). When both associated attributes can be confidently quantified from corporate data, manual coding of these would no longer be required (delivering cost dividends).

3.4 Next Steps and Future Technology

The ARRB team suggests that corporate data could be used for some, although currently not all, attributes for producing AusRAP / ANRAM data sets. With a bit more investigation and fine tuning, additional attributes could potentially be extracted from the corporate database and would cover most of the critical attributes relating to crash risk and crash estimation outlined in Table 2.4 (likely attributes include speed, grade, curvature, skid resistance, pavement condition, lane and shoulder width, intersection type and median type). It is noted that additional work would need to be undertaken to improve alignment and to ensure confidence in the corporate data; for instance, continuing to rate these attributes to facilitate ongoing comparison, analysis and refinement of the corporate extraction approach until there is confidence that the corporate data can be consistently




and accuratey extracted. While use of corporate data is anticipated to deliver relatively small savings over the short-term, further cost saving opportunities would arise over time. It is also noted that roadside hazard distance and object attributes (also critical attributes relating to crash risk and crash estimation) are currently difficult to systematically or automatically ascertain; however, it is likely to become possible as technology in this space continues to evolve.

New and evolving technologies will continue to lead to improvements in accuracy of coding of data attributes (in both traditionally coded and corporate data sets), as well as efficiencies in producing such data. Computer recognition, LiDAR technology and machine learning may be used to accurately train available data to predict the target variable. For instance, categorisation of roadside object type and distance measurement (as well as many other attributes such as lane width, delineation and access points) will eventually be able to replace the visual assessment for these attributes. Work is currently being undertaken in both the private and public sectors in Australia and overseas to progress this (for example, HERE’s work on mapping the road environment to assist automated vehicle navigation would be relevant and translatable for preparation of AusRAP and ANRAM data sets). ARRB’s @Lab team is currently undertaking preliminary work in this space with promising results. Next steps for this work would look to use larger data sets for training the machine learning algorithms, as well as undertaking testing to ensure all required categories can be reliably identified through this process (including in different road environments).

Main Roads WA requested ARRB to indicate a ballpark estimate for AusRAP and ANRAM coding for the remainder of the Main Roads network. Assuming a network carriageway length of up to 19 000 km, and less than 10% of the network being urban, the ballpark figure for AusRAP and ANRAM coding would be in the order of $1 million. Note, this does not include data collection (however, data is being collected as part of the iPave network data collection).




REFERENCES

ANRAM 2014, version 1.04, Australian National Risk Assessment Model software.

iRAP 2018, ViDA software, iRAP, London, UK.

iRAP 2014, iRAP star rating and investment coding manual, August 2014, International Road Assessment

Progamme (iRAP), London, UK.

Karpinski, J 2014, ‘Main Roads trial of Austroads National Risk Assessment Model (ANRAM)’, ARRB

conference, 26th, 2014, Sydney, NSW, ARRB Group, Vermont South, Vic, 15 pp.

AusRAP maps and tables throughout this report were created using VIDA software by iRAP. Copyright ©

iRAP. All rights reserved. For more information about VIDA software, please visit

<https://vida.irap.org/en-gb/home>.

Map data - Google Maps (2018), ‘Western Australia’, map data, Google, California, USA.




APPENDIX A DERIVING IRAP AND ANRAM ATTRIBUTES FROM IRIS

This appendix outlines the underlying approach employed by Main Roads WA for extracting the relevant information from the corporate data base.

Table A 1: Deriving IRAP and ANRAM attributes from IRIS

iRA

P A

ttri

bu

te N

o.

iRA

P C

olu

mn

Re

f

AN

RA

M C

olu

mn

Re

f

iRA

P C

od

ing

Ma

nu

al R

ef

iRAP Coding Attribute

IRIS

Datatype

Do

ma

in

Description

1 A N/A 4.1 Coder Name

CODER_NAME

Y VARCHAR(4) N ▪ Default to Main Roads WA.

▪ Subject to further consideration if v ideo coding is used.

2 B N/A 4.2 Coder date

CODING_DATE

Y DATE(DD/MM/

YYYY)

N ▪ Date at which Segment (data) was created or modified.

▪ Update Coder date on change to underly ing network or asset data (as specified within this

extract). Note; ex isting IRIS methodology will regenerate data for the entire road and not

incrementally (i.e. only Segments where changed has occurred).


3 C N/A 4.3 Road survey date

SURVEY_DATE

Y DATE(DD/MM/

YYYY)

N ▪ Equals Coder date.


4 D BG 4.4 Image reference

IMAGE_REF

Y VARCHAR(4) N ▪ Default to NULL.


5 E A 4.5 Road name

ROAD_NO

Y VARCHAR(4) N ▪ Main Roads road number.

▪ Note, this extract is limited to the Main Roads network only (i.e. where NE_NT_TYPE = ‘M’). It is

not applicable to the Local Road, Proposed or Principal Shared Path networks.

6 F B 4.6 Section

SECTION_NO

Y VARCHAR(30)

N ▪ The iRAP Coding Manual indicates Section could be a name to differentiate between ‘sections of

a road’, i.e. if the road agency has some internal naming or referencing method. Optionally , we

could use State Link Number/name or some another means to identify /group Segments within a

Section.




NUMBER

▪ Doesn’t have to be a unique number. Can also be words to differentiate segments of road (e.g.

Perth to Williams or Perth to Albany, Link 37 etc” .

▪ Should be relevant to the direction of travel. Useful for div ided carriageways.

▪ This attribute is used by ViDA to determine how it will handle the smoothing of data by referring to

the “Section” or by “Length” as defined in the project data set setup paramenters within ViDA.

▪ Different rules apply depending on the selection type.

▪ Unique number for each Segment within the extract, maintained on change to data (as per

Coding date), will not persist over time and will not be sequential (within a road or data set).

▪ Ref to attribute 79.

7 G D 4.7 Distance

SEGMENT_CHAINAGE

Y NUMBER(5) N ▪ Distance from start of road. Distance is a running chainage not road distance, it is consecutive

with no gaps. Distance Breaks that have a length > 0 must be taken into consideration so that

each Segment is continuous.

▪ Each road will start from 0.

▪ Start True (not SLK) of the Segment

▪ iRAP coding manual specifies that a Segment ‘should not be less than 0.1km’ however a

Segment may break at a change in carriageway (e.g. from A to U or B to U and v ice versa) hence

may be less than 0.1km and each segment will have a unique Section. The last Segment on a

road or at change in Carriageway may not equal 100m.

▪ Otherwise, if continuous 100m segments are required, then derive Carriageway based on first

occurrence.

8 H E 4.8 Length

SEGMENT_LENGTH

Y NUMBER(3,2) N ▪ Length of Segment in kilometres.

▪ Segment length (report interval) = 100m.

9 I F 4.9 Latitude

LATITUDE

Y NUMBER(8,6) N ▪ Latitude of Segment start in WGS84.

10 J G 4.9 Longitude

LONGITUDE

Y NUMBER(9,6) N ▪ Longitude of Segment start in WGS84.

11 K H 4.10 Landmark

LANDMARK

Y VARCHAR(100

)

N ▪ May allow locations to be referenced relative to landmarks or points of interest.

▪ Use precedence rules as below;

− If bridge is located within Segment, report Bridge Number and Description as where

Description is Bridge name otherwise Crossing name. Use single location of Bridge and not

Bridge Length.




− Otherwise, return intersecting road names, exclude name of road being processed. Is

Intersection Number of any value?

− If any Landmark is coincident with start or end of a Segment, report in first.

12 L AU 4.11 Comments

COMMENT

Y VARCHAR(100

)

N ▪ Default to NULL.

13 M C 4.12 Carriageway

CWAY

Y NUMBER(1) Y ▪ Main Roads left carriageway = 1 (Carriageway A) , Main Roads right carriageway = 2

(Carriageway B) Otherwise = 3 (Carriageway U or undiv ided – Main Roads single carriageway).

▪ Code 4 and 5 are not used.

▪ Refer to Distance. As the output Segments are 100m in length it is foreseeable than a Segment

will span a change in Carriageway. iRAP specifies that the first occurrence of an attribute value

should be used. Use the first occurrence of Carriageway in network connectiv ity order when

processing Left and Single. For Right, further discussion required.

▪ Recommend moving away from iRAP specification and splitting all Segments at change in

Carriageway, then determine if unique Section is required for each part that will make up the

100m Segment.

Code Category Main Roads Carriageway

1 Carriageway A of a div ided carriageway road Left

2 Carriageway B of a div ided carriageway road Right

3 Undiv ided road Single

4 Carriageway A of a motorcycle facility

5 Carriageway B of a motorcycle facility

14 N AT 4.13 Upgrade cost

UPGRADE_COST

N NUMBER(1) Y ▪ Records the influence that the surrounding land-use, env ironment and topography will have on the

cost of major works.

▪ Default value is 2 (ARAM User Guide).

Code Category

1 Low

2 Medium

3 High

15 O N/A 4.14 Motorcycle observed flow

MOTORCYCLE_FLOW

N NUMBER(1) Y ▪ Records the number of motorcycles in use within a 100m Segment.

▪ Default to 1 (Main Roads WA Not currently considering, unless able to measure).




Code Category

1 None

2 1 motorcycle observed

3 2 to 3 motorcycles observed



6 8+ motorcycles observed

16 P K 4.15 Bicycle observed flow

Bike flow

BICYCLE_FLOW

N NUMBER(1) Y ▪ ANRAM Attribute name;

− Bike flow.

▪ Records the number of bicyclists observed within a 100m Segment.


▪ Some Metropolitan roads do not permit cyclists.

▪ Future consideration.

Code Category

1 None

2 1 bicycle observed

3 2 to 3 bicycles observed



6 8+ bicycles observed

17 Q L 4.16 Pedestrian observed flow across the road

Pedestrian flow across the road

PEDESTRIAN_FLOW

N NUMBER(1) Y ▪ Records the number of pedestrians crossing or about to cross the road within a 100m Segment,

acknowledged as a random sample.

▪ Coding manual acknowledges this is a random sample.

▪ Some sections of Metropolitan Highways prohibit pedestrians.



Code Category

1 None




2 1 pedestrian crossing observed

3 2 to 3 pedestrians crossing observed



6 8+ pedestrians crossing observed

18 R M 4.17 Pedestrian observed flow along the road

driver-side

Pedestrian flow along the road driver-side

PEDESTRIAN_FLOW_DRIVER

N NUMBER(1) Y ▪ Records the number of pedestrians walking along the right side of the road within a 100m

Segment, acknowledged as a random sample.




Code Category

1 None

2 1 pedestrian along driver-side observed

3 2 to 3 pedestrians along driver-side observed



6 8+ pedestrians along driver-side observed

19 S BO 4.18 Pedestrian observed flow along the road

passenger-side

PEDESTRIAN_FLOW_PASSENGER

N NUMBER(1) Y ▪ Records the number of pedestrians walking along the left side of the road within a 100m Segment,

acknowledged as a random sample.




Code Category

1 None

2 1 pedestrian along passenger-side observed

3 2 to 3 pedestrians along passenger-side

observed





observed


observed

6 8+ pedestrians along passenger-side observed

20 T AE 4.20 Land use – driver side

LANDUSE_DRIVER

N NUMBER(1) Y ▪ Records the type of roadside development that is observed on the right side of the road.

▪ Default to 5.

Code Category

1 Undeveloped areas

2 Farming and agricultural

3 Residential

4 Commercial

5 Not Recorded

6 Educational

7 Industrial and manufacturing

21 U AD 4.19 Land use – passenger side

LANDUSE_PASSENGER

N NUMBER(1) Y ▪ Records the type of roadside development that is observed on the right side of the road.

▪ Default to 5.

Code Category

1 Undeveloped areas

2 Farming and agricultural

3 Residential

4 Commercial

5 Not Recorded

6 Educational

7 Industrial and manufacturing

22 V N 4.21 Area type

AREA_TYPE

N NUMBER(1) Y ▪ Defines the level of roadside development through which the road is passing

▪ Could be based on speed limit. Option might be to coarsely define Urban/Rural using speed limit

held in IRIS then rev iew with intention to store in IRIS.




▪ Town site boundary data set rev iewed but not considered appropriate.

Code Category

1 Rural / open area

2 Urban / rural town or v illage

23 W P 4.22 Speed limit

SPEED_LIMIT

Y NUMBER(2) Y ▪ Posted numerical speed limit.

▪ Return max (SPLI.IIT_SPEED_LIMIT) for Segment.

Code Category

1 <30km/h

2 35km/h

3 40km/h

4 45km/h

5 50km/h

6 55km/h

7 60km/h

8 65km/h

9 70km/h

10 75km/h

11 80km/h

12 85km/h

13 90km/h

14 95km/h

15 100km/h

16 105km/h

17 110km/h

18 115km/h

19 120km/h

20 125km/h




21 130km/h

22 135km/h

23 140km/h

24 145km/h

25 >=150km/h

24 X BE 4.23 Motorcycle speed limit

SPEED_LIMIT_MOTORCYCLE

Y NUMBER(2) Y ▪ Posted numerical speed limit for motor cycles.

▪ Equals Speed limit.

▪ Same List of Values as Speed limit.

25 Y BF 4.24 Truck speed limit

SPEED_LIMIT_TRUCK

Y NUMBER(2) Y ▪ Posted numerical speed limit for trucks.

▪ Same List of Values as Speed limit.

▪ If Speed limit = 110 km/h then set to 100km/h, otherwise equals Speed limit.

▪ Or alternatively ;

− if a Speed Limit Condition ex ists on any RAV Network (as stored in IRIS), return this value.

Where more than one condition is present, return the lowest speed

− otherwise, as above.

26 Z BB 4.25 Differential speed limits

SPEED_LIMIT_DIFFERENTIAL

Y NUMBER(1) Y ▪ Records the difference in either the operating speed or speed limit between cars and trucks when

it exceeds 20 km/h.

▪ If Speed limit – Truck speed limit > 20 then Differential speed limit is Present (2).

▪ Otherwise Not present (1).

Code Category

1 Not present

2 Present

27 AA AS 4.26 Median Type

MEDIAN_TYPE

Y NUMBER(1) Y ▪ Records road infrastructure feature that separates the two opposing traffic flows.

▪ Medians, barriers and line marking held in IRIS, just need to rev iew and agree on business logic to

map data to codes.

▪ Return first occurrence within Segment;

− If road is a Ramp or Rotary then return One way (13). Note; assuming a Left or Right

carriageway is not considered one way.

− If MEDI.IIT_MEDIAN_TYPE is Kerbed or Raised or Free Draining or Depressed then Median

Type code is 3, 4, 5, 6 or 7 as determined by MEDI.IIT_WIDTH.




− If MEDI.IIT_MEDIAN_TYPE is Painted or Other Level Treatment and MEDI. IIT_WIDTH >

1.0m then code is Central hatching (10).

− If LINE.LINE_MARKING is not Edge Line then a Centreline is present (11).

− Cannot recall many instances where a physical div ider is used to separate opposing traffic

flows and none of the above ex ists? GEH back to back kerb?

Code Category

1 Safety barrier – metal

2 Safety barrier – concrete

3 Physical median width >= 20.0m

4 Physical median width >= 10.0m to < 20.0m



7 Physical median width >= 0m to < 1.0m

8 Continuous central turning lane

9 Flex ipost

10 Central hatching (>1m)

11 Centre line

12 Safety barrier – motorcycle friendly

13 One way

14 Wide centre line (0.3m to 1m)

15 Safety barrier – wire rope

28 AB AY 4.27 Centreline rumble strips

CENTRELINE_RUMBLE

N NUMBER(1) Y ▪ Textured markings running along the centre of the road whose function is to warn drivers crossing

the median.

▪ Default to 1 (JK)?

Code Category

1 Not present

2 Present

29 AC AN 4.30 Roadside severity – driver-side distance N NUMBER(1) Y ▪ Distance to nearest object to the edge line likely to be reached which could result in serious or

fatal injury to road users.




SEVERITY_DISTANCE_DRIVER ▪ Objects were determined and stored in IRIS (Road Hazard Rating – RHRA) as part of the last

rating exercise. The object, which side of road it is located, and offset were recorded.

▪ The suitability of this data and its currency needs to be assessed in the context of significant

network changes or maintenance.

▪ The below commentary in D11#120567 summarises the broad findings from the prev ious rating

exercise.

Offset Comment

0 to <1 m

Within the RHR data there are few objects within this range other than

barriers.

1 to <5 m

The majority of high severity hazards on the Main Roads WA network are

within this range.

5 to <10 m Approx imately 25% (by length) of RHR data has hazards within this offset.

>=10 m

As part of the RHR typically hazards were only recorded to 10 m due to

limitation of accuracy of measured.

Code Category

1 0 to <1m

2 1 to <5m

3 5 to <10m

4 >= 10m

30 AD AO 4.30 Roadside severity – driver-side object

SEVERITY_OBJECT_DRIVER

N NUMBER(2) Y ▪ Nearest object likely to be reached which could result in serious or fatal injury to road user.

▪ Similar comments as per Roadside severity – driver-side distance, the business needs to

consider what is the cost/benefit of utilising the prev ious rating data set, undertaking a gap filling

exercise or resurvey.

Code Category

1 Safety barrier – metal

2 Safety barrier – concrete

3 Safety barrier – motorcycle friendly

4 Safety barrier – wire rope

5 Aggressive vertical face




6 Upwards slope – rollover gradient

7 Upwards slope – no rollover gradient

8 Deep drainage ditch

9 Downwards slope

10 Cliff

11 Tree >=10cm dia.

12 Sign, post or pole >= 10cm dia.

13 Rigid structure/bridge or building

14 Semi-rigid structure or building

15 Unprotected safety barrier end

16 Large boulders >=20cm high

17 None

31 AE AL 4.28 Roadside severity – passenger-side distance

SEVERITY_DISTANCE_PASSENGER

N NUMBER(1) Y ▪ See Roadside severity – driver-side distance.

32 AF AM 4.29 Roadside severity – passenger-side object

SEVERITY_OBJECT_PASSENGER

N NUMBER(2) Y ▪ See Roadside severity – driver-side object.

33 AG V 4.31 Shoulder rumble strips

SHOULDER_RUMBLE

Y NUMBER(1) Y ▪ Textured markings running along a road whose function is to warn drivers leav ing the travelled

way on the left side of the roadway.

▪ For each Segment, if inventory type LINE. IIT_EDGELINE_PRESENT = 1 ex ists (Audio tactile

edge line), then Shoulder rumble strip is Present (2), otherwise 1.

▪ Note, the edgeline present inventory does not need to ex ist along the entire Segment (it can be

partial and return the Code 2).

Code Category

1 Not present

2 Present

34 AH T 4.33 Paved shoulder – drivers’ side

SHOULDER_DRIVER

Y NUMBER(1) Y ▪ Refers to the safe and drivable section of the road to the side of the edge line on the right-hand

side of the carriageway.

▪ Measured from the centre of the shoulder marking (edge line) to the edge of the paving.




▪ If no edge line, then no paved shoulder ex ists. Refer iRAP Coding Manual 4.32.

▪ Return value of PASH. IIT_WIDTH and assign Code as per below table.

▪ For Carriageway A or Left (1) use XSP = R for sealed shoulder.

▪ For Carriageway B or Right (2) use XSP = L for sealed shoulder.

▪ For Carriageway U or Single use XSP = R for sealed shoulder.

▪ Note; ‘safe drivable section’ has been interpreted as only the sealed portion of the shoulder

measured from the centre of the edgeline to the edge of the paving. If an unsealed shoulder is to

be included then the above logic will alter. Similarly , are we prepared to accept that a sealed

shoulder will predominantly have an edge line or do we incorporate the IRIS line marking

inventory?

Code Category

1 Wide (>= 2.4m)

2 Medium (>= 1.0m to < 2.4m)

3 narrow (>= 0m to < 1.0m)

4 None

35 AI V 4.32 Paved shoulder left – passenger side

SHOULDER_PASSENGER

Y NUMBER(1) Y ▪ Refers to the safe and drivable section of the road to the side of the edge line on the left-hand side

of the carriageway.

▪ Measured from the centre of the shoulder marking (edge line) to the edge of the paving.

▪ If no edge line, then no paved shoulder ex ists. Refer iRAP Coding Manual 4.32.

▪ Refer Paved shoulder right – drivers’ side.

▪ For Carriageway A use XSP = L for sealed shoulder.

▪ For Carriageway B use XSP = R for sealed shoulder.

▪ For Carriageway U use XSP = L for sealed shoulder.

36 AJ AP 4.34 Intersection type

INTERSECTION_TYPE

Y NUMBER(2) Y ▪ Records presence and type of intersection.

▪ Subject to determination of business rules, high confidence that this could be derived from IRIS

using a mix of intersection data, where ramps meet through roads, gaps in medians, rail crossings

and traffic sign data.

Code Category

1 Merge lane

2 Roundabout




3 3–leg (unsignalised) with protected turn lane

4 3–leg (unsignalised) with no protected turn lane

5 3–leg (signalised) with protected turn lane

6 3–leg (signalised) with no protected turn lane

7 4–leg (unsignalised) with protected turn lane

8 4–leg (unsignalised) with no protected turn lane

9 4–leg (signalised) with protected turn lane

10 4–leg (signalised) with no protected turn lane

11 Do not use this code

12 None

13 Railway Crossing – passive (signs only)

14 Railway Crossing – active (flashing lights / boom gates)

15 Median crossing point – informal

16 Median crossing point – formal

17 Mini roundabout

37 AK BN 4.35 Intersection channelization

INTERSECTION_CHANNELISATION

▪ Raised or coloured islands that designate intended vehicle path.

▪ Turn pockets or slip lanes.

Code Category

1 Adequate

2 Poor

3 Not applicable

38 AL AR 4.37 Intersection quality

INTERSECTION_QUALITY

N ▪ Could use ex istence of signage from IRIS.

39 AM AQ 4.36 Intersecting road volume

INTERSECTING_ROAD_VOLUME

Y ▪ Could use mix of Main Roads and Local Government traffic data where available.

▪ For many rural intersection values will be low, perhaps use road hierarchy to estimate vehicles per

day, maybe consider Special Use Category or whether intersecting road is sealed (surface type)?




40 AN AK 4.38 Property access points

Access points

ACCESS_POINTS

▪ ANRAM attribute name;

− Access points.

▪ IRIS stores Control of Access.

▪ For vast majority of rural network intersecting roads could be used. Note; access to mine sites

have recently been added to support HVS needs.

▪ Roadside stopping paces to be considered.

▪ Investigating use of property address database or whether Landgate hold access information in

any of the topographical data sets.

41 AO O 4.39 Number of lanes

NUMBER_OF_LANES

Y ▪ Should record the predominant character of the road and changes over short lengths of road (less

than 400m) should not be recorded.

▪ Based on single direction of travel or both. Exclude overtaking lanes less than 400m in length and

auxiliary lanes such as turn pockets or dedicated bus lanes etc. Assume gravel road has 1 lane in

either direction?

▪ Count ex istence of L1 to L9 or R1 to R9 or LO and RO respectively , return value based on below

table.

Code Carriageway Category

1 A or B or U One

2 A or B or U Two

3 A or B or U Three

4 A or B or U Four or more

5 U Two and one

6 U Three and two

42 AP S 4.40 Lane width

LANE_WIDTH

Y ▪ Get the minimum lane width within each Segment for lanes as defined in Number of lanes,

assign value as per below table.

▪ Methodology for gravel roads to be considered?

Code Category

1 Wide (>= 3.25m)

2 Medium (>= 2.75m to < 3.25m)

3 narrow (>= 0m to < 2.75m)

43 AQ W 4.41 Curvature Y ▪ Records horizontal alignment of road.




CURVATURE ▪ If road is a roundabout, then set value to 1 (‘straight or gently curv ing’).

▪ Get smallest curve radius within section and assign value as;

Code Category Curve radius (r)

1 Straight or gently curv ing r >900m

2 Moderate 499m < r 901m

3 Sharp 199m < r <500m

4 Very sharp r < 200m

44 AR X 4.42 Quality of curve

QUALITY_OF_CURVE

Y ▪ How easy it is to judge how sharp a curve is and if it can be driven safely .

▪ Some historic logic is used based on radius and presence (or lack of) of warming signs, needs

clarification.

Code Category

1 Adequate

2 Poor

3 Not applicable

45 AJ Z 4.43 Grade

GRADE

Y ▪ Gradient of road.

▪ Will obtain from new inertial data.

Code Category

1 >= 0% to <7.5%

2 Not applicable

3 Not applicable

4 >= 7.5% to <10%

5 >= 10%

46 AT AA 4.44 Road condition

ROAD_CONDITION

Road surface condition

Y ▪ ANRAM attribute name;

− Road surface condition.

▪ Ability of road to prov ide a level, even running surface, free from major defects.

▪ Ignore medium (historic approach) unless new methodology is proposed.

▪ If roughness OR rutting below Main Roads WA intervention level by link category. Roughness:

M>3.44, A>3.82, B>4.2, C>5.33, D>5.33. OR Rutting: M/A>20, B/C/D>30 then set value to 3,

otherwise 1.




Code Category

1 Good

2 Medium

3 Poor

47 AV AW 4.45 Skid resistance / grip

SKID_RESISTANCE-GRIP

Y ▪ Could use mix of Textured data (macrotexture) and seal type.

▪ Use requirements outlined within the “Guidelines for Managing Surface Friction on.

▪ Main Roads in WA” Extract included below;

High Speed Roads and Roads Surfaced with a Sprayed Seal

This section is applicable to roads with a posted speed limit of 90 km/hr or greater and all roads

surfaced with a sprayed seal or microsurfacing (slurry seal) regardless of the posted speed limit.

Investigatory levels for texture are shown in the table below based on sand patch measured textured

depth (or a SPTD equivalent from laser measured road data). Sections of road below the investigatory

level should be assessed using a risk based process. The Road Surface Inspection Sheet included at

Appendix A may assist with this process. If there is doubt about which level applies to a section of road

the default limit should be 1.0mm.

Site Description Investigatory level

Manoeuvre-free and relatively flat or low vertical gradient 0.8mm

ALL other areas including:

Traffic light controlled intersections

Roundabout approaches

Curves with tight radius ≤ 250 m

Gradients 5% and 50 m long

1.0mm

Roads Surfaced with Asphalt and with a Posted Speed Limit less than 90 km/hr

Where a road is surfaced with asphalt and the posted speed limit is less than 90 km/hr the approach

for maintaining surface friction will be to use a skid resistance approach using the SCRIM or BPT.

Investigatory levels for different site descriptions of a road are shown in the following table. Sections of

road below the investigatory level should be assessed using a risk based process. The Road Surface

Inspection Sheet included at Appendix A may assist with this process.

Site Description Investigatory level of

SRV at 20ºC (SCRIM)

Investigatory level

of SRV at 20ºC (BPT)




Manoeuvre-free and relatively flat or low

vertical gradient 0.40 40

ALL other areas including:

Traffic light controlled intersections

Roundabout approaches

Curves with tight radius ≤ 250 m

Gradients 5% and 50 m long

0.45 45

Code Category

5 Unsealed poor

4 Unsealed adequate

3 Sealed poor

2 Sealed medium

1 Sealed adequate

48 AV Y 4.46 Delineation

DELINIATION

Y ▪ Records the road attributes which informs drivers of road conditions to keep them within the driven

lane and aware of the road ahead. It is based on a combination of factors including edge lines,

guideposts and signage.

▪ Consideration for gravel roads where signage and post may ex ist but no edge lines.

▪ If line marking types 3 or 7 (separation + edge or separation + lane + edge) ex ist within Segment

the set value to 1, otherwise 2.

Code Category

1 Adequate

2 Poor

49 AW AX 4.47 Street Lighting

STREET_LIGHTING

N Code Category

1 Not present

2 Present

50 AX AG 4.48 Pedestrian crossing facilities

PEDESTRIAN_CROSSING_FACILITIES

N ▪ Could use presence of signals at intersection, or ex istence of sign in median or presence of

median or pedestrian bridge over road etc.

Code Category




1 Grade separated facility

2 Signalised with refuge

3 Signalised without refuge

4 Unsignalised marked crossing with refuge

5 Unsignalised marked crossing without a refuge

6 Refuge only

7 No facility

14 Unsignalised raised marked crossing with refuge

15 Unsignalised raised marked crossing without refuge

16 Raised unmarked crossing with refuge

17 Raised unmarked crossing without refuge

51 AY AH 4.49 Pedestrian crossing quality

PEDESTRIAN_CROSSING_QUALITY

N ▪ Three primary factors to be considered:

− signing

− marking or raised crossing

− good sight distance.

Code Category

2 Poor

1 Adequate

3 Not applicable

52 AZ BM 4.50 Pedestrian crossing facilities – side road

PEDESTRIAN_CROSSING_FACILITIES-

SIDE_ROAD

N ▪ Presence of purpose built pedestrian crossing facilities.

▪ When crossing at a signalised intersection the pedestrian crossing should only be considered

signalised if the signals have a pedestrian phase.

Code Category

1 Grade separated facility

2 Signalised with refuge

3 Signalised without refuge

4 Unsignalised marked crossing with refuge

5 Unsignalised marked crossing without a refuge




6 Refuge only

7 No facility

14 Unsignalised raised marked crossing with refuge

15 Unsignalised raised marked crossing without refuge

16 Raised unmarked crossing with refuge

17 Raised unmarked crossing without refuge

53 BA BL 4.51 Pedestrian fencing

PEDESTRIAN_FENCING

Y ▪ Fence present to restrict pedestrian crossing flow.

▪ Only required one side of the road.

▪ Wall and fence data stored in IRIS, not certain if applicable or of value?

Code Category

1 Not present

2 Present

54 BB BA 4.54 Speed management / traffic calming

SPEED_MANAGEMENT-TRAFFIC_CALMING

N ▪ Presence of road infrastructure features that will typically reduce the operating speed by 5 to

10lm/h below the speed limit.

▪ Possibly derived from sign data held in IRIS, requires business logic to derive.

Code Category

1 Not present

2 Present

55 BC N/A 4.55 Vehicle parking

VEHICLE_PARKING

▪ Extent to which vehicle parking along the side of road is present.

▪ Parking spaces. Bus stops and general encroachment on the road within 2m of the outside edge

of the drivable lane.

Code Category

3 Two sides

2 One side

1 None

56 BD AC 4.57 Sidewalk – drivers’ side

SIDEWALK–DRIVERS_SIDE

N Code Category

4 Non-physical separation 0m to <1.0m




3 Non-physical separation 1.0m to <3.0m

2 Non-physical separation ≥ 3.0m

1 Physical barrier

7 Informal path 0m to <1.0m

6 Informal path ≥ 1.0m

5 None

57 BE AB 4.56 Sidewalk – passenger side

SIDEWALK–PASSENGER_SIDE

N Code Category

4 Non-physical separation 0m to <1.0m

3 Non-physical separation 1.0m to <3.0m

2 Non-physical separation ≥ 3.0m

1 Physical barrier

7 Informal path 0m to <1.0m

6 Informal path ≥ 1.0m

5 None

58 BF AZ 4.58 Service road

SERVICE_ROAD

N ▪ Really? Suspect 99% of network will not have road adjacent to main road. Can recall some

instances on Forest Hwy.

Code Category

1 Not present

2 Present

59 BG AJ 4.59 Facilities for motorised two wheelers

FACILITIES_FOR_MOTORISED_TWO_WHEE

LERS

N ▪ Presence of purpose built facilities for motorcycles and other motorise two-wheelers.

Code Category

5 Inclusive motorcycle lane on roadway

2 Exclusive one way motorcycle path without barrier

1 Exclusive one way motorcycle path with barrier

4 Exclusive two way motorcycle path without barrier

3 Exclusive two way motorcycle path with barrier

6 None




60 BH AI 4.60 Bicycle facility

BICYCLE_FACILITY

N ▪ Could consider determining if width of slow lane (nearest to kerb) is ≥ 4.2m.

▪ Or if PSP is adjacent to road (PSP’s are stored in IRIS), predominantly Metropolitan area.

▪ Lanes dedicated for cyclists are not recorded in IRIS.

Code Category

5 Extra wide outside (≥4.2m)

3 On-road lane

2 Off-road path

1 Off-road path with barrier

6 Signed shared roadway

7 Shared use path

4 None

61 BI AV 4.61 Roadworks

ROADWORKS

Y ▪ Major road construction or road works in progress.

▪ Road side bookings are stored in IRIS but for Metropolitan area only .

Code Category

3 Major road works in progress

2 Minor road works in progress

1 No road works

62 BJ BC 4.62 Sight distance

SIGHT_DISTANCE

N ▪ Horizontal or vertical alignment or physical obstructions such as roadside objects and vegetation

that may reduce sight distance to less than 100m.

▪ Very simplistic rating.

▪ Don’t appreciate need to use geometry data (as was past practice).

Code Category

2 Poor

1 Adequate

63 BK I 5.1 Vehicle flow (AADT)

VOLUME

Y NUMBER(6) ▪ For what year? Need to formalise requirement and methodology to factor prev ious years based on

fixed growth rate (prev iously 2.5% ) or by associating network sections to growth rates derived from

Network Performance Sites (NPS).




▪ In addition, define methodology to fill gaps when data is not available (consider filling gap at traffic

section level).

▪ Presume directional volumes for A and B Carriageways and combined for U?

64 BL J 5.2 Motorcycle %

MOTORCYCLE_%

▪ Percentage of Vehicle flow classed as motorised 2 wheel or light 3 wheel vehicle

▪ Coding manual specifies category values based on class ranges.

▪ If less than 1% then Category 2, 1% ≤ Category 3 ≤ 5% , 5% ≤ Category 4 ≤ 10% ,

▪ Analyse NPS data to derive either single rural average or on a road by road basis. Highly unlikely

motorcycles comprise more than 10% of traffic.

Code Category

10 100%

9 81% – 99%

8 61% – 80%

7 41% – 60%

6 21% – 40%

5 11% – 20%

4 6% – 10%

3 1% – 5%

2 0%

1 Not recorded

65 BM BP 5.3 Pedestrian peak hour flow across the road

PEDESTRIAN_PEAK_HOUR_FLOW_ACROS

S_THE_ROAD

N ▪ If < 1 then Category 1, If ≥ 1 and ≤ 5 then Category 2, If ≥ 6 and ≤ 25 then Category 3, If ≥ 26

and ≤ 50 then Category 4, If ≥ 51 and ≤ 100 then Category 5, If ≥ 101 and ≤ 200 then Category

6, If ≥ 201 and ≤ 300 then Category 7, If ≥ 301 and ≤ 400 then Category 8, If ≥ 401 and ≤ 500

then Category 9, If ≥ 501 and ≤ 900 then Category 10, If > 900 then Category 11.

Code Category

11 900+

10 501 to 900

9 401 to 500

8 301 to 400

7 201 to 300




6 101 to 200

5 51 to 100

4 26 to 50

3 6 to 25

2 1 to 5

1 0

66 BN BQ 5.4 Pedestrian peak hour flow along the road

driver-side

PEDESTRIAN_PEAK_HOUR_FLOW_ALONG

_THE_ROAD_DRIVER-SIDE





Code Category

11 900+

10 501 to 900

9 401 to 500

8 301 to 400

7 201 to 300

6 101 to 200

5 51 to 100

4 26 to 50

3 6 to 25

2 1 to 5

1 0

67 BO BR 5.5 Pedestrian peak hour flow along the road

passenger-side

PEDESTRIAN_PEAK_HOUR_FLOW_ALONG

_THE_ROAD_ PASSENGER-SIDE





Code Category

11 900+




10 501 to 900

9 401 to 500

8 301 to 400

7 201 to 300

6 101 to 200

5 51 to 100

4 26 to 50

3 6 to 25

2 1 to 5

1 0

68 BP N/A 5.6 Bicycle peak hour flow

BICYCLE_PEAK_HOUR_FLOW

N ▪ Coding manual acknowledges this is a random sample.

▪ Some sections of Metropolitan Highways prohibit cyclists.

▪ Previous decision was to default value to 1 (1 bicycle observed per 100m)?

▪ If < 1 then Category 1, If ≥ 1 and ≤ 5 then Category 2, If ≥ 6 and ≤ 25 then Category 3, If ≥ 26




Code Category

11 900+

10 501 to 900

9 401 to 500

8 301 to 400

7 201 to 300

6 101 to 200

5 51 to 100

4 26 to 50

3 6 to 25

2 1 to 5




1 0

69 BQ Q 5.7 Operating speed (85th percentile)

OPERATING_SPEED_(85TH_ PERCENTILE)

Y ▪ Speed at which 85% of vehicles are travelling.

▪ Define methodology when not available, i.e. add 10kmph to speed limit.

▪ Note, if required for Metropolitan area then should peak periods be excluded from the calculation?

Code Category

1 <30km/h

2 35km/h

3 40km/h

4 45km/h

5 50km/h

6 55km/h

7 60km/h

8 65km/h

9 70km/h

10 75km/h

11 80km/h

12 85km/h

13 90km/h

14 95km/h

15 100km/h

16 105km/h

17 110km/h

18 115km/h

19 120km/h

20 125km/h

21 130km/h

22 135km/h




23 140km/h

24 145km/h

25 >=150km/h

31 <20mph

32 25mph

33 30mph

34 35mph

35 40mph

36 45mph

37 50mph

38 55mph

39 60mph

40 65mph

41 70mph

42 75mph

43 80mph

44 85mph

45 >=90mph

70 BR R 5.8 Operating speed (mean)

OPERATING_SPEED_(MEAN)

Y ▪ Average operating speed.

▪ Define methodology when not available.

▪ Note, if required for Metropolitan area then should peak periods be excluded from the calculation?

Code Category

1 <30km/h

2 35km/h

3 40km/h

4 45km/h

5 50km/h




6 55km/h

7 60km/h

8 65km/h

9 70km/h

10 75km/h

11 80km/h

12 85km/h

13 90km/h

14 95km/h

15 100km/h

16 105km/h

17 110km/h

18 115km/h

19 120km/h

20 125km/h

21 130km/h

22 135km/h

23 140km/h

24 145km/h

25 >=150km/h

31 <20mph

32 25mph

33 30mph

34 35mph

35 40mph

36 45mph




37 50mph

38 55mph

39 60mph

40 65mph

41 70mph

42 75mph

43 80mph

44 85mph

45 >=90mph

71 BS N/A 5.9 Roads that cars can read

ROADS_THAT_CARS_CAN_READ

▪ Meets specification for vehicles to be able to recognise the delineation (markings and signs).

▪ Default “Does not meet specification”.

Code Category

1 Meets specification

2 Does not meet specification

72 BT BH 5.10 Car star rating policy target

CAR_STAR_RATING_POLICY_TARGET

N ▪ Records the minimum policy Star Rating target.

▪ Currently not used in ViDA. However, will be included at a later date.

Code Category

1 1 Star

2 2 Star

3 3 Star

4 4 Star

5 5 Star

6 Not applicable

73 BU BI 5.10 Motorcycle star rating policy target

MOTORCYCLE_STAR_RATING_POLICY_TA

RGET



Code Category

1 1 Star




2 2 Star

3 3 Star

4 4 Star

5 5 Star

6 Not applicable

74 BV BJ 5.10 Pedestrian star rating policy target

PEDESTRIAN_STAR_RATING_POLICY_TAR

GET



Code Category

1 1 Star

2 2 Star

3 3 Star

4 4 Star

5 5 Star

6 Not applicable

75 BW BK 5.10 Bike Bicycle star rating policy target

BIKE_BICYCLE_STAR_RATING_POLICY_TA

RGET



Code Category

1 1 Star

2 2 Star

3 3 Star

4 4 Star

5 5 Star

6 Not applicable

76 BX N/A N/A Annual fatality growth multiplier

ANNUAL_FATALITY_GROWTH_MULTIPLIER

Number (1) ▪ Default to 1.

▪ As per iRAP Coding Manual.

77 BY N/A 4.52 School zone warning

SCHOOL_ZONE_WARNING

▪ Presence of flashing signs & speed limits.

▪ Static signs or road markings.




Code Category

1 School zone flashing beacons

2 School zone static signs or road markings

3 No school zone warning

4 Not applicable (no school at the location)

78 BZ N/A 4.53 School zone crossing supervisor

SCHOOL_ZONE_CROSSING_SUPERVISOR

▪ Presence of crossing attendant.

▪ Refer Police for data source.

Code Category

1 School zone crossing superv isor present at school start and finish times

2 School zone crossing superv isor not present

3 Not applicable (no school at the location)

79 CA B N/A Smoothed section ID

Section

SECTION

▪ ANRAM attribute name:

− Section.

▪ This field is an iRAP attribute generated after upload and processing by ViDA. The iRAP data

produced from ViDA is then subsequently used within the ANRAM file as the “Section” attribute at

column “B”. NOTE: Smoothing ca be variable depending on how the user chooses to smooth the

result in iRAP. It can be selected using the default assumptions, or it can be specified break point

along the road to manage changes over time to cater for specific long-term projects.

BS N/A Jurisdiction

JURISDICTION

Y ▪ Default to 5 (WA)?

BT N/A Road type

ROAD_TYPE

Y ▪ Subject to definition of business rules.

Source: This content produced by Main Roads WA team

Notes:

▪ Attribute Descriptiv e Text ATTRIBUTE_NAME (Blue = ANRAM Attribute Name (where diff to iRAP) Red = Mandatory or NOT NULL, Grey = Optional or NULL allowed)

▪ Segment – means the output 100m data




APPENDIX B EXAMPLE ROAD SEGMENT LENGTH APPROACH

The Main Roads WA proposed approach for addressing the varying segment length issue is illustrated in this appendix.

The segment length for both the traditional and corporate data sets was calculated using the trigonometric formula in Equation A1 below. The results are presented in Table B 1 for sample of segments. Table B 1 shows the calculated segment lengths were similar for both corporate and traditional approaches, and also shows the ‘Applied length’ (which corresponds with the calculated length rounded to one decimal place).

Trigonometric formula for calculating segment length

= = ACOS(COS(RADIANS(90-LatitudeStart))×COS(RADIANS(90-LatitudeEnd))+SIN(RADIANS(90-LatitudeStart))×SIN(RADIANS(90-LatitudeEnd))×COS(RADIANS(LongitudeStart-LongitudeEnd)))×6371

A1

Table B 1: Calculation and comparison of segment length traditional and corporate data sets

Image reference Road Name Section

Traditional (TSD) coded data set Main Roads WA corporate data set

Diff in Dist

Main Roads WA

Code Extract Length

Corporate Data set Length

Using Trig Formula

Traditional Data set Length

Using Trig Formula

Applied Length Latitude

Longitude

Distance Latitude Longitud

e Distance

Great Eastern Hwy Fr Llyod St Frame 97 Great Eastern Highway Perth-Coolgargie –31.8917 116.0134 16.165 –31.89184028 116.0134676 16.159 –0.006 0.088

Great Eastern Hwy Fr Llyod St Frame 107 Great Eastern Highway Perth-Coolgargie –31.89192164 116.0144094 16.265 –31.8920073 116.0144628 16.247 –0.018 0.093 0.096 0.098 0.1









Great Eastern Hwy Fr Llyod St Frame 197 Great Eastern Highway Perth-Coolgargie –31.89354048 116.0236667 17.165 –31.89374714 116.0248308 17.245 0.08 0.104 0.199 0.099 0.2




APPENDIX C CODING COMPARISON BETWEEN DATA

SETS

The tables in this appendix present a coding comparison between the traditionally coded (TSD) and corporate AusRAP data sets.

C.1 Critical Attributes

Coding comparison for the attributes was considered critical to the estimation of risk (as per Table 2.4). The green shaded cells indicate where the coding is the same.

Table C 1: Speed limit


Traditional 7 9 11 13 15 17

5 10

7 115 36 23 2

9 47 3 15

11 4 1 323 11 49 30

13 1 2 9 95 33

15 3 237 29

17 4 1 5169

Same: 95.7%, Different: 4.3%.

Table C 2: Road condition


Traditional 1 2 3

1 2668 2307 1172

2 14 57 34


Table C 3: Grade


Traditional 1 2

1 6246 5

4 1





Pavement width

Table C 4: Lane width

Main Roads WA

corporate

Traditional 1 2

1 5440 55

2 78 679


Table C 5: Paved shoulder drivers’ side

Main Roads WA

corporate

Traditional 1 2 3

1 13 5 2

2 1 778 37

3 8 2678 2400

4 9 39 282

Same: 51%, Different: 49%.

Table C 6: Paved shoulder passengers’ side

Main Roads WA

corporate

Traditional 1 2 3

1 834 55 32

2 24 728 72

3 32 2045 2174

4 13 30 213


Table C 7: Curvature


Traditional 1 2 3 4

1 5054 250 48 2

2 705 105 88


Table C 8: Roadside distance drivers’ side


Traditional 1 2 3 4

1 10 27 35

2 12 462 281 1471

3 2 191 458 1407

4 4 135 147 1610





Table C 9: Roadside object drivers’ side


Traditional 1 2 4 6 8 9 10 11 12 15 17

1 31 2 2 1 10

2 36 4 3 6 14

4 2 2 24 1 3 3 36

5 8

6 10 1 2 41 10 4 31 8 65

7 5 5 38 2 1 21 2 26

8 4 6 29 76 6 38 19 128

9 2

11 22 38 306 287 97 1213 386 4 1563

12 30 5 4 12 1 16 57 326

13 1 1 10

15 9 2 1 4

16 4

17 45 1 82 35 47 7 1 72 117 5 673


Table C 10: Roadside distance passengers’ side


Traditional 1 2 3 4

1 4 9 1 49

2 6 630 188 963

3 445 621 1338

4 159 128 1711





Table C 11: Roadside object passengers’ side


Traditional 1 2 4 6 8 9 10 11 12 15 17

1 65 3 2 15

2 5 3 4 4

4 3 71 1 10

5 2 18 2 3 2 28

6 2 1 46 4 5 22 17 1 163

7 5 2 33 4 2 11 7 139

8 4 6 27 65 5 39 20 177

9 1 3

11 43 1 56 223 163 85 1 811 152 7 2128

12 49 4 18 4 23 4 50 444 3 442

13 2 3 1 1 11

15 2 1 1 2 3

16 6

17 12 10 3 27 3 24 7 1 369


Table C 12: Intersection type


Traditional 1 3 4 5 7 8 9 12 14 16

1 3 5 6 21 1

3 28 1 1 1 34

4 1 28 60 1 104 1

5 1 1

6 1 1

7 6 3 6

8 1 7 7 17

9 1 1 1 7

10 1

12 19 58 93 2 6 7 5686 1 1

14 1 3 3

15 10

16 2 2





Table C 13: Intersecting road volume


Traditional 1 2 3 4 5 6 7

3 1 1 2 5

4 1 1 23 32 30 71

5 1 1 1 14 22 71

6 1 42 59

7 5 7 28 34 113 5686


Table C 14: Median type


Traditional 1 2 3 4 5 6 10 11 14 15

1 5 1 8 4

2 12 4 43 4

3 1 99 4 4 310 13

4 1 47 43 76 1 632 11

5 15 56 8 129 1

6 1 14 32 1 68 2

8 2 3

10 52

11 1 1879 2603

15 1 2 17 42


C.2 Attributes where Coding is the Same

The following attributes are coded the same in both data sets (Same: 100%, Different: 0%):

▪ pedestrian fencing

▪ speed management / traffic calming

▪ motorcycle facility

▪ car star rating policy target

▪ motorcycle star rating policy target

▪ pedestrian star rating policy target

▪ bicycle star rating policy target

▪ annual fatality growth multiplier

▪ school zone warning

▪ school zone crossing supervisor

▪ sight distance

▪ AADT (same AADT data – provided by Main Roads WA – used in both data sets).




C.3 Attributes where Coding is Different

Table C 15: Divided / undivided


Traditional 1 2 3

1 690 8

2 1019

3 3 4532


Table C 16: Area type


Traditional 1 2

1 5143 737

2 131 241


Table C 17: Speed limit



5 10

7 115 36 23 2

9 47 3 15

11 4 1 323 11 49 30

13 1 2 9 95 33

15 3 237 29

17 4 1 5169


Table C 18: Truck speed limit



5 10

7 115 36 23 2

9 47 3 15

11 4 1 323 11 49 30

13 1 2 9 95 33

15 7 1 237 5198





Table C 19: Median type


Traditional 1 2 3 4 5 6 10 11 14 15

1 5 1 8 4

2 12 4 43 4

3 1 99 4 4 310 13

4 1 47 43 76 1 632 11

5 15 56 8 129 1

6 1 14 32 1 68 2

8 2 3

10 52

11 1 1879 2603

15 1 2 17 42


Table C 20: Centreline rumble strips


Traditional 1 2

1 3597 616

2 2039


Table C 21: Roadside distance drivers’ side


Traditional 1 2 3 4

1 10 27 35

2 12 462 281 1471

3 2 191 458 1407

4 4 135 147 1610





Table C 22: Roadside object drivers’ side


Traditional 1 2 4 6 8 9 10 11 12 15 17

1 31 2 2 1 10

2 36 4 3 6 14

4 2 2 24 1 3 3 36

5 8

6 10 1 2 41 10 4 31 8 65

7 5 5 38 2 1 21 2 26

8 4 6 29 76 6 38 19 128

9 2

11 22 38 306 287 97 1213 386 4 1563

12 30 5 4 12 1 16 57 326

13 1 1 10

15 9 2 1 4

16 4

17 45 1 82 35 47 7 1 72 117 5 673


Table C 23: Roadside distance passengers’ side


Traditional 1 2 3 4

1 4 9 1 49

2 6 630 188 963

3 445 621 1338

4 159 128 1711





Table C 24: Roadside object passengers’ side


Traditional 1 2 4 6 8 9 10 11 12 15 17

1 65 3 2 15

2 5 3 4 4

4 3 71 1 10

5 2 18 2 3 2 28

6 2 1 46 4 5 22 17 1 163

7 5 2 33 4 2 11 7 139

8 4 6 27 65 5 39 20 177

9 1 3

11 43 1 56 223 163 85 1 811 152 7 2128

12 49 4 18 4 23 4 50 444 3 442

13 2 3 1 1 11

15 2 1 1 2 3

16 6

17 12 10 3 27 3 24 7 1 369


Table C 25: Shoulder rumble strips


Traditional 1 2

1 3488 245

2 1602 917


Table C 26: Paved shoulder drivers’ side


Traditional 1 2 3

1 13 5 2

2 1 778 37

3 8 2678 2400

4 9 39 282





Table C 27: Paved shoulder passengers’ side


Traditional 1 2 3

1 834 55 32

2 24 728 72

3 32 2045 2174

4 13 30 213


Table C 28: Intersection type


Traditional 1 3 4 5 7 8 9 12 14 16

1 3 5 6 21 1

3 28 1 1 1 34

4 1 28 60 1 104 1

5 1 1

6 1 1

7 6 3 6

8 1 7 7 17

9 1 1 1 7

10 1

12 19 58 93 2 6 7 5686 1 1

14 1 3 3

15 10

16 2 2


Table C 29: Intersection channelization


Traditional 1 2

1 5865 281

2 60 46





Table C 30: Intersecting road volume


Traditional 1 2 3 4 5 6 7

3 1 1 2 5

4 1 1 23 32 30 71

5 1 1 1 14 22 71

6 1 42 59

7 5 7 28 34 113 5686


Table C 31: Intersection quality


Traditional 1 2 3

1 95 78 206

3 92 95 5686


Table C 32: Property access


Traditional 1 2 4

1 48 385

2 34

3 9 198

4 4 58 5516


Table C 33: Number of lanes


Traditional 1 2 3 4

1 7 4349 25

2 1 1599 84 14

3 17

5 11 145


Table C 34: Lane width

Main Roads WA

corporate

Traditional 1 2

1 5440 55

2 78 679





Table C 35: Curvature


Traditional 1 2 3 4

1 5054 250 48 2

2 705 105 88


Table C 36: Curve quality


Traditional 1 2 3

1 594 14 290

3 811 18 4525


Table C 37: Grade


Traditional 1 2

1 6246 5

4 1


Table C 38: Road condition


Traditional 1 2 3

1 2668 2307 1172

2 14 57 34


Table C 39: Skid resistance


Traditional 1 2 3 4

1 1867 83 46 18

2 4045 96 94 1

3 2


Table C 40: Delineation


Traditional 1 2

1 1274 4885

2 12 81





Table C 41: Street lighting


Traditional 1

1 5594

2 658


Table C 42: Pedestrian crossing


Traditional 7

2 6

3 4

4 1

6 41

7 6200


Table C 43: Pedestrian crossing quality


Traditional 3

1 46

3 6206


Table C 44: Pedestrian crossing – side road


Traditional 7

2 10

6 23

7 6219


Table C 45: Parking / side friction


Traditional 1

1 6135

2 115

3 2





Table C 46: Sidewalk drivers’ side


Traditional 5

2 32

3 24

4 11

5 1773

6 51

7 4361


Table C 47: Sidewalk passengers’ side


Traditional 5

1 4

2 40

3 26

4 66

5 369

6 1463

7 4284


Table C 48: Service road


Traditional 1

1 6244

2 8


Table C 49: Bicycle facility


Traditional 4

1 2

2 3

3 1043

4 5149

5 14

6 13

7 28





Table C 50: Roadworks


Traditional 1

1 5882

2 370


Table C 51: 85th%ile speed and Mean speed



5 10

7 123 35 18

9 53 2 10

11 11 2 326 10 44 25

13 3 7 10 96 24

15 6 234 29

17 14 4 5156


Coding comparison for attributes ‘85th%ile speed’ and ‘mean speed’ were the same.

C.4 Attributes for which Default Values Applied by Main Roads WA

The following attributes are coded the same in both data sets (Same: 100%, Different: 0%):

▪ motorcycle %

▪ pedestrian peak hour flow across road

▪ pedestrian peak hour flow along road drivers’ side

▪ pedestrian peak hour flow along road passengers’ side

▪ bicycle peak hour flow.

For the attribute ‘Roads that cars can read’, default values were used for both coding approaches, however different defaults were applied (i.e. Same: 0%, Different: 100%).

Table C 52: Upgrade impact cost


Traditional 1

1 4307

2 1448

3 497





Table C 53: Motorcycle observed


Traditional 1

1 6244

2 8


Table C 54: Bicycle observed


Traditional 1

1 6250

2 2

Same: 100%, Different: 0% (note 2 records different).

Table C 55: Pedestrian observed flow across road


Traditional 1

1 6250

2 2

Same: 100%, Different: 0% (note 2 records different).

Table C 56: Pedestrian observed flow along road drivers’ side


Traditional 1

1 6237

2 9

3 6


Table C 57: Pedestrian observed flow along road passengers’ side


Traditional 1

1 6220

2 19

3 9

4 2

5 1

6 1





Table C 58: Land use driver side


Traditional 5

1 4036

2 1688

3 311

4 214

7 3


Table C 59: Land use passengers’ side


Traditional 5

1 4263

2 1531

3 235

4 223





APPENDIX D CODING COMPARISON BY J KARPINSKI

The table below shows the coding comparison undertaken by Jan Karpinski between traditionally coded AusRAP and corporate data sets.

Table D 1: Comparison of AusRAP data and Main Roads WA data for Eyre Highway (H003), Great Eastern Highway (H005), Great Northern Highway (H006), Coolgardie Norseman Highway (H010) and Victoria Highway (H011)

Source: Karpinski (2014).




APPENDIX E GLOSSARY

The report assumes the reader has prior knowledge of ANRAM v1.0 and AusRAP and their basic terminology. For those not familiar with either risk assessment system, the following definitions may be of assistance.

ANRAM – the current working version of ANRAM used in this project was v1.04.

AusRAP – the Australian version of iRAP (identical). The current version is v3.02.

ViDA – software used by AusRAP to upload and process AusRAP road attribute data and produce star ratings based on risk scores, road investment programs and program-level safety benefits.

AusRAP Star Rating Score (SRS) – AusRAP SRSs are measure of individual risk of severe casualty crash based on AusRAP risk scores. Available only through ViDA at this time.

ANRAM risk score – risk scores calculated by ANRAM for each of the four vehicle crash types (run-off-road, head-on, intersection and other). They represent the individual road user’s risk of a severe (fatal or serious injury) casualty crash. It is referred to as individual risk. There is a general correlation between ANRAM and AusRAP risk scores although the absolute values are quite different. ANRAM v1.0 also uses slightly older risk algorithms than AusRAP.

ANRAM FSI crashes – estimate of the severe (fatal or serious injury) casualty crashes based on road stereotype, engineering features, speed, potential conflicting traffic and severe crash history. It is determined per road section per 5 years produced for each vehicle crash type and then aggregated. It can be expressed as a rate per km as a standardised measure of collective risk, i.e. risk of severe crashes to all road users on that section (vehicle-based crashes).

Individual risk – a measure of risk to an individual road user expressed per kilometre of travel by a vehicle, or vehicle-kilometres travelled (VKT). It is collective risk adjusted for road user exposure. Individual risk is often measured by fatal and serious injury crash rates, or risk assessment scores. AusRAP SRSs and ANRAM risk scores averaged for a given road section are an approximation of individual risk.

Collective risk – a measure that indicates crash frequency as experienced by the community, that is the number of crashes per unit of time. It can be expressed per section, per kilometre, or per intersection. ANRAM outputs Total ANRAM FSI crashes per road section which is converted to Total ANRAM FSI crashes per kilometre to provide a comparison rate of collective risk.

Road attribute – a characteristic of a road, e.g. sealed shoulder width, curvature, AADT. Each attribute has two or more categories with risk values assigned to each, based on safety research evidence.

Road type – ANRAM v1.04 divides roads into six categories, e.g. urban undivided, urban divided. This determines the algorithms used to estimate FSI crashes.

TSD – Traffic Speed Deflectometer (TSD) is now known as iPAVE in Australia.

Documents

Use of Main Roads WA Corporate Data for Road Safety Risk Data …€¦ · Corporate Data for Road Safety Risk Data Sets Exploring the use of Corporate Data for AusRAP/ANRAM Data Sets