Content uploaded by Shane Turner
Author content
All content in this area was uploaded by Shane Turner on Aug 07, 2019
Content may be subject to copyright.
Glossary
AS Site-specific crash rate (reported injuries)
AT Typical crash rate (predicted injuries)
AADT Annual average daily traffic
Austroads National Association of Australian Road Authorities
C Daily cycle volume
CAS Ministry of Transport’s Crash Analysis System
CMF Crash modifying factor
CRF Crash reduction factor
DSi Number of deaths and serious injury casualties. May be reported, estimated or
predicted. The term FSi is used in this manual instead of DSi as crash numbers are
used in the NZ Economic Evaluation Manual.
EEM Economic Evaluation Manual
FSi Number of fatal and serious injury crashes that involve at least one death or serious
injury. May be reported, estimated or predicted. A crash may involve several
deaths and serious injuries. Crash numbers are used in economic evaluation.
HRIG High-risk intersection guide
HRRRG High-risk rural roads guide
Intersection For the purposes and clarity for using the guide an intersection is:
Where two or more streets or roads join or cross, or
Where a major public driveway joins a street or road and is constructed
as an intersection. (Note: it is easy to overlook these when searching in
CAS.)
Mid-block Road sections ≥ 50m from an intersection.
NZTA New Zealand Transport Agency
ONRC One Network Road Classification
P Daily pedestrian crossing volume
Q Daily traffic volume
Severity Factors The expected ratio of FSi crashes to all injury crashes.
Contents
Crash Estimation Compendium.............................................................................................................................................. 1
Glossary 3
1.0 Introduction ..................................................................................................................................................................... 6
2.0 Methodology .................................................................................................................................................................. 8
2.1 Model predictions ............................................................................................................................................................ 8
2.1.1 Methodology by site and crash type ................................................................................................................................ 8
2.1.2 Crash model types ......................................................................................................................................................... 9
2.2 Crash reduction factors and crash modifying factors ....................................................................................................... 11
2.3 Severity factors ............................................................................................................................................................. 12
3.0 Rural Roads (≥ 80 km/h) ............................................................................................................................................... 13
3.1 Rural two-lane roads ≥ 80/km ........................................................................................................................................ 13
3.2 Rural isolated curves ≥ 80km/h ..................................................................................................................................... 17
3.3 Single-lane rural bridges ≥ 80km/h ................................................................................................................................ 19
3.4 Two-lane rural bridges, ≥ 80km/h.................................................................................................................................. 19
4.0 Urban Roads (≤ 70 km/h) ............................................................................................................................................. 20
4.1 Urban mid-block – injury crashes .................................................................................................................................... 20
4.2 Urban mid-block – pedestrian and cyclist crashes ........................................................................................................... 21
6.0 Intersections – Product of Flow Models .......................................................................................................................... 24
6.1 Urban Priority and Signalised Cross roads and T-junctions 50-70km/h ............................................................................ 24
6.2 Urban Roundabouts 50-70 km/h ................................................................................................................................... 25
6.3 High-speed (Rural) Priority and Signalised Cross roads and T-junctions (≥ 80km/h on main road) ................................... 26
6.4 High-speed (Rural) Roundabouts (≥ 80km/h on main road) ........................................................................................... 29
6.5 Urban and Rural Railway Crossings ................................................................................................................................ 29
7.0 Intersections - Conflicting Flow Models .......................................................................................................................... 32
7.1 Urban signalised crossroads <80km/h ............................................................................................................................ 32
7.2 Urban roundabouts (<80km/h) ..................................................................................................................................... 36
7.3 Urban Priority T-junctions (<80km/h on main road) ....................................................................................................... 39
7.4 High Speed Priority Cross Roads (≥ 80km/h on main road)............................................................................................. 42
7.5 High-speed priority T-junctions (≥ 80km/h on main road) .............................................................................................. 44
8.0 Crash modification factors ............................................................................................................................................. 50
8.1 Introduction ................................................................................................................................................................... 50
8.2 Typical crash reductions ................................................................................................................................................ 50
9.0 Severity factors ............................................................................................................................................................. 72
References and Bibliography ................................................................................................................................................ 74
Appendix 1 .......................................................................................................................................................................... 76
Figure 1: Injury crashes per 100 million vehicles for rural curves for type B, C, and D crashes (2015) .................................................... 18
Table 1: Crash rates and crash prediction model types ........................................................................................................................................... 9
Table 2: Rural (State Highways) two-lane roads by horizontal terrain type ................................................................................................... 13
Table 3: Rural (local) two-lane road coefficients by horizontal terrain type ................................................................................................... 15
Table 4: Horizontal Alignment Classification .......................................................................................................................................................... 16
Table 5: Cross-section crash modifying factors (CMFs) ...................................................................................................................................... 16
Table 6: Rural bridge type k values ............................................................................................................................................................................. 19
Table 7: Urban mid-block land-use coefficients ..................................................................................................................................................... 20
Table 8: Urban mid-block land-use k values ........................................................................................................................................................... 20
Table 9: Urban mid-block – Pedestrian and Cyclist crash variables and CAS movement categories ...................................................... 21
Table 10: Urban mid-block – pedestrian and cyclist facilities models (model references 6 and 16). ....................................................... 21
Table 11: Four-lane divided rural roads coefficients ................................................................................................................................................ 23
Table 12: Four-lane divided rural roads k values ...................................................................................................................................................... 23
Table 13: General cross-road and T-junction urban intersections (50-70km/h) coefficients (reference 21) ....................................... 25
Table 14: General cross-road and T-urban intersections 50-70km/h k values ............................................................................................. 25
Table 15: General urban roundabouts 50-70km/h coefficients (reference 5) .............................................................................................. 26
Table 16: General urban roundabouts 50-70km/h k values ............................................................................................................................... 26
Table 17: General high-speed cross roads and T-junctions ≥ 80km/h coefficients (reference 8) .......................................................... 28
Table 18: General high-speed cross and T-intersections ≥ 80km/h k values ............................................................................................... 28
Table 19: High-speed roundabout coefficients (reference 8) ............................................................................................................................. 29
Table 20: High-speed roundabout k values ............................................................................................................................................................. 29
Table 21: Urban and rural railway crossings coefficients ...................................................................................................................................... 31
Table 22: Urban and rural railway crossings k values ............................................................................................................................................ 31
Table 23: Urban signalised cross roads (<80km/h) variables and CAS movement categories ............................................................... 34
Table 24: Urban signalised crossroads (<80km/h) crash prediction models (reference 6 and 16) ........................................................ 35
Table 25: Urban roundabouts (<80km/h) variables and CAS movement categories ................................................................................. 37
Table 26: Urban roundabouts (<80km/h) crash prediction models (reference 5) ..................................................................................... 38
Table 27: Urban priority T-junctions (<80km/h on main road) variables ...................................................................................................... 39
Table 28: Urban priority T-junction (<80km/h on main road) models (reference 8) ................................................................................ 40
Table 29: High speed priority cross roads (≥ 80km/h on main road) variables ........................................................................................... 42
Table 30: High speed priority cross roads (≥ 80km/h on main road) models (reference 8) ................................................................... 44
Table 31: High speed priority T-junctions (≥ 80km/h on main road) variables ............................................................................................ 45
Table 32: High speed priority T-junction (≥ 80km/h on main road) models (reference 8) ...................................................................... 47
Table 33: Common rural midblock crash reduction/modification factors....................................................................................................... 53
Table 34: Common urban midblock crash reduction/modification factors ................................................................................................... 59
Table 35: Common Motorway Crash Reduction/Modification Factors ........................................................................................................... 61
Table 36: Common intersection crash modification/reduction factors (urban and rural) ........................................................................ 64
Table 37: Common Urban Cyclist Crash Reduction/Modification Factors (apply only to crashes involving cyclists) ...................... 68
Table 38: Common Urban Pedestrian Crash Reduction/Modification Factors (applies only to pedestrian crashes) ....................... 69
Table 39: Urban Intersection (less than 80 km/h) FSi Severity Factors on all roads. .................................................................................. 72
Table 40: Rural Intersection (80 km/h plus on one intersecting road) FSi Severity Factors ..................................................................... 73
Table 41: Mid-blocks and Special Sites FSi Severity Factors ................................................................................................................................ 73
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 6
1.0 Introduction
This manual presents methods for estimating (police) reported injury crash predictions for various road and site
elements in New Zealand. A full list of road and site types currently covered by this manual are outlined, including
the transport modes covered by these models and factors. This is the first amendment of the Compendium.
There are known gaps in the crash models, rates and crash reduction factors that are currently available for use in
New Zealand. The intention is to address these gaps in future versions of the manual.
While this manual has been prepared as a compendium to the NZ Economic Evaluation Manual (EEM) the
models, rates and reduction rates, along with severity factors have a wider range of use under a safe system
approach, than just economic evaluation. For example the underlying crash rate at a site or along a route, and
especially the risk of fatal and serious crashes, can be estimated with more reliability using the models and
severity ratios. Historical crash data, especially for more severe crashes and fatalities can be very variable, and
the crash predictions allow an analyst to assess whether the history reflects an underlying crash risk or are just
showing a spike in crash risk that is unlikely to be repeated. A safe system approach needs to focus on the areas
of high underlying risk of such crashes, rather than respond to a one off crash occurrence.
A key role of the manual is to allow an assessment of the effectiveness of safety improvement works. Crash
reduction factors (CRF) and crash modification factors (CMF) have been provided for a variety of different road
features and safety improvement countermeasures. The factors have been developed in evaluation studies using
police reported injury crashes. The crash reduction factors have been developed for different crash types, level of
severity and different transport modes (e.g. crashes involving pedestrians only). CMFs have been derived for all
injury crashes or for all injury crashes involving a transport mode. Many of the CMFs and CRFs have been
developed or collated as part of Austroads research.
Under-reporting of police reported crashes is not considered in this manual. While fatal crashes are often
assumed to be 100% reported, for minor and serious crashes reporting rates are at best 50% and often lower.
Reporting rate factors are found in Appendix A6 of the NZ EEM. The crash rates and models also do not consider
non-injury crashes. Further advice on non-injury crashes can be found in Appendix A6.
The manual also includes severity factors for different routes and site types. These factors allow the risk of fatal
and serious injury crashes to be estimated from predictions of total injury crashes (fatal, serious injury and minor
injury). These factors are new and limited due to sample size restrictions for some sites and crash types. Care
should be taken in their use.
The crash rates, crash prediction models, CRFs, CMFs and severity factors presented here are not exhaustive and
analysts are permitted to use other research that is available, as long as the robustness of this research can be
demonstrated in the New Zealand (and Australian) context. Crash reduction and crash modifying factors used
from outside of the compendium need to be fully referenced (for example papers, research reports or unpublished
material), along with information on sample size, modelling technique, goodness-of-fit statistics, and confidence
levels stated. Alternative crash rates and crash prediction software may also be used provided they are calibrated
to New Zealand conditions.
For intersection and mid-block crash prediction models, analysts are referred to the appropriate research report
on crash prediction models in the reference section. The crash prediction models in these reports are more
extensive than provided in the manual and may be useful when looking at some crash counter-measures.
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 7
However the model parameters may need to be adjusted given the downwards trends in crashes and because
many of the models predict crashes over five years rather than one year.
This document has been formed as a compendium to the EEM, and is to be used in conjunction with the EEM
when applying EEM appraisal method B (crash analysis) and method C (weighted crash analysis).
Section 2 of the manual provides an outline of the methodology that is used to calculate crash predictions using
the various analysis tools. Sections 3 to 7 provide the crash rates and crash prediction models that can be used for
rural links, urban links, intersections, railways crossings, curves and narrow bridges. Section 8 includes common
CRFs and CMFs for different link and site types that can be used to assess the effectiveness of various safety
countermeasures. Section 9 includes the severity factors that are used to estimate the risk of serious injury and
fatal crashes at a site.
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 8
2.0 Methodology
2.1 Model predictions
The crash rate and crash prediction models in this manual, unless otherwise stated, have been developed for the
most common types of site in each category. For example, traffic signal models were generally developed for two
and three phase signals, and are therefore are not as accurate for signals with four or more phases, or where there
are a lot of phase changes during set periods of the day. The models and rates are most valid within the flow
ranges provided. Analysts should exercise caution when using the models and rates outside these ranges.
The more unusual a site is from the typical site type, the less appropriate the general models and equations will be
for predicting the typical crash rate. In most cases where there is a feature of a site, such as the site’s layout, that
has a significant effect on the crash rate, the rates and models in this manual are not likely to be appropriate.
The models presented here deal with (reported) injury crashes only. Crashes and casualties have a close
statistical relationship. There are a number of factors; such as the number of vehicle occupants; that can be used
to determine casualty numbers using the established crash numbers. Refer to the HRIG (NZ Transport Agency
2013) and HRRRG (NZ Transport Agency 2011) for more information on this relationship.
Generally all flow models are suitable for most mid-block or intersection types indicated. Where a breakdown of
crashes by crash type or road user type is required; or, in the case of intersections, where the proportion of turning
vehicles is high compared to through vehicles, then more detailed conflicting flow models by crash type and
movement should be used.
2.1.1 Methodology by site and crash type
Many projects are made up of multiple site types, including links (of different traffic volume and speed),
intersections, bridges, curves and railway crossings (see figure example below). To estimate the total number of
crashes at a site the predictions for each site type must be calculated and added together (ATOTAL = AT(LINK1) + A
T(CURVE1) + AT(INT1) …).
For intersections, crashes that are 50 metres up each leg are attributed to the intersection. In a similar way,
crashes around bridges and railways crossing extend up to approximately 50 metres from the site. Mid-block
crash rates generally exclude ‘major’ intersection crashes. Midblock crash rates and crash prediction models do
include crashes at accesses and lower volume intersections. It is acknowledged that the cause of a crash may not
always be contained within the 50 metre buffer. At major intersections traffic queuing may at times extend
beyond 50 metres from the limit lines and cause crashes. Likewise there may be mid-block type crashes that do
occur within the intersection buffer area that are not attributed to the intersection. These limitations of the crash
rates and crash prediction models need to be considered in analysis.
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 9
For some improvement projects it is necessary to predict crashes at a site by type and/or mode (for example
intersections (AT(INT))). At a high level this may be separating out crashes involving pedestrian and cyclists from
motor-vehicle only crashes (e.g. AT(INT) = AT(PED) + AT(CYCLE) + AT(MOTOR VEH)). This is required when different
improvements are focused on different transport modes (e.g. the installation of a new pedestrian crossing facility
or a new cycle lane).
It may also be necessary to look at specific crash types (see Appendix A for NZ crash codes) for a particular
mode. Some improvements, such as the installation of a right turn bay at a rural intersection or installation of
right turn signal phase at an urban signalised intersection only impact on some crash types. Models by crash type
are called conflicting-flow crash prediction models (for example motor vehicles (AT(MOTOR VEH))). Several
conflicting flow models are available by site type for different transport modes. The crash predictions by crash
type and approach need to be added together to produce total crashes for each mode (e.g. AT(MOTOR VEH) = AT(HA
App 1) + AT(HA App 2) + AT(HA App 3) + AT(HA App 4) + AT(LB App 1) + AT(LB App 2) + AT(LB App 3) + AT(LB App 4) + AT(F App 1) + AT(F App
2) + AT(F App 3) + AT(F App 4) ….)
2.1.2 Crash model types
The five models groups that presented in the compendium are shown in Table 1.
Table 1: Crash rates and crash prediction model types
Rural Roads (2 and 3 lane
mid-blocks sections) ≥
80km/h
Rural two-lane roads (by ONRC and terrain type)
Two-lane roads with passing lanes
Rural isolated curves
Single lane rural bridges
Two-lane rural bridges
Multi-lane High Speed
Roads
Motorways
Four lane divided rural roads (expressways – with either wide grass medians
or physical median barriers)
Urban Roads (Mid-blocks)
50-70km/h
Urban mid-blocks (by road hierarchy)
Urban Arterials with ≥ 6 lanes
Product of Flow Models -
Intersections
General urban cross and T-junction intersection 50-70km/h
General urban roundabouts 50-70km/h
General high speed roundabout ≥ 80km/h on one approach
Urban and Rural railway crossings
Conflicting Flow Models -
Intersections
Urban signalised cross roads <80km/h
Urban roundabouts <80km/h
High speed priority crossroads >70km/h
High-speed priority T-junctions >70km/h
The rates and models present in this compendium have either been developed exclusively for the NZ Transport
Agency EEM (1) or as part of a research project. In the latter case reference of the relevant research report has
been provided. In many cases the original models have been modified for this compendium to include the
downward trend in crashes since the models were developed.
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 10
2.1.2.1 Rural Road Mid-block Crash Rates
General rural road crash rates are suitable for most rural mid-block analysis, except those with continuous four or
more lanes. For multiple-lane roads use the crash prediction models provided for motorways and 4-lane divided
roads. Passing lane and short 4-laned sections (double passing lanes), can be assessed using a crash modifying
factor (CMF). For bridges, isolated out-of-context curves, railway crossings and major intersections use the other
crash models provided.
The rural 2-lane mid-block crash rate has the following form:
Injury crashes per year (AT) = crash rate (b0) x Exposure (volume) x ∑CMFs
∑Crash Modifying Factor (CMF) = CMF1 * CMF2 * … (e.g. lane and shoulder width)
Exposure (mid-blocks) = L x AADT x 365 / 108
Where: AADT = annual average daily traffic
L = length (km)
Crash prediction models are also available for rural roads in New Zealand. Refer to research by Turner et al (19)
and Cenek and Davis (14). While these models maybe useful for evaluating rural realignments, they have not as
yet been fully assessed for use in economic evaluation. Once this process is completed these models may be
added to future versions of this guideline.
2.1.2.2 Urban Road Mid-blocks
Crash prediction models are used to estimate injury crashes at urban mid-block sites. The reported injury crashes
per year is dependent on roadside development. Separate pedestrian and cyclist injury crash models are also
available.
The urban 2 and 4 lane mid-block crash prediction model has the following form:
Injury crashes per year (AT) = b0 x Qb1 x L x ∑CMFs
∑Crash Modifying Factor (CMF) = CMF1 * CMF2 * … (e.g. solid and flushed medians)
Where: b0 and b1 = model parameters
Q = annual average daily two-way traffic volume
L = length (km)
Major intersections and railway crossings should be assessed separately using either the product-of-flow or
conflicting flow crash prediction models.
2.1.2.3 Product of Flow Models – Intersections
Two types of crash prediction model are available for intersections. High level product-of-flow models predict
total injury crashes based on the product of the traffic volumes on the two roads that are intercepting. Separate
models are available for different forms of control and for cross roads and T-junctions. These models should only
be used when analysing intersection changes that impact on all injury crashes or for project feasibility analysis.
Changes that often impact on all injury crashes include changing form-of-control (e.g. priority control to traffic
signals) and traffic volume increases (possibly as a result of a new development). These models are also useful
for calculating the injury crash rate at new intersections. For more detailed analysis of intersections conflicting
flow models should be applied.
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 11
The product-of-flow intersection models have the following general form:
Injury crashes (priority and traffic signals) = b0 x Qmajorb1 x Qminorb2 x ∑CMFs
Injury crashes (roundabouts) = b0 x Qapproachb1 x ∑CMFs
∑Crash Modifying Factor (CMF) = CMF1 * CMF2 * … (e.g. lighting and splitter island)
Where: b0 and b1 = model parameters
Qmajor = annual average daily two-way traffic volume on highest
volume road (signals) or priority road
Qminor = annual average daily two-way traffic volume on lowest volume
road (signal) or side-road
Qapproach = annual average daily two-way traffic volume on each
roundabout approach
L = length (km)
Product-of-flow crash prediction models are also available for different railway crossing control types. These
models include both traffic volume and the typical number of train services per day.
2.1.2.4 Conflicting Flow Models - Intersections
Conflicting flow models provide a breakdown of the predicted crashes by road user type (e.g. pedestrian and
cyclists) and crash type (refer to Appendix A). Crash type models are usually only available for the major crash
types at each intersection. The total number of injury crashes at an intersection is calculated by adding up the
crashes by each type and approach and then using either a general/other crash prediction model or a factor to
take into account the crashes not modelled.
Conflicting flow models are typically used in analysis when there are a high proportion of vehicles making turning
movements, especially right turns and when treatment impacts on particular crash types or crash modes.
Examples of the latter include installing a right turn bay at a rural priority intersection and right turn signal phasing
at urban traffic signals.
This manual contains a large number of conflicting flow models. The New Zealand research available also has a
large number of other crash prediction models. Many of the models include non-flow variables, like speed and
road layout factors. Even with the large number of models available there are some major gaps in the range of
models provided. In the case that detailed models are not available then analysts may have to use the product of
flow models.
Generally CMFs should not be applied to these model predictions, as the CMFs normally apply only to all injury
crashes. It is not possible to present a general model form, but two examples are given:
Right turn against crashes (rural priority) = b0 x qxb1 x qyb2 x RTB factor
Where: b0, b1and b2 = model parameters
qx and qy = various daily turning movement volumes (of which there are
twelve at a X-roads and six at a T-junction)
RTB factor = adjustment to crash prediction (CMF) if right turn bay provided
Entering versus circulating cycle crashes (roundabouts) = b0 x Qeb1 x Ccb2 x Speed b3
Where: b0, b1, b2 and b3 = model parameters
Qe and Cc = daily entering volume for motor-vehicle and circulating volume
for cyclists (Cc)
Speed = Mean speed of traffic entering from each approach
2.2 Crash reduction factors and crash modifying factors
A Crash Reduction Factor (CRF) indicates the expected percentage reduction in crashes following the
introduction of a treatment. Crash reduction factors can apply to all injury crashes, crash of a particular severity
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 12
(e.g. fatal and serious injury), specific crash types (e.g. loss-of-control crashes), by a particular mode (e.g.
pedestrian crashes) or by environmental conditions (e.g. night-time and wet-weather crashes). These factors are
typically applied to historical crashes to estimate future crash numbers after an intervention. In economic
evaluation they are used for Method A – Crash-by- Crash analysis.
The effectiveness of traffic engineering countermeasures in Australia and New Zealand has traditionally been
presented using Crash Reduction Factors (CRFs), which presents the expected percentage reduction in crashes.
The term Crash Modifying Factor (CMF) is now used more widely overseas, although both terms are used in this
manual (Austroads, 2012).
A Crash Modification Factor (CMF) is used to adjust a crash prediction from a crash rate or crash prediction
model to reflect a road feature or safety improvement measure that is not reflected in the rate or model. In this
manual CMFs are provided for all injury crashes or all injury crash involving a specific mode. Hence they are only
applied to models that predict all injury crashes, not to conflicting flow models. Refer to general model forms
provided above for how CMFs can be applied in crash prediction.
CMFs have been included in the manual for use in economic evaluation. CMFs should be used for Method B
(Crash Rate Analysis) and Method C (Weighted Crash Procedure).
2.3 Severity factors
Severity factors are used to estimating the expected number of deaths and serious injury crash equivalents (AFSi)
based on reported injury crashes at a site. To predict the equivalent FSi multiply the all injury predictions
calculated by the various crash rates and crash prediction models in this guide by the severity factors
ADSi = SF x ATOTAL
Where, SF is the Severity Factor (from tables provided)
ATOTAL is a site’s predicted injury crash rate
The expected number of FSi by mode type for a site can also be estimated (for example FSi (pedestrian crashes) =
SF (ped) x APED).
The severity outcome of crashes is influenced by vehicle speeds, intersection and link types, transport mode
involved and the crash movement types. The New Zealand Crash Analysis System (CAS) has been used to
determine the severity factors of all movement types by vehicle speed, mode and site type. Severity factors by
crash type have not been provided in this guide due to accuracy issues associated with sample size.
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 13
3.0 Rural Roads (≥ 80 km/h)
This section includes crash rates for rural 2-lane mid-blocks, isolated out-of-context curves and narrow two lane
and single lane bridges. Crash prediction models for rural intersections and railway crossings are found in
Sections 6 and 7.
3.1 Rural two-lane roads ≥ 80/km
For two-lane rural roads in 80 and 100km/h speed limit areas, the typical crash rate (reported injury crashes per
year) is calculated using the exposure-based equation:
Injury crashes per year = crash rate (b0) x Exposure(X) x ∑CMFs
∑Crash Modifying Factor (CMF) = CMF1 * CMF2 * … (e.g. lane and shoulder width)
X, Exposure (mid-blocks) = L x AADT x 365 / 108
Where: AADT = annual average daily traffic
L = length (km)
Coefficient b0 is provided in
Table 2 and Table 3 for the various levels of the One Network Road Classification (ONRC). The ONRC is the
national categorisation of roads based on their functions; refer to (20) for details on each road category. The
horizontal alignment category is based on bendiness and is defined in Table 4. The alignment ranges should
generally be maintained throughout the road section. The k-value is used in economic evaluation (Method C).
The coefficient b0 is applicable to a given mean seal width. The CMFs for seal widths are provided in Table 5, and
varies according to three road types (grouping of various one network classification types), seal shoulder width
and lane width. For road type one, two and three the seal width is assumed to be 9.5 metre, 8.2 metre and 6.7
metre respectively. Other CMFs for rural roads (e.g. for providing shoulder and median barriers) are provided in
Section 8.
Operating speed is an important consideration in rural road crashes and the severity of these crashes. The crash
rates provided in
Table 2 and Table 3 do include the effects at a high level of operating speed. Operating speeds on a tortuous
alignment are generally a lot lower than on a straight alignment, due to the constraints of the curves. What the
crash rates don’t consider is the consistency of the alignment. A consistent alignment is less likely to catch
drivers out, as they can maintain a constant speed. Out-of-context curves occur where there is a large speed
change required to negotiate the curve or series of curves. For isolated curves the rates in the next section can be
used to predict the impact on crash occurrence. For more complicated alignments including a variety of curves
and straights analysts need to use a rural road crash prediction model if a more accurate crash prediction of injury
crashes and serious and fatal crashes is required (refer to 14 and 19).
The speed limit on a rural road can impact on operating speed and the associated change in injury crash rates and
crash severity (i.e. the proportion that are serious or fatal). Speed limit reductions rarely reduce speeds by the full
reduction applied (e.g. a 10 km drop in speed limit may only reduce operating speeds by 3 to 5 km/h). The speed
reduction can be particularly low or zero when the speed limit is still above the roads normal operating speed. The
power models developed by Elvik et al (11) can be used to assess the crash benefits of reducing operating speeds
by speed limit reductions.
Table 2: Rural (State Highways) two-lane roads by horizontal terrain type
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 14
One Network
Classification Horizontal Alignment b0
K
National Strategic
(High Volume)*
Straight 8 0.9
Curved 16 0.9
Winding 29 1.2
Tortuous 35 1.2
National Strategic
Straight 13 0.9
Curved 19 0.9
Winding 29 1.2
Tortuous 35 1.2
Regional Strategic
Straight 13 3.0
Curved 18 3.0
Winding 31 1.2
Tortuous 35 1.2
Arterial
Straight 13 3.0
Curved 22 3.0
Winding 31 1.2
Tortuous 35 1.2
Primary Collector
Straight 18 3.0
Curved 23 3.0
Winding 34 4.2
Tortuous 35 3.0
Secondary Collector
Straight 18 3.0
Curved 29 3.0
Winding 34 3.0
Tortuous 35 3.0
*As outlined in section 5.0 the crash rate for well-designed 4-lane motorways and four lane divided road is in the order of 3 to 8 crashes per 100MVKT. For 6 plus lane
motorways refer to the models in Section 5.0.
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 15
Table 3: Rural (local) two-lane road coefficients by horizontal terrain type
One Network
Classification Horizontal Alignment b0
k
National Strategic (Including
High Volume)
Straight 8
0.9
Curved 16
0.9
Winding 29
1.2
Tortuous 35
1.2
National Strategic
Straight 14
0.9
Curved 19
0.9
Winding 29
3.0
Tortuous 35
3.0
Regional Strategic
Straight 18
1.2
Curved 23
3.0
Winding 31
1.2
Tortuous 35
3.0
Arterial
Straight 20
2.3
Curved 23
3.0
Winding 31
1.2
Tortuous 35
1.2
Primary Collector
Straight 25
3.0
Curved 29
3.0
Winding 37
4.2
Tortuous 37
3.0
Secondary Collector
Straight 24
3.0
Curved 29
3.0
Winding 34
3.0
Tortuous 35
3.0
Access
Straight 24
3.0
Curved 33
3.0
Winding 33
3.0
Tortuous 34
3.0
Access (Low volume)
Straight 24
3.0
Curved 33
3.0
Winding 33
3.0
Tortuous 34
3.0
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 16
Table 4: Horizontal Alignment Classification
Horizontal alignment type Degrees/km
Straight 0-50
Curved 50-150
Winding 150-300
Tortuous >300
Table 5 provides modification factors for two-lane rural crash rates for various combinations of seal widths that
differ from the mean seal widths assumed for that road type. First, the overall seal width, shoulder width and lane
width are determined. Then, look up CMF that corresponds to the Road Type/One Network Road Classification,
shoulder width and lane width in Table 4. Adjust b0 by multiplying with the modification factor and use this value to
calculate the typical crash rate. In the case of shoulder widening, different modification factors would be used for
the do-minimum and option.
Table 5: Cross-section crash modifying factors (CMFs)
CMFs for Road Type 3 (Secondary Collector and Access)
Seal shoulder width
Lane width
2.75m 3.00m 3.25m 3.50m 3.60m
0m 1.17 1.10 1.03 0.96 0.93
0.25m 1.10 1.03 0.96 0.89 0.86
0.50m 1.03 0.96 0.89 0.82 0.79
0.75m 0.89 0.82 0.75 0.68 0.66
1.00m 0.75 0.68 0.61 0.55 0.52
1.50m 0.61 0.55 0.48 0.41 0.41
2.00m 0.48 0.41 0.41 0.41 0.41
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 17
CMFs for Road Type 2 (Arterial and Primary Collector)
Seal shoulder width
Lane width
2.75m 3.00m 3.25m 3.50m 3.60m
0m 1.47 1.38 1.30 1.21 1.17
0.25m 1.38 1.30 1.21 1.12 1.09
0.50m 1.30 1.21 1.12 1.03 1.00
0.75m 1.20 1.13 1.01 0.87 0.83
1.00m 1.07 1.01 0.85 0.71 0.65
1.50m 0.77 0.69 0.60 0.54 0.51
2.00m 0.60 0.51 0.51 0.51 0.51
CMFs for Road Type 1 (National and Regional Strategic)
Seal shoulder width
Lane width
2.75m 3.00m 3.25m 3.50m 3.60m
0m 2.11 2.01 1.90 1.79 1.74
0.25m 2.01 1.90 1.79 1.67 1.58
0.50m 1.90 1.79 1.67 1.45 1.36
0.75m 1.79 1.67 1.45 1.22 1.18
1.00m 1.67 1.45 1.22 1.11 1.07
1.50m 1.22 1.11 1.00 0.89 0.85
2.00m 1.00 0.89 0.78 0.66 0.66
3.2 Rural isolated curves ≥ 80km/h
Figure 1and the equation below provide typical crash rates for reported injury loss-of-control and head-on crashes
on rural curves, adjusted for the general trends in crashes (see Jacket, 13, for original crash rates). They should be
used only for an isolated curve that is replaced with a single curve of a higher design speed.
The data for typical injury crash rates has been based on sealed rural state highways. An underlying assumption
is that the road section under consideration is not affected by ice or other adverse factors such as poor visual
conditions.
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 18
The typical crash rate (reported injury crashes per year, by CAS movement categories B, C and D) for an isolated
rural curve is calculated using the equation:
AT = b0 X e(b
1
S)
where: b0 = 3.55
b
1 = 2.0
X is the exposure in 100 million vehicles (in one direction) passing through the curve
S = 1 –
AT must be calculated for both directions, and S is likely to vary between the two directions (a k value of 1.1 is used in
the weighted crash procedure). If the design speed is approximately equal to the approach speed then the equation
reduces to:
AT = b0 X
The following assumptions apply when using the equation or Figure 1:
For Figure 1 the rate is in terms of injury crashes per 100 million vehicles, and for the equation the rate is in
injury crashes per year through the curve
The design speed of the curve should be determined from a standard design reference
The approach speed to the curve is the estimated 85th percentile speed at a point prior to slowing for the
curve (for longer tangents this would approximate the speed environment).
Figure 1: Injury crashes per 100 million vehicles for rural curves for type B, C, and D crashes
(2015)
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 19
3.3 Single-lane rural bridges ≥ 80km/h
The typical crash rate (reported injury crashes per year) of a single-lane bridge on a rural road (≥ 80km/h) is
determined by the equation:
AT = b0 X
where: X is the exposure in 100 million vehicles crossing the bridge per year
b
0 = 8.7 (QT)0.3 (2015 analysis year)
QT is the two-way daily traffic volume (AADT)
This equation does not take into account low design speed approach curves (65km/h advisory speed or less), traffic
signal control or adjoining intersections within 200 metres of the bridge.
3.4 Two-lane rural bridges, ≥ 80km/h
The typical crash rate (reported injury crashes per year) of a two-lane bridge on a rural road (≥ 80km/h) is
determined by the equation:
AT = b0 X
where: X is the exposure in 100 million vehicles crossing the bridge per year
b
0 = 0.83 × c × (0.5 – 0.25 RW + 0.025 RW2) (2015 analysis year)
With RW being the difference between the seal width across the bridge and the total sealed lane width in metres
(both directions) on the bridge approaches (normally 7 metres on state highways). A narrow bridge seal width
leads to a negative value for RW. The limits of RW are governed by the limiting width for single-lane operation
(for the maximum negative value of RW) and 2.5 metres (maximum positive value of RW). The value of c is given
by the formula:
c = e(3.5 – Q
T
/ 7,500)
where: QT is the two-way daily traffic volume (AADT)
This model does not take into account low design speed approach curves (65km/h advisory speed or less) or
adjacent intersections within 200 metres of the bridge. In this situation the combined effects of different road
elements (bridge, curve and intersection) can be greater or less than the effects of that predicted using the
various crash rates and crash prediction models for each road element. The use of crash history through the
weighted crash analysis procedure can enable the combined crash effect to be better understood, although the
crash history in turn is heavily influenced by the random occurrence of injury crashes.
In the weighted crash procedure, use the k-values provided in Table 6Error! Reference source not
found..
Table 6: Rural bridge type k values
Rural bridge type k value
Single-lane bridge 0.3
Two-lane bridge 0.2
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 20
4.0 Urban Roads (≤ 70 km/h)
Crash prediction models are available for all injury crashes on urban mid-blocks and for pedestrian and cyclists
involved in injury crashes at mid-blocks. Crash prediction models for urban intersections are found in Sections 7
and 8.
4.1 Urban mid-block – injury crashes
Crash prediction models are used to estimate total injury crashes at urban midblock sites. The typical crash rate
(reported injury crashes per year) is dependent on roadside development. Separate pedestrian and cyclist models
are also available. All reported injury crashes are calculated using the model:
Injury crashes per year = b0 x Qb1 x L x ∑CMFs
∑Crash Modifying Factor (CMF) = CMF1 * CMF2 * … (e.g. solid and flushed medians)
Where: b0 and b1 = model parameters (Table 7)
Q = annual average daily two-way traffic volume
L = length (kilometres)
Table 7: Urban mid-block land-use coefficients
Land-use Commercial Other
Mid-block road t
yp
e b0 b1 b0 b1
Access (Local) 2.19 × 10-4 0.98 2.19 × 10-4 0.98
Primary and Secondary
Collectors
2.99 × 10-5 1.08 2.99 × 10-5 1.08
National and Regional
Strategic and Arterial (2
and 4 lane)
6.63 × 10-6 1.20 1.16 × 10-4 0.88
Table 8 shows the traffic volume range over which the crash prediction models should be applied and also the ‘k’
values to use in economic evaluation (using Method C). There is less certainty in crash estimation when a route
has a traffic volume outside this flow range.
Table 8: Urban mid-block land-use k values
Mid-block type Speed limit Flow range AADT
k value
Commercial Other
Access (Local) 50km/h < 3,000 0.6 0.6
Primary and Secondary
Collectors
50km/h 2,000 – 8,000 10.0 10.0
National and Regional
Strategic and Arterial (2 and
4 lane)
50 or 60km/h 3000 – 24,000 8.5 10.8
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 21
There is currently no New Zealand information available for six or more lane arterials. Six-lane roads are likely to
have a greater proportion of weaving-related crashes, particularly where intersections are closely spaced.
4.2 Urban mid-block – pedestrian and cyclist crashes
Pedestrian and cyclist crash prediction models are provided for estimating injury crashes that involve crossing
pedestrians and through cyclists on road mid-block (Table 9). These models can be used to assess the benefits of
a new or improved pedestrian or cyclist facility by applying a CMF. These models are for urban (speed limit
<80km/h) areas and do not include any pedestrian or cyclist crashes that occur at side roads. However, driveway
crashes are included. The number of reported injury crashes per year for each crash type is calculated using the
models in Table 10.
Table 9: Urban mid-block – Pedestrian and Cyclist crash variables and CAS movement categories
Crash types Variables CAS movement
categories
All mid-block pedestrian
crashes
NA-NO, PA-PO
All mid-block cyclist crashes
All
Table 10: Urban mid-block – pedestrian and cyclist facilities models (model references 6 and 16).
Crash types Model k value
(mid-point)
All mid-block pedestrian crashes AT = 1.27 × 10-4 × Q0.69 × P0.26 × L -
All mid-block cyclist crashes AT = 2.36 x 10-4 x Q0.84 x L0.30 x No_Parking
(Parking = 1 and No_Parking = 0.25) -
Q = Two-way vehicle flow in veh/day
P = Pedestrian crossing volume per
100 metres in ped/100m/day
L = Segment length in km
Q = Two-way vehicle flow in veh/day
C = Two-way cycle flow in
veh/day/100m
L = Segment length in km
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 22
5.0 Multi-lane High Speed Roads (including Motorways)
The typical two-way crash rate (reported injury crashes per year) for 4-lane motorways and four-lane divided rural
roads is calculated using the model:
AT = b0 × QTb
1
× L
where: QT is the daily two-way traffic volume (AADT) on the link
L is the length of the motorway link
b
0 and b1 are given in Table 11
The main difference between crash rates on four-lane divided rural roads and four lane motorways is the presence
of at-grade intersections and accesses and on some routes cyclists. In New Zealand the mid-block crash rates for
motorways and four lane divided roads are similar. Hence a single crash prediction model for mid-blocks can be
used for both. When assessing four-lane divided roads additional analysis is required to predict the crash risk
associated with at-grade intersection and accesses (using intersection models) and bicycles.
An analysis of crash rates on motorways and four-lane divided roads indicates that the crash rate typically varies
between 3 and 11 crashes per 100 million vehicle kilometres, with most being under 9. The exception is on 6+ lane
motorways and motorway sections with steep grades (often with climbing lanes), where in some cases the rates
exceed 11.
Table 11 shows the model parameters. The b1 value is much greater than 1 indicating that the rate of injury rates
per vehicle increases as traffic volumes (and number of lanes) increase. This explains the higher rates found on
motorways with more than four lanes, including the addition of climbing lanes. A similar result has been found in
a number of other countries. This increase is likely to be due to an increase in lane changing and also traffic
congestion in peak periods on the higher volumes motorway sections.
Table 12 shows the range of one-way flows over which the crash prediction models should be applied and the k
values for use in the weighted crash procedure.
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 23
Table 11: Four-lane divided rural roads coefficients
b0 b1
Motorway and four-lane divided roads. 2.56 × 10-7 1.45
Table 12: Four-lane divided rural roads k values
Flow range AADT k value
Motorway and four-lane divided roads. 15,000 – 68,000 10.2
Motorway link crash prediction models are also available by crash type in Turner (2001). New Zealand crash
prediction models are not currently available for motorway interchanges and other grade-separated intersections.
Interchange models are available for a variety of different interchange layouts, including motorway to motorway
links, in the USA. The USA interchange models are included in the ISAT software that is available through the
Federal Highway Authority (FHWA).
Some calibration of the ISAT models has been done for several interchanges. The calibration shows that these
models work well for the Auckland motorway network (the USA predictions being a little higher), but less so for
other grade separated intersections around New Zealand. It is preferable than using crash rates and models for
standard intersections and urban links within this manual. For the Auckland motorway a calibration factor of 0.85
(15% reduction) should be applied to ISAT urban motorway predictions (this factor is based on analysis
undertaken in the early 2010’s). We recommend caution when using ISAT outside of the Greater Auckland area.
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 24
6.0 Intersections – Product of Flow Models
Product of flow models use road link traffic volumes to estimate the number of crashes occurring at cross roads
and T-intersections. The typical models used are:
Injury crashes (priority and traffic signals) = b0 x Qmajorb1 x Qminorb2 x ∑CMFs
Injury crashes (roundabouts) = b0 x Qapproachb1 x ∑CMFs
∑Crash Modifying Factor (CMF) = CMF1 * CMF2 * … (e.g. lighting and splitter island)
Where: b0, b1 and b2 = model parameters
Qmajor = annual average daily two-way traffic volume on highest
volume road (signals) or priority road
Qminor = annual average daily two-way traffic volume on lowest volume
road (signal) or side-road
Qapproach = annual average daily two-way traffic volume on each
roundabout approach
L = length (km).
6.1 Urban Priority and Signalised Cross roads and T-junctions 50-70km/h
The ‘general’ model is suitable for most urban cross roads (four leg) and T-junctions (three leg) types and uses
two-way link volumes where the posted speed limit is 50–70km/h. Where a breakdown by crash type and road
user type is required, or where the proportion of turning vehicles is high compared with through vehicles, then the
appropriate conflicting flow models (in Section 7) should be used.
For urban intersections on the primary road network (excluding roundabouts), the typical crash rate (reported
injury crashes per year) is calculated using:
AT = b0 × Qmajorb1 × Qminor/sideb2
where: Qmajor is the highest two-way link volume (AADT) for cross roads and the primary road volume
for T-junctions.
Qminor/side is the lowest of the daily two-way link volumes (AADT) for cross roads and the side
road flow for T-junctions
b
0, b1 and b2 are given in Table 13.
Table 14 shows the range of flows over which the crash prediction models should be applied. The k values are for
use in the weighted crash procedure.
Caution should be exercised when using the prediction models for intersections where opposing approach flows
(on Qmajor or Qminor) differ by more than 25%. In such cases, the conflicting flow models in Section 7 should be
used.
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 25
Table 13: General cross-road and T-junction urban intersections (50-70km/h) coefficients (reference 21)
Intersection type b0 b1 b2
Uncontrolled – T 2.19 × 10-3 0.36 0.19
Priority – cross 1.08 × 10-3 0.21 0.51
Priority – T 4.89 × 10-5 0.76 0.20
Traffic signals – cross 2.81 × 10-3 0.46 0.14
Traffic signals – T 1.31 × 10-1 0.04 0.12
Table 14: General cross-road and T-urban intersections 50-70km/h k values
Intersection type Range Qmajor AADT Range Qminor AADT k value
Uncontrolled – T 3000 – 30,000 500 – 4,000 2.6
Priority – cross 5000 – 22,000 1500 – 7000 2.3
Priority – T 5000 – 26,000 1000 – 5000 3.8
Traffic signals – cross 10,000 – 32,000 5000 – 16,000 4.8
Traffic signals – T 11,000 – 34,000 2000 – 9000 4.6
6.2 Urban Roundabouts 50-70 km/h
Often roundabouts do not have the roads with the highest or lowest volumes on opposing arms, or if they have
three arms these are seldom in a ‘T’. Therefore, crashes are calculated for each arm of the roundabout, and the
total obtained by adding these together. The typical crash rate (reported injury crashes per approach per year) is
calculated using the model:
AT = b0 × Qapproachb1
where: Qapproach is the two-way link volume (AADT) on the approach being examined.
b
0, and b1 are given in Table 15.
This model can be applied for roundabouts with three, four or five approaches. Table 16 shows the range of flows
over which the crash prediction model should be applied. The k values are for use in the weighted crash
procedure.
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 26
Table 15: General urban roundabouts 50-70km/h coefficients (reference 5)
Number of entry lanes per approach Single Multiple
b0 b1 b0 b1
Roundabout 4.81 × 10-4 0.58 7.95 × 10-4 0.58
Table 16: General urban roundabouts 50-70km/h k values
Number of entry lanes per approach Single Multiple
Flow range
AADT
k value Flow range
AADT
k value
Roundabout 170 – 25,000 2.2 800 – 42,000 2.2
6.3 High-speed (Rural) Priority and Signalised Cross roads and T-junctions
(≥ 80km/h on main road)
The ‘general’ model is suitable for most high-speed (rural) cross roads and T-junctions and use two-way link
volumes. High speed intersections are those where the speed limit on the main road is 80km/h or greater. The
side-road can be any speed limit. Where a breakdown of crashes by crash and road user type is required, or
where the proportion of turning vehicles is high compared with through vehicles then conflicting flow models in
Section 7 should be used.
For high-speed cross roads and T-junctions, the typical crash rate (reported injury crashes per year) is calculated
using the model:
AT = b0 × Qmajorb1 × Qminor/sideb2
where: Qmajor is the highest two-way link volume (AADT) for cross roads and the primary road volume for
T-junctions.
Q
minor/side is the lowest of the daily two-way link volumes (AADT) for cross roads and the side road
flow for T-junctions.
b
0, b1 and b2 are given in
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 27
Table 17.
Table 18 shows the range of flows over which the crash prediction models should be applied. The k values are for
use in the weighted crash procedure.
Caution should be exercised when using the prediction models for intersections where opposing approach flows
(on Qmajor or Qminor) differ by more than 25%. In such cases, the conflicting flow models in Section 7 should be
used.
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 28
Table 17: General high-speed cross roads and T-junctions ≥ 80km/h coefficients (reference 8)
Intersection type b0 b1 b2
Priority – cross 3.74 × 10-4 0.39 0.50
Priority – T 3.52 × 10-4 0.18 0.57
Traffic signals – cross 3.15 × 10-4 0.52 0.19
Traffic signals – T 4.41 × 10-2 0.37 -0.10
Table 18: General high-speed cross and T-intersections ≥ 80km/h k values
Intersection type Range Qmajor AADT Range Qminor AADT k value
Priority – cross 50 – 24,000 50 – 3500 2.6
Priority – T 50 – 26,000 50–- 9000 4.7
Traffic signals – cross 19,000 – 46,000 11,000 – 20,000 4.7
Traffic signals – T 10,000 – 54,000 1700 – 17,000 2.0
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 29
6.4 High-speed (Rural) Roundabouts (≥ 80km/h on main road)
Often roundabouts do not have the roads with the highest or lowest volumes on opposing arms, or if they have
three arms these are seldom in a ‘T’. Therefore, crashes are calculated for each arm of the roundabout, and the
total obtained by adding these together. The typical crash rate (reported injury crashes per approach per year) is
calculated using the model:
AT = b0 × Qapproachb1
where: Qapproach is the two-way link volume (AADT) on the approach being examined.
b
0, and b1 are given in Table 19.
This model can be applied for roundabouts with three or four approaches. Table 20 shows the range of flows over
which the crash prediction model should be applied. The k values are for use in the weighted crash procedure.
Table 19: High-speed roundabout coefficients (reference 8)
b0 b1
Roundabout 4.33 × 10-4 0.53
Table 20: High-speed roundabout k values
Flow range AADT k value
Roundabout 800 – 29,000 2.1
6.5 Urban and Rural Railway Crossings
For urban and rural railway crossings, the typical crash rate (reported injury hit train and rear-end crashes per
year) is calculated using the model:
AT = b0 × Tb1 × QTb2
where: T is the number of trains per day
Q
T is the daily two-way traffic volume (AADT)
b
0, b1 and b2 are given in
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 30
Table 21
Table 22 shows the range of traffic volumes and trains over which the crash prediction models should be applied.
The k values are for use in the weighted crash procedure.
A large number of railway crossings are located in close proximity to low design speed curves. Low design speed
approach curves are often caused by the route having to deviate sharply when crossing the railway line. In such
circumstances separate predictions of the typical crash rates on these approach curves need to be made using the
model for rural isolated curves (≥ 80km/h). Analysts should be aware that the combined crash rate for both the
railway crossing and approach curves may be different than the sum of the two element predictions. In such
cases the weighted crash analysis procedure can be useful as the actual crash history is also used in the
calculation of the crash rate.
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 31
Table 21: Urban and rural railway crossings coefficients
Control type b0 b1 b2
Half-arm barriers 4.18 ×10-4 0.27 0.33
Flashing lamps and bells 6.22 ×10-4 0.61 0.32
No control 1.44 ×10-3 0.31 0.36
Table 22: Urban and rural railway crossings k values
Control type
Traffic volumes
k value
QT AADT Trains AADT
Half-arm barriers <13,000 <40 1.8
Flashing lamps and bells <6000 <30 0.7
No control <1000 <20 2.7
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 32
7.0 Intersections - Conflicting Flow Models
Conflicting flow models provides a breakdown of the predicted crashes by road user type (e.g. pedestrian and
cyclists) and crash type (refer to Appendix A). Crash type models are usually only available for the major crash
types at each intersection. The total number of injury crashes at an intersection is calculated by adding up the
crashes by each type and approach and then using either a general/other crash prediction model or a factor to
take into account the crashes not modelled.
Conflicting flow models are typically used in analysis when there are a high proportion of vehicles making turning
movements, especially right turns and when treatments impact on particular crash types or crash modes.
Examples of the latter include installing a right turn bay at a rural priority intersection and right turn signal phasing
at urban traffic signals.
There is no general model form for conflicting flow models. Some include only flows while others have many
other variables. The sections that follow demonstrate the models that are available for each intersection type.
7.1 Urban signalised crossroads <80km/h
There have been several research studies in New Zealand that have developed crash prediction models for urban
traffic signals. This varies from very basic product-of-flow models (as in Section 6) through to detailed models
with a large number of variables for each road user type, by city (across New Zealand) and by time of day (e.g.
morning and evening peaks). In this case ‘national models’ by key crash type for each transport model (motor-
vehicles, pedestrians and cyclists) have been presented. For more detailed analysis by city type, day of week or
for more complex intersections it is recommended that analysts utilise the models provided in the various
research studies of traffic signals listed in the reference section (in particular reference 18).
The conflicting flow models for signalised crossroads are suitable for situations where a breakdown of crashes by
crash and road user type is required, or where the proportion of turning vehicles is high compared with through
vehicles. For urban (speed limit <80km/h) signalised crossroads on the primary road network the typical crash
rates can be calculated for the six crash types (13, 19) in Table 23. The number of reported injury crashes per year
for each crash type on each approach can be calculated using the models in
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 33
Table 24. These models calculate the number of crashes per approach and therefore must be used for each
approach to the intersection for which the crash type can occur (e.g. at signalised cross roads the crossing (HA)
and right-turn-against (LB) crash types shown can occur on all four approaches).
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 34
Table 23: Urban signalised cross roads (<80km/h) variables and CAS movement categories
Crash types Variables CAS movement
categories
Crossing
(no turns, motor vehicle
only)
HA
Right turn against
(motor-vehicle only)
LA, LB
Others
(motor-vehicle only)
-
Pedestrian versus motor
vehicle
NA-NO, PA-PO
Right turn against
(cyclist travelling
through)
LA, LB
Others
(cyclist versus motor
vehicle)
-
q2/11 = Through vehicle flows in
veh/da
y
q2 = Through vehicle flow in veh/day
q7 = Right-turning vehicle flow in
veh/day
Qe = Entering vehicle flow in veh/day
Qe = Entering vehicle flow in veh/day
P = Pedestrian crossing volume in
ped/day
q7 = Right-turning vehicle flow in
veh/day
c2 = Through cycle flow in cyc/day
Qe = Entering vehicle flow in veh/day
Ce = Entering cycle flow in cyc/day
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 35
Table 24: Urban signalised crossroads (<80km/h) crash prediction models (reference 6 and 16)
Crash types Model k value
Crossing
(no turns, motor vehicle only) AT = 9.17 × 10-5 × q20.36 × q110.38 1.1
Right turn against
(motor vehicle only) AT = 5.61 × 10-5 × q20.49 × q70.42 1.9
Others
(motor vehicle only) AT = 2.12 × 10-4 × Qe0.59 5.9
Pedestrian versus motor vehicle AT = 2.79 × 10-2 × Qe-0.05 × P 0.03 1.4
Right turn against
(cyclist travelling through) AT = 3.01 × 10-4 × q70.34 × c20.20 1.3
Others
(cyclist versus motor vehicle) AT = 1.23 × 10-3 × Qe0.28 × Ce0.03 1.1
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 36
7.2 Urban roundabouts (<80km/h)
The conflicting flow models for roundabouts are suitable for situations where a breakdown of crashes by crash
and road user type is required, such as roundabouts with high proportions of cyclists. For urban (speed limit
<80km/h) roundabouts on the primary road network the typical crash rates can be calculated for the seven crash
types (15) in Table 25. The number of reported injury crashes per year for each crash type on each approach can
be calculated using the models in
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 37
Table 26. These models calculate the number of crashes per approach and therefore must be applied at all
approaches to the roundabout.
Table 25: Urban roundabouts (<80km/h) variables and CAS movement categories
Crash types Variables CAS movement
categories
Entering-vs-circulating
(motor-vehicle only)
HA, JA-JO KA-KO, LA-
LO
Rear-end
(motor-vehicle only)
FA-FO, GA, GD
Loss-of-control
(motor-vehicle only) CA-CO, DA-DO, AD, AF
Other
(motor-vehicle only)
-
Pedestrian
NA-NO, PA-PO
Entering-vs-circulating
(cyclist circulating)
HA, JA-JO KA-KO, LA-
LO
Other (cyclist)
-
Qe = Entering vehicle flow in veh/day
Qc = Circulating vehicle flow in
cyc/day
Sc = Mean free speed of circulating
vehicles
Qe = Entering vehicle flow in veh/day
Qe = Entering vehicle flow in veh/day
V10 = Visibility 10 metres back from the limit
line to vehicles on the approach to the right
Qe = Entering vehicle flow in veh/day
Qe = Entering vehicle flow in veh/day
P = Pedestrian crossing volume in
ped/day
Qe = Entering vehicle flow in veh/day
Cc = Circulating cycle flow in cyc/day
Se = Mean free speed of entering
vehicles
Qe = Entering vehicle flow in veh/day
Ce = Entering cycle flow in cyc/day
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 38
Table 26: Urban roundabouts (<80km/h) crash prediction models (reference 5)
Crash types Model k value
Entering-vs-circulating
(motor-vehicle only) AT = 6.12 × 10-8 × Qe0.47 × Qc0.26 × Sc2.13 1.3
Rear-end (motor-vehicle only) AT= 9.63 × 10-2 × Qe-0.38 × e0.00024 × Qe 0.7
Loss-of-control
(motor-vehicle only) AT = 6.36 × 10-6 × Qe0.59 × V100.68 3.9
Other (motor-vehicle only)
AT = 1.34 × 10-5 × Qe0.71 × ΦMEL
ΦMEL = 2.66 (if multiple entry lanes)
ΦMEL = 1.00 (if single entry lane)
-
Pedestrian AT= 3.14 × 10-4 × P 0.60 × e0.000067 × Qe 1.0
Entering-vs-circulating
(cyclist circulating) AT= 3.88 × 10-5 × Qe0.43 × Cc0.38 × Se0.49 1.2
Other (cyclist) AT = 2.07 × 10-7 × Qe1.04 × Ce0.23 -
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 39
7.3 Urban Priority T-junctions (<80km/h on main road)
The conflicting flow models for priority T-junctions in urban areas are suitable for situations where a breakdown
of crashes by major crash type is required. Currently crash models are only available for the two main crash types
which are 1) Crossing vehicle turning (JA crashes) and 2) Right turn against (LB crashes). The predictions from
these models should be treated with caution until further research explores in more detail the new design
variables introduced in the design index. The models are provided in
Table 27 with parameters in Table 28.
Table 27: Urban priority T-junctions (<80km/h on main road) variables
Crash types Variables
CAS
movement
categories
Crossing – vehicle turning
(major road approach to right of
side road)
JA
Right turn against
(motor-vehicle only) – LA, LB
q5 = Through vehicle flow along major road to right of minor
road vehicles in veh/day
q1 = Right-turning flow from minor road in veh/day
MRSL = main road (through road) speed limit
DI = Design Index, as defined in Table 28
q4 = Through vehicle flow in veh/day
q3 = Right-turning vehicle flow in veh/day
MRSL = main road (through road) speed limit
DI = Design Index, as defined in Table 28
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 40
Table 28: Urban priority T-junction (<80km/h on main road) models
Where the variables in the two design indices (DI) are as follows:
RTBTL – Right turn bay taper length (in metres).
MRMW – Main Road Median width. Equals 1 for painted line, 2 when median <0.5m, 3 when between
0.5 and 1m, 4 when between 1 and 2m and 5 when >2m.
NSNTL – Near side number of through lanes. Equals 1 for one lane or 2
for two lanes.
DFSUF – Distance to far side upstream feature (to left of side-road). Equals 1 when distance is 0-49m, 2
when 50 to 99m, 3 when 100m to 199m and 4 when 200m plus.
SRNL – Side road number of lanes. Equals 1 for left turn & right turn, 2 for left-right stacked side by side
in single lane and 3 for combined left and right in one lane.
SRMW – Side Road Median width. Equals 1 if no centreline, 2 if painted line, 3 if <0.5m width, 4 if
between 0.5 and 1m, 5 if between 1 and 2m and 6 if >2m.
GMRRS – Gradient of main road, right side. Equals 1 if flat, 3 if moderate and 5 if steep.
UMIT – Upstream median island type. Equals 1 for painted line, 2 for hit posts, 3 for solid barrier, 4 for
painted island and 5 for solid island.
WAL – Width of acceleration lane (in metres).
CP – Car parking. Equals 1 for none, 2 for one of three sides, 3 for two of three sides and 4 for three (or
all) of three sides.
DNSUF – Distance to near side upstream feature (to right of side-road). Equals 1 when distance is 0-
49m, 2 when 50 to 99m, 3 when 100m to 199m and 4 when 200m plus.
SRMI – Side road median island. Equals 1 when present, 2 when not present.
SL – Street lighting. Equals 1 when none, 2 when one at the top of T-Junction, 3 when one at the side of
approach road and 4 when full.
TTCB – Top of T-junction chevron board. Equals 1 when present, 2 when not present.
UMIW – Upstream median island width. Equals 1 when <0.5m, 2 when 0.5m-1m, 3 when 1m-2m and 4
when >2m.
WDL – Width distraction to left. Equals 2 when none present and 4 when present (e.g. bus stop).
TMRW – Total main road width (in metres).
Crash types Model k value
Crossing – Vehicle turning
(major road approach to right of side road)
AT= 1.96 × 10-17 × q10.025× q50.13 × MRSL3.80× DI5.8
DI = (0.88*RTBTL+6.49*(6-MRMW)+17.86*NSNTL
+ 1.50*(19-4*DFSUF)+30.30*(7-
2*SRNL)+1.41*(4*SRMW+1)+7.69*(2*GMRRS-
1)+18.52*(6-UMIT)+1.53*(19-4*WAL)+2.15*(19-
4*CP))/10
50
Right Turn Against
AT= 3.35 × q30.40× q50.21 × MRSL-4.53× DI3.07
DI = (2.11*(4*DNSUF-1)+11.98*(3-
SRMI)+15.87*SRMW+2.14*(4*SL-
1)+24.69*TTCB+9.00*(4*UMIW-
1)+8.55*WDL+0.88*TMRW)/8
50
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 41
Seagull Layouts
Crash Prediction Models are also available for priority tees with seagull shaped intersections (as shown below)
with either painted or raised islands. The Seagull models include LTSLs. The performance of Seagull layouts
depends on the design of the intersection. Research indicates that well designed Seagulls have a good safety
record.
Seagull intersection treatments are rarely used in New Zealand in part due to poor road safety experience at a
number of such intersections in the past. There are however locations where seagulls are an ideal treatment in
terms of improving efficiency and due to their relatively low construction costs, compared to other options. They
are popular in urban areas where site constraints do not permit a roundabout or traffic signals to be built. Recent
research has indicated that seagulls can be safer in some situations than traditional T-intersections, but only if
designed correctly.
Some of the Key design factors that need to be avoided:
A. Locating such intersections on moderate to sharp bends or on crests and dips especially when it is
difficult for drivers to read the intersection layout.
B. Four or more lanes with high traffic flows, due to difficulties picking a gap in traffic and there is two lanes
to cross before the safety of the central median.
C. Where the speed limit is high (greater than 60km/h) and/or the right turn out movement is high.
D. On wide median roads where the right-turn-in lane is between a 15 and 45 degree angle to the through
lane. It should be as close as possible to parallel to the through lane, as occurs at traditional painted
right turn lanes.
E. There are nearby intersections (including railway crossings and pedestrian crossings), major accesses,
parking and busy bus stops or other distractions that may divert drivers attention.
The research indicates that well-designed seagull intersections may perform better than standard, non-chanelised
T-intersections (note the long list of variables that can impact on the crash rate). A ‘Beta version’ spreadsheet
calculator has been developed for assessing urban and rural seagull intersections. This is available through the
Transport Agency.
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 42
7.4 High Speed Priority Cross Roads (≥ 80km/h on main road)
The conflicting flow models for priority crossroads in high-speed areas are suitable for situations where a
breakdown of crashes by crash type is required, or where the proportion of turning vehicles is high compared with
through vehicles. For high-speed (speed limit ≥ 80km/h on main road) priority cross roads on two-lane, two-way
roads the typical crash rates can be calculated for the five crash types in Table 29. The number of reported injury
crashes per year for each crash type is calculated in Table 30: These models calculate the number of crashes per
approach for both ‘major road’ and ‘minor road’, with the minor road being the road with stop or give way control.
Table 29: High speed priority cross roads (≥ 80km/h on main road) variables
Crash types Variables CAS movement
categories
Crossing – hit from right
(major road approaches only)
HA
Crossing – hit from right
(minor road approaches only)
HA
Right turning and following vehicle
(major road approaches only)
GC, GD, GE
Other
(major road approaches only)
-
Other
(minor road approaches only)
-
q2/5 = Through vehicle flows in
veh/day
Qe = Entering vehicle flow on
major road in veh/day
q2/11 = Through vehicle flows in
veh/da
y
q5 = Through vehicle flow along
major road in veh/day
q4 = Right-turning flow from major
road in veh/day
Qe = Entering vehicle flow on
minor road in veh/day
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 43
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 44
Table 30: High speed priority cross roads (≥ 80km/h on main road) models (reference 8)
Crash types Model k value
Crossing – hit from right
(major road approaches only) AT = 1.2 × 10-4 × q20.60 × q50.40 0.9
Crossing – hit from right
(minor road approaches only) AT = 2.05 × 10-4 × q20.40 × q110.44 2.0
Right turning and following vehicle
(major road approaches only)
AT = 1.08 × 10-6 × q40.36 × q51.08× ΦRTB
ΦRTB = 0.22 (if right-turn bay present)
ΦRTB = 1.00 (if right-turn bay absent)
2.6
Other
(major road approaches only) AT = 1.14 × 10-4 × Qe(Major)0.76 1.1
Other
(minor road approaches only) AT = 3.44 × 10-3 × Qe(Minor)0.27 0.2
7.5 High-speed priority T-junctions (≥ 80km/h on main road)
The conflicting flow models for priority T-junctions in high-speed areas are suitable for situations where a
breakdown of crashes by crash type is required, where one turning movement from the side road is greater than
the other, or where the intersection has a visibility and other design deficiencies. For high-speed (speed limit
80km/h on main road) priority T-junctions on two-lane and four-lane, two-way roads the typical crash rates can
be calculated for the five crash types in Table 31.
The typical crash rate (number of reported injury crashes) per year for each crash type is calculated using Table
32. Two models are provided for the first crash type, crossing – vehicle turning. The first model (which includes
measured approach speed, rather than speed limit, and multiple design factors) is the preferred model, as it was
developed more recently. The second model has been included for situations were visibility may be an issue and
where approach speed and many of the design variables are not available. Unlike models for other intersections,
these models are each for a specific approach.
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 45
Table 31: High speed priority T-junctions (≥ 80km/h on main road) variables
Crash types Variables
CAS
movement
categories
Crossing – vehicle turning
(major road approach to right of
side road)
JA
Right-turning and following
vehicle
(major road approach to left of
side road)
GC, GD, GE
q5 = Through vehicle flow along major road
to right of minor road vehicles in veh/day
q1 = Right-turning flow from minor road in
veh/day
VD = Sum of visibility deficiency in both
directions when compared with Austroads
SISD (3). Note: if there is no visibility
deficiency then a default value of 1 should
be used for VD
MRAS = main road (through road)
approach speed (measured)
DI = Design Index, as defined in Table 32
q5 = Through vehicle flow along major road
to right of minor road vehicles in veh/day
q3 = Right-turning flow from major road in
veh/day
SL = Mean free speed of vehicles
approaching from the left of vehicles minor
road
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 46
Other
(major road approach to left of
side road)
-
Other
(major road approach to right of
side road)
-
Other
(side road approach)
-
q5 = Through vehicle flow along major road
to right of minor road vehicles in veh/day
q3 = Right-turning flow from major road in
veh/day
q5 = Through vehicle flow along major road
to left of minor road vehicles in veh/day
q6 = Left-turning flow from major road in
veh/day
q1 = Right-turning flow from minor major
road in veh/day
q2 = Left-turning flow from minor road in
veh/day
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 47
Table 32: High speed priority T-junction ( 80km/h on main road) models (reference 8)
Where the variables in the design index (DI) are as follows:
RTB – Right turn bay. Equals 1 if Yes and 2 if No.
LWRTMR – Lane width of right turn from main road (in metres).
RTBS – Right turn bay stacking. Equals number of vehicles, assuming one vehicle = 6m.
MRMW – Main road median width. Equals 0 when none, 1 when a painted line, 2 when <0.5m, 3 when
0.5m to 1m, 4 when 1m to 2m and 5 when >2m.
PNSUF – presence of a near-side (side-road side) upstream feature (to right of intersection). Equals +1 if
Yes and -1 if No.
RTAVLL – Right approach visibility two metres from limit line (in metres).
Crash types Model k value
Model 1 for Crossing – Vehicle turning
(major road approach to right of side road)
AT= 6.46 × 10-14 × q10.51× q50.27 × MR
AS3.97× DI1.58
DI = (34.48*(6-2*RTB)+90.91*(2*LWRTMR-
3)+22.32*RTBS+20*(4-
2*MRMW)+45.45*(PNSUF + 3)+11.49*(17/3-
4*RTAVLL))/6
50
Model 2 for Crossing – Vehicle turning
(major road approach to right of side road) AT= 4.39 × 10-6 × q11.33× q50.15 ×VD0.33 8.1
Right-turning and following vehicle
(major road approach to left of side road) AT = 4.39 × 10-27 × q30.46× q40.67 ×SL11 0.2
Other
(major road approach to right of side road) AT = 1.32 × 10-5 × (q5 + q6)0.91 1.0
Other
(major road approach to left of side road) AT = 2.48 × 10-4 × (q3 + q4)0.51 3.0
Other
(side road approach) AT = 1.22 × 10-2 × (q1 + q2)-0.02 0.6
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 48
Left Turn Slip Lanes and Seagull Layouts
Crash Prediction Models are also available for priority tees with a left turn slip lane (LTSL) into the side-road and
for seagull shaped intersections (as shown below) with either painted or raised islands. The Seagull models
include LTSLs. In rural and high speed areas we recommend use of raised seagull islands. The performance of
LTSLs and Seagulls layout depends on the design of the intersection. Research indicates that in some situations
well designed LTSL and Seagulls have a good safety record.
Left Turn Slip Lanes (LTSLs)
LTSL are commonly used to improve the efficiency of priority controlled intersections, by providing an area/lane
of various dimensions for vehicles to decelerate within when turning left into a side-road. While they may reduce
the likelihood of relatively rare rear-end crashes involving through and left turning traffic some designs do appear
to increase the risk of the more severe and common crash type involving vehicles turning right out of the side-
road being hit by through vehicles from there right (‘JA’ crashes). Problems occur when the left turners block the
visibility to following through vehicles on the through lane(s). The location of the side-road limit line, and hence
location of driver, the volume of left turners and the design of the LTSL has an impact on these crashes. The crash
risk can vary by time of day depending on the various turning movement volumes.
Best practice is to either 1) start the left turn lane early and provide a painted or raised island that create adequate
separation of through and left turn lanes so that right turn out drivers can clearly see the through traffic or 2)
provided a short left turning area close to the intersection such that through vehicles are unable to overtake left
turners (see figures below). For 2 through lanes we recommend use of option 1 only. At well-designed
intersections research indicates that crash reduction of 50% or more can be achieved for LTSLs. A ‘Beta version’
spreadsheet calculator has been developed for assessing rural LTSLs. This is available through the Transport
Agency.
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 49
Example of LTSL separation (a) and late LTSL (b)
Seagull (Chanelised) Treatments
Seagull intersection treatments are rarely used in New Zealand in part due to poor road safety experience at a
number of such intersections in the past. There are however locations where seagulls are an ideal treatment in
terms of improving efficiency and due to their relatively low construction costs, compared to other options. For
example, they are popular on higher speed two to four-lane divided highways where side-road volumes are low.
Recent research has indicated that seagulls can be safer in some situations than traditional T-intersections, but
only if designed correctly.
Key design factors that need to be avoided:
F. Locating such intersections on moderate to sharp bends or on crests and dips especially when it is
difficult for drivers to read the intersection layout.
G. Four or more lane roads where the left turning vehicles on main road obstruct the visibility for right turn
out drivers of through vehicles. This can be addressed by a suitable LTSL design.
H. Where the right turn out movement is high (greater than 400 vehicles per day)
I. On wide median roads where the right-turn-in lane is between a 15 and 45 degree angle to the through
lane. It should be as close as possible to parallel to the through lane, as occurs at traditional painted
right turn lanes.
J. There are nearby intersections or other distractions (e.g. commercial land-use) that may divert drivers
attention.
The research indicates that well-designed seagull intersections may perform better than standard, non-chanelised
T-intersections. A ‘Beta version’ spreadsheet calculator has been developed for assessing urban and rural seagull
intersections. This is available through the Transport Agency.
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 50
8.0 Crash modification factors
8.1 Introduction
The following section provides average crash modification factors for treatments or improvements in urban and
rural areas. These modifications can be applied to the crashes and crash rate calculated using any of the three
crash analysis methods. Key references for CMF and CRF include Austroads (7), and The Handbook of Road
Safety Measures (2). Before and after New Zealand studies of treatments have also been included. Other
international sources of CMFs and CRFs include the Highway Safety Manual (22), and the CMF Cleaning house
(23)
A CMF’s / CRF’s typical area of influence of intersections extends 50 metres along each leg, and similarly an area
of influence 50 metres from either side of a bridge and railway crossing should generally be adopted. However,
analysts are cautioned that at some sites the area of influence can be affected by vehicle speeds, and road
geometry. Judgement is also required to assess when the effect of a CMF may extend beyond the area of
treatment (for example passing lanes). In rural areas, crash migration should also be considered; this issue is
explained in more detail within Appendix A6.
The modification factors are only a guide to possible modification rates and the evaluation documentation will
need to substantiate all claimed crash modifications, particularly if they are expected to be greater than indicated
here.
Relative confidence level categories of low, medium and high have been assigned to each treatment. The
confidence level is based upon the level, location, date, and type of research available to corroborate the
CMF/CRF. A low level of confidence may also indicate that the benefit can range significantly depending on the
environment in which it is applied. We would recommend that users perform sensitivity analysis when there are
low levels of confidence in the CMF/CRF particularly when most of the project benefits are from such treatments.
In such circumstances the use of more localised research on the project location may also be valid.
8.2 Typical crash reductions
The following tables (Table 33 to
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 51
Table 38) provide a typical range of injury crash modification factors (CMFs) and crash reduction factors (CRFs)
for mid-block and intersection treatments. The tables are ordered to correspond with each crash model type;
rural mid-blocks and motorways, urban mid-blocks, and product of flow intersection models (urban and rural).
The crash modifying factors should be applied to total crash predictions for each intersection and mid-block
length. (where CMFs are available). CMFs cannot be used for specific crash types e.g. as predicted by conflicting
flow models) or for other crash subcategories (e.g. night crash). They are however provided for total pedestrian
and cyclist crash predictions where relevant.
For crash prediction models by conflict type key non-flow factors are usually included with crash prediction
models. Some treatments such as delineation are common to several site environments and are shown under
each model where commonly used. Treatments for cyclists and pedestrians transcend all models and are shown
separately.
When there is more than one measure the CMFs should be multiplied together.
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 52
CRFs are provided by crash type and crash sub-category (e.g. night) where the treatment impacts on a specific
crash type. CRFs are provided for crash-by-crash analysis.
When using multiple CRFs for each crash type it is not appropriate to add all of the reduction factors together. In
these cases judgement should be exercised in determining the likely overall effectiveness of multiple measures on
each crash type.
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 53
Table 33: Common rural midblock crash reduction/modification factors
Common rural midblock crash reduction/modification factors
Treatment Sub type Crash Reduction
Factor
Crash
Modification
Factor
Confidenc
e
Comment
Install overtaking lanes
25% All crashes 0.75 Low Reduce these crashes linearly to zero for crashes following the passing lane up
to 5km away. Ensure loss of control crashes do not increase due to design.
50% of head-on
crashes
N/A Low
30% of overtaking
crashes.
N/A Low
Install no overtaking
markings
35% All crashes 0.65 Medium Where no-overtaking lines missing and are required due to poor visibility
50% of head-on
crashes
N/A Medium
40% of overtaking
crashes
N/A Medium
Install edge-line 10% 0.9 Low
Install centreline 20% 0.8 Low
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 54
Common rural midblock crash reduction/modification factors
Treatment Sub type Crash Reduction
Factor
Crash
Modification
Factor
Confidenc
e
Comment
Install rural wide centreline
(NEW - described in
HRRRG)
20% of all injury
crashes
0.80
Low
Wide centrelines are particularly effective at reducing deaths and serious
injuries and head-on and run-off road crashes where traffic volumes are
greater than 14,000 vpd. Care should be taken applying this treatment at
locations with high numbers of intersection and ‘other’ crash types and where
volumes are less than 14,000 vpd.
40% of cross
centreline crashes
N/A Low
Edge-line and centreline
combination (NEW)
30% 0.7 Low
Painted speed limits
(NEW)
0% 1 Low A 0% crash reduction factor is allocated based on conflicting overseas
research, and the lack of effect detected in the Australasian context.
Provide traverse rumble
strips (NEW)
25% 0.75 Low Traverse rumble strips are rarely used in New Zealand. They are only
applicable in a few locations. Before trialing this measure please contact the
NZ Transport Agency.
Install edge marker posts
5% of all injury
crashes
0.95 Low Edge marker posts are more effective on curves than on straight sections of
road. They are normally applied at the same time or after the installation of
centrelines and edge-lines.
40% of loss-of
control on curve
crashes
N/A Low
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 55
Common rural midblock crash reduction/modification factors
Treatment Sub type Crash Reduction
Factor
Crash
Modification
Factor
Confidenc
e
Comment
Install raised reflective
pavement markings
(RRPMs)
All 5% 0.95 Low This reduction applies to centre-line RRPMs. CRFs are not currently available
for shoulder RRPM.
Install audio-tactile profiled
line markings
Profile edge line 20% of all crashes 0.8 Medium An increase in bicycle and motorcycle crashes may occur when these users
are prevalent in the subject area.
30% of run-off-
road crashes
N/A Low
Profile centre
line
15% of all crashes 0.85 Medium
30% of head-on
crashes
N/A Low
Resurfacing of curves
Various
Compare injury crash rate at site with typical crash rate and injury crash rates
at other local sites that are considered satisfactory.
Consistent super-elevation
on a curve (NEW)
40% 0.6 Low When super-elevation is very inconsistent on a curve.
Sealing unsealed shoulders
(NEW)
30% 0.7 High Factors are based on typical shoulder widths of greater than 0.75m.
Consideration must be given to the impact of increased vehicle speeds that
may result and mitigate effects. Widening is likely to be more effective on
curves than on straights.
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 56
Common rural midblock crash reduction/modification factors
Treatment Sub type Crash Reduction
Factor
Crash
Modification
Factor
Confidenc
e
Comment
Sealing gravel road (NEW)
0% 1.0 Low Can cause an increase in crashes where steep grades and out of context
curves are present, due to increased speeds. In such circumstances road
improvements are needed to mitigate such hazards (e.g. curve advisory
signage).
Install bridge signs (NEW)
30% of crashes
associated with
bridges
N/A Low
Install chevron signs on
horizontal curves (NEW)
25% of curve
related crashes
only
N/A High
Speed cameras (NEW) Mobile overt 40% 0.6 Medium Where speeding is identified as a problem.
Covert speed camera evaluations are typically conducted on an area-wide
basis so cannot be compared to overt evaluations which are conducted at or
near camera sites.
The effectiveness of speed cameras is related to how frequently they are
implemented.
Mobile covert -
rural
20% 0.8 Medium
Fixed overt –
rural
30% 0.7 Low
Install w-section guardrail
(around roadside hazards)
30% of all injury
crashes
0.7
High This CMF only applies over isolated sections of guardrail. For continuous
guardrail refer to following CRFs and CMFs.
40% of all
fatalities
N/A
High
30% of all serious
injury crashes
N/A
High
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 57
Common rural midblock crash reduction/modification factors
Treatment Sub type Crash Reduction
Factor
Crash
Modification
Factor
Confidenc
e
Comment
10% of all minor
injury crashes
N/A
High
Install continuous
combined roadside and
median wire rope
improvements (NEW)
65% of all injury
crashes
0.35 Low
80%of all fatal
and serious injury
crashes
N/A
Low
Install continuous flexible
median barrier (NEW)
50% of all injury
crashes
0.5 Low
60% of all fatal
and serious injury
crashes
N/A
Low
90% of fatal and
serious head on
crashes
N/A
Low
Install continuous flexible
roadside barrier (NEW)
15% of all injury
crashes
0.85
Low
45% of run off
road injury
crashes
N/A
Low
65% of fatal and
serious injury run
off road crashes
N/A
Low
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 58
Common rural midblock crash reduction/modification factors
Treatment Sub type Crash Reduction
Factor
Crash
Modification
Factor
Confidenc
e
Comment
Install clear zones to 6
metres where there are
significant hazards
35% of loss-of-
control crashes
N/A Low In many situations roadside barriers (continuous or around hazards) are likely
to be more effective than clear-zones.
Install vehicle activated
signs (for example speed
activated warning signs)
(NEW)
All 35% N/A Medium Treatment is typically used near curves, bridges, schools, work-sites, speed
limit changes and intersections. Crash reduction applies to crashes
associated with site of treatment
Install route lighting two lane roads
(levels V1-V3)
two lane roads
(level V4)
15% of night-time
crashes
12% of night time
crashes
0.95 High
Medium
Crash reduction factor based on night crashes only. CMFs based on 32% of
crashes occurring at night. Where there is sufficient evidence (from the crash
history) that a site has a higher or lower proportion than this then a site
specific CMF should be developed.
CRFs for pedestrian crashes are higher than presented here (see Table 36).
Research indicates that lighting has very little effect on loss-of-control
crashes. Where the majority of crashes at a site are loss-of-control then the
installation of lighting will have a much lower crash benefit than indicated by
these factors.
Lighting luminance levels are as follows (refer to AS/NZ standard 1158.1.1 for
further details)
V1 >=1.5 cd/m2
V2 >=1.0 cd/m2
V3 >= 0.75 cd/m2
V4 >= 0.50 cd/m2
dual
carriageway
(levels V1-V3)
dual
carriageway
(level V4)
25% of night-time
crashes
20% of night-time
crashes
0.90 High
Medium
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 59
Common rural midblock crash reduction/modification factors
Treatment Sub type Crash Reduction
Factor
Crash
Modification
Factor
Confidenc
e
Comment
These factors can be used when upgrading lighting that is below category V4
(i.e. luminance of less than 0.50 cd/m2).
Table 34: Common urban midblock crash reduction/modification factors
Common Urban Midblock Crash Reduction/Modification Factors
Treatment Sub type Crash
Reduction
Factor
Crash
Modification
Factor
Confidenc
e
Comment
Medians Flush median 15% 0.85 Low
Solid median 45% 0.55 Medium
Parking ban (both
sides of the street)
Midblock 20% 0.8 Low Research indicates that banning parking on one side only may increase crashes.
Parking - convert
angle to parallel
(NEW)
All environments 40% 0.6 Low There is a lack of Australasian research on this treatment and there is a significant
discrepancy between the results. Hence, this is only an indication of the likely level
of crash reduction that could be expected from this treatment.
Road diet: Four
lanes to two lanes
plus flush median
All 35% 0.65 Low
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 60
Common Urban Midblock Crash Reduction/Modification Factors
Treatment Sub type Crash
Reduction
Factor
Crash
Modification
Factor
Confidenc
e
Comment
New route Lighting
New route lighting to:
-Subcategory V4
-Subcategory V3
-Subcategory V2 / V1
20%
30%
40% of night-
time crashes
0.95
0.91
0.88
High
Crash reduction factor based on night crashes only. CMFs based on 29% of
crashes occurring at night. Where there is sufficient evidence (from the crash
history) that a site has a higher or lower proportion than this then a site specific
CMF should be developed.
CRFs for pedestrian crashes are higher than presented here (see Table 36).
Research indicates that lighting has very little effect on loss-of-control crashes.
Where the majority of crashes at a site are loss-of-control then the installation of
lighting will have a much lower crash benefit than indicated by these factors.
Lighting luminance levels are as follows (refer to AS/NZ standard 1158.1.1 for
further details)
V1 >=1.5 cd/m2
V2 >=1.0 cd/m2
V3 >= 0.75 cd/m2
V4 >= 0.50 cd/m2
When upgrading lighting from one category to another (e.g. from V4 to V2) then
pro rata the factors provided. (e.g. upgrading from V4 to V2 gives a CRF of (1-
0.20) x 0.40 = 32%)
New lighting - railway
level crossing (NEW)
–V4 to V1
20% of night-
time crashes
N/A High
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 61
Common Urban Midblock Crash Reduction/Modification Factors
Treatment Sub type Crash
Reduction
Factor
Crash
Modification
Factor
Confidenc
e
Comment
Traffic calming All environments 20% 0.8 Medium Where available use CMFs and CRFs that are specific to each treatment used in
traffic calming.
Bus lanes (taxis
permitted)
All 25% increase 1.25 Low There is no Australasian research available on this treatment. This risk may be
mitigated by suitable design.
High occupancy
vehicle lanes
All 60% increase 1.60 Low There is no Australasian research available on this treatment. This risk may be
mitigated by suitable design.
Table 35: Common Motorway Crash Reduction/Modification Factors
Common Motorway Crash Reduction/Modification Factors
Treatment Sub type Crash Reduction
Factor
Crash
Modification
Factor
Confidence Comment
Install w-section guardrail
(around roadside hazards)
40% of all
fatalities
N/A
High These CRF only applies over isolated sections of guardrail. For
continuous guardrail refer to following CRFs and CMFs.
The factors where developed from primarily two-lane rural roads. If
motorway factors do become available then these should be used.
30% of all
serious injury
crashes
N/A
High
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 62
Common Motorway Crash Reduction/Modification Factors
Treatment Sub type Crash Reduction
Factor
Crash
Modification
Factor
Confidence Comment
10% of all minor
injury crashes
N/A
High
Install continuous combined
roadside and median wire
rope improvements (NEW)
65% of all injury
crashes
0.35 Low The factors where developed from primarily two-lane rural roads. If
motorway factors do become available then these should be used.
80%of all fatal
and serious injury
crashes
N/A
Low
Install continuous flexible
median barrier (NEW)
50% of all injury
crashes
0.5 Low The factors where developed from primarily two-lane rural roads. If
motorway factors do become available then these should be used.
60% of all fatal
and serious injury
crashes
N/A
Low
90% of fatal and
serious head on
crashes
N/A
Low
Install continuous flexible
roadside barrier (NEW)
15% of all injury
crashes
0.85
Low The factors where developed from primarily two-lane rural roads. If
motorway factors do become available then these should be used.
45% of run off
road injury
crashes
N/A
Low
65% of fatal and
serious injury run
off road crashes
N/A
Low
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 63
Common Motorway Crash Reduction/Modification Factors
Treatment Sub type Crash Reduction
Factor
Crash
Modification
Factor
Confidence Comment
Install impact attenuators
(NEW)
All 50% of all injury
crashes
N/A Medium Research on CRFs and CMFs included assessments of attenuators
located at tunnel portals, fixed objects, bridge pillars, and gore areas.
70% of fatal
crashes
N/A High
Street lighting (NEW)
New lighting –
motorway and
interchange to V3 level
or better
31% of night-time
injury crashes
0.91 High Crash reduction factor based on night crashes only. CMFs based on
30% of crashes occurring at night. Where there is sufficient
evidence (from the crash history) that a site has a higher or lower
proportion than this then a site specific CMF should be developed.
V3 lighting luminance level is >= 0.75 cd/m2 (refer to AS/NZ
standard 1158.1.1 for further details)
47% of night-
time fatal and
serious injury
crashes
N/A Medium
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 64
Table 36: Common intersection crash modification/reduction factors (urban and rural)
Common Intersection Crash Modification/Reduction Factors (Urban and Rural)
Treatment Sub type Crash Reduction
Factor
Crash Modification
Factor
Confidenc
e
Comment
Traffic Signals (urban). Install traffic signals Factors shall be determined using the priority,
roundabout and signal prediction models outlined
in ‘Section 6.0 Intersections – Product of Flow
Models’.
Research indicates that installation of traffic signals at three
leg intersections are less beneficial than four legged
intersections.
Linked / Coordinated
signals (urban) (NEW).
Linking existing
signals
15% 0.85 Medium
Signal visibility (urban) Replace a pedestal
mount with a mast
arm mount signal
(NEW)
35% per treated
approach
0.65 per approach Low This level of crash reduction will only occur at high volume
intersections, especially where there are high proportions of
trucks. Master arms are not normally used at lower volume
traffic signals (as they will have a reduced effect).
Increase lens size to
twelve inches (NEW)
5% per treated
approach
0.95 per approach Low Additional safety benefits may also be gained through the use
of LEDs to improve signal visibility especially in areas prone to
sunstrike.
Provide additional
signal head (NEW)
20% per treated
approach
0.8 per approach Medium Only applicable where the number of signal heads is below the
desirable
Install median (throat)
island on side-road
(rural)
35% per side-road
approach
0.65 per approach Medium Crash reduction likely to be higher at cross-roads than T-
junctions
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 65
Common Intersection Crash Modification/Reduction Factors (Urban and Rural)
Treatment Sub type Crash Reduction
Factor
Crash Modification
Factor
Confidenc
e
Comment
Install right-turn lane Install right-turn lane
– signalised
intersection (urban)
(NEW)
30% per approach 0.7 per approach Medium
Install right-turn
lane(s) - unsignalised
intersection (urban)
(NEW)
35% 0.65 Medium
Install right-turn lane
- rural unsignalised T-
intersections (NEW)
40% 0.6 Low
Install right-turn lanes
- rural unsignalised
cross road
intersections (NEW)
30% 0.7 Medium
Install left-turn lane
(NEW)
Urban intersections 20% per approach 0.8 per approach Low Additional crash reductions may be gained for cyclists if a
cycle lane is installed between left and through lane.
Rural intersections 0% 1.0 Low The research and the benefits of left turn lanes on high speed
intersections is inconclusive. While most research indicates
that left turn slip lanes reduce crashes there are also studies
that show that crashes may increase. A key issue with these
lanes is that vehicles in the left turn lane may restrict visibility
to through vehicles. This treatment should be applied with
caution.
Staggered junctions –
rural (converting cross
With minor road
traffic < 15% of main
road
35% 0.65 Low Note that various stagger elements such as the stagger depth,
alignment, and layout may significantly affect the potential
benefits.
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 66
Common Intersection Crash Modification/Reduction Factors (Urban and Rural)
Treatment Sub type Crash Reduction
Factor
Crash Modification
Factor
Confidenc
e
Comment
road junctions to two T
– junctions) (NEW) With minor road
traffic 15-30% of
main road
25% 0.75 Low
With minor road
traffic > 30% of main
road
35% 0.65 Low
Intelligent active
warning signs at rural
intersections (e.g.
RIAWS) (NEW)
35% 0.65 Medium Crash reductions are likely to be higher for serious injury and
fatal crashes due to reductions in operating speeds.
Static advance warning
of rural intersections -
where it is deemed
necessary
All 7% 0.93 Low
Install red light camera
at signalised
intersections (NEW)
5% 0.95 High
Street lighting New lighting – rural
intersection
30% 0.9 Medium Crash reduction factor based on night crashes only. CMFs
based on 29% and 32% of crashes occurring at night in urban
and rural intersections respectively. Where there is sufficient
evidence (from the crash history) that an intersection has a
higher or lower proportion than this then a site specific CMF
should be developed.
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 67
Common Intersection Crash Modification/Reduction Factors (Urban and Rural)
Treatment Sub type Crash Reduction
Factor
Crash Modification
Factor
Confidenc
e
Comment
New lighting - urban
intersection (NEW)
35% 0.9 Low
CRFs for pedestrian crashes are higher than presented here
(see Table 36). Research indicates that lighting has very little
effect on loss-of-control crashes.
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 68
Table 37: Common Urban Cyclist Crash Reduction/Modification Factors (apply only to crashes involving cyclists)
Common Cyclist Crash Reduction/Modification Factors
Treatment Sub type Crash
Reduction
Factor
Crash
Modification
Factor
Confidenc
e
Comment
On-road cycle lanes Standard
10%
0.9
Low
Less than 1.4 metres wide
Wide
(NEW)
20% 0.8 Low Greater than 1.4 metres wide
Advanced cycle stop
boxes
Intersections 35% 0.65 Low Advanced stop boxes need to be to depths specific in cycling guidelines. Research
indicates that the crash reduction is less when inadequate depth is provided.
Separated cycle paths
alongside roads (NEW)
– one way for cyclists
All crashes 0% 1.0 Low The limited research available on cycle paths indicates that intersection and access
crashes may increase as a result of these treatments, and may cancel the benefits that
occur along mid-block sections. Where paths can be provided away from intersections
and accesses crash benefits are likely. Where there are a lot of intersections and accesses
without suitable mitigation of crash risk there may be an increase in cycle crashes. The
main benefits of such facilities are a reduction in the perceived risk of cycling by the
public.
European experience indicates that two-way cycle paths have a much higher crash rate
than one-way facilities. This is in part due to crossing motorists not expecting cyclists
from both directions. The effect is exacerbated on one-way streets.
As research becomes available on different cycle facilities these factors will be revisited.
Shared path (cycle and
pedestrian) alongside
roads (NEW) – one way
for cyclists
All crashes 0% 1.0 Low
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 69
Table 38: Common Urban Pedestrian Crash Reduction/Modification Factors (applies only to pedestrian crashes)
Common Pedestrian Crash Reduction/Modification Factors
Treatment Sub type Crash
Reduction
Factor
Crash
Modification
Factor
Confidenc
e
Comment
Improved lighting
(NEW) at mid-blocks
and intersections
Level V4 55% N/A Medium Lighting luminance levels are as follows (refer to AS/NZ standard 1158.1.1 for
further details)
V1 >=1.5 cd/m2
V2 >=1.0 cd/m2
V3 >= 0.75 cd/m2
V4 >= 0.50 cd/m2
When upgrading lighting from one category to another (e.g. from V4 to V2) then
pro rata the factors provided (e.g. upgrading from V4 to V2 gives a CRF of (1-0.55)
x 0.80 = 36%
Level V3 70%
N/A Medium
Level V1 & 2 80% N/A Medium
Add exclusive
pedestrian phase at
signals (Barnes dance)
(NEW)
All 55% 0.45 Low Should only be applied to intersections with high pedestrian volume in major
commercial areas (like city centres)
Improve signal timing to
reduce pedestrian
delays (NEW)
All 35% 0.65 Low Only applicable if major reductions in pedestrian delay can be gained.
Install pedestrian
overpass
All 85% 0.15 Low Where there are strong at grade desire-lines the benefit may be less.
Install raised platform All 20% 0.8 Low Treatment unsuitable for major roads. Normally introduced as part of area wide
traffic calming schemes. The 80% reduction specific in the previous version of EEM
was an error.
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 70
Common Pedestrian Crash Reduction/Modification Factors
Treatment Sub type Crash
Reduction
Factor
Crash
Modification
Factor
Confidenc
e
Comment
Install pedestrian refuge When kerbside
parking
15% 0.85 Low Higher reductions may be achieved on high volume roads. Crash reduction is likely
to be lower when traffic lanes are 4m wide or greater (excluding cycle lanes).
Based on lane width of around 3.5m.
When no
kerbside parking
45% 0.55 Low
Install kerb extensions 35% 0.65 Low Kerb extension must bring waiting pedestrians out beyond the line of parked
vehicles, where inter-visibility between through traffic and pedestrians is adequate.
Based on a traffic lanes of around 3.5m (excluding cycle lane where present).
Crash reductions are likely to be reduced as traffic lanes width increase beyond 4m.
Install pedestrian refuge
and kerb extensions
45% 0.55 Medium Based on urban traffic lanes of around 3.5m (excluding marked cycle lanes). Crash
reductions are likely to be reduced as traffic lanes width increase beyond 4m.
Install zebra crossing Two-lane roads 0% 1.0 Low Where speed limit is 50km/h or less. An increase in crash risk is likely on 2-lane
roads with speed limits in excess of 50km/h
Multi-lane roads
(NEW)
90% increase
in pedestrian
crashes
1.90 Low Research indicates that crash rates increase on multi-lane roads when the AADT is
12,000 or greater. Also, that the difference in pedestrian crash risk is not
significant different in marked zebra crossings vs unmarked crossings on multi-lane
roads with an AADT below 12,000.
Install mid-block traffic
signals
All 45% 0.55 Low Benefits are lower on multilane roads and where speed limit is above 50km/h.
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 71
Common Pedestrian Crash Reduction/Modification Factors
Treatment Sub type Crash
Reduction
Factor
Crash
Modification
Factor
Confidenc
e
Comment
Install fencing and
barriers (NEW) to
direct pedestrians
All 20% 0.8 Medium Not applicable in all circumstances. Where pedestrian crossing desire-lines are
strong pedestrians may jump the fence and crash reductions will be lower.
Traffic signals rest on
red (NEW).
All 50% 0.5 Low
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 72
9.0 Severity factors
The severity factors by intersection, midblock and by other site type are provided in Table 39 to Table 41 by
transport mode involved. The total number of FSi for a subject site is calculated by aggregating the FSi equivalents
for each mode type. For example the FSi for an urban roundabout (less than 80 km/h on all roads) with five motor
vehicle injury crashes, and three cyclist injury crashes is calculated as follows:
5 (motor-vehicle crashes) * 0.13 = 0.65
3 (cyclists crashes) * 0.22 = 0.66
Total FSi = 1.31
For rural mid-blocks the terrain and alignment types will impact on the operating speed. For example, rural tortuous
alignments are likely to have mean speeds of 50-70 km/h. The severity factor can be estimated by interpolating
between mid-block factors of 50 km/h and 70 km/h.
Table 39: Urban Intersection (less than 80 km/h) FSi Severity Factors on all roads.
Urban Intersection (less than 80 km/h on all roads)
Location and Mode FSi Severity Factors
Signalised Cross roads (motor vehicles) 0.13
Signalised T-junctions (motor vehicles) 0.14
Roundabouts (motor vehicles) 0.13
Priority Cross roads (motor vehicles) 0.14
Priority T-junctions (motor vehicles) 0.15
Pedestrians 0.23
Cyclists 0.22
Motor-cyclists 0.24
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 73
Table 40: Rural Intersection (80 km/h plus on one intersecting road) FSi Severity Factors
Rural Intersection (80 km/h plus on one intersecting road)
Location and Mode FSi Severity Factors
Signalised Cross roads (motor vehicles) 0.27
Signalised T-junctions (motor vehicles) 0.20
Roundabouts (motor vehicles) 0.18
Priority Cross roads (motor vehicles) 0.35
Priority T-junctions (motor vehicles) 0.32
Pedestrians 0.48
Cyclists 0.32
Motorcyclists 0.47
Table 41: Mid-blocks and Special Sites FSi Severity Factors
Special Sites
Location, Mode, and Operating Speed FSi Severity Factors
Bridges (all speeds) 0.30
Mid-blocks 50km/h (motor vehicles) 0.15
Mid-blocks 50km/h (pedestrians) 0.26
Mid-blocks 70km/h (motor vehicles) 0.22
Mid-blocks 70km/h (pedestrians) 0.52
Mid-blocks 100km/h (motor vehicles) 0.26
Mid-blocks 100km/h (pedestrians) 0.64
Rail crossings (all speeds) 0.53
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 74
References and Bibliography
No Document/reference Website information (if any)
1 NZ Transport Agency. Economic evaluation http://nzta.govt.nz/resources/economic-evaluation-
manual/economic-evaluation-manual/docs/eem-
manual.pdf
2 Elvik, Hoye, Vaa, & Sørensen (2009). The
handbook of road safety measures (second
edition)
https://books.google.co.nz/books/about/The_Handbo
ok_of_Road_Safety_Measures.html?id=JuTAZmIseeAC
3 NZ Transport Agency (2013). High-risk
intersection guide
http://www.nzta.govt.nz/resources/high-risk-
intersections-guide/docs/high-risk-intersections-
guide.pdf
4 NZ Transport Agency (2011). High-risk rural
roads guide
http://www.nzta.govt.nz/resources/high-risk-rural-
roads-guide/docs/high-risk-rural-roads-guide.pdf
5 Turner, S; Roozenburg, A; and Smith A
(2009). “Roundabout Crash Prediction
Models” NZTA Research Report 386.
Wellington, NZ
http://www.nzta.govt.nz/resources/research/reports/
386/docs/386.pdf
6 Turner, S, Binder, S and Roozenberg, A
(2009). “Cycle Safety: Reducing the Crash
Risk”. NZ Transport Agency Research Report
389
http://www.nzta.govt.nz/resources/research/reports/
389/docs/389.pdf
7 Austroads (2012). Effectiveness of Road Safety
Engineering Treatments (AP-R422-12)
https://www.onlinepublications.austroads.com.au/ite
ms/AP-R422-12
8 Turner, S and Roozenberg A (2007). “Crash
Rates at Rural Intersections”. Road Safety
Trust, Wellington NZ
http://www.nzta.govt.nz/resources/crash-rates-and-
rural-intersections/docs/crash-rates-and-rural-
intersections.pdf
9 International Road Assessment Program
(2013). iRAP Methodology Fact Sheet #12 -
Multiple Countermeasures
http://www.irap.net/en/about-irap-
3/methodology?download=140:irap-methodology-
fact-sheet-12-multiple-countermeasures.
10 NZ Transport Agency (2015). National Roads
and Roadsides Business Case
11 Elvik, R, Christensen, P, Amundsen, A,
(2004). Speed and Road Accidents, Institute of
Transport Economics (TOI) Report 740
12 Jackett, M J (1992). On which curves do
accidents occur? A policy for locating advisory
speed signs. Volume 1 Proceedings. IPENZ
Annual Conference
13 Jackett, M (1993). Accident Rates on Urban
Routes – 1992 Update. IPENZ Transactions
Vol 20 (1) pp 10-16
14 Cenek, P; Henderson R; and Davies R (2012).
Modelling crash risk on the New Zealand state
highway network. NZTA Research Report
477, Wellington, NZ
http://www.nzta.govt.nz/assets/resources/research/r
eports/477/docs/477.pdf
15 Turner, S (1995). Estimating Accidents in a
Road Network PhD thesis. Department of Civil
Engineering, University of Canterbury, NZ
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 75
16 Turner S, Roozenburg A, and Francis, A
(2006). Predicting Accident Rates for
Pedestrians and Cyclists. Land Transport NZ
Research Report No. 289, Land Transport
NZ, Wellington, NZ
17 Turner S, Singh R, Allatt T and Nates G
(2010). Effectiveness and Selection of
Treatments for Cyclists at Signalised
Intersections. Austroads Research Report, NZ
18 Turner S, Singh R and Nates G (2012). Crash
Prediction Models for Signalised Intersections :
Signal Phasing and Geometry. NZ Transport
Agency Research Report 483
19 Turner S, Singh R and Nates G (2012). The
Next Generation of Rural Road Crash prediction
models : Final Report). NZ Transport Agency
Research Report 509
20 NZ Transport Agency. One Network Road
Classification
https://www.nzta.govt.nz/roads-and-rail/road-
efficiency-group/one-network-road-classification/
21 Turner S (2001). Accident Prediction Models.
Transfund NZ Research Report 192
23 Highway Safety Manual (AASTO) http://www.highwaysafetymanual.org/Pages/default.a
spx
24 CMF Cleaninghouse http://www.cmfclearinghouse.org/index.cfm
25 iRAP Safety Toolbox http://toolkit.irap.org/
The NZ Transport Agency’s Crash Estimation Compendium
First edition, Amendment 1
Effective from 01/06/2018 Page 76
Appendix 1
