nsm - advancing the model for safety improvement of the road

Research Collection

Master Thesis

NSM - Advancing the Model for Safety Improvement of the RoadNetwork in the City of Zurich

Author(s): Rothenfluh, Marco

Publication Date: 2015

Permanent Link: https://doi.org/10.3929/ethz-a-010530679

Rights / License: In Copyright - Non-Commercial Use Permitted

This page was generated automatically upon download from the ETH Zurich Research Collection. For moreinformation please consult the Terms of use.

ETH Library

https://doi.org/10.3929/ethz-a-010530679

http://rightsstatements.org/page/InC-NC/1.0/

https://www.research-collection.ethz.ch

https://www.research-collection.ethz.ch/terms-of-use

NSM - Advancing the Model for Safety Improvement of the Road Network in the City of Zurich

Marco Rothenfluh

Supervision:

Dr. Monica Menendez, Institute for Transport Planning and Systems, ETH Zurich

Qiao Ge, Institute for Transport Planning and Systems, ETH Zurich

In cooperation with the City of Zurich (DAV) and Federal Roads Office (FEDRO)

Master thesis Spatial Development and Infrastructure Systems March 2015

Legend

Perimeter ZH

Intersection

Reclassified roads

Existing Traffic oriented roads

New classified Traffic oriented roads

Residential zone

Non living area

±0 1 2 3 4

Kilometers

Advancing the Model for Safety Improvement of the Road Network in the City of Zurich March 2015

I

Acknowledgments

First of all I would like to thank Dr. Monica Menendez who gave me the opportunity to write

this thesis and for the great support despite her maternity leave. Further thanks go to Qiao Ge,

my assistant of charge, who helped me a lot with his specific knowledge in the field of sensi-

tivity analysis and all the advice about its implementation.

Many thanks go to Ursin Decurtins for providing me with all his primary data and background

information about the existing network model.

I would also like to say thank you to Wernher Brucks from the Dienstabteilung Verkehr of

Zurich. The provision of information and the possibility of doing this thesis in cooperation with

the Dienstabteilung Verkehr was very fruitful.

Another thank you goes out to Hagen Schüller from the PTV Group for assisting me with spe-

cial information about the theoretical issues of this thesis.

I am also very grateful to FEDRO and all the members of the NSM pilot project for granting

me the chance to take part in all their meetings, and the option of continuing to further develop

NSM in Switzerland. Special gratitude goes to Stevan Skeledzic from the Beratungsstelle für

Unfallverhütung for his very cooperative exchange of ideas and his endurance in explaining

diverse insights.

Lastly, I would like to thank my family, who always supported me in a great manner throughout

the entire duration of my studies.


II

Table of Contents

1 Introduction ......................................................................................................... 1

1.1 Intention of Network Safety Management (NSM) ..................................................... 1

1.2 NSM model for the city of Zurich .............................................................................. 5

1.3 Objective of this Master thesis .................................................................................. 6

2 Improving the Model of the Network of Zurich ..................................................... 7

2.1 Existing Network ....................................................................................................... 7

2.2 Change of different network elements .................................................................... 12

2.3 Improved model ...................................................................................................... 18

3 Results of various analyses ................................................................................21

3.1 Prioritization with the given input parameters ......................................................... 21

3.2 Variation of different parameters ............................................................................ 29

3.3 Sensitivity analysis (SA) ......................................................................................... 47

3.4 Network density of residential zones ...................................................................... 54

3.5 Accident patterns of human powered mobility ........................................................ 58

3.6 Impact of tram ......................................................................................................... 60

3.7 Result overview ...................................................................................................... 61

4 Overlay with Black Spot Management (BSM) .....................................................63

5 Discussion ..........................................................................................................67

5.1 Accuracy of parameters .......................................................................................... 67

5.2 Constraints of the network model ........................................................................... 70

5.3 Limitation of the implemented NSM method........................................................... 71

6 Relevance for practice ........................................................................................73

7 Conclusion and recommendations .....................................................................75

7.1 Conclusion .............................................................................................................. 75

7.2 Recommendations .................................................................................................. 76

8 References .........................................................................................................77


III

List of Tables

Table 1 Cost factors of inner urban areas ............................................................... 4

Table 2 Elements with the highest infrastructure potential of the existing network .11

Table 3 Comparison of the section creation for traffic oriented roads .....................17

Table 4 Comparison of the aggregated residential zones ......................................18

Table 5 Legend of accident patterns of highly prioritized network elements ...........21

Table 6 Differences of the results between the two models....................................29

Table 7 Casualty cost rates of Switzerland and Germany ......................................30

Table 8 Own calculated ACR of accident severity categories .................................32

Table 9 Own calculated ACR of accident types ......................................................33

Table 10 Comparison of the basic rates of FEDRO and the average rates of

Zurich ........................................................................................................36

Table 11 Values of the coefficient baACR approach ................................................37

Table 12 Comparison of different basic rates ...........................................................39

Table 13 Comparison of node size 50 m and 30 m ..................................................42

Table 14 Example of a five series sample ................................................................52

Table 15 Minimal and maximal relation of ACR........................................................52

Table 16 Network elements most influenced by tram accidents ...............................61

Table 17 Overview of the highest rankings of each NSM analysis ...........................62

Table 18 Comparison of rankings of BSM and NSM ................................................66


IV

List of Figures

Figure 1 Taxonomy of factors affecting road safety .................................................. 1

Figure 2 Input parameters of the GIS model ............................................................ 3

Figure 3 Methodology for calculating the infrastructure potential .............................. 4

Figure 4 Main sequences of the network model procedure for urban areas ............. 8

Figure 5 Existing network of NSM ...........................................................................10

Figure 6 Inner city of Zurich in the improved model .................................................14

Figure 7 New digitalized situation of two level nodes ..............................................15

Figure 8 Improved model for NSM ..........................................................................19

Figure 9 Results of traffic oriented roads .................................................................22

Figure 10 Results of nodes .......................................................................................24

Figure 11 Results of residential zones ......................................................................25

Figure 12 Merging results of the rankings of traffic oriented roads and nodes ...........27

Figure 13 Distribution of the accident rate of traffic oriented roads ............................35

Figure 14 Change of priorities with AADT from the year 2030 ...................................41

Figure 15 Rankings of nodes with the accident cost digit (ACDI) method ..................44

Figure 16 Change in priorities when assigning the accidents of nodes to traffic oriented

roads 46

Figure 17 Significant parameters as a function of accident severity and accident

frequency ..................................................................................................48

Figure 18 Correlation analysis of traffic oriented road sections .................................49

Figure 19 Process of model sampling for traffic oriented roads .................................51

Figure 20 Impacts of different parameters on the results for traffic oriented roads ....53

Figure 21 Impacts of different parameters on the results for nodes ...........................54


V

Figure 22 Percentage of dedicated road space of each zone ....................................55

Figure 23 Priorities of zones when including network density ....................................57

Figure 24 Priorities including pedestrian and bicycle accidents only..........................59

Figure 25 Comparison of BSM and NSM ..................................................................64


VI

Abbreviations

AADT……………………………………….…………….Annual Average Daily Traffic

AC……………………………………………………………..……………Accident cost

acc…………………………………..…………………………………………….accident

ACD…………………………………………….………………….Accident cost density

ACDI……………………………………………………...…………..Accident cost digit

ACR………………………………………………………..…………..Accident cost rate

AR…………………………………………………………………....……..Accident rate

avAC……………………………………………………….........avoidable accident cost

baACD…………………………..……………………………basic accident cost density

baACR…………………………………………………..………..basic accident cost rate

baAR……………………………………………....…………………..basic accident rate

BSM…………………………………………………………….Black Spot Management

DAV………………………………………………Dienstabteilung Verkehr Stadt Zürich

FEDRO………………………………………...……….…………..Federal Roads Office

FSI…………………………………………………...………...…..Fatal or serious injury

GIS………………………………………………………Geographic information system

GVM……………………………………………..……………………Total traffic model

InfraPo………………………………………………..…………..Infrastructure potential

MI…………………………………………………………………..………..Minor injury

NSM…………...………………………….……………….Network Safety Management

PDO………………………………………………………..………Property damage only


VII

RIA………………………………..…………………….Road Safety Impact Assessment

RSI……………………………………………..…………………Road Safety Inspection

SA………………………………………………………………..…..Sensitivity Analysis

SNR………………………………………………………………………..…..Swiss Rule

VSS…………..……Research and standardization in the field of road and transportation

veh…………………………………….………………..…………………………vehicle

ZH…………………………………………………………………………….……Zurich


VIII

Master thesis Spatial Development and Infrastructure Systems

NSM - Advancing the Model for Safety Improvement of the

Road Network in the City of Zurich

Marco Rothenfluh

ETH Zurich

Schumacherweg 3

CH-8046 Zurich

Phone: +41 79 798 76 51

Mail: [email protected]

March 2015

Abstract

Network Safety Management (NSM) is a road safety tool, based on the number of accidents, to

identify locations of the highest potential for improving road infrastructure. The aim of this work

is to analyse the present results for the city of Zurich and to test different variations of input

parameters. For this, the existing road network is explored, and on the basis of the assumption of

different accident cost rates, the avoidable accident cost per year is calculated and a prioritization

of the three network elements traffic oriented roads, nodes and residential zones is made. Due to

a great sensitivity of the model parameters, different samples of inputs are tested, and specific

cost rates for the city of Zurich are calculated.

Furthermore an overlay with the widely adopted road safety tool Black Spot Management (BSM)

is performed to verify the detected accident clusters. The findings are then used to make pragmatic

recommendations on the use of NSM as a road safety analysis tool.

Keywords

Network Safety Management (NSM), infrastructure potential, accident cost, robustness, sensi-

tivity analysis

Preferred citation style

Rothenfluh, M. (2015) NSM - Advancing the Model for Safety Improvement of the Road Net-

work in the City of Zurich, Master thesis Spatial Planning and Infrastructure Systems, Institute

for Transport Planning and Systems (IVT), ETH Zurich, Zurich.


IX

Masterarbeit Raumentwicklung und Infrastruktursysteme

NSM - Advancing the Model for Safety Improvement of the

Road Network in the City of Zurich

Marco Rothenfluh

ETH Zurich

Schumacherweg 3

CH-8046 Zurich

Phone: +41 79 798 76 51

Mail: [email protected]

Januar 2015

Zusammenfassung

Network Safety Management (NSM) ist ein Infrastruktur-Sicherheitsinstrument, welches auf-

grund des Unfallgeschehens, Orte mit hohem Verbesserungspotential an Infrastruktur detektiert.

Das Ziel der Arbeit ist eine Analyse der bisherigen Resultate der Stadt Zürich und die Variation

verschiedener Input Parameter. Dazu wird das Strassennetzwerk untersucht und mit Hilfe von

Unfallkostenraten werden die jährlich vermeidbaren Unfallkosten berechnet. Dies resultiert in

einer Priorisierung der drei Netzwerkelemente verkehrsorientierte Strassen, Knoten und Sied-

lungszonen. Angesichts der grossen Sensitivität der Modell Parameter, werden verschiedene In-

put Stichproben getestet und spezifische Kostenraten für die Stadt Zürich kalkuliert.

Ausserdem wird eine Überlagerung mit den Resultaten des Infrastruktur-Sicherheitsinstruments

Black Spot Management (BSM) gemacht, um die entdeckten Unfallmuster zu verifizieren. Die

Resultate sollen zur praktischen Anwendung von NSM als Sicherheitsinstrument beitragen.

Schlüsselwörter

Network Safety Management (NSM), Infrastrukturpotential, Unfallkosten, Robustheit, Sensiti-

vitätsanalyse

Bevorzugter Zitierstil

Rothenfluh, M. (2015) NSM - Advancing the Model for Safety Improvement of the Road Net-

work in the City of Zurich, Master-Arbeit Raumentwicklung und Infrastruktursysteme, Institut

für Verkehrsplanung und Transportsysteme (IVT), ETH Zürich, Zürich.


1

1 Introduction

Although the number of accidents has decreased over the last decade in Switzerland, there are

still various approaches of improving safety for different road users (FEDRO, 2014d). Beside

improvements in the construction of vehicles and more training courses for drivers, the existing

infrastructure is still one of the most important factors in reducing the impacts of road accidents.

Thereby, the aim is not only to lower the absolute number but also to reduce the severity of the

individual accidents. Figure 1 shows a taxonomy of factors that affect road injuries.

Figure 1 Taxonomy of factors affecting road safety

Source: ELVIK, HØYE, VAA, & SØRENSEN, 2009

This work focuses on the improvements to infrastructure. Due to large networks it is essential

for the road administrators to know where the hotspots of accidents are, in order to plan future

investments as economically as possible.

1.1 Intention of Network Safety Management (NSM)

In the past, road administrators have defined a hazardous part of the network based on a certain

amount of accidents. These parts, called black spots, were usually mitigated by improving the

local traffic situation. Due to declining accident rates the threshold for defining a black spot

either had to be adjusted downwards or the size of a specific section was enlarged (SCHERMERS,

CARDOZO, & ELVIK, 2011). The widely used Black Spot Management (BSM) is a reactive tool

that detects individual lacks of safety and provides specific treatments.


2

In order to analyse the accidents on a larger scale, BSM is supplemented by another tool that

focuses more on the entire network. This network safety management (NSM) is a method to

assess the crash reduction potential of locations in a road network (SCHERMERS, CARDOZO, &

ELVIK, 2011). It is one of the six infrastructure safety tools that ASTRA is implementing to

improve road network safety (FEDRO, 2013). The objective of NSM is to prioritize different

network elements such as road sections or nodes to guarantee that investments in infrastructure

are as highly cost efficient as possible (BAST, SÉTRA, 2005). Although the two methods are

widely adapted, the quality of the results differs a lot due to variations of data and a lack of

standardized definitions (SØRENSEN & ELVIK, 2008).

The pilot projects of NSM in Switzerland have started in 2013. The Federal Roads Office

(FEDRO) in collaboration with the “Research and Standardization in the field of road and trans-

portation” (VSS) created a Swiss rule (SNR 641 725) which defines the process of implement-

ing NSM in Switzerland. Additionally, the most relevant parameters were determined so that

the first results of four different pilot sites, the two rural areas Berne and Aargau and the two

urban areas Basel and Zurich, were presented at the end of 2014.

According to the Swiss rule, the main goal of NSM is to calculate a so called “infrastructure

potential”. This means, the possible optimization which can be done by the infrastructural side,

compared to a best practice design. Or in other words – what are the avoidable accident costs

based on a non-optimal configuration of a certain network element.

To calculate this infrastructure potential two main steps are necessary: First of all, it is important

to have a well-prepared network model that represents the actual road situation as realistically

as possible. The network configuration is done by a geographic information system (GIS) from

ESRI (Product ArcGIS Desktop, Version 10.0). For that several input parameters are necessary.

Basic inputs for all the own created figures in this report are the provided accident data from

STADT ZÜRICH, 2013 and road network data from TIEFBAUAMT STADT ZÜRICH, 2014.

Figure 2 shows the input data of the GIS and the process for a later implementation into the

NSM model. For this work only accidents in the city of Zurich for the time period 2009 to 2013

are considered. Owing to a good data basis, this work also includes accidents that resulted in

property damages only.


3

Figure 2 Input parameters of the GIS model

Source: own research

The accuracy of NSM parameter is dependent on the quality of the input data and the different

processing steps. A critical analysis of the given input data is exposed in chapter 5.1.

The second step is implementing the network parameters to the NSM model. Figure 3 shows

schematically the process of calculating the infrastructure potential. It is divided into two parts.

The left part represents the actual accident situation whereas the right part indicates a hypothet-

ical prediction of accidents, based on the existing road design. When comparing both parts, the

crucial term is the accident cost density (ACD), respectively the basic accident cost density

(baACD). This ensures that each entity is broken down to a comparable length and period of

time. The specific formula for calculating the different values can be found in appendix A 1.


4

Figure 3 Methodology for calculating the infrastructure potential

Source: SNR 641 725, 2013

For calculating the infrastructure potential and therefore to know what costs can be avoided in

the future, different cost values are necessary. These cost factors are generally admitted values

specifically for Switzerland and are reviewed by VSS and FEDRO. Since on different network

elements the baACR differ, various values are provided. Table 1 only lists specific terms for

inner urban areas.

Table 1 Cost factors of inner urban areas

Accident

severity

Accident cost rate

(ACR) [CHF/acc]

Network element Basic accident cost rate

(baACR) [CHF/(1000*veh*km)]

FSI 696’500 Road 175

MI 84’000 nodes (3 inlets) 22.61

PDO 45’000 nodes (>3 inlets) 49.71

Source: SNR 641 725, 2013, FEDRO, 2014b

1 For nodes different basic accident cost rates are used because their network length is not defined.

InfraPo

Infrastructure

potential

Number of ac-

cidents per year

Accident cost

per year

Accident cost

density per

year

ACD

Avoidable acci-

dent cost density

[CHF/(km*a)]

baACD baACR

Basic accident

cost density per

year

Basic accident cost

rate

x cost ratio / length x AADT

basic risk (target figure) accident frequency (actual accident situation)

N AC

[U/a]

[CHF/a]

[1000 CHF/(km/a)]

[1000 CHF/(km/a)]

[CHF/(1000 veh/km)]


5

The categorization of accidents is based on the originated cost of each individual accident. It

includes not only the direct accident costs themselves (medical cost, property damage, admin-

istrative cost), but also follow-up costs such as personal immaterial costs, existing costs due to

inability to work, costs of replacing work force etc. (BUNDESAMT FÜR RAUMENTWICKLUNG

(ARE), 2006). A further discussion of the different accident costs of each category is given in

chapter 5.1.1. The basic accident cost rate is a theoretical value of how many accidents are

expected per thousand driven vehicles kilometer under the condition of a best practice design

(CHAMBON, LEMKE, & GANNEAU, 2006).

1.2 NSM model for the city of Zurich

As one of the chosen pilot sites the Dienstabteilung Verkehr (DAV) of the city of Zurich in

collaboration with the ETH started to implement NSM for Zurich. The results are properly

documented in the master thesis of DECURTINS, 2014. An explanation of the model application

and a short summary of the existing results are given in chapter 2.1.

Although NSM is a broadly used method in road safety management, input data have to be

adapted individually for each investigation. The heterogeneity of a network accounts more for

different accident patterns in a city than in rural areas. Hence, the generally valid parameters

have to be calibrated for each perimeter. There are not only the generated accident cost that

vary massively among different countries, but also the availability and the application of data

lead to different results. There are several approaches to dealing with these uncertainties, but

no general rule is valid so far (AURICH, 2012).

Zurich with over 13’000 registered accidents of all severity categories between 2009 and 2013

has a large and dense inner urban road network, compared to the rest of Switzerland. However,

there are more conflict zones and more interactions with other road users. Therefore not only

the number of accidents are different but also the cause and severity of them are too. To deal

with these circumstances, it is essential to know how these parameters affect the final results

and how the model can be optimized.


6

1.3 Objective of this Master thesis

Based on the results of DECURTINS, 2014, this master thesis aims to examine the influence of

different parameters to the given approach of NSM, by analysing the existing network and by

altering various input parameters. The outcomes of this work shall be used by FEDRO as results

of the pilot study in Zurich. In a further step it can be conducted as a decision tool to convince

policy makers to invest in road safety infrastructure. Moreover, it should contribute to future

establish more accurate parameters for inner urban areas.

The thesis is organized as follows: In chapter 2 there is a proper analysis of the developed model

by DECURTINS, 2014 including a description of the limitations of the used GIS model is shown.

To overcome some of these inaccuracies an improved digitalized model is presented at the end

of this chapter. Chapter 3 contains several analyses of certain model input parameters in order

to find out what the most relevant parameters for the outcome of NSM are. An overlay with the

existing results of BSM is given in chapter 4. Section 5 provides a reflexion of the results in-

cluding a critical discussion about the inputs and definition of different parameters. Chapter 6

describes the relevance for practice, and finally chapter 7 summarizes the major findings and

gives suggestions for future research.


7

2 Improving the Model of the Network of Zurich

This chapter is about the first described step of getting a proper model of the road network with

the help of a GIS. The procedure requires numerous data (see chapter 1.2). Although the differ-

ent steps are given by the Swiss rule, the scope of how data are processed is still very wide. The

following points present a brief overview of why modifications of the input model can result in

dissimilar results for NSM:

• Existence of data

• Actuality and accuracy of input data

• Methodology of creating a road network in a geographic information system (GIS)

• Level of aggregation of different network elements

To minimize these origins of inexactness it is essential to have a consistent approach that either

includes the given guidelines or is also adjustable to a variation of certain parameters.

2.1 Existing Network

In Switzerland, the first attempt of NSM in an urban area was done in Zurich by DECURTINS,

2014. The basis of that project is determined in the Swiss rule SNR 641 725, with the intention

of using a different road network hierarchy due to a more complex road network in urban envi-

ronment. Figure 4 shows briefly the four main steps of the model application in Zurich.

The first two steps depend on the existence and availability of data and therefore cannot be

influenced by a NSM operator. Switzerland, especially the canton of Zurich has recorded a lot

of different data about traffic so that a complete analysis can be done. But there is always the

question about the methodology of collecting and reporting the data as a certain variation of

errors must always be considered.

Apart from the uncontrollable factors it is crucial how to proceed data. The assignment of dif-

ferent entities into the given network element categories can really change the final results. The

following section explains how the specific network elements are categorized and implemented

to a GIS model of Zurich.


8

Figure 4 Main sequences of the network model procedure for urban areas

Source: DECURTINS, 2014

2.1.1 Model assumptions for different network elements

In fact, different definitions of network elements are given in SNR. However, with the intention

of having a proper differentiation, some assumptions still have to be made by the NSM operator.

Traffic oriented roads

From the received data, there is no given category of a traffic oriented roads. SNR does not give

a proper definition and states that traffic oriented roads are between the borders of residential

zones and surround sub ordinary road categories. To simplify this DECURTINS, 2014 defined a

traffic oriented road as a road with an AADT greater than 5’000 veh/day and a speed limit

equals or greater than 50 km/h. Secondly, a cross-comparison between the assigned traffic ori-

ented roads and the road status of swisstopo helped to further adjust the classification. In doing

so, the classification results in lots of small sections so that a future comparison of the individual

infrastructure potential would only depend on the individual section length. Consequently SNR

provides requirements of 0.5 km and 2.0 km for the minimum and maximum section length.

Nodes

A node is defined when two traffic oriented roads cross each other. The default value for the

size is a 50 m radius around the point of intersection of two road axis in order to cover also the

approaching area of a node (SNR 641 725, 2013). This simplification allows calculating a net-

work length within each node by the following term:

Collecting input data

Different attributes of the necessary input parameter

Processing data

Clipping to the perimeter and referencing and visualizing different network elements

Creating network elements

Differentiation of traffic oriented roads, nodes, residential zones

Level of aggregation

Min and max network length of road sections and residential zones


9

𝐿𝑒𝑛𝑔𝑡ℎ𝑛𝑜𝑑𝑒 = 50𝑚 ∗ 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑖𝑛𝑙𝑒𝑡𝑠 (1)

For larger nodes, a more accurate digitalization of the dimension is made manually. The ap-

proach of multiplying it by the amount of inlets stays the same. The AADT per node is deter-

mined by the traffic volume of the incoming inlets in the following way:

𝐴𝐴𝐷𝑇𝑛𝑜𝑑𝑒 =∑ 𝐴𝐴𝐷𝑇𝑖

𝑛𝑖=1

2 (2)

i inlet

n Number of inlets

Sources: SNR 641 725, 2013, DECURTINS, 2014

Residential zones

The definition of residential zones was made according to the received information of traffic

zones. There is information about zones with shared space and a speed limit of maximum 30

km/h. These are indicators of having residential oriented roads within each zone but neverthe-

less give no inference of what the land is allocated for in these zones The existing model as-

sumes that in an urban area, next to the space for traffic oriented roads, all the remaining area

is defined as residential zones. Only the areas for the Lake of Zurich, the two rivers Sihl and

Limmat and the widespread area of rail tracks are excluded from the zone building.

No specifications are given for the size of a residential zone in the Swiss rule. DECURTINS, 2014

tried to build the zones as homogenously as possible and set the minimum values to 0.22 km2

for the area and 0.6 km for the inner zonal network length for minor roads. An upper limit of

the size or network length is not defined.

Assignment of the accidents to the appropriate network element

All the different network elements are digitalized in form of polygons. This ensures that there

is no overlap between the three classes and that the location of each accident can be assigned

properly to one network element. Otherwise they had to be referred to the next located element

which could cause problems of a unique allocation of accidents to one specific element.

2.1.2 Results

After processing the input data with the given definitions, the existing network for the analysis

of the city of Zurich looks as depicted in Figure 5. It is divided into the three different network


10

elements in which each accident of the period 2009 – 2013 can be assigned to one entity. There

are a total of 145 sections of traffic oriented roads, 156 nodes and 71 aggregated residential

zones generated for the NSM analysis.

Figure 5 Existing network of NSM


Using this network of analysis as input for the calculation of the infrastructure potential (see

Figure 3), the following elements are the ones with the highest infrastructure potential. Notice

Legend

Perimeter ZH

Traffic oriented road

Intersection

Residential zone

±0 1 2 3 4

Kilometers

Legend

Perimeter ZH


Intersection

Residential zone

±0 1 2 3 4

Kilometers

Legend

Perimeter ZH


Intersection

Residential zone

±0 1 2 3 4

Kilometers


11

that for each category only the elements with the highest level of prioritization2 are listed in

Table 2.

Table 2 Elements with the highest infrastructure potential3 of the existing network

Ranking Traffic oriented roads Nodes Residential zone

1 Badenerstrasse West Bellevueplatz Industry Altstetten

2 Bahnhofquai, Mühlegasse,

Uraniastrasse, Sihlstrasse

Bucheggplatz Niederdorf

3 Bahnhofstrasse/Bleicherweg Heimplatz Industry Herdern

4 Albisstrasse Central

5 Schaffhauserstrasse Limmatplatz

6 Badenerstrasse Ost Pfingstweidstrasse/

Hardstrasse

7 Limmatstrasse Bürkliplatz

8 Escherwyssplatz


2.1.3 Limitations of the model

Even though the existing model generated reasonable results that correspond with the

knowledge of local experts, it has some restrictions. During the process of digitalization, some

simplifications have to be made. The following points can typically be sources of errors:

• Level of detail in digitalization

• Amount of included roads

• Assignment of the accidents to different elements

• Proper representation of specific road network configuration

• Aggregation of inhomogeneous network elements

In the following chapter 2.2, an attempt of minimizing the influence of these restrictions is

presented. Due to a lack of available data and no proper definition of certain categories, it is

obvious that not all of these limitations can be solved. Nevertheless, an exact analysis of the

2 The highest level of prioritization is defined by the sum of the avoidable cost of each element. All the elements

that belong to the upper 20% of the total avoidable accident cost per year are prioritized as highest rank

3 Calculated infrastructure potential including FSI, MI and PDO accidents


12

network configuration in combination with personal knowledge of the author and the help of

Google Street View ensures a better quality of the GIS model.

Another disadvantage of the existing model is the lack of automatic processing. Due to digital-

izing all the network elements as polygons, a change of the shape of one element leads to an

adjustment of the geometry of a neighbouring element in order to make sure that there is no

overlay between different polygons. Other approaches of GIS models for NSM with more au-

tomatization can be found for the canton of Aargau (SKELEDZIC, 2014).

2.2 Change of different network elements

This work punctually tries to improve the given network for analysis. The methodology and

approach of digitalization continues in the same way as the primary one, done by DECURTINS,

2014. The following chapters explain the changes of the GIS model that are made with the

purpose of representing the available network in a more realistic way.

2.2.1 Assignment of road categories

The canton of Zurich provided a datasheet with the property of all the road elements. Since all

the highways belong to FEDRO, only roads that are operated by the canton or the city of Zurich

are taken into account. On top, some specific sections such as Milchbucktunnel and all the on-

and off-ramps to and from the highways are not considered in this analysis. The network con-

tains all the other roads that have an attribute vmax>0 km/h and are either in the ownership of

the canton or the city of Zurich (TIEFBAUAMT KANTON ZÜRICH, 2014).

One of the most important parts is the classification of traffic oriented or residential roads. As

written in chapter 2.1.1, there is no proper distinction between them and it depends on several

criteria. The intention of a traffic oriented road is to bundle the traffic and provide a sufficient

capacity for most through-going traffic streams. In contrast, residential or also called minor

roads should have as little traffic as possible to guarantee a certain safety standard (FEDRO,

2014a). They exhibit mostly origin and destination traffic flows and from the infrastructural

side, they are often supplied with obstacles (e.g. parking slots, bumps) to prevent through-going

traffic (TOURING CLUB SCHWEIZ, 2002). In addition to that, all the roads signalised as 30 km/h

speed limit are classified as residential roads.

As seen, the classification depends on how the owner wants to operate the roads. Therefore the

classification made in the improved model includes a variation of different criteria:

• Road owner


13

• AADT

• Speed limit (vmax)

• Presence of on-street parking

• Presence of pedestrian crossings

• Local knowledge

• Google street view

An example of the inner city of Zurich should illustrate the distinction of road categories (Figure

6). Whereas traffic oriented roads (blue tagged in the figure) in general have a higher AADT

and therefore another geometric design, residential roads are usually smaller and their main

function is to access a certain quarter. Notice that only roads with a dedicated speed limit greater

than zero for motorized vehicles are classified as residential roads. For instance, Bahnhofstrasse

or Limmatquai are not shown in the figure because they are only accessible for human powered

mobility and public transport. So the inner city does not really have a dense network of traffic

oriented roads. Traffic streams are concentrated around the main station such as on Urani-

astrasse and on the Quaibrücke.

In comparison to the existing model, two roads (Bahnhofstrasse, Falkenstrasse) are relegated to

residential roads. However, several new roads such as Tièchestrasse, Wallisellenstrasse and

Baslerstrasse are classified as traffic oriented roads. The total network of traffic oriented roads

has increased by 25 km to total 178 km. An overview of the newly classified roads is given in

chapter 2.3.

The basis for AADT is the total traffic model (GVM) of the city of Zurich (VRTIC, WEIS, &

FRÖHLICH, 2012). Thereby, only AADTs on certain road segments are measured so that on all

the other roads the AADT is approximated. The AADTs of roads with low traffic volume es-

pecially involve a certain inaccuracy.


14

Figure 6 Inner city of Zurich in the improved model


2.2.2 Crossing network on different levels

The definition indicates that wherever two traffic oriented roads cross each other, there must be

a node. The given methodology assumes that all the inlets are on the ground level and ignores

the fact of having network elements on different heights. Mainly nodes of heavy traffic or roads

leading to nearby motorways are sometimes built beneath or over ground level. Examples

thereof can be found at the following places in Zurich4:

• Hardbrücke: Hardplatz, Eschwerwyssplatz, Wipkingerplatz

• Hirschwiesentunnel, Bucheggtunnel, Schöneichtunnel

• Europabrücke: Hohlstrasse, Aargauerstrasse

4 List of location does not claim for completeness

Bellevue Bürkliplatz

Zurich main station

Central

Limmat

Sihl

Lake of Zurich

Legend

Intersection


Residential road vmax>0km/h

Residential zone

Open space

0 100 200 300 400 500

Meters

±


15

• Sihlquai, Kornhausbrücke

• Bahnhofquai

• Alfred-Escher-Strasse, Tunnelstrasse

• Off-ramp Brunau, Interchange Zurich East, Allmendstrasse

The model in GIS cannot allocate the accidents to the appropriate section because the attribute

table of accidents only includes X and Y coordinates. Although every location of accidents is

protocolled, it has to be assigned manually to each road section afterwards.

Since AADT has a main influence on the outcome of NSM, it is important not to combine traffic

volumes of different inlets which do not have conflicting zones. To deal with this, two different

approaches are made: On the one hand, nodes are properly analysed and only the AADT of the

merging inlets is summarized. The AADT of the road section crossing over or beneath the node

is omitted and assumed as a non-interacting traffic stream. On the other hand, some of the in-

frastructure is newly digitalized to enable a more realistic representation of the actual situation.

An example is given in Figure 7. It represents Escherwyssplatz and Wipkingerplatz with the

overcrossing Hardbrücke.

Figure 7 New digitalized situation of two level nodes


Hardbrücke

42’455 veh/day

Wipkingerplatz

18’480 veh/day

Nordstrasse

60’526 veh/day

Escherwyssplatz

18’724 veh/day


16

Instead of assigning all the accidents to nodes, the red circles of Escherwyssplatz and Wipking-

erplatz are divided in two parts. The through-going line represents Hardbrücke with an AADT

more than double of the AADT of the two nodes beneath. All the accidents within the blue

polygon are assigned to Hardbrücke, the rest to the node. At Nordstrasse the Hardbrücke comes

back to ground level, therefore this node has a much higher traffic volume than Escherwyssplatz

and Wipkingerplatz. Consequently, the partition of the infrastructure in combination with the

proper traffic volume enables a more realistic allocation of the accidents.

The assumption of both approaches is that network parts without interaction have less accidents

than interacting inlets at nodes (ELVIK, HØYE, VAA, & SØRENSEN, 2009). In other words, all the

accidents of that location are assigned to the conflicting area (node) and not to the bridge above,

respectively the tunnels beneath. This assumption is independent of the AADT of the different

sections.

2.2.3 Section creation for traffic oriented roads

The accident cost density depends on the length of sections. Small sections with lots of acci-

dents would lead to a very high accident cost density. In order to reduce the effect of small road

sections, the minimum length of 0.5 km respectively the maximum length of 2.0 km for the

section creation has been observed according to the Swiss rule. Nevertheless the geometric

design of road sections in an inner city can change massively within these 500 meters. Due to

a lack of more information about geometric characteristics of roads, the two main criteria for

aggregating are the smallest difference of the AADT and the amount of accidents on individual

road sections (Vengels & Weinert, 2008). Therefore, the radius for searching sections with sim-

ilar AADT and accidents patterns is extended. A further discussion of creating aggregated road

sections for safety analysis can be found in (EBERSBACH & SCHÜLLER, 2008).

The improved model has, in comparison to the existing model, more dedicated traffic oriented

road sections and generally has a lower average length (Table 3).


17

Table 3 Comparison of the section creation for traffic oriented roads

Information Existing model Improved model

Total number of sections 145 208

Total network length [km] 153 178

Mean length [km] 1.06 0.86

Median [km] 0.97 0.76

Number of sections greater 1 km 47 68

Percentage of sections greater 1 km [%] 32.4 32.7


2.2.4 Residential zones based on land use plan

Although it is in discussion SNR does not actually determine minimum values for residential

zones. Analogous to the assumptions of DECURTINS, 2014, a minimal network length (0.6 km)

respectively minimal size of a zone (0.22 km2) is taken into account. In comparison to the earlier

network, the basic layer for the zones differs. Although the perimeter of Zurich is mostly cov-

ered by urban space, there are big areas where there are no houses or other buildings (e.g. forest,

lakes, agriculture zone, parks etc). The supposition of DECURTINS, 2014 is that every polygon

which is not assigned as a traffic oriented road or as a node belongs to the residential zone (apart

from the Lake of Zurich, the rivers Sihl and Limmat and the railway tracks Zurich main station).

The improved model only considers residential zones from the land use plan of the city itself

(AMT FÜR RAUMENTWICKLUNG (ARE), 2014). Since there are some roads with a dedicated speed

greater than zero that are not located in the residential zone (e.g. concrete roads through forests),

a new network length for the residential road is calculated. The effect of the prioritization of

the infrastructure potential can be seen in chapter 0.

The aggregation of individual zones is mainly based on similar classes of the land use plan and

the observable structure of a quarter from a satellite picture of google maps. Similarly to the

traffic oriented roads the difference between the primary and recent model are given in Table

4.


18

Table 4 Comparison of the aggregated residential zones

Information Existing model Improved model

Total number of zones 71 72

Total network length [km] 429 384

Total area of zones [km2] 81.3 42.4

Mean network length [km] 6.1 5.3

Mean area of zones [km2] 1.2 0.6


2.3 Improved model

Implementing all changes to the network elements mentioned above, a new developed model

is created in Figure 8. Whereas from the first view, it has not changed a lot, the model represents

the real situation with more detail and is a better basis for calculating the infrastructure poten-

tial.

The subsequent points sum up briefly the most relevant changes in the improved model.

• Instead of assigning the entire area as residential roads, an overlay with the land use

plan of the city of Zurich nearly halves the designated area of the residential zones.

• Diverse roads with through-going traffic are newly classified as traffic oriented

roads.

• All the highway sections which are operated by FEDRO are omitted.

• Only a few roads in the city are reclassified as residential roads.

• Diverse new nodes are added to the network including a more detailed adaption of

the AADT of the feeding inlets.

• In fact, the section aggregation cannot be seen, but the sections are shorter and better

consolidated referring to similar AADT’s


19

Figure 8 Improved model for NSM


Legend

Perimeter ZH

Intersection

Reclassified roads



Residential zone

Non living area

±0 1 2 3 4

Kilometers

Legend

Perimeter ZH

Intersection

Reclassified roads



Residential zone

Non living area

±0 1 2 3 4

Kilometers

Legend

Legend

Perimeter ZH

Intersection

Reclassified roads



Residential zone

Non living area

±0 1 2 3 4

Kilometers


21

3 Results of various analyses

This chapter presents the results of the NSM model with different input parameters. In all the

analyses the improved basic network model from chapter 2.3 is applied. The priority of the

network elements is illustrated by different colours. A broader overview of the results is given

in form of tables and further figures in the corresponding appendixes.

For a better understanding of the maps the following legend in Table 5 explains the different

accident patterns. A legend is only enclosed for network elements with the highest priority.

Table 5 Legend of accident patterns of highly prioritized network elements

Accident type Accident cause Participation


Different accident patterns are given in the following figures. All the symbols in the following

figures depict the most frequently protocolled accident pattern. The symbols are independent

and do not stand for conspicuous accident patterns. The column labelled participation shows

the mode of transport that, apart from the cars, is second most involved in accidents. The digit

next to the accident represents the number of accidents that led to fatal or serious injuries.

3.1 Prioritization with the given input parameters

With the exception of the change of the basic network model, this chapter follows exactly the

same assumptions and input parameter from the primary model of DECURTINS, 2014.

Rear-end collision

Collision with pedestrian

Turning collision

Lane changing

Grazing collision

Obstacle collision

Bicycle

Tram

Heavy traffic

Pedestrian Alcohol suspected

Right of way

Disregarded pedestrian

Lane changing

Disregarded tram

Crossing

Deflection


22

3.1.1 Traffic oriented roads

In Figure 9 the resulting map of the prioritization of traffic oriented roads is displayed.

Figure 9 Results of traffic oriented roads


60 – 100% of the avoidable accident cost per year


Upper 20% of the avoidable accident cost per year

No infrastructure potential

Legend

±0 1 2 3 4

Kilometers

16

13

8

9

7

14

Categories: FSI, MI, PDO

Number of accidents: 5'781


23

Six sections manifest a cumulative 20% of the total avoidable accident cost per year and con-

sequentially are ranked as highest priority. In comparison to other traffic oriented road sections,

all these have a high number of accidents in relation to their actual traffic volume. According

to the given formulas of NSM (see appendix A 1) that results in a high ACD and relatively low

baACD. Rear-end collisions are the most recorded accident type among four of these road sec-

tions. This indicates that there is often congestion and therefore the inattention of car drivers

can lead to a collision with the car in front. With the exception of Albisstrasse, all of these

highly prioritized roads are surrounded by commercial shops, so that there is a lot of interaction

with not only motorized traffic but also human powered mobility.

By trend, roads with a high AADT cause less avoidable accident cost. At first glance it seems

illogical due to the positive relationship of AADT and ACD (ELVIK, HØYE, VAA, & SØRENSEN,

2009). But on the other hand, roads with a high traffic volume often have better infrastructure

facilities and might have more overpasses or underpasses in order to separate different road

users (RETTING, FERGUSON, & MCCARTT, 2003). This leads to less conflicting points among road

users and therefore a higher specific road safety.

An overview of the ranking of all sections resulting in high or medium priority including the

corresponding number of accidents is shown in appendix A 2.

3.1.2 Nodes

Figure 10 presents the priorities of InfraPo for the nodes. There are 10 nodes with high, 50 with

medium and 150 with low priority. Only four nodes have no infrastructure potential. More de-

tails can be seen in appendix A 3.

The assigning of the individual accidents depends on the size of a node. Therefore, the digital-

ization of large nodes is compared to the actual size of the intersection and sometimes nodes

nearby are merged into one big node. Because of this, bigger nodes usually include more acci-

dents. The calculation divides the accident cost to the extent of the node, which is represented

by the number of feeding inlets. But this often causes large nodes to have a higher value of

avoidable accident costs.

The given results confirm this issue. Bellevue and Bucheggplatz, as the two largest digitalized

nodes, are ranked as first and second priority, respectively. Mainly rear-end collisions while

entering the node and accidents caused by lane changing maneuvers are the most depicted ac-

cidents. An interesting fact is that both at Limmatplatz as well as at Central, the road users most

frequently involved in accidents, are the pedestrians. Both nodes have a high volume of pedes-

trians and are operated without traffic signals or separated pedestrian crossing facilities.


24

Figure 10 Results of nodes






Legend

±0 1 2 3 4

Kilometers

5

6

8

7

10

3

5

5

6

5




25

Residential zones

Similarly, the map of priorities for residential zones is illustrated in Figure 11.

Figure 11 Results of residential zones


±0 1 2 3 4

Kilometers

60 – 100% of the accident cost density per year

20 – 60% of the accident cost density per year

Upper 20% of the accident cost density per year

No recorded accidents

Legend



16

12


26

In contrast to the calculation for traffic oriented roads and nodes, the analysis for residential

zones consists only a ranking of the existing ACD. An AADT-based analysis is not necessary

owing to the aim of residential roads, to minimize traffic as far as possible and rather provide

an optimal road design for the existing local traffic. As a consequence, a theoretical value for

the baACD cannot be calculated. Hence, it gives more of an indication where most accidents

on residential roads happen, and as soon as one accident in a zone is recorded, the zone appears

with a potential for improvement.

The distribution of the prioritized residential zones shows a clear focus on the inner city part of

Zurich. The two old city zones along the river Limmat make up a total of 17% of the entire

residential zone’s ACD. The few traffic oriented roads, the density of the public transport sys-

tem and a high modal split of human powered mobility are, amongst other things, reasons for

this. Primarily, Bahnhofstrasse and Limmatquai, which are forbidden for cars, have a high

amount of accidents and heavily influence the final results. Most accidents are driving accidents

affected by collision with obstacles (tram tracks, signals, parked vehicles) or other road users.

Further zones with a substantial cumulative percentage of the total ACD are located around the

main station and in the area of Sihlfeld, Altstetten and the industry zones. This industry requires

a special focus: Despite having only a few accidents, these zones result in relatively high ACD.

Owing to a dispersed road network with only a short total length, the value for the ACD rises

and suggests an immoderately high network priority.

3.1.3 Traffic oriented roads and nodes merged

Taking the sum of priorities of adjacent roads and nodes, it seems as if a high priority in one

element causes another high value of a different network element. Although SNR intends that

traffic oriented roads and nodes should be treated separately, infrastructural improvement

measures are often planned in a combination of roads and nodes. Besides, the traffic regime at

nodes is always influenced by the inlets. In Figure 12, the merging results of the sum of the

ranking of the roads as well as of the nodes are marked by different colours.


27

Figure 12 Merging results of the rankings of traffic oriented roads and nodes


At least 1x low priority

2x medium priority

1x high, 1x medium priority

Node and traffic oriented road with high priority

Legend

±0 1 2 3 4

Kilometers




28


3.1.4 Differences in the ranking to the existing model

In face of the implemented changes to the existing model of DECURTINS, 2014, it is essential to

know how these improvements have affected the final outputs. The results of the primary NSM

analysis are attached in appendix A 5.

It can be said that for the two network elements traffic oriented roads and nodes, the results

look pretty similar. Small changes among the ranking of road sections can be found within the

inner city and around Utoquai. The road with the highest priority in both models remains

Badenerstrasse5. The two highest prioritized nodes in both approaches are Bellevue and

Bucheggplatz. Instead of six nodes with the highest priority, the improved model results in a

total of ten nodes which comprise the upper 20% of the accident cost density per year.

As a result of different input layers, major changes are apparent for the residential zones. Alt-

hough zones such as the industrial parts with their very small network length of minor roads

still have a high ranking, they are no longer the ones with the highest priority. Instead, zones in

the inner city with a widespread distribution of different accident types are ranked on top. Table

6 displays the most relevant dissimilarities between the existing and improved model.

5 Due to a large network length, Badenerstrasse is split into two sections in both models. These two sections are

ranked as first respectively second in the improved model.


29

Table 6 Differences of the results between the two models

Traffic oriented roads Nodes Residential zone

Results Existing

model

Improved

model

Existing

model

Improved

model

Existing

model

Improved

model

Number of entities 145 208 156 214 71 72

Number of highest

priority

7 6 8 10 3 2

Number of medium

priority

19 24 36 50 16 14

Number of entities

without InfraPo

66 97 11 4 2 1

Max value of avAC

[1’000CHF/a]

2’465 2’124 2’644 2’565 1’6436 1’1216


Besides an explicit increase in the absolute number of different entities for roads and nodes, the

outputs are concordant with both the location of the hotspots as well as the classification of the

allocation of priorities. The slight decline in the maximum value of avAC can be traced back to

the more detailed aggregation level. In both models, the only factor that has not changed is the

number and location of accidents. The similar results show that despite the creation of a differ-

ent section, the absolute number of each accident category within a certain network element has

a major influence on the final results. A further discussion of the effects of changing input

parameters can be found in the subsequent chapter.

3.2 Variation of different parameters

Apart from the change of the network elements, the results of the previous chapter are computed

with the given input parameters of SNR and FEDRO, 2014b. As mentioned above, the outcome

of the model depends not only on the basis of the network model but also on the previously as

fixed assumed accident cost rates (ACR) and basic accident cost rates (baACR). Below are the

different parameters tested in altered variations.

6 Maximum value for ACD is used instead of avAC.


30

3.2.1 Accident cost rate (ACD)

In 2014 FEDRO presented a document in collaboration with the PTV group which elaborates

the different ACR and baACR for the two road safety tools NSM and Road Safety Impact As-

sessment (RIA). A distinction between cost factors of different road categories and rural and

urban areas are thereby exposed by FEDRO, 2014b. The relevant accident cost factors for the

input model are already listed in Table 1.

The category urban area is still an abstract term that does not distinguish between an urban area

in a city (big and dense road network) and urban areas of smaller towns or villages (only a few

traffic oriented roads). The dedicated space for different road users is often limited in inner

urban areas, so that more conflict points occur. As a result, accident patterns are different among

various urban areas (HOLZ-RAU & SCHEINER, 2013).

Own accident costs are calculated on the basis of the available number of accidents in the city

of Zurich in the period from 2009 till 2013. Thereby, a distinction between both the accident

cost for three accident severity categories as well as different accident types is calculated. In

order to do this, the average costs that each casualty of an accident generates are required. Table

7 lists the cost rates for different casualty categories for Switzerland and Germany. For each

person in an accident they state an according casualty cost rate. The more people involved in

an accident, the higher the casualty cost rate per accident. In contrast, the ACR stands for a

mean value of AC caused by one crash, independent of the number of people involved.

Table 7 Casualty cost rates of Switzerland and Germany

Accident severity category Switzerland: Casualty cost rate

[CHF/casualty]

Germany: Casualty cost rate

[Euro/casualty]

Fatal 3’191’421 1’161’892

Serious injury 488’907 116’151

Minor injury 33’471 4’829

Property damage only 44’824 5’951 / 20’8087

Source: FEDRO, 2014b, BUNDESANSTALT FÜR STRASSENWESEN (BAST), 2014

The definition of the severity category serious injury includes hospital stay of more than 24

hours. Accordingly, a minor injury is categorized as an accident as a result of which the injured

7 Germany distinguishes between two different property damage categories


31

person is able to leave hospital within one day. The cost itself includes the direct accident costs

and the follow-up costs.

A comparison with the casualty cost rates of Germany gives an indication of how cost rates can

vary among different sources. The more expensive living costs in Switzerland cannot be the

only reason for the noticeably lower cost rates of the neighbouring country. Rather, differences

in data collecting and data processing make it difficult to have comparable values. More details

of the data generation can be found in (BUNDESAMT FÜR RAUMENTWICKLUNG (ARE), 2006) and

(BAUM, KRANZ, & WESTERKAMP, 2010). The following calculations are all based on the casualty

cost rates of Switzerland.

Own accident costs of different accident severity categories

According to ELVIK, HØYE, VAA, & SØRENSEN, 2009 accident severity varies among different

network elements. To test this with the available accident data of Zurich own accident costs of

severities are calculated. Thereto the following steps are made:

• Assigning the number of accidents to the three network elements

• Multiplying the number of involved persons by the according ACR

• Summing up the individual AC of each network element

• Dividing the total sum by the total number of accidents within each element

Similarly to the calculation of the infrastructure potential, the assumption that in each accident

involving people, there is also property damage remains valid. Table 8 shows the calculated

ACR for each accident severity differentiated by the network elements. The number of exam-

ined accidents is 13’0118. More details are given in appendix A 6.

8 The number slightly differs from the total number of accidents. Some accidents could not be assigned to one of

the three network elements.


32

Table 8 Own calculated ACR of accident severity categories

Network element ACR(FSI)

[CHF/acc]

ACR(MI)

[CHF/acc]

ACR(PDO)

[CHF/acc]

Traffic oriented road 692’000 84’144 44’824

Node 650’564 83’918 44’824

Residential zone 641’935 82’707 44’824

Road & node 675’132 84’055 44’824

Average 664’908 83’706 44’824


One of the main reasons for dividing the accidents into the three elements was an expected

decrease in the ACR in residential zones due to a generally lower speed limit than on traffic

oriented roads. Although there is a slight decrease in ACR, it is not as evident as anticipated.

Accordingly, there are about the same percentages of accident severity categories in all the three

network elements in the past five years. The similarities for MI accidents are even greater and

for PDO accidents no distinction could be made due to the condition that in each accident a

maximum of one PDO can be noted.

Generally, it is remarkable how concordant the calculated ACR are with the given ACR of the

Swiss norm from Table 1. Therefore, no further analyses of the NSM model are made. A critical

discussion of the calculated ACR is presented in chapter 5.1.3.

Own accident costs of different accident types

Similar to the accident severity categories, own ACR are calculated for the different accident

types. Each protocolled accident gets assigned an accident type. FEDRO distinguishes between

11 main categories, whereas accidents with animals hardly exist in Zurich. The number of cat-

egories is thus reduced to ten. Although it is guaranteed that each category holds more than 100

accidents, statistical uncertainty can still occur. To further minimize this variation, a higher

number of mainly personal accidents should be included. Of the roughly 13’000 accidents, the

minimal number of accidents in a category is 102. However, around 60% of these are PDO

accidents. Table 9 lists the ACR for different accidents types and separates them for roads and

nodes, and zones. Notice that for each accident type all the accident severities are included. A

more detailed table including the percentages of each accident type can be seen in appendix A

6.


33

Table 9 Own calculated ACR of accident types

Network element Traffic oriented roads & Nodes

ACR(FSI+MI+PDO)

Zones

ACR(FSI+MI+PDO)

Driving accident 110'248 114'485

Lane changing 67'242 70'657

Collision 74'352 65'447

Turn off 110'510 110'094

Turn into 104'848 89'607

Crossing 105'155 69'049

Frontal collision 80'046 73'874

Parking 48'622 45'444

Pedestrian 287'057 224'446

Other 101'393 91'684

Average 108‘947 95‘479


This time traffic oriented roads and nodes are combined in one category. This is due to the

presence of smaller nodes of incoming residential roads that are already assigned to traffic ori-

ented roads. Both elements are specifically built for a certain traffic capacity and differences in

accident cost of individual accident types would not be as significant as compared to residential

roads.

In contrast to the previous table there are no obvious discrepancies among the accident types.

The following points try to interpret the most relevant results of the calculation:

• The lowest ACR occur in parking accidents. Only 21 individuals were injured in 856

accidents. The remaining accidents ended solely in property damages.

• Lane changing and collision accidents also most often end in property damage only.

These accidents are often affected by congestion, because of the speed and therefore

the severity of the crash situation is low.

• Compared to the turn off accidents there is a difference between turn into accidents

of different network elements. A turning maneuver into a traffic oriented road gener-

ates on average higher accident costs than turning into a residential road.

• Crossing accidents on traffic oriented roads and nodes also result in a higher ACR

than on residential roads


34

• The most severe accidents are the ones with pedestrian participation. In nearly 2% of

pedestrian accidents on traffic oriented roads and nodes, a pedestrian gets killed. Ad-

ditionally, around 30% of these accidents result in serious injuries for the unprotected

pedestrians.

• The difference of more than CHF 14’000 between the averages of roads and nodes,

and zones per accident confirms that accidents on traffic oriented infrastructures tend

to result in more severe accidents. This corresponds with the given positive correla-

tion of the traffic volume and accident frequency (see also chapter 3.3.1).

3.2.2 Basic accident cost rate (baACR)

Besides the actual accident cost rates, NSM requires a comparable value in order to find out

what the avoidable cost of each network entity is (Figure 3). This equivalent term is the basic

accident cost rate (baACR), which stands for a monetary value of the expected accident cost

per one thousand kilometers driven. It is an abstract value that represents a best practice design

of each network element. According to FEDRO, 2014b, an ideally built road section is located

outside of a central area and includes, amongst other things, a constructional lane separation.

Numerous measures for creating an optimal road network configuration can be found in the

literature (ELVIK, HØYE, VAA, & SØRENSEN, 2009), (YANNIS, EVGENIKOS, & PAPADIMITRIOU,

2008). The basic approaches of making roads and nodes more safe are often in combination

with reducing the speed limit and operating nodes in form of roundabouts or traffic signals.

The calculation of baACR demands a lot of input attributes and needs advanced knowledge in

the analysis of multiple criteria. For this reason, first a simplified approach is presented by

calculating an own baACR on the basis of the average baAR of Zurich.

Average baACR of Zurich

Separated by traffic oriented roads and nodes, a basic accident rate (baAR9) of each element is

calculated by the following formula:

𝑏𝑎𝐴𝑅𝑛𝑒𝑡𝑤𝑜𝑟𝑘 𝑒𝑙𝑒𝑚𝑒𝑛𝑡 =𝑁∗106

𝐴𝐴𝐷𝑇∗𝐿∗365∗𝑎 (3)

with

baAR Accident rate [acc/(106*veh*km)]

N number of accident(FSI, MI, PDO)

9 Although the literature calls it accident rate (AR) to clarify that it is in this case a matter of calculating a theoret-

ical value, it is from here on basic accident rate (baAR).


35

AADT annual average daily traffic [veh/day]

L length [km]

a number of years

Source: SN 640 009a, 2006

The term baAR does not distinguish between the severities of accidents. Consequently, some

entities with a high number of accidents can distort the result. For traffic oriented roads, Figure

13 shows a distribution of the baAR skewed to the right.

Owing to them not having a normal distribution, all the baAR values greater than five are

skipped. As a result, a mean baAR value of 1.78 accidents per million driven vehicle kilometers

is calculated for the city of Zurich. The same analysis is made for the nodes (appendix A 7).

For nodes, baAR values greater than seven are not taken into account and the mean value is

2.98 accidents per million driven vehicle kilometers.

Figure 13 Distribution of the accident rate of traffic oriented roads

0

2

4

6

8

10

12

14

16

18

20

0

0.4

0.8

1.2

1.6 2

2.4

2.8

3.2

3.6 4

4.4

4.8

5.2

5.6 6

6.4

6.8

7.2

7.6 8

8.4

8.8

9.2

9.6 10

Fre

qu

ency

baARtraffic oriented roads [acc/(106*veh*km)]

Histogram baARtraffic oriented roads


36


By multiplying these baAR values with the given ACR from the Swiss rule, it is possible to

calculate the corresponding baACR. Table 10 lists the given basic rates presented by FEDRO,

compared to the calculated average baAR and baACR of Zurich.

Table 10 Comparison of the basic rates of FEDRO and the average rates of Zurich

Network element Basic rate(FEDRO) Basic rate(ZH)

baAR10 baACR11 baAR baACR

Traffic oriented roads 1.84 175 1.78 170

Nodes(all) 0.47 42 2.98 248

Nodes(3 inlets) 0.27 22.6 3.00 250

Nodes(>3 inlets) 0.60 49.7 2.9 245

Source: FEDRO, 2014b, own research

It can be seen that the baAR respectively the baACR of traffic oriented roads both of Zurich

and FEDRO are very similar. This means that despite a typical inner urban network, the risk of

accidents on traffic oriented roads is not higher than on other traffic oriented roads in Switzer-

land. For nodes it is the complete opposite. Nodes are additionally divided according to their

number of inlets. The baAR and thus the baACR of nodes in Zurich are much higher than in

the rest of Switzerland. The difference of a higher accident risk at nodes with more than 3 inlets

cannot be approved for Zurich with the average approach. A proper conclusion for the consid-

erable increase in the baAR respectively baACR at nodes cannot be given, though it seems

reasonable that the shorter average node distance and thus the higher node density are reasons

for this rigorous increase. In addition, there are more pedestrians and more complex nodes in

cities than in other urban areas (FEDRO, 2014a).

Coefficient approach

The method above represents a very rough approach of calculating an average baAR, respec-

tively baACR of all traffic oriented road and node entities. To advance this approach, the avail-

able input data is applied to a method, developed by SCHÜLLER, 2015, which results in specific

10

The unit of baAR is [acc/(106*veh*km)]

11 The unit of baACR is [CHF/(1000*veh*km)]


37

outputs of coefficients for each network element. These coefficients are subdivided by the cor-

responding accident severity category. Table 11 demonstrates the coefficients for traffic ori-

ented roads and nodes.

Table 11 Values of the coefficient baACR approach

Coefficient Roads Nodes

FSI MI PDO FSI MI PDO

k 0.0225 0.0249 0.0362 0.0011 0.0208 0.0026

y 0.3369 0.4862 0.5284 0.5757 0.4299 0.7206

Inlet12 - - - 0.6999 0.5667 0.6754

Source: SCHÜLLER, 2015

By taking these coefficients, the accidents per year (acc/a) for each accident severity category

can be calculated by the following terms.

(𝑎𝑐𝑐/𝑎)𝑛𝑜𝑑𝑒𝑠 = 𝑘 ∗ 𝑖𝑛𝑙𝑒𝑡 ∗ 𝐴𝐴𝐷𝑇𝑦 (4)

(𝑎𝑐𝑐/𝑎)𝑟𝑜𝑎𝑑𝑠 = 𝑘 ∗ 𝐿 ∗ 𝐴𝐴𝐷𝑇𝑦 (5)

with

acc/a accidents per year

k coefficient (Table 11)

y coefficient (Table 11)

inlet number of inlets to a node

AADT annual average daily traffic [veh/day]

L length [km]

The acc/a(nodes/roads) can be converted to the common unit of baAR by the following term.

𝑏𝑎𝐴𝑅 = 106∗(𝑎𝑐𝑐/𝑎)𝑛𝑜𝑑𝑒𝑠,𝑟𝑜𝑎𝑑𝑠

365∗𝑎 (6)

12

If the number of inlets to a node is equal to 3, the value for the inlets stays 1.


38

with

baAR basic accident rate [acc/(106*veh*km)]

a number of years

Sources: SCHÜLLER, 2015

Analogous to the approach of the average rates, the baAR can be multiplied by the given ACR

to get the baACR. The extended table below summarizes the values of the different baAR re-

spectively baACR from FEDRO, the average rate and the coefficient approach (Table 12). No-

tice that, for the coefficient approach, the aggregated values baAR and baACR of all the acci-

dent severity categories are presented.


39

Table 12 Comparison of different basic rates

Network element Basic rate(FEDRO) Average rate

approach(ZH)

Coefficient

approach(ZH)

baAR13 baACR14

baAR baACR baAR baACR

Traffic oriented roads 1.84 175 1.78 170 2.06 215.8

Nodes(all) 0.47 42 2.98 248 0.59 59.2

Nodes(3 inlets) 0.27 22.6 3.00 250 0.51 51.8

Nodes(>3 inlets) 0.60 49.7 2.9 245 0.72 71.6

Source: FEDRO, 2014b, own research, SCHÜLLER, 2015

In contrast to the average rate approach, the results of the coefficient approach predict a higher

risk on traffic oriented roads than the other values. For nodes however, the baAR is similar to

the given values from FEDRO. The difference of nodes with three inlets, respectively nodes

with more than three inlets can, in contrast to the previous approach, be approved by this

method. The coefficient approach calculates a higher risk on nodes with more than three inlets

in Zurich. This is plausible due to the fact that greater nodes usually have more conflict points

and are often operated by several mode of transports.

Notice that for all the calculations above, no additional infrastructure attributes are used. Hence,

both approaches represent more of an arithmetic value of the accident rates than real basic ac-

cident costs, which are usually characterized by a best practice design.

Consequences of different baACR

The influence on the NSM ranking for traffic oriented roads and nodes is given in appendix A

7. Both approaches lead to a decline in priority for some of the previously highest prioritized

nodes. But there are no remarkable changes for the rankings. For traffic oriented roads however,

the differences between the coefficient approach and the given parameters by FEDRO are sig-

nificantly more obvious. Three out of the six highest prioritized road sections were previously

ranked as medium priority. It can be said that, by trend, the ranking of roads with a higher traffic

volume has increased. As a result the heavily used Rosengartenstrasse, for which originally no

InfraPot is predicted, arises for the first time in the higher rankings (see Table 9 in appendix A

7).

13

The unit of baAR is [acc/(106*veh*km)]

14 The unit of baACR is [CHF/(1000*veh*km)]


40

3.2.3 AADT of 2030

The results of chapter 3.1 show a considerable dependence on the AADT. For the city of Zurich,

a growth of traffic volumes both for cars and public transport is predicted (Tiefbauamt Kanton

Zürich, 2013). Due to a positive correlation of AADT and the number of accidents (see also

chapter 3.3.1), a further increase of the absolute number of accidents is expected.

The GVM of the canton of Zurich includes data for the estimated traffic volume for the year

2030. Only for public transport is a 30% growth forecasted. Although the city tries to encourage

public transport as best as possible, an increase in motorized vehicles is expected. The growth

will not occur everywhere at the same rate. A greater growth is expected in the northern and

western parts of Zurich (VBZ, 2013). The traffic volume for 2030 is given by the average annual

weekdaily traffic volume. To convert this to AADT, it is divided by the factor 1.083

(BUNDESAMT FÜR RAUMENTWICKLUNG (ARE), 2010).

When replacing the actual traffic volume with the predicted one of 2030, changes in the prior-

itization of traffic oriented roads and zones can be observed (Figure 14). For six traffic oriented

roads and one node (junction Herdernstrasse/Bullingerstrasse), the model predicts an increase,

respectively for 28 road sections and six nodes a decrease in priority.

Notice, that the figure only represents a rough estimation of predicted changes. It can discover

evidence of inappropriate road designs in the future, but for a more detailed analysis, other

parameters have to be adjusted, too. To reduce this randomness of results, model-based safety

tools such as RIA are more suitable.


41

Figure 14 Change of priorities with AADT from the year 2030


3.2.4 Node size 30 meters

Usually the extent of nodes should consider the influence of the inlets. Choosing the right lane

and braking maneuvers are typically made before entering a node. As a result, accidents not

only happen at conflicting points in the node itself but also in the approaching area. Further

information about the dimension of influential inlets on nodes can be found in AURICH, 2012.

±0 1 2 3 4

Kilometers

Decrease of priority

Unchanging priority

Increase of priority

Legend




42

In spite of the importance of inlets, an analysis with an assumed node size of 30 m is examined.

On the one hand, it should better represent the impact of conflicting points in a node itself and

on the other hand, the approaching lengths of nodes in inner urban areas are often reduced due

to lower speed limits. Notice, that in all the analyses, nodes between one traffic oriented road

and one or more residential roads are excluded. Instead, these nodes are assigned to traffic

oriented roads.

All the results of reducing the node size to 30 m are presented in appendix A 8. In Table 13,

only nodes with the highest priority are listed. The right side shows the primary rankings re-

spectively the priorities of the evaluation from chapter 3.1.2.

Table 13 Comparison of node size 50 m and 30 m

Name Node size 30 m Node size 50 m

Ranking No of

accidents

Priority Ranking No of

accidents

Priority

Bellevue 1 107 high 1 135 high

Bucheggplatz 2 85 high 2 99 high

Heimplatz 3 134 high 3 140 high

Schwamendingerstrasse/

Dörflistrasse

4 32 high 10 35 high

Langstrasse/Lagerstrass

e

5 40 high 8 50 high

Bürkliplatz 6 45 high 6 61 high

Limmatplatz 7 27 high 5 51 high

Schaffhauserstrasse/

Seebacherstrasse

8 20 high 15 24 medium

Stauffacher 9 38 high 16 46 medium


In general, the impact of changing the extent of nodes is pretty low. Although the ranking has

changed slightly, the priority has stayed the same in the majority of cases. For node size 30 m,

only two nodes are newly classified as high priority, whereas for three nodes a lower priority

can be obtained. One of the latter, Escherwyssplatz, has recently even been downgraded to low

priority. This is due to an explicit decrease in the absolute number of accidents when assuming

a node size of 30 m.


43

Again, this analysis only serves to obtain approximated evidence about the sensitivity of NSM

rankings for nodes under different node sizes. In reality, the situation is still better represented

with the primary approach of a node size of 50 m. Moreover, it shows that the extent of the

approaching area of inlets is not decisive for the final node rankings.

3.2.5 Nodes ranking based on incoming vehicles

So far, the calculation of nodes’ ACD has always been dependent on the number of inlets,

respectively the network length within a node’s diameter (see chapter 2.1.1). In fact, the net-

work configuration of nodes cannot be represented accurately enough by assuming circles of

50 m. So, a division by the individual network length does not really make sense and can cause

inaccuracy in the calculation itself.

To compare the results with the primary rankings from chapter 3.1, another method is applied.

It focuses more on the number of incoming vehicles and distinguishes between main und minor

traffic streams. The corresponding formula is

𝐴𝐶𝐷𝐼 = 𝐴𝐶∗1000

√𝐴𝐴𝐷𝑇𝑚𝑎𝑖𝑛∗𝐴𝐴𝐷𝑇𝑚𝑖𝑛𝑜𝑟∗365∗𝑎 (7)

with

ACDI Accident cost digit [CHF/(1’000*veh*km)]

AC Accident cost [CHF/acc]

AADTmain Annual average daily traffic volume of the main road [veh/d]

AADTminor Annual average daily traffic volume of the minor road [veh/d]

a number of years

Source: MATTHEWS, 2009

When applying this formula, a proper identification of the main and minor roads is essential.

Therefore, a verification of the allocated inlet traffic streams had to be done manually. The

results are presented in appendix A 9 and illustrated as an overview in Figure 15.


44

Figure 15 Rankings of nodes with the accident cost digit (ACDI) method


±0 1 2 3 4

Kilometers





Legend




45

It stands out that the formerly top ranked node Bellevue has dropped down to medium priority.

The reason for this is the heavy total traffic volume of nearly 60’000 vehicles per day. This

method generally gives more weight to the absolute traffic volumes, so that rankings of other

nodes with a high demand have decreased (Bürkliplatz, Manessestrasse/Tunnelstrasse). In con-

trast, the only moderately used node of Wallisellenstrasse/Saatlenstrasse with its total of 10

accidents experienced a dramatic increase in avAC. By trend, the results are quite congruent

with the primary ranking of nodes’ infrastructure potential.

3.2.6 Assigning accidents of nodes to traffic oriented roads

In chapter 3.1.3, the resulting priorities are merged in order to see if traffic oriented roads are

leading to a similar prioritization as nodes and vice versa. It cannot be fully approved though

by trend, road inlets of high prioritized nodes also exhibit an increased InfraPo (see Figure 12).

In order to continue this approach, an analysis is done by ignoring the partition of the two

network elements traffic oriented roads and nodes. Instead, the length of both ends of road

sections is extended, so that all the accidents on the traffic oriented road network are assigned

to roads. The main problem with this method is an existing overlap of the polygons of the

extended road inlets on nodes, so that some accidents are doubly assigned on road inlets. Nev-

ertheless, this method gives an idea of the significance of a node’s approaching area compared

to the road section itself. For example, if a road section only has a few accidents but on the

recent extended part (the one of the previous node area) there are lots of accidents, the AC and

also the ACD increases. Thus, the InfraPo changes and there might be a change in priority (see

complete list in appendix A 10). Figure 16 only highlights traffic oriented roads that either got

an increase due to a plain growth of assigned accidents or a decrease in priority. Generally,

there are more roads that have recently received a higher priority than with the primary approach

of a clear distinction between nodes and traffic oriented roads.





Legend


46

Figure 16 Change in priorities when assigning the accidents of nodes to traffic oriented roads


±0 1 2 3 4

Kilometers

Decrease of priority

Unchanging priority

Increase of priority

Legend




47

3.3 Sensitivity analysis (SA)

The calculations of the cost terms in the previous chapter have shown that, depending on which

input cost rates the NSM model is based on, different results can be expected. Extending this

approach in the coming chapter, the required NSM model parameters are examined by a sensi-

tivity analysis. These are notably the following parameters:

• Network length [km]

• AADT [veh/d]

• ACR [CHF/acc]

• baACR [CHF/(1’000*veh*km)]

On the one hand, it can be interesting to see what the absolute numbers of the ranges of the

input parameters are, but on the other hand, testing different relations between individual factors

can emphasize the importance of certain parameters. Due to a lack of sufficiently accurate data

for AADT, and therefore no indication of a baACR for residential zones, the following analyses

are made for traffic oriented roads and nodes.

3.3.1 Correlation of parameters

Variations of input parameters require prior knowledge about the relationship between different

factors. Figure 17 illustrates the effects of several significant factors for accident severity, re-

spectively accident frequency. The research is based on the accidents from the whole of Swit-

zerland in the period from 2009 to 2012. Traffic signals and roundabouts are the only parame-

ters that have a negative influence on both accident severity and accident frequency. The exist-

ence of tram and a high commercial use are often in combination with higher pedestrian volume

and accordingly show a positive relationship (see also chapters 0 and 3.6).


48

Figure 17 Significant parameters as a function of accident severity and accident frequency

Source: FEDRO, 2014a

The factor AADT shows a negative influence on the accident frequency. The less accidents, the

more severe they are. One reason for this is an increased percentage of driving accidents, which

often cause severe injuries.

Regarding the connection between AADT and the absolute number of accidents, AADT shows

a positive relationship. A correlation analysis of traffic oriented roads between AADT and the

present number of accidents is illustrated in Figure 18. As mentioned in the literature, AADT

does not show a linear relationship to the accident frequency (AURICH, 2012), (FEDRO, 2014a).

Acc

iden

t se

veri

ty

Accident frequency

-

- +

+

Traffic signal/

Roundabout AADT

High commerical

use Tram

> 3 Inlets Density of nodes

< 3 inlets

Parking

> 2 Lanes


49

Figure 18 Correlation analysis of traffic oriented road sections


The graph above demonstrates that an increase in AADT leads to a declining increase rate in

the number of accidents. Hence, among roads with a low traffic volume, an increase in AADT

has a more serious influence on the corresponding ACD, compared to an AADT increase in

more heavily used roads. Similar results can be found for the accidents of the entire urban area

of the canton of Zurich in the period of 2009 till 2012 (FEDRO, 2014a).

The Pearson correlation coefficient of r = 0.42 indicates a medium strong positive relationship

between this two parameters. The same correlation analysis is performed for the length of traffic

oriented road sections and the corresponding accidents. The correlation of r = 0.27 in that case

is explicitly lower. Generally, the relationship between length of road sections and accidents

tends to be linear, so that an increase of section length induces a proportional increase in acci-

dents (SCHÜLLER, 2009).

The same correlation analysis is made for nodes. Referring to formula (1), only a limited num-

ber of network lengths exist and therefore a correlation analysis is not meaningful. The plots of

the remaining two correlation analyses are shown in appendix A 11.

0

20

40

60

80

100

120

140

0 10'000 20'000 30'000 40'000 50'000 60'000

No o

f acc

idet

ns

AADT

Correlation AADT/Accidents

r = 0.42

Road section

logarithmic trendline


50

3.3.2 Testing different samples of parameters

The outcome of NSM relies on the four described parameters described above. To test the sig-

nificance of the individual factors, a SA on the basis of the model from GE & MENENDEZ, 2014

is implemented. The model consists of an efficient and qualitative SA approach. It is able to

screen the most important parameters of a model based on a computation and comparison of

the SA indexes. Considering the given distribution of data, the model creates different data

samples and aims to identify the most influential parameters. For that purpose, different as-

sumptions have to be made.

• AADT: Similarly to the calculation of the baACR, the model omits extreme values in

order to approximate the data to a normal distribution. The data range is split into 10

sections. The threshold value of AADT that identifies a jump to the next section is

taken as the changing input variable for the model of GE & MENENDEZ, 2014.

• Length: Identically to AADT, the data range is again split into 10 sections and the

threshold values are taken as input values. Consider that the distribution of the length

has many numbers in the lower region, so that the absolute deviation of the first two

values is very small.

• ACR(PDO) = CHF 45’000: This value is taken as the basis value and therefore not

changed.

• ACR(MI) = CHF 84’000: This value is varied on the basis of the ACR(PDO). The factor

of variation is [1,10] so as to enable that the minimum value is also CHF 84’000 and

the maximum value can be a maximum of 10*84’000 = CHF 840’000.

• ACR(FSI) = CHF 696’000: Identically, it is varied on the basis of ACR(PDO). To ensure

this cost rate is equal or higher than the primary one, the range of variation is [10,20].

• baACD: This value is used as the comparable value to the varied input parameters of

the actual accident data and therefore not changed.

These assumptions guarantee a variation of the input parameters, restricted to the given ranges.

However, the focus is not on the absolute values but rather on the differences in the relationships

between the parameters.

The model itself can only vary one parameter at a time. In each step, either the length, the

AADT, the ACR(MI) or the ACR(FSI) is varied randomly. As a consequence of four changing

parameters, the model creates five series samples. By repeating this a thousand times, you get

a total of 5’000 combinations of inputs for calculating the infrastructure potential. Figure 19

tries to summarize the process of creating different input combinations in a simplified way.


51

Figure 19 Process of model sampling for traffic oriented roads


Taking these 5’000 samples as the inputs for the NSM model, there is a problem with assigning

the accidents. As is generally known, the accidents are real data and consequently not changed

in the process of testing input parameters. For each network entity they have to be assigned

either to the AADT or the network length. Due to a stronger correlation with AADT (see Figure

18), the accidents are allocated to the corresponding AADT. An example of a five series sample

demonstrates this problematic (Table 14).


52

Table 14 Example of a five series sample

Length*

[km]

AADT*

[veh/d]

Acc(total) Factor*

ACR(MI)

Factor *

ACR(FSI)

ACR(PDO)

[CHF/acc]

ACR(MI)

[CHF/acc]

ACR(FSI)

[CHF/acc]

0.5 3’671 17 8 20 45’000 360’000 900’000

0.5 3’671 17 3 20 45’000 135’000 900’000

0.5 3’671 17 3 15 45’000 135’000 675’000

0.5 11’773 56 3 15 45’000 135’000 675’000

0.796 11’773 56 3 15 45’000 135’000 675’000


In every row, one of the four varied parameters (labeled with *) is altered. The value in the

column of Acc(total) is linked to the AADT and changes as soon as the AADT is altered (jump

in the fourth row). The two varied factors of MI and FSI are in each row multiplied with the as

fixed assumed ACR of PDO.

As mentioned above, the aim of the robustness test is not to quantify the impact of each input

parameter. Rather, the differences in the relationship between parameters are crucial. With the

help of the defined ranges of the ACR parameters, Table 15 shows the minimal and maximal

differences in ACR relations.

Table 15 Minimal and maximal relation of ACR

Relation ACR(accident) [CHF/acc] Relative Factor

ACR(PDO) ACR(MI) ACR(FSI) Factor(PDO) Factor(MI) Factor(FSI)

relation(original) 45’000 84’000 696’000 1 1.87 15.46

relation(min) 45’000 45’000 450’000 1 1 10

relation(max) 45’000 450’000 900’000 1 10 20


Compared to the given ACR from SNR, the test includes a broader scope of the relations be-

tween accident severity categories. This ensures that this approach already takes the impact of

prospectively changing ACR into consideration. An ACR smaller than CHF 45’000 for PDO is

excluded.


53

Results for traffic oriented roads

Applying these steps to the traffic oriented roads, the results in Figure 20 show an unambiguous

dominance of the AADT. Notice that the figure should be interpreted in a qualitative way. De-

spite the value of the axis, it cannot be said that the AADT has for example about three times

more impact on the results than the other parameters. It stands more for the relative differences

of the altered parameters. The units of the axis are dependent on the context of the inputs and

have no physical meaning in following figures (GE & MENENDEZ, 2014).

Figure 20 Impacts of different parameters on the results for traffic oriented roads


The model creates two different outputs, one for the absolute mean and another for the mean

values. Only the results for the mean value are illustrated above. In both, the AADT is the

dominant parameter, yet the significance of the length changes. Whereas for the absolute mean

a positive relationship is indicated, the evaluation of the mean value shows that the higher the

network length is, the less avAC arises. The impact of the cost factors stays the same for both

evaluations and always has a marginally positive impact.

Consider that the AADT has the greatest standard deviation of all inputs. Therefore, the AADT

does not increase every time the output of the model (avAC) also increases. The variation of this

AADT

Length

Factor(MI)

Factor(FSI)

0

500'000

1'000'000

1'500'000

2'000'000

2'500'000

-500'000 0 500'000 1'000'000 1'500'000

stand

ard

dev

iati

on

mean

Traffic oriented roads: Impact of different parameters

(mean)


54

parameter can cause strong non-linearity effects. A review of the generated sample dataset verifies

this aspect.

Results for nodes

For nodes, the same assumptions are considered. The only difference is that the parameter of

network length is omitted due to a restricted range of different values (see formula (1) and

chapter 3.3.1). The results for the evaluation of the absolute mean are illustrated in Figure 21.

Figure 21 Impacts of different parameters on the results for nodes


The graph looks similar to the previous one. AADT is dominating with a high standard devia-

tion, and the ACR of minor injury is more relevant than the one for FSI.

3.4 Network density of residential zones

The methodology of SNR 641 725 determines that residential zones are not assessed by a dis-

tance-based traffic volume and are only ranked according to their ACR (see chapter 0). Alt-

hough the improved network model from chapter 2.3 considers only the residential and working

zones on the basis of the land use plan of the city of Zurich, no information is provided about

AADT

Factor(MI)

Factor(FSI)

0

500'000

1'000'000

1'500'000

2'000'000

2'500'000

0 200'000 400'000 600'000 800'000 1'000'000 1'200'000 1'400'000

stand

ard

dev

iati

on

mean

Nodes: Impact of different parameters

(mean)


55

the area and particularly the density of the road network within each zone. However, the density

of a road network can give information about the urbanistic structure and because of this, dif-

ferent accident patterns may exist. Figure 22 represents a density map of the percentage of

residential road space in each zone. Remember, all the roads that are not assigned as traffic

oriented roads but have a speed limit greater than 0 km/h are classified as residential roads.

Figure 22 Percentage of dedicated road space of each zone

Source: TIEFBAUAMT STADT ZÜRICH, 2014, own research

±0 1 2 3 4

Kilometers

Legend

0.2 - 2.5%

2.5 - 5.0%

5.0 - 8.0%

8.0 - 9.1%


56

The map shows a different distribution than the NSM results of Figure 11. Due to many roads

being closed for cars and numerous pedestrian zones, the two inner city zones with the highest

ACD are not the ones with the highest residential road density. Instead, the zones around Lang-

strasse and Zurich Unterstrass are the ones with the highest road network density. The range of

densities varies from 0.2% (Zurich Manegg and industry Herdern) up to 9.1% (Beckenhof).

With the help of these road network densities, an enhanced analysis of residential zones can be

done according to the methodology of (Skeledzic, 2014). It consists of a comparison of each

zone in relation to a mean value of all zones together. To be more precise, it ranks the proportion

of the existing accident cost to the total cost of all the zones according to the ratio of network

length and zone area. The different steps of calculation and the ranking of the enhanced analysis

are given in appendix A 12.

In Figure 23 the zones are classified by different colours according to their priorities. In com-

parison to Figure 11, the zone of Leutschenbach was latterly prioritized as high. The other two

highest ranked zones in the inner city stayed top priority. Diverse other zones, which are located

in more remote areas, have resulted in an increase in priority. In contrast, the densest zones as

described above all have a low priority.

One of the main limitations of the method of ranking the ACD is the neglect of the zone size.

Assuming the same number of accidents respectively the same accident costs, smaller zones

are ranked significantly higher than larger zones. The enhanced model takes this into account

by creating a ratio between network length and zone size. The aim is to give larger zones more

weight due to the fact that zones with a longer road network but the same ACD cause effectively

more accident costs.

However, also the enhanced model cannot provide information about the distribution of acci-

dents within a zone. They could be either clustered at a specific location or could be spread over

the entire zone. In that case, the most effective impact on the ranking would be the elimination

of that specific cluster. By assuming the given zone size as the level of aggregation, these acci-

dent hotspots cannot be identified by NSM.


57

Figure 23 Priorities of zones when including network density


60 – 100% of the total ratio value

20 – 60% of the total ratio value

Upper 20% of the total ratio value

No recorded accidents

Legend

±0 1 2 3 4

Kilometers




58

3.5 Accident patterns of human powered mobility

The existence of pedestrian and bicycles has a major impact on road safety. As seen for the

calculated ACR of each accident type in chapter 3.2.1, the participation of non-motorized road

user provokes more severe accidents and therefore more accident cost. One of the main diffi-

culties is the quantification of pedestrian and bicycle volume. The traffic streams are typically

widespread and distances are shorter (WALTER & WEIDMANN, 2013). According to AURICH, 2012

there is also a significant relationship between accident patterns of human powered mobility

and its urbanistic surrounding. Mainly residential zones with a high density of living and com-

mercial use have an increased accident potential for pedestrians and bicycles.

Due to a lack of human powered mobility data, it cannot be determined where the highest po-

tential for improving the traffic situation for non-motorized traffic volumes is. Nevertheless, an

analysis of the accidents with the participation of pedestrians or bicycles can give evidence as

to where the hotspots of these accidents are. There are totally 2’522 accidents in which at least

one pedestrian or cyclist is involved. Of these, only 235 accidents resulted in PDO accidents,

which leads to the conclusion that around 90% resulted in minor, respectively serious or fatal

injuries to the involved people.

When only considering accidents of human powered mobility, the accident costs and therefore

ACD can be calculated. Figure 24 depicts the entities of the highest respectively medium rank-

ings for all the three network elements. Owing to an explicit decrease in the total number of

accidents, a few accidents within a network element of a small area (zone) or short length (traf-

fic oriented roads) can cause an excessive ACD. Hence, only entities with five or more acci-

dents are taken into account.

Despite having the highest population density, the zones Wiedikon, Langstrasse and Zurich

Unterstrass did not particularly result in the highest ACDs. In contrast, there are again the zones

in the inner city with a less dense network of traffic oriented roads but a particularly high per-

centage of workplaces are ranked on top. The most protocolled accident type is the collision

with obstacles inside or outside of the roadway.

For traffic oriented roads, Langstrasse is the one that has the highest ACD regarding human

powered mobility accidents only. The categories of high and medium priority are generally

more distributed for traffic oriented roads. This is compared to zones, where accidents are more

concentrated within a few zones.


59

Figure 24 Priorities including pedestrian and bicycle accidents only


Finally the most hazardous nodes for pedestrians and bicycles are often characterized by high

ACDs of traffic oriented roads and zones. The only exception is Bucheggplatz, which has a

difficult traffic routing for bicycles and a crossover for pedestrians, which is often ignored by

them.

20 – 60% of the ACD of bicycle and pedestrian accidents per year

Upper 20% of the ACD of bicycle and pedestrian accidents per year

Legend

±0 1 2 3 4

Kilometers




60

3.6 Impact of tram

Some of the inner city parts, such as Bahnhofstrasse and Limmatquai used to be classified as

traffic oriented roads. In order to attract more people and to support commercial shops, they

were redesigned into pedestrian passages where no cars are allowed to drive. Although cars are

by far the mode of transport most involved in accidents, these car-free roads still cause acci-

dents, but with a different accident pattern. Together with accidents on the minor road network,

tram accidents are mainly responsible for the high value of ACD of residential zones in the

inner city (Figure 11).

With regard to road safety, it is proven that traffic oriented roads with trams have an augmented

accident potential when trams are operated on the same lane as cars (CURRIE & REYNOLDS,

2010). Additionally, the availability of trams is usually combined with an increased flow of

pedestrians and reduced traffic space for other road users. According to FEDRO, 2014a both of

them are main reasons for a relatively high accident risk.

Regarding the improved road network from chapter 2.3 a total of 68 out of the 208 road sections

and 87 of the 214 nodes have accidents with the participation of trams. To approximate the

impact of tram availability on road safety, firstly, all of the tram accidents assigned to these two

network elements are excluded. Secondly, a ranking of ACD with the remaining accidents is

examined. A significantly decrease in the ranking, compared to the primary order, indicates a

high accident impact of trams. The corresponding tables for traffic oriented roads and nodes are

given in appendix A 13.

Table 16 summarizes network elements with the highest ratios of tram accidents respectively

ACD. For example Universitätsstrasse has about one third of all accidents involving trams.

Although a high percentage of tram accidents usually result in a higher ACD of tram accidents,

this does not specifically count for Sihlstrasse/Bahnhofstrasse. The 7% of tram accidents cause

33% of the total ACD for that node. Therefore, the priority has changed from medium priority

to high priority. The result could be even more evident due to the fact that this road section is

aggregated over two different roads of which only the short section of Bahnhofstrasse has trams

on it. Generally, for Bahnhofstrasse, which is mostly dedicated as a residential oriented road, a

very high number of tram accidents has been identified. The analysis of nodes has given similar

outcomes for the corresponding percentages. Though Central has the highest percentage of

ACD(tram_acc), it is the only node whose priority has also changed from high to medium.


61

Table 16 Network elements most influenced by tram accidents


Name Tram accidents ACD(tram_acc) Primary Rank Priority change

Badenerstrasse East 18% 40% 1 No

Sihlstrasse/Bahnhofstrasse 7% 33% 6 Yes

Hardturmstrasse West 17% 26% 12 No

Universitätsstrasse South 34% 26% 9 No

Nodes

Name Tram accidents ACD(tram_acc) Primary Rank Priority change

Bellevue 8% 15% 1 No

Limmatplatz 20% 28% 5 No

Central 14% 31% 4 Yes

Schaffhauserstrasse/

Seebacherstrasse

29% 28% 15 No

Schmiede Wiedikon 12% 24% 24 No


3.7 Result overview

To provide a brief overview of all the analyses, the results are summarized in Table 17. For

each of them, the first three ranked network elements are listed. In general, the results corre-

spond very well with each other very well. Therefore, it can be concluded that, despite the

change of different input parameters, the only non-varied factor - the number and severity of

accidents - impacts the final results of NSM the most.


62

Table 17 Overview of the highest rankings of each NSM analysis

Analysis Traffic oriented roads Nodes Residential zones

Improved

network

Badenerstrasse East

Badenerstrasse West

Limmatstrasse

Bellevue

Bucheggplatz

Heimplatz

Bahnhofstrasse/Rennweg

Uni quarter/Niederdorf

Stadelhofen

Average rate Badenerstrasse East

Badenerstrasse West

Albisstrasse

Bucheggplatz

Bellevue

Heimplatz

Coefficient

approach

Badenerstrasse East

Langstrasse North

Uraniastrasse

Bellevue

Bucheggplatz

Heimplatz

AADT 2030 Badenerstrasse East

Limmatstrasse

Albisstrasse

Bellevue

Bucheggplatz

Heimplatz

Nodes 30 m Bellevue

Bucheggplatz

Heimplatz

ACDI method Bucheggplatz

Wipkingerplatz

Wallisellenstrasse/

Saatlenstrasse

Roads+ nodes Limmatstrasse

Rämistrasse

Sihlstrasse/Bahnhofstrasse

Zone density Uni quarter/Niederdorf

Leutschenbach


Pedestrians

and bicycle

impact

Langstrasse/

Kornhausbrücke

Uraniastrasse

Birmensdorferstrasse East

Bellevue

Limmatplatz

Pfingstweidstrasse/

Hardstrasse

Stadelhofen


Uni quarter/Niederdorf

Tram impact Badenerstrasse East

Sihlstrasse/Bahnhofstrasse

Hardturmstrasse West

Bellevue

Bürkliplatz

Heimplatz



63

4 Overlay with Black Spot Management (BSM)

As mentioned in the introduction, traditionally BSM was the common tool to detect hazardous

parts of the network. The main difference to NSM is that BSM tries to identify hazardous loca-

tions based on a certain threshold value of accidents within a defined radius. Therefore, hotspots

are usually situated more locally.

Luckily, accidents and especially serious or fatal accidents happen quite rarely. But for BSM,

the absolute number of accidents is often statistically not relevant enough, and only some black

spots can be treated by BSM (FEDRO, 2013). In order to neutralize these shortcomings, BSM

is often combined with other road safety tools. Nowadays BSM has either been completely

replaced (France, England, Sweden, Finland) or supplemented (Norway, Denmark, Germany)

by NSM (EUROPEAN COMMISSION, 2003), (SØRENSEN & ELVIK, 2008). In Switzerland BSM is

still in use and NSM has just started as a pilot project on different sites (see chapter 1.1).

According to SØRENSEN & ELVIK, 2008, it is highly recommended to combine the results of BSM

and NSM because of similar data inputs and a common comparison of a best practice design.

They propose to start first with the implementation of BSM, due to more experience and an

intuitively better understanding of their results. Moreover, BSM should be implemented as long

as it continues to improve road safety. As soon as the majority of black spots are disarmed,

BSM detects only a small potential of improvement. In that case, it is required to supplement

BSM with other road safety approaches such as NSM.

For the city of Zurich, a total of 144 black spots were identified in the period from 2011 till

201315. A black spot in an inner urban area is defined as a minimum of five personal accidents16

within a radius of 50 meters (SNR 641 724, 2014). In order to compare the generated results of

chapter 3.1 with BSM, an overlay of the individual locations is made. Thereby, only entities of

traffic oriented roads and nodes with a priority of high or medium are taken into account. Figure

25 displays black spots that are either located within NSM elements of high or medium priority

(green) or have not been discovered by NSM (red).

15

Data input for NSM were the accidents in the period 2009 to 2013

16 FSI accidents are doubly weighted as MI accidents


64

Figure 25 Comparison of BSM and NSM

±0 1 2 3 4

Kilometers

Black spots that are not discovered by NSM


Medium priority of traffic oriented roads

Black spots within NSM elements of high or medium priority

High priority of traffic oriented roads

Low priority of traffic oriented roads

Legend


65


The figure shows that in total 88 of 144 black spots are located on network elements with high

or medium priority. This leads to a coverage rate of about 61%. When also including low pri-

ority, only nine uncovered black spots remain. This means that black spots also exist on network

element where NSM predicts no InfraPo.

This discrepancy can be explained by the distribution of accidents. As soon as there is a cluster

of a certain number of accidents, a black spot is identified. However, NSM normalize accidents

according to the length of each network element, so that the effective density of accidents and

therefore the ACR can be quite low for NSM. Compared to the expected number of accidents

(baAR), it can result in no predicted avoidable accident costs.

NSM not only normalize accidents according to the corresponding length but also to the given

AADT (see formulas in appendix A 1). As a result, roads with high traffic volume have by trend

a lower priority. On these sections some of the black spots cannot be discovered by NSM. Ex-

amples for this can be found on Mythenquai, Seebahnstrasse and Hardbrücke

Regarding the rankings of black spots, Table 18 lists the ten most highly prioritized black spots,

compared to the primary rankings of NSM.


66

Table 18 Comparison of rankings of BSM and NSM

Black spot location Results BSM Results NSM

Ranking Value17 Ranking Priority

Pfingstweidstrasse/Hardstrasse 1 24 9 high

Sihlstrasse/St. Annagasse18 2 20 6 high

Limmatplatz 3 19 5 high

Langstrasse/Militärstrasse 4 19 8 high

Heimplatz 5 16 3 high

Schwamendingerstr./Dörflistr. 6 16 10 high

Walchebrücke/Neumühlequai 7 13 17 medium

Goldbrunnenplatz 8 13 46 medium

Kornhausbrücke 9 13 7 medium

Bürkliplatz18 10 13 6 high


Although the rankings differ between the two approaches, the table shows a distinct similarity

between priorities. Seven of the ten highest ranked black spots resulted in NSM as highest pri-

ority. BSM does not distinguish between the severities of accidents. Therefore, the black spot

Pfingstweidstrasse/Hardstrasse with a total of 24 accidents between 2011 and 2013 is ranked

on top. Yet, NSM weighs the accidents not only according to the severity, but also includes the

approaching area of inlets. As a result, for the data from 2009 till 2013 it exhibits a total of 33

accidents.

17

This represents the number of accidents within 2011 and 2013, weighted by the accident severity category.

18 On NSM this black spot is assigned to roads. Therefore the ranking/priority of road is declared.


67

5 Discussion

The following chapter reflects the presented results and gives a critical analysis of the chosen

methodology. Furthermore, it depicts the limitations of NSM regarding input data sets and net-

work model configuration.

5.1 Accuracy of parameters

As described above, the outcome of the proposed NSM method is highly dependent on the

existence and the accuracy of inputs. The presented work includes as much data as was able to

be arranged in the context of this master thesis. Although some hints of data inaccuracy are

described in the individual sections themselves, it is necessary to clarify the background of the

used data and its challenges throughout the working process.

5.1.1 Accident data

Input data for the analyses are all the accidents protocolled by the police in the period from

2009 till 2013. The classification of the three accident severity categories is done by filtering

the accidents according to the protocolled attributes of accident consequence. However, at the

scene of the accident itself, it is not always feasible to detect the cause or the accident’s severity

with regard to the involved persons. According to FEDRO, 2014b the classification of accident

severities is mainly defined by the length of a person’s hospital stay. In order to have only two

accident categories in which persons are involved, the range of assigning an accident to MI or

FSI severity is quite broad. For example, as soon as one involved person has to stay in hospital

for more than 24 hours, the ACR rises to CHF 696’000. This ACR is per definition a mean

value of the occurring costs per accident and disregards the number of involved persons or a

further classification into fatal, invalid or other serious injuries. Consequently, it is important

to know what these ACR are composed of. Other countries include different aspects, which is

why it can be problematic to compare different countries’ ACRs (BUNDESANSTALT FÜR

STRASSENWESEN (BAST), 2014).

The basic information needed for calculating own ACR and baACR is the number of involved

persons per accident. The primary accident data of STADT ZÜRICH, 2013 did not include these

attributes so that an export from the FEDRO database had to be done (FEDRO, 2014c). By


68

using the same filters19, the total number of accidents and the classification into specific acci-

dent severity categories was not perfectly adequate in comparison with the primary one. A rea-

son for this cannot definetely be determined, but the rearrangement of the accident data respon-

sibility from cantonal databases to one comprehensive database, managed by FEDRO, might

be an explanation. Nevertheless, the exported datasheet is only applied for the calculation of

own ACR/baACR in chapter 3.2.1 and 3.2.2. With regard to comparable findings, all the other

analyses are based on the primary data sheet, received from DAV.

Another issue is the estimated number of unreported accidents. In reality, a non-negligible num-

ber of MI and especially PDO accidents are still not recorded by the police. Estimations of the

BUNDESAMT FÜR RAUMENTWICKLUNG (ARE), 2006 predict a factor of 3.64 for the number of

unreported accident injuries. For PDO the factor is assumed to be even higher (MATTHEWS,

2009). The ACRPDO of CHF 45’000 already includes this uncertainty.

5.1.2 AADT

As seen in the SA in chapter 0, the parameter AADT impacts, in comparison to the other

changeable factors, the outcome of NSM most. The basis of the traffic volume is the GVM of

Zurich, which enables an AADT value for every single road section with a speed limit greater

than 0 km/h. The GVM calculates the AADT on the basis of loop detectors widely distributed

within the canton of Zurich (VRTIC, WEIS, & FRÖHLICH, 2012). Thereby, major roads are more

often provided with loop detectors, so that minor roads are more exposed to inaccuracies.

Hence, no baACD is applied for the minor road network of residential zones.

The same problematic has to be taken into account when interpreting the results of the analysis

with the AADT predicted for the year 2030. It represents more a trend of where future accident

hotspots might occur. Additionally, some hazardous network elements may be disarmed until

2030 and urban changes could cause shifts of traffic origins within the total network (VENGELS

& WEINERT, 2008).

5.1.3 Cost rates

NSM requires baAR, which stand for a best practice design. The own calculated cost rates did

not consider any additional road design issues, such as number of lanes, road width or horizontal

and vertical road planning. Despite of representing more of an arithmetic mean value, it can be

shown that the accident patterns in a city are different from those in the rest of Switzerland.

19

Accidents filtered by the BFS-No. 261 (city of Zurich) and the period 2009 – 2013


69

There is a much higher accident density and nodes in particular have an increased expectation

of accidents for the same amount of driven vehicle kilometers.

As mentioned above, the main impact on the presented cost rates are the cost values per casu-

alty. This work assumed these cost rates as given, and knew that inaccuracies exist. Mainly, the

very high casualty cost rate of a deceased individual (CHF 3’191’421) can distort individual

results, when only having a few accidents as exposition.

The calculations above considered a total of more than 13’000 accidents. Though the results

provide a reasonable first impression, no other specific cost rates of other cantons exist so far.

For a well-grounded analysis, more accidents involving people should be included or the sam-

ple size ought to be expanded to include all the accident data of Switzerland. FEDRO tries to

force other cities such as Basel and Berne to create more comparable ACR and baAR/baACR.

5.1.4 Assumptions for SA

The applied SA model works best when parameters are not correlated among themselves. As

seen in Figure 18, there is a non-negligible positive correlation between AADT and the number

of accidents. Therefore, accidents are always assigned to the referring AADT term (see chapter

0). The model itself enables an endless spectrum of different combinations so that some as-

sumptions about the ratio of cost factors, the number of intervals for the distribution of AADT

and network length had to made.

The outputs of the model do not only depend on the correlation of factors but also on their

distribution. Due to extreme values of high AADT and network length, a better distribution is

aspired. The output of the model is in fact able to determine which factor has the greatest im-

pact, but it cannot provide any quantitative values. Therefore, it cannot be taken for granted that

an increase in AADT always leads to a growth of avAC. The reason for this is the extremely

high standard deviation, which is caused by different sample combinations. The samples them-

selves do not consider a variation of the number of each accident severity category. These com-

binations do not follow any distribution and have a significant influence indeed on the variation

of avAC.


70

5.2 Constraints of the network model

5.2.1 Network elements

All the network elements of both the existing and the improved network model are digitalized

as polygons. The advantages are the simplified assignment of an accident’s location to the in-

dividual network elements and the guarantee that, in the case of no polygon overlaps, the entire

perimeter of study is covered. In contrast, polygons do not allow an easy extending of the enti-

ties’ shape. For example, when assigning all the accidents to traffic oriented roads (chapter

3.2.5), the polygons are prolonged by a user-defined buffer length. This buffer also increases

the width of a road. Due to the fact that, by the majority, no parallel residential roads exist

within this buffer length, the effect of having an unrealistic number of accidents for the extended

polygons is marginal.

For nodes a radius of 50 m is defined. Comparable NSM studies also use this threshold value

in order to cover the conflicting area of nodes and the approaching area of inlets at the same

time. According to AURICH, 2012, approaching areas have a higher accident risk than other parts

of a road section, so that the accident occurrence is better represented when assigning them to

nodes. The impact of approaching areas increases with higher speed limits. As a consequence,

the threshold values have to be adjusted in case rural parts are also included.

In cities there are quite often network parts that are built on different levels (e.g. tunnels,

bridges, overpasses). There are difficulties in properly assigning accidents to the corresponding

network entity due to a lack of a third dimension’s coordinate. Indeed, the changed digitaliza-

tion of these parts in the improved network model resulted in a better outcome, but still cannot

represent the effective road design adequately enough. In this case, approaches of digitalizing

road sections as lines instead of polygons can dissolve the problem. Continuing with the same

network model, it is hardly recommended to further spatial join the accidents according to the

attribute labelled street name. This attribute was missing in the provided accident data sheet,

but in future can easily be exported from the FEDRO database.

For zones, the continued idea of setting minimum values for network length and zone area

might have to be adjusted in future times. By trend, zones with a small network length result in

an excessively high ACD right. Moreover the results in total are not very robust. On these

grounds, it is recommended to raise up the minimum values for further studies


71

5.2.2 Level of aggregation

The guidelines in SNR 641 725 limit the scope of the minimal and maximal length of inner

urban roads. The intention is to ensure that too short road sections do not result in an unduly

high ACD. VENGELS & WEINERT, 2008 and EBERSBACH & SCHÜLLER, 2008 propose that section

creation should not only consider the homogeneity of road segments but should also refer to the

existing accident situation. Roads with a similar number of accidents ought to be merged more

frequently.

The situation in Zurich has shown that the minimum threshold value of 500 m per road section

might be too high. There are numerous nodes within this distance on which the inlet’s AADT

and traffic regime varies a lot. Merging these sections together does in fact reduce the problem-

atic of increased ACDs, but cannot properly assign a correct AADT value for every single road

section.

5.3 Limitation of the implemented NSM method

As one of the six infrastructure safety tools, NSM focuses on a network wide scope and has its

main strengths lie in the identification of network elements with a high infrastructure potential.

However, NSM has to be combined with other safety tools such as RSI and BSM which focus

more on a local scale. NSM does not provide suitable measures for road sections with a high

safety potential. Thus a detailed analysis of the accident characteristic has to be carried out

individually for the particular section under review.

In spite of the improvements to the existing road network model, the process of creating a more

realistic network model has not ended yet. On the one hand, the distinction of traffic respec-

tively residential oriented roads still depends on subjective criteria and on the other hand, the

level of aggregation is bounded by the given guidelines from SNR and can lead to an inadequate

representation of the real road network. Both aspects affect the outcomes of NSM massively.

Furthermore, the creation of the entire network model is very time-consuming. In order to re-

duce the efforts of digitalizing each network element manually, it can be further developed by

automatizing some working steps. The section creation could be made on the basis of the given

road axis and the assignment of accidents to one specific element by a nearest neighbour anal-

ysis.

Regarding the parameters, AADT has the biggest impact of all (see chapter 0). Due to digitali-

zation of traffic oriented roads as polygons, the segments of the given input file of AADT are

not congruent with the own created road sections. Therefore, an approximated AADT value of

all input segments has to be assigned on each individual section. The same has to be done for


72

nodes whose AADT is only based on the traffic volume of inlets. To verify these AADT values,

a comparison with the existing GVM of Zurich would be meaningful.

The comparison of the effective accident occurrence and the one estimated by a best practice

design requires diverse input attributes. The own calculated ACR respectively baAR and

baACR typify the situation in the city of Zurich in a more realistic way, but indeed they repre-

sent more of an arithmetic mean than very well-founded cost rates. Mainly for the baAR of both

the average and the coefficient approach, it was not possible within the scope of this work to

organise more infrastructural input data.

Finally, the analyses of the two specific topics human powered mobility and impact of trams

only roughly estimates those impacts. Due to the exclusion of some of the data the impacts

cannot be represented in a comprehensive way. For advancing these two approaches, more data

about routes and traffic flows of bicycles and pedestrians, densities of workplaces or catchment

areas of public facilities are required. Having this information would allow a more sophisticated

approximation of the interaction between these modes of transport and the frequency and se-

verity of accidents to be performed.


73

6 Relevance for practice

In Switzerland, the implementation of NSM on rural and urban networks is still in the beginning

stages. The on-going process of the four chosen pilot studies are supported by FEDRO which

tries to further develop a methodology that fits best for Swiss standards. This work represents

advanced analyses for the pilot site of the city of Zurich, based on the primary results from

DECURTINS, 2014. The work contributes to a comprehensive understanding of model parameters

by varying and testing different inputs. The main findings of the analyses have already been

presented at a meeting at FEDRO in Berne and should further assist other pilot projects.

The first step after the pilot phase is going to be the announcement of NSM to a broader audi-

ence. Therefore it is intended to primarily publish a brief summary of all the pilot studies in

order to promote this safety tool. In a further step the goal is both, a supplementary improvement

of road infrastructure designs and an enhancement of prioritization when rebuilding or main-

taining the existing infrastructure.

In cases where there is an intact road infrastructure, an increased accident occurrence is often

not relevant enough for rebuilding an individual road section or node. Although monetary val-

ues of safety potential are provided, a cost-benefit analysis does not often legitimate invest-

ments for primary road safety. Rather, extensions of capacity or the proceeding deterioration of

a certain network element are habitually main activators of constructional projects. However,

NSM can support these measures by enhancing specific network elements, which have an aug-

mented infrastructure potential. It is in this way also a decision tool of prioritizing road networks

elements concerning their safety aspect.

In order to continue the evaluation of infrastructural deficits, the presented results need to be

adjusted regularly. Therefore a monitoring and actualization of data inputs is required. To fur-

ther involve other circumstances, such as the importance of pollution restrictions or noise pre-

vention, an expansion of the NSM methodology can contribute to an extensively development

of an optimal road network design.


75

7 Conclusion and recommendations

7.1 Conclusion

The implemented methodology of NSM for the city of Zurich has shown that network elements,

where an improvement of infrastructure is expected to be highly cost efficient, can be identified.

For that purpose, proper input data is required. On the one hand, the road network has to be

adjusted in order to represent the effective road network as accurately as possible. Not only a

proper distinction of traffic oriented and residential roads is necessary but also the different

road sections’ level of aggregation can lead to different outcomes. On the other hand, the impact

of variable input parameters, such as network length, AADT and the provided cost rates are

attempted to be prioritized. A sensitivity analysis (SA) has yielded that AADT is the parameter

with the strongest impact. Despite a widespread standard deviation, the SA model predicts an

increase in avoidable accident cost (avAC) per year when raising the AADT.

The results depict that, independently of input changes, the results for the calculated infrastruc-

ture potential (InfraPo) are similar for all the analyses. The highest InfraPo for traffic oriented

roads can mainly be found at Badenerstrasse, Limmatstrasse and Langstrasse/Kornhausbrücke.

For nodes, the highest rankings belong to Bellevue, Bucheggplatz and Heimplatz. Due to non-

existing basic accident cost rates for minor roads, residential zones are ranked by their specific

value of accident cost densities (ACD). The two inner urban zones Bahnhofstrasse/Rennweg

and University quarter/Niederdorf are ranked on top. Regarding the results in general, it is con-

spicuous that on the one hand, network elements with a high AADT tend to have a lower In-

fraPo. On the other hand, when analysing nodes of comparable AADTs, the ones operated by

an uncontrolled traffic regime have a distinctly higher value for avAC per year.

For the analyses, all three categories of accident severity – called fatal or serious injury (FSI),

minor injury (MI) and property damage only (PDO) – within the period from 2009 until 2013

are taken into consideration. The total of more than 13’000 accidents is further examined in

order to calculate specific accident cost rates (ACR) and basic accident cost rates (baACR) for

the city of Zurich. It resulted in minor differences between the two traffic oriented network

elements roads respectively nodes and the more domestic oriented residential zones. By classi-

fying the accidents according to their accident type, the dissimilarities of cost rates are far more

significant.


76

Moreover, an overlay with the 144 locations, identified by Black Spot Management (BSM), has

exposed that 61% of the black spots are covered by medium or high prioritized traffic oriented

roads or nodes predicted by NSM. This shows that a discrete consideration of only one road

safety tool is not adequate in order to fully understand the impact of accident occurrence within

the city of Zurich.

7.2 Recommendations

In spite of the improved network, there are still limitations regarding several aspects. The defi-

nition of a traffic oriented road and especially the distinction between them and residential

roads, cannot be made by objective criteria. Moreover, the effective dimension of node size,

including its approaching area, is unable to be adequately represented by a node size of 50 m.

Additionally, the aggregation of individual traffic oriented sections should not only be based

on the similarities of road characteristics but also on the existing accident occurrence. When

implementing this road section creation to an inner urban area, such as for the city of Zurich,

the provided minimum threshold value of 0.5 km does not seem to be short enough.

The main work load of NSM is firstly to create a consistent road network in a GIS. Having

achieved this, it allows you to make various analyses in order to gain an individual focus on

specific subjects. For further analyses, it is hardly recommended to imply a wider spectrum of

geometric road data. Particularly, infrastructure attributes such as the number of lanes, the ex-

istence of a constructional lane separation, the presence of tram tracks or the present traffic

operation at nodes can contribute to more sophisticated analyses. This data is specifically re-

quired for calculating own basic rates in order to approximate them to a best practice design.

After the refinement of the pilot study’s guidelines, the findings of NSM should further be

employed in combination with other subjects, so as to give road safety more priority in the

planning processes of rebuilding road facilities. Therefore, questions such as the followings

should be taken into consideration:

• Are there significant parts of a road network that can be renovated at the same time?

• Is it possible to have a specific geometric design that fits best when having similar

road characteristics?

• Are there special accident patterns that can be referred to a non-optimal infrastruc-

ture design?

• How does an improvement of one or more a highly prioritized network entities af-

fect the rakings of the other elements?


77

8 References

AMT FÜR RAUMENTWICKLUNG (ARE). (2014). ÖREB Nutzungsplanung (kantonal/kommunal).

Zürich: Kanton Zürich.

AURICH, A. (2012). Modelle zur Beschreibung der Verkehrssicherheit innerörtlicher

Hauptverkehrsstrassennetze unter besonderer Berücksichtigung der Umfeldnutzung.

Dresden: Fakultät Verkehrswissenschaften "Friedrich List", Technische Universität

Dresden.

BAUM, H., KRANZ, T., & WESTERKAMP, U. (2010). Volkswirtschaftliche Kosten durch

Strasssenverkehrsunfälle in Deutschland. Bremen: Berichte der Bundesanstalt für

Strassenwesen (Bast), Reihe "Mensch und Sicherheit" Heft M 208, Carl Schünemann

Verlag Gmbh.

BUNDESAMT FÜR RAUMENTWICKLUNG (ARE). (2006). Unfallkosten im Strassen- und

Schienenverkehr der Schweiz, Aktualisierung für die Jahre 1999 bis 2004. Bern.

BUNDESAMT FÜR RAUMENTWICKLUNG (ARE). (2010). Nationales Personen- und

Güterverkehrsmodell des UVEK, Durchschnittlicher Tagesverkehr 2010 für den

Personen- und Güterverkehr. Bern.

BUNDESANSTALT FÜR STRASSENWESEN (BAST). (2014). Volkswirtschaftliche Kosten von

Strassenverkehrsunfällen in Deutschland. Bergisch Gladbach.

BUNDESANSTALT FÜR STRASSENWESEN (BAST), SERVICE D'’ÉTUDES TECHNIQUE DE ROUTES ET

AUTOROUTES (SÉTRA). (2005). Network Safety Management (NSM). Retrieved December

2014, from

http://www.sure.equipement.gouv.fr/IMG/pdf/NSM_V_FD_final_cle55ec71-1.pdf

CHAMBON, P., LEMKE, K., & GANNEAU, F. (2006). Network Safety Managment - A French-

German Co-operation. Begisch Gladbach, Bagneux Cedex: Bundesanstalt für

Strassenwesen (Bast), Service d'’Études technique de routes et autoroutes (Sétra).

CURRIE, G., & REYNOLDS, J. (2010). Vehicle and Pedestrian Safety at Light Rail Stops in Mixed

Traffic. Washington: Transportation Research Record, Vol. 2146, pp. 26-34.


78

DECURTINS, U. (2014). Prioritization of Road Sections in Zurich for Safety Improvement Using

Network Safety Management. Zürich: Institute for Transport Planning and Systems

(IVT), ETH Zürich.

EBERSBACH, D., & SCHÜLLER, H. (2008). Praktische Anwendung der Empfehlungen für die

Sicherheitsanalyse von Strassennetzen (ESN) - Erfahrungen mit den Verfahren der

Abschnittsbildung. Bonn: Fachzeitschrift Strassenverkehrstechnik Jahrgang 52,

Ausgabe 9, Kirschbaum Verlag GmbH.

ELVIK, R., HØYE, A., VAA, T., & SØRENSEN, M. (2009). The Handbook of Road Safety Measures

second edition. Bingley UK: Emerald Group Publishing Limited.

EUROPEAN COMMISSION. (2003). Road Infrastructure Safety Management - Report of the

Working Group on Infrastructure Safety. DG Energy and Transport, High Level Group

- Road Safety.

FEDERAL ROADS OFFICE (FEDRO). (2013). ISSI Infrastruktur-Sicherheitsinstrumente

Vollzugshilfe. Bern.

FEDERAL ROADS OFFICE (FEDRO). (2014a). Einflüsse von Situation und Infrastruktur auf das

Unfallgeschehen: Phase 1. Bern: Forschungspaket VeSPA, Teilprojekt 2.

FEDERAL ROADS OFFICE (FEDRO). (2014b). Ergebnisbericht Ermittlung von Kennzahlen für

NSM (Netzeinstufung) und RIA (Folgeabschätzung). Bern.

FEDERAL ROADS OFFICE (FEDRO). (2014c). Verkehrsunfälle. Bern: Management-

Informationssystem Strasse und Strassenverkehr (MISTRA), Export Unfälle BFS-Nr

0261 für die Jahre 2009- - 2013.

FEDERAL ROADS OFFICE (FEDRO). (2014d). VSS-Standardreports nach SN 641 725. Bern.

GE, Q., & MENENDEZ, M. (2014). An Efficient Sensitivity Analysis Approach for

Computationally Expensive Microscopic Traffic Simulation Models. International

Journal of Transportation, Vol. 2 No 2 (2014) pp. 49-69.

HOLZ-RAU, C., & SCHEINER, J. (2013). Geographical patterns in road safety: literature review

and a case study from Germany. EJTIR Issue 13(2) pp. 99-122.

MATTHEWS, W. (2009). Unfall- und Unfallkostenraten im Strassenverkehr. Bern:

Forschungsauftrag VSS 2000/343 auf Antrag des Schweizerischen Verbands der

Strassen- und Verkehrsfachleute (VSS).


79

RETTING, R., FERGUSON, S., & MCCARTT, A. (2003). A review of evidence-based traffic

engineering measures designed to reduce pedestrian-motor vehicle crashes. Amer. J.

Public Health, vol 93, no.9, pp. 1456-1463.

SCHERMERS, G., CARDOZO, J., & ELVIK, R. (2011). Road Infrastructure Safety Management

Evaluation Tools (RISMET), Guidelines for development and application of Evaluation

Tools for road infrastructure safety management in the EU. ERA-NET Road.

SCHÜLLER, H. (2009). Modelle zur Beschreibung des Gechwindigkeitsverhaltens auf

Stadtstrassen und dessen Auswirkungen auf die Verkehrssicherheit auf Grundlage der

Strassengestaltung. Dresden: Dissertationsschrift, Fakultät Verkehrswissenschaft

"Friedrich List", Technische Universität Dresden.

SCHÜLLER, H. (2015). Kennzahlenermittlung für die Stadt Zürich im Rahmen des Projekts

Infrastruktur-Sicherheitsinstrumente ISSI im Auftrag des Bundesamt für Strassen

(ASTRA). Stuttgart: PTV Group.

SKELEDZIC, S. (2014). Pilot Network Safety Management. Bern: Beratung für Unfallverhütung

bfU.

SN 640 009a. (2006). Strassenverkehrsunfälle - Lokalisierung und Rangierung von

Verkehrsunfällen. Zürich: Schweizerischer Verband der Strassen- und

Verkehrsfachleute VSS.

SNR 641 724. (2014). Strassenverkehrssicherheit Unfallschwerpunkt-Management. Zürich:

Schweizerischer Verband der Strassen- und Verkehrsfachleute VSS.

SNR 641 725. (2013). Strassenverkehrssicherheit Netzeinstufung. Zürich: Schweizerischer

Verband der Strassen- und Verkehrsfachleute VSS.

SØRENSEN, M., & ELVIK, R. (2008). Black Spot Management and Safety Analysis of Road

Networks - Best Practice Guidelines and Implementation Steps. Norwegen: Publikation

des Institute of Transport Economics, Norwegina Centre for Tranport Research.

STADT ZÜRICH. (2013). VUSTRA (Verkehrsunfallstatistik 2009 - 2013). Zürich: Dienstabteilung

Verkehr.

TIEFBAUAMT KANTON ZÜRICH. (2013). Gesamtverkehrsmodell des Kantons Zürich. Zürich.

TIEFBAUAMT KANTON ZÜRICH. (2014). Strassennetz Kanton Zürich. Kanton Zürich.


80

TIEFBAUAMT STADT ZÜRICH. (2014). Data sets Vmax, AADT, speed profiles, hectars of

population density, zones. Zürich.

TOURING CLUB SCHWEIZ. (2002). Sicherheit in Quartieren, Eine Informationsschrift über

Strassengestaltung und Verkehrsmassnahmen. Vernier.

VBZ. (2013). Züri-Linie 2030. Zürich: VBZ-Netzentwicklungsstrategie Synthesebericht.

VENGELS, S., & WEINERT, R. (2008). Pilotanwendung der Empfehlungn für die

Sicherheitsanalyse von Strassennetzen (ESN), Bericht zum Forschungsprojekt FE

01.165/2004/CRB. Bergisch Gladbach: Bundesanstalt für Strassenwesen.

VRTIC, M., WEIS, C., & FRÖHLICH, P. (2012). Gesamtverkehrsmodell für den Personenverkehr

des Kantons Zürich, Modellkalibration 20111. Olten und Wädenswil: TransOptima

GmbH and TransSol Gmbh.

WALTER, U., & WEIDMANN, U. (2013). Einführung Langsamverkehr. Zürich: ETH Zürich,

Institut für Verkehrsplanung und Transportsysteme.

YANNIS, G., EVGENIKOS, P., & PAPADIMITRIOU, E. (2008). Best Practice for Cost-Effective Road

Safety Infrastructure Investments. Athens: National Technical University of Athens,

Conference of European Directors of Roads.


A-1

Appendix

A 1 Formulas for calculation of NSM terms ....................................................... 2

A 2 Results of traffic oriented roads .................................................................. 3

A 3 Results of nodes ........................................................................................ 5

A 4 Results of residential zones ....................................................................... 8

A 5 Results of the existing model ..................................................................... 9

A 6 Own calculated accident cost rates (ACR) ............................................... 12

A 7 Own calculated basic accident rates (baAR) ............................................ 14

A 8 Ranking of nodes with radius 30 meters ................................................... 22

A 9 Nodes ranking based on incoming vehicles ............................................. 25

A 10 Assigning accidents on nodes to traffic oriented roads ............................. 27

A 11 Correlation analysis .................................................................................. 29

A 12 Network density of residential zones ........................................................ 30

A 13 Influence of Tram ..................................................................................... 31


A-2

A 1 Formulas for calculation of NSM terms

Table 1 Formulas for NSM

Term Formula Unit

Accident Cost

(AC) 𝐴𝐶 = 𝐴𝑐𝑐(𝐹𝑆𝐼) · 𝐴𝐶𝑅(𝐹𝑆𝐼) + 𝐴𝑐𝑐(𝑀𝐼) · 𝐴𝐶𝑅(𝑀𝐼) + 𝐴𝑐𝑐(𝑃𝐷𝑂) · 𝐴𝐶𝑅(𝑃𝐷𝑂) CHF/acc

Accident Cost

Density (ACD) 𝐴𝐶𝐷 =

𝐴𝐶

1000 ∗ 𝐿 ∗ 𝑇 CHF*1000/(km*a)

Basic Accident

Cost Density

(baACD) 𝑏𝑎𝐴𝐶𝐷 =

𝑏𝑎𝐴𝐶𝑅 · 365 · 𝐴𝐴𝐷𝑇

106 CHF*1000/(veh*km)

Infrastructure

Potential

(InfraPo) 𝐼𝑛𝑓𝑟𝑎𝑃𝑜 = 𝑈𝐾𝐷 − (

𝑏𝑎𝐴𝐶𝑅 ∗ 365 ∗ 𝐴𝐴𝐷𝑇

106) CHF*1000/(km*a)

Avoidable

Accident cost

(avAC)

𝑎𝑣𝐴𝐶 = 1000 ∗ 𝐼𝑛𝑓𝑟𝑎𝑃𝑜 ∗ 𝐿 CHF/a

Source: SNR 641 725, 2013


A-3

A 2 Results of traffic oriented roads

Table 2 Prioritization of traffic oriented roads

Ranking Name AADT

[veh/d]

Acc(FSI)

[Acc/5a]

Acc(MI)

[Acc/5a]

Acc(PDO)

[Acc/5a]

avAC

[CHF/a]

Priority Accident type

Accident cause Participation

1 Badenerstrasse East 9'579 13 37 39 2'123'840 high Rear-end collision Alcohol Bicycle

2 Badenerstrasse West 10'474 16 26 26 1'913'193 high Rear-end collision Alcohol Bicycle

3 Limmatstrasse 2'008 8 23 36 1'653'926 high Turning collision Disregarded tram Tram

4 Albisstrasse 11'280 14 45 40 1'623'780 high Rear-end collision Deflection Pedestrian

5 Uraniastrasse 14'571 9 18 51 1'417'607 high Rear-end collision Lane changing Bicycle

6 Sihlstrasse 11'048 7 24 39 1'375'755 high Obstacle collision Lane changing Pedestrian

7 Langstrasse North/

Kornhausbrücke

18'924 8 49 37 1'366'848 medium

8 Thurgauerstrasse/

Binzmühlestrasse

11'656 10 16 19 1'135'667 medium

9 Universitätstrasse South 13'271 9 19 13 1'095'620 medium

10 Rämistrasse 14'321 8 32 33 1'092'905 medium

11 Hofwiesenstrasse/

Franklinstrasse

5'716 6 17 22 1'059'207 medium

12 Hardturmstrasse West 3'393 6 16 20 1'022'409 medium

13 Binzmühlestrasse 9'081 7 10 15 981'575 medium

14 Sihlquai East 12'326 7 32 54 961'883 medium

15 Dreikönigstrasse/

Bleicherweg

5'451 4 14 40 905'835 medium

16 Langstrasse South 3'798 4 16 20 884'301 medium

17 Schaffhauserstrasse North 13'466 4 31 26 875'509 medium

18 Kornhausstrasse/

Rötelstrasse

7'768 7 22 7 857'728 medium


A-4

Ranking Name AADT

[veh/d]

Acc(FSI)

[Acc/5a]

Acc(MI)

[Acc/5a]

Acc(PDO)

[Acc/5a]

avAC

[CHF/a]



19 Birmensdorferstrasse East 14'134 6 21 13 853'595 medium

20 Militärstrasse 3'895 4 18 23 820'392 medium

21 Manessestrasse/

Steinstrasse

6'188 5 17 16 806'626 medium

22 Altstetterstrasse North 2'165 4 14 11 791'570 medium

23 Weinbergstrasse 4'507 6 11 10 761'084 medium

24 Hohlstrasse West 5'692 3 21 22 740'801 medium

25 Stauffacherstrasse 7'722 5 11 17 721'084 medium

26 Birmensdorferstrasse

West

11'089 8 7 16 715'764 medium

27 Brandschenkenstrasse 4636 5 12 18 713'430 medium

28 Hardtrumstrasse East/

Sihlquai West

18'200 6 27 40 703'667 medium

29 Giesshübelstrasse/

Bederstrasse

14'812 5 30 27 689'891 medium

30 Schaffhauserstrasse South 6'878 6 17 12 658'986 medium



A-5

A 3 Results of nodes

Table 3 Prioritization of nodes

Ranking Name AADT

[veh/d]

Acc(FSI)

[Acc/5a]

Acc(MI)

[Acc/5a]

Acc(PDO)

[Acc/5a]

avAC

[CHF/a]



1 Bellevue 58'182 10 40 85 2'565'137 high Rear-end collision Lane changing Bicycle

2 Bucheggplatz 22'085 8 39 52 2'116'610 high Rear-end collision Lane changing Bicycle

3 Heimplatz 37'234 5 25 110 1'937'139 high Grazing collision Crossing Panel Truck

4 Central 23'645 7 12 45 1'473'767 high Crossing pedestrian Lane changing Pedestrian

5 Limmatplatz 22'996 6 18 27 1'297'168 high Crossing pedestrian Disregarded

pedestrian

Pedestrian

6 Bürkliplatz 49'590 5 14 42 1'247'840 high Rear-end collision Deflection Pedestrian

7 Escherwyssplatz 18'724 3 18 59 1'166'084 high Lane changing Lane changing Panel Truck

8 Langstrasse/Lagerstrasse 22'190 5 17 28 1'132'966 high Rear-end collision Alcohol Bicycle

9 Pfingstweidstrasse/Hardstrasse 18'860 6 10 17 1'087'774 high Rear-end collision Alcohol Bicycle

10 Schwamendingerstrasse/

Dörflistrasse

21'014 5 17 13 1'072'598 high Rear-end collision Right of way Bicycle

11 Bahnhofstrasse/

Muesumstrasse

37'893 3 5 49 895'713 medium

12 Wipkingerplatz 18'480 4 11 22 855'791 medium

13 Albisriederplatz 27'104 3 14 32 842'464 medium

14 Römerhof 12'193 4 11 12 834'513 medium

15 Schaffhauserstr/

Seebacherstrasse

16'778 4 10 10 794'040 medium

16 Stauffacher 25'020 3 16 27 770'544 medium

17 Walchebrücke/Neumühlequai 35'452 2 9 52 768'977 medium

18 Bahnhofplatz/Bahnhofquai 23'170 2 14 31 763'931 medium

19 Zwinglihaus 11'296 4 10 7 746'817 medium

20 Kreuzplatz 23'826 4 8 17 736'146 medium


A-6

Ranking Name AADT

[veh/d]

Acc(FSI)

[Acc/5a]

Acc(MI)

[Acc/5a]

Acc(PDO)

[Acc/5a]

avAC

[CHF/a]



21 Letzigrund 15'127 3 20 3 725'718 medium

22 Schaffhauserstr/Glatttalstrasse 20'812 3 6 23 699'648 medium

23 Bahnhofplatz/Löwenstrasse 21'032 2 12 26 687'976 medium

24 Schmiede Wiedikon 16'886 3 13 8 687'106 medium

25 Kanonengasse/Militärstrasse 9'263 3 10 14 677'993 medium

26 Utoquai/Kreuzstrasse 35'484 3 9 17 677'894 medium

27 Schaffhauserstrasse/

Franklinstrasse

11'878 3 4 27 673'932 medium

28 Kornhausbrücke/

Rousseaustrasse

28'530 3 12 12 623'690 medium

29 Schimmelstrasse/

Manessestrasse

70'090 1 9 45 608'674 medium

30 Talstrasse/Bleicherweg 16'440 2 8 24 608'458 medium

31 Sternen Oerlikon 1'468 4 2 2 603'074 medium

32 Birmensdorferstrasse/

Aemtlerstrasse

14'510 2 15 9 593'446 medium

33 Talstrasse/Sihlstrase 33'410 2 8 33 588'585 medium

34 Thurgauserstrasse/

Dörflistrasse

16'372 3 7 8 586'942 medium

35 Kasernenstrasse/Lagerstrasse 9'986 3 4 12 580'444 medium

36 Bahnhofstrasse/Uraniastrasse 25'314 3 7 8 575'878 medium

37 Tobelhofstrasse/

Dreiwiesenstrasse

16'130 3 7 6 569'242 medium

38 Sihlhölzlistrasse/

Manessestrasse

42'383 2 8 23 567'357 medium


Bucheggstrasse

18'848 3 6 12 558'018 medium

40 Weststrasse/Manessestrasse 63'285 2 7 26 551'694 medium

41 Hohlstrasse/Herdernstrasse 19'865 2 9 21 546'528 medium

42 Farbhof 21'106 2 10 14 546'284 medium


A-7

Ranking Name AADT

[veh/d]

Acc(FSI)

[Acc/5a]

Acc(MI)

[Acc/5a]

Acc(PDO)

[Acc/5a]

avAC

[CHF/a]



43 Hegibachplatz 22'544 3 7 10 543'408 medium

44 Zehntenhausplatz 28'635 2 7 27 535'109 medium

45 Utoquai/Falkenstrasse 40'644 2 3 27 521'509 medium

46 Goldbrunnenplatz 23'420 2 10 16 505'430 medium

47 Bahnhof Oerlikon Ost 24'227 1 12 28 504'902 medium

48 Stampfenbachplatz 13'276 2 9 10 503'173 medium

49 Hubertus 24'848 2 12 15 502'311 medium

50 Furttalstrasse/

Wehntalerstrasse

26'180 2 9 11 496'206 medium

51 Krematorium Sihlfeld 15'248 2 9 8 482'733 medium

52 Schaffhauserplatz 18'534 3 5 7 480'546 medium

53 Albisriederstrasse/

Birmensdorferstrasse

16'991 2 8 9 472'776 medium

54 Seebahnstrasse/

Kalkbreitestrasse

35'324 1 15 23 470'041 medium

55 Badenerstrasse/

Kalkbreitestrasse

12'460 2 6 9 444'783 medium


Schaufelbergstrasse

18'096 2 10 7 443'746 medium

57 Seilbahn Rigiblick 27'081 1 13 13 441'091 medium

58 Rämistrasse/Tannenstrasse 17'328 3 1 3 439'959 medium

59 Wasserwerkstrasse/

Neumühlequai

34'692 1 7 25 438'874 medium

60 Wehntalerstrasse/

Hofwiesenstrasse

25'342 2 8 13 437'857 medium



A-8

A 4 Results of residential zones

Table 4 Prioritization of residential zones

Ranking Name Acc(FSI)

[Acc/5a]

Acc(MI)

[Acc/5a]

Acc(PDO)

[Acc/5a]

ACD

[CHF/(a*km)]



1 Bahnhofstrasse/Rennweg 16 34 84 1'121'264 high Obstacle collision Deflection Tram

2 University quarter/Niederdorf 12 44 84 1'027'114 high Obstacle collision Deflection Bicycle

3 Stadelhofen 8 13 34 875'306 medium

4 Förrlibuckstrasse 1 3 23 529'197 medium

5 Löwenstrasse/Bahnhofplatz 3 13 47 510'824 medium

6 Kasernenareal/Europaallee 6 22 66 431'483 medium

7 Industry Herdern 1 4 3 389'000 medium

8 Letzipark/Hohlstrasse 4 11 43 362'183 medium

9 Bäkeranlage 7 15 65 347'322 medium

10 Station Oerlikon 5 17 48 339'411 medium

11 Leutschenbach 5 9 43 302'019 medium

12 Industry Altstetten 1 4 11 265'766 medium

13 Giesshübel 1 6 16 250'945 medium

14 Hardplatz/Bullingerplatz 4 12 35 242'633 medium

15 Limmatplatz/Museumsstrasse 2 12 39 221'817 medium

16 Zurich West/Technopark 0 3 24 214'307 medium



A-9

A 5 Results of the existing model


Figure 1 Results of traffic oriented roads with PDO accidents



A-10

Nodes

Figure 2 Results of nodes with PDO accidents



A-11

Residential zones

Figure 3 Results of residential zones with PDO accidents



A-12

A 6 Own calculated accident cost rates (ACR)

Accident severity

Table 5 Own calculated ACR of accident severity categories

FSI MI PDO

Network element accident Number of persons CHF accident Number of persons CHF accident CHF

No_acc fatal SI MI ACR(FSI) No_acc fatal SI MI ACR(MI) No_acc ACR(PDO)

Line 418 21 404 58 692'000 1751 0 0 2057 84'144 3‘636 44'824

Node 287 9 288 43 650'564 1137 0 0 1328 83'918 2‘871 44'824

Zone 172 5 173 13 641'935 660 0 0 747 82'707 2‘079 44'824

Road & node 705 30 692 101 675'132 2888 0 0 3385 84'055 6‘507 44'824

Average 108‘947 83‘706 44‘824



A-13

Accident types

Table 6 Own calculated ACR of accident types

Traffic oriented road & node Zone

Accident type accident Accident percentage CHF accident Accident percentage CHF

No_acc fatal SI MI PDO ACR(acc_type) No_acc fatal SI MI PDO ACR(acc_type)

Driving accident 1791 0.5% 7.7% 17.3% 74.5% 110'248 931 0.6% 7.8% 16.2% 75.3% 114'485

Lane changing 1359 0.2% 2.2% 11.8% 85.7% 67'242 232 0.0% 4.3% 10.3% 85.3% 70'657

Collision 2252 0.0% 2.5% 35.6% 61.9% 74'352 410 0.0% 1.0% 33.2% 65.9% 65'447

Turn off 838 0.0% 8.5% 41.6% 49.9% 110'510 178 0.0% 6.7% 35.4% 57.9% 110'094

Turn into 611 0.0% 9.2% 34.2% 56.6% 104'848 125 0.0% 6.4% 32.8% 60.8% 89'607

Crossing 543 0.0% 8.3% 37.0% 54.7% 105'155 131 0.0% 1.5% 34.4% 64.1% 69'049

Frontal collision 123 0.0% 4.9% 15.4% 79.7% 80'046 102 0.0% 4.9% 7.8% 87.3% 73'874

Parking 640 0.0% 0.6% 2.0% 97.3% 48'622 216 0.0% 0.0% 1.9% 98.1% 45'444

Pedestrian 552 1.8% 30.1% 66.5% 1.6% 287'057 187 0.5% 26.2% 71.7% 1.6% 224'446

Other 1386 0.5% 6.7% 12.5% 80.3% 101'393 397 0.3% 3.3% 7.3% 89.2% 91'684

Average 108‘947 95‘479



A-14

A 7 Own calculated basic accident rates (baAR)

Distribution of the basic accident rates (baAR) for nodes

Figure 4 Histogram of nodes


0

2

4

6

8

10

12

14

16

18

0

0.4

0.8

1.2

1.6 2

2.4

2.8

3.2

3.6 4

4.4

4.8

5.2

5.6 6

6.4

6.8

7.2

7.6 8

8.4

8.8

9.2

9.6 10

Fre

qu

ency

baARnodes [acc/(106*veh*km)]

Histogram baARnodes


A-15

Average rate approach: ranking of traffic oriented roads

Table 7 Ranking of traffic oriented roads with the average rate approach (red = increase in priority)

Ranking Name Sum(Acc)

[Acc/5a]

ACD)

[CHF/(1000*veh*km)]

baACD)

[CHF/(1000*veh*km)]

avAC

[CHF/a]

Priority Primary

Ranking

Primary

Priority

1 Badenerstrasse East 89 2'586 594 2'142'650 high 1 high

2 Badenerstrasse West 68 1'969 650 1'941'330 high 2 high

3 Albisstrasse 99 1'532 700 1'664'952 high 4 high

4 Limmatstrasse 67 1'376 125 1'658'785 high 3 high

5 Uraniastrasse 78 3'142 904 1'434'652 high 5 high

6 Langstrasse North/Kornhausbrücke 94 3'039 1'174 1'392'647 high 7 medium

7 Sihlstrasse 70 3'457 686 1'385'836 medium 6 medium

8 Thurgauerstrasse/Binzmühlestrasse 45 1'959 723 1'155'557 medium 8 medium

9 Rämistrasse 73 2'084 889 1'117'342 medium 10 medium

10 Universitätstrasse South 41 2'413 823 1'112'574 medium 9 medium

11 Hofwiesenstrasse/Franklinstrasse 45 1'855 355 1'066'624 medium 11 medium

12 Hardturmstrasse West 42 1'064 211 1'029'883 medium 12 medium

13 Sihlquai East 93 1'518 765 991'486 medium 14 medium

14 Binzmühlestrasse 32 2'505 563 990'027 medium 13 medium

15 Dreikönigstrasse/Bleicherweg 58 1'629 338 912'868 medium 15 medium

16 Schaffhauserstrasse North 61 2'587 836 887'968 medium 17 medium

17 Langstrasse South 40 2'011 236 887'767 medium 16 medium

18 Kornhausstrasse/Rötelstrasse 36 1'271 482 873'421 medium 18 medium

19 Birmensdorferstrasse East 40 2'610 877 866'493 medium 19 medium

20 Militärstrasse 45 1'079 242 827'415 medium 20 medium

21 Manessestrasse/Steinstrasse 38 1'395 384 815'740 medium 21 medium

22 Altstetterstrasse North 29 1'239 134 794'411 medium 22 medium

23 Weinbergstrasse 27 916 280 771'053 medium 23 medium

24 Hohlstrasse West 46 1'547 353 747'304 medium 24 medium


A-16


[Acc/5a]

ACD)

[CHF/(1000*veh*km)]

baACD)

[CHF/(1000*veh*km)]

avAC

[CHF/a]

Priority Primary

Ranking

Primary

Priority

25 Birmensdorferstrasse Wes 31 1'477 688 734'605 medium 26 medium

26 Hardtrumstrasse East/

Sihlquai West 73

2'028 1'129 730'671 medium 28 medium

27 Stauffacherstrasse 33 1'631 479 730‘019 medium 25 medium

28 Brandschenkenstrasse 35 906 288 723'321 medium 27 medium

29 Giesshübelstrasse/Bederstrasse 62 1'813 919 711'409 medium 29 medium

30 Schaffhauserstrasse South 35 947 427 675'266 medium 30 medium

31 Baslerstrasse/Bullingerstrasse 54 2'586 1'312 637'225 medium 32 medium

32 Seilergraben 41 711 310 636'326 medium 31 medium



A-17

Average rate approach: ranking of nodes

Table 8 Ranking of nodes with the average rate approach (green = decline in priority)


[Acc/5a]

ACD)

[CHF/(1000*veh*km)]

baACD)

[CHF/(1000*veh*km)]

avAC

[CHF/a]

Priority Primary

Ranking

Primary

Priority

1 Bucheggplatz 99 7'456 1'975 1'644'315 high 2 high

2 Bellevue 135 11'316 5'203 1'528'269 high 1 high

3 Heimplatz 140 8'424 3'330 1'273'587 high 3 high

4 Central 64 6'324 2'114 1'052'386 high 4 high

5 Limmatplatz 51 6'903 2'056 969'317 high 5 high

6 Escherwyssplatz 80 5'004 1'674 832'402 medium 7 high

7 Pfingstweidstrasse/Hardstrasse 33 5'781 1'687 818'889 medium 9 high


Dörflistrasse

35 7'324 1'918 810'971 medium 10 high

9 Langstrassse/Lagerstrasse 50 4'934 1'984 737'515 medium 8 high

10 Römerhof 27 5'664 1'113 682'708 medium 14 medium

11 Bürkliplatz 61 8'728 4'525 630'437 medium 6 high

12 Zwinglihaus 21 3'939 1'010 585'771 medium 19 medium 13 Schaffhauserstrasse/

Seebacherstrasse

24 5'432 1'531 585'151 medium 15 medium

14 Sternen Oerlikon 8 3'042 131 582'145 medium 31 medium 15 Kanonengasse/Militärstrasse 27 3'558 828 545'931 medium 25 medium 16 Wipkingerplatz 37 3'758 1'653 526'457 medium 12 medium 17 Letzigrund 26 3'903 1'353 510'054 medium 21 medium 18 Schmiede Wiedikon 24 4'720 1'541 476'873 medium 24 medium 19 Bahnhofplatz/Bahnhofquai 47 5'284 2'114 475'461 medium 18 medium 20 Schaffhauserstrasse/Franklinstrasse 34 2'911 1'062 462'252 medium 27 medium 21 Kasernenstrasse/Lagerstrasse 19 3'952 911 456'117 medium 35 medium 22 Albisriederplatz 49 4'704 2'424 456'045 medium 13 medium


A-18

Ranking Sum(Acc)

[Acc/5a]

ACD)

[CHF/(1000*veh*km)]

baACD)

[CHF/(1000*veh*km)]

avAC

[CHF/a]

Priority Primary

Ranking

Primary

Priority

23 Schaffhauserstrasse/Glatttalstrasse 32 4'836 1'899 440'536 medium 22 medium 24 Bahnhofplatz/Löwenstrasse 40 4'760 1'919 426'125 medium 23 medium 25 Bahnhofstrasse/Museumstrasse 26 6'284 3'458 423'940 medium 11 medium 26 Birmensdorferstrasse/

Aemtlerstrasse

34 4'076 1'324 412'794 medium 32 medium

27 Talstrasse/Bleicherweg 18 4'192 1'500 403'778 medium 30 medium 28 Thurgauerstrasse/Dörflistrasse 99 4'048 1'494 383'108 medium 34 medium



A-19

Coefficient approach: ranking of traffic oriented roads

Table 9 Ranking of traffic oriented roads with the average rate approach (red = increase in priority, green = decline in priority)


[Acc/5a]

ACD)

[CHF/(1000*veh*km)]

baACD)

[CHF/(1000*veh*km)]

avAC

[CHF/a]

Priority Primary

Ranking

Primary

Priority

1 Badenerstrasse East 89 2'586 786 1'936'222 high 1 high

2 Langstrasse North/Kornhausbrücke 94 3'039 732 1'723'154 high 7 medium

3 Uraniastrasse 78 3'142 561 1'654'686 high 5 high

4 General-Guisan-Quai/Quaibrücke 102 3'510 835 1'631'896 high 37 low

5 Sihlstrasse 70 3'457 388 1'534'408 high 6 high

6 Universitätsstrasse South 41 2'413 588 1'277'179 high 9 medium

7 Badenerstrasse West 68 1'969 1'118 1'252'992 medium 2 high

8 Rämistrasse 73 2'084 812 1'188'955 medium 10 medium

9 Limmatstrasse 67 1'376 501 1'159'657 medium 3 high

10 Thurgauerstrasse/Binzmühlestrasse 45 1'959 743 1'136'932 medium 8 medium

11 Rosengartenstrasse 122 2'638 1'112 1'112'168 medium 189 No infrapot

12 Schaffhauserstrasse North 61 2'587 429 1'094'202 medium 17 medium

13 Binzmühlestrasse 32 2'505 364 1'091'639 medium 13 medium

14 Birmensdorferstrasse East 40 2'610 432 1'089'107 medium 19 medium

15 Kreuzbühlstrasse 54 2'586 514 1'036'060 medium 32 low

16 Hofwiesenstrasse/Franklinstrasse 45 1'855 417 1'022'361 medium 11 medium

17 Hardturmstrasse East/Sihlquai West 73 2'028 783 1'012'137 medium 28 medium

18 Giesshübelstrasse/Bederstrasse 62 1'813 701 884'661 medium 29 medium

19 Langstrasse South 40 2'011 247 882'257 medium 16 medium

20 Dreikönigstrasse/Bleicherweg 58 1'629 406 864'727 medium 15 medium

21 Stauffacherstrasse 33 1'631 423 765'927 medium 25 medium

22 Hohlstrasse West 46 1'547 366 739'012 medium 24 medium

23 Manessestrasse/Steinstrasse 38 1'395 489 730'626 medium 21 medium

24 Birmensdorferstrasse West 31 1'477 724 700'858 medium 26 medium


A-20


[Acc/5a]

ACD)

[CHF/(1000*veh*km)]

baACD)

[CHF/(1000*veh*km)]

avAC

[CHF/a]

Priority Primary

Ranking

Primary

Priority

25 Altstetterstrasse North 29 1'239 280 689'451 medium 22 medium

26 Badenerstrasse (Altstetten) 49 1'492 635 655'637 medium 42 low

27 Kalkbreitestrasse 32 1'622 377 622'838 medium 47 low

28 Hardturmstrasse West 42 1'064 568 598'524 medium 12 medium

29 Hohlstrasse East 47 1'694 510 592'153 medium 79 low

30 Kornhausstrasse/Rötelstrasse 36 1'271 740 588'260 medium 18 medium



A-21

Average rate approach: ranking of nodes

Table 10 Ranking of nodes with the average rate approach (green = decline in priority)


[Acc/5a]

ACD)

[CHF/(1000*veh*km)]

baACD)

[CHF/(1000*veh*km)]

avAC

[CHF/a]

Priority Primary

Ranking

Primary

Priority

1 Bellevue 135 2'829 947 1'882'099 high 1 high 2 Bucheggplatz 99 2'237 535 1'701'835 high 2 high 3 Heimplatz 140 2'106 727 1'379'110 high 3 high 4 Central 64 1'581 557 1'024'266 medium 4 high 5 Limmatplatz 51 1'381 548 832'851 medium 5 high 6 Eschwerwyssplatz 80 1'251 486 765'128 medium 7 high 7 Schwamendingerstrasse/

Dörflistrasse

35 1'099 343 755'490 medium 10 high

8 Bürkliplatz 61 1'309 571 737'833 medium 6 high 9 Langstrasse/Lagerstrasse 50 1'234 536 697'151 medium 8 high

10 Pfingstweidstrasse/Hardstrasse 33 1'156 488 668'277 medium 9 high 11 Römerhof 27 850 249 600'267 medium 14 medium 12 Schaffhauserstr/Seebacherstrasse 24 815 301 514'246 medium 15 medium 13 Sternen Oerlikon 8 608 115 493'875 medium 31 medium 14 Wipkingerplatz 37 940 482 457'423 medium 12 medium 15 Bahnhofstrasse/Muesumstrasse 57 943 487 456'019 medium 11 medium 16 Bahnhofplatz/Bahnhofquai 47 793 363 429'140 medium 18 medium 17 Zwinglihaus 21 788 363 425'227 medium 19 medium 18 Schmiede Wiedikon 24 708 302 406'311 medium 24 medium 19 Kanonengasse/Militärstrasse 27 712 323 388'135 medium 25 medium 20 Schaffhauserstr/Glatttalstrasse 32 725 341 384'238 medium 22 medium



A-22

A 8 Ranking of nodes with radius 30 meters

Table 11 Prioritization of nodes (red = increase in priority; green = decline in priority)

Ranking Name AADT

[veh/d]

Acc(FSI)

[Acc/5a]

Acc(MI)

[Acc/5a]

Acc(PDO)

[Acc/5a]

avAC

[CHF/a]

Priority Primary Ranking Primary Priority

1 Bellevue 58'182 7 36 71 2'565'137 high 1 high

2 Bucheggplatz 22'085 8 31 46 2'116'610 high 2 high

3 Heimplatz 37'234 5 23 106 1'937'139 high 3 high

4 Schwamendingerstr/Dörflistrasse 21'014 5 14 13 1'473'767 high 10 high

5 Langstrasse/Lagerstrasse 22'190 4 15 21 1'297'168 high 8 high

6 Bürkliplatz 49'590 3 10 32 1'247'840 high 6 high

7 Limmatplatz 22'996 4 9 14 1'166'084 high 5 high

8 Schaffhauserstr/Seebacherstrasse 16'778 4 9 7 1'132'966 high 15 medium

9 Stauffacher 25'020 3 12 23 1'087'774 high 16 medium

10 Zwinglihaus 11'296 4 8 4 1'072'598 medium 19 medium

11 Kanonengasse/Militärstrasse 9'263 3 10 12 895'713 medium 25 medium

12 Bahnhofplatz/Bahnhofquai 23'170 2 10 24 855'791 medium 18 medium

13 Römerhof 12'193 3 10 7 842'464 medium 14 medium

14 Kreuzplatz 23'826 3 7 15 834'513 medium 20 medium

15 Walchebrücke/Neumühlequai 35'452 2 6 33 794'040 medium 17 medium

16 Utoquai/Kreuzstrasse 35'484 3 7 10 770'544 medium 26 medium

17 Pfingstweidstrasse/Hardstrasse 16'886 3 10 2 768'977 medium 9 high

18 Schmiede Wiedikon 16'440 2 8 21 763'931 medium 24 medium

19 Talstrasse/Bleicherweg 18'860 3 6 12 746'817 medium 30 medium

20 Bahnhofplatz/Löwenstrasse 21'032 2 11 15 736'146 medium 23 medium

21 Tobelhofstrasse/Drewiesenstrasse 16'130 3 7 5 725'718 medium 37 medium

22 Schaffhauserstr/Glatttalstrasse 20'812 2 6 22 699'648 medium 22 medium

23 Bahnhofstrasse/Uraniastrasse 25'314 3 7 5 687'976 medium 36 medium

24 Schaffhauserstrasse/Franklinstrasse 11'878 3 1 16 687'106 medium 27 medium

25 Schimmelstrasse/Manessestrasse 70'090 1 7 35 677'993 medium 29 medium


A-23

Ranking Name AADT

[veh/d]

Acc(FSI)

[Acc/5a]

Acc(MI)

[Acc/5a]

Acc(PDO)

[Acc/5a]

avAC

[CHF/a]


26 Zehntenhausplatz 28'635 2 5 24 677'894 medium 44 medium

27 Hubertus 24'848 2 11 13 673'932 medium 49 medium

28 Kornhausbrücke/Rousseaustrasse 28'530 2 12 10 623'690 medium 28 medium

29 Wipkingerplatz 18'480 3 4 8 608'674 medium 12 medium

30 Sihlhölzlistrasse/Manessestrasse 42'383 2 7 15 608'458 medium 38 medium

31 Goldbrunnenplatz 23'420 2 8 15 603'074 medium 46 medium

32 Central 23'645 3 5 6 593'446 medium 4 high

33 Utoquai/Falkenstrasse 40'644 2 2 23 588'585 medium 45 medium

34 Seebahnstrasse/Kalkbreitestrasse 35'324 1 13 22 586'942 medium 54 medium

35 Stampfenbachplatz 13'276 2 8 8 580'444 medium 48 medium

36 Schaffhauserplatz 18'534 3 4 4 575'878 medium 52 medium

37 Krematorium Sihlfeld 15'248 2 8 7 569'242 medium 51 medium

38 Birmensdorferstrasse/Schaufelbergstrasse 18'096 2 9 7 567'357 medium 56 medium

39 Bernerstrasse Süd/Bändlistrasse 16'654 2 9 5 558'018 medium 63 low

40 Bahnhof Oerlikon Ost 24'227 1 8 24 551'694 medium 47 medium

41 Furttalstrasse/

Wehntalerstrasse

26'180 2 6 8 546'528 medium 50 medium

42 Letzigrund 15'127 2 11 0 546'284 medium 21 medium

43 Hardstrasse/Bullingerstrasse 10'145 2 6 8 543'408 medium 62 low

44 Albisriederstrasse/Birmensdorferstrasse 16'991 2 7 5 535'109 medium 53 medium

45 Badenerstrasse/Kalkbreitestrasse 12'460 2 5 8 521'509 medium 55 medium

46 Birmensdorferstrasse/Aemtlerstrasse 14'510 1 14 6 505'430 medium 32 medium

47 Bahnhofquai/Museumsstrasse 37'893 1 2 30 504'902 medium 11 medium

48 Tessinerplatz 17'516 2 5 7 503'173 medium 66 low

49 Feldeggstrasse/Seefeldstrasse 6'240 2 4 7 502'311 medium 64 low

50 Lagerstrasse/Kasernenstrasse 25'342 2 5 10 496'206 medium 35 medium

51 Wehntalerstrasse/Hofwiesenstrasse 33'410 1 5 27 482'733 medium 60 medium

52 Talstrasse/Sihlstrase 9'986 2 3 8 480'546 medium 33 medium

53 Binzmühlestrasse/Fronwaldstrasse 10'637 2 3 8 472'776 medium 65 low

54 Sihlstrasse/Uraniastrasse 36'908 2 7 13 470'041 medium 67 medium

55 Thurgauserstrasse/Dörflistrasse 16'372 2 4 6 444'783 medium 34 medium


A-24

Ranking Name AADT

[veh/d]

Acc(FSI)

[Acc/5a]

Acc(MI)

[Acc/5a]

Acc(PDO)

[Acc/5a]

avAC

[CHF/a]


56 Wasserwerkstrasse/Neumühlequai 34'692 1 6 19 443'746 medium 59 medium

57 Altstetterstrasse/Rautistrasse 10'280 1 11 8 441'091 medium 77 low



A-25

A 9 Nodes ranking based on incoming vehicles

Table 12 Prioritization of nodes (red = increase in priority; green = decline in priority)

Ranking Name Main stream

[veh/d]

Minor stream)

[veh/d] ACD

20 baACD

20 avAC

[CHF/a]

Priority Primary

Ranking

Primary

Priority

1 Bucheggplatz 11'619 3'017 1035 133 5'414'342 high 2 high

2 Wipkingerplatz 6'920 5'047 443 109 2'344'005 high 12 medium

3 Wallisellenstrasse/Saatlenstrasse 2'355 435 702 12 2'071'555 high 109 low

4 Langstrasse/Lagerstrasse 12'282 3'847 492 146 1'727'204 high 8 high

5 Limmatplatz 18'609 2'008 619 187 1'727'088 high 5 high

6 Pfingstweidstrasse/Hardstrasse 13'610 2'871 540 149 1'562'911 medium 9 high

7 Central 13'912 5'636 489 177 1'559'449 medium 4 high

8 Escherwyssplatz 18'200 4'042 419 202 1'523'234 medium 7 high

9 Heimplatz 16'048 7'872 513 217 1'481'942 medium 3 high

10 Römerhof 10'221 2'473 468 115 1'410'990 medium 14 medium

11 Schaffhauserstrasse/Franklinstrasse 5'716 5'157 372 99 1'366'020 medium 27 medium

12 Sternen Oerlikon 5'683 2'842 415 77 1'349'903 medium 31 medium

13 Kanonengasse/Militärstrasse 5'639 3'895 416 86 1'318'078 medium 25 medium

14 Stauffacher 11'984 5'158 324 155 1'178'780 medium 16 medium

15 Schwamendingerplatz 9'939 435 435 43 1'176'880 medium 96 low

16 Seefeldstrasse/Höschgasse 5'479 2'132 341 69 1'087'965 medium 64 low

17 Feldstrasse/Hohlstrasse 3'895 1'948 363 24 1'016'364 medium 87 low

18 Bellevue 33'639 7'161 502 168 1'002'276 medium 1 high

19 Zwinglihaus 10'290 3'717 349 127 887'792 medium 19 medium

20 Regensbergstrasse/Hofwiesenstrasse 6'513 1'031 304 31 818'235 medium 103 low

21 Aargauerstrasse/Würzgrabenstrasse 7'027 529 273 69 817'189 medium 151 low

22 Birmensdorferstrasse/Aemtlerstrasse 14'134 1'859 327 66 782'600 medium 32 medium

20

The unit of ACD and baACD is [CHF/(1000*veh*km)]


A-26

Ranking Name Main stream

[veh/d]

Minor stream)

[veh/d] ACD

21 baACD

20 avAC

[CHF/a]

Priority Primary

Ranking

Primary

Priority

23 Schaffhauserstrasse/Seebachstrasse 13'466 3'657 331 71 781'884 medium 15 medium

24 Schwamendingerstrasse/Dörflistrasse 11'426 7'567 326 78 744'073 medium 10 high

25 Kasernenstrasse/Lagerstrasse 7'179 4'330 291 47 731'491 medium 35 medium

26 Letzigrund 10'474 4'347 317 134 730'061 medium 21 medium

27 Thurgauerstrasse/Dörflistrasse 11'426 2'842 292 59 699'330 medium 34 medium


21

The unit of ACD and baACD is [CHF/(1000*veh*km)]


A-27

A 10 Assigning accidents on nodes to traffic oriented roads

Table 13 Difference of priority when assigning all the accidents on nodes to roads (red = increase in priority; green = decline in priority)


[Acc/5a]

Acc(FSI)

[Acc/5a]

Acc(MI)

[Acc/5a]

Acc(PDO)

[Acc/5a]

avAC

[CHF/a]


1 Limmatstrasse 223 17 62 144 4'464'480 high 3 high

2 Rämistrasse 245 21 68 156 4'204'861 high 10 medium

3 Sihlstrasse 207 18 67 122 4'134'161 high 6 high

4 Badenerstrasse West 170 25 73 72 3'983'912 high 2 high

5 Uraniastrasse 231 18 60 153 3'943'087 high 5 high

6 Militärstrasse 203 15 66 122 3'934'468 high 20 medium

7 Hofwiesenstrasse/Franklinstrasse 171 17 48 106 3'723'412 high 11 medium

8 Stauffacherstrasse 188 16 56 116 3'714'140 high 25 medium

9 Universitätsstrasse South 196 18 71 107 3'708'467 high 1 high

10 Schaffhauserstrasse South 190 17 60 113 3'586'772 high 30 medium

11 Langstrasse North/Kornhausbrücke 197 18 84 95 3'382'342 medium 7 medium

12 Leonhardstrasse/Weinbergstrasse 138 17 38 83 3'331'575 medium 43 low

13 Birmensdorfstrasse East 145 15 71 59 3'050'551 medium 19 medium

14 Langstrasse South 137 13 39 85 3'025'242 medium 16 medium

15 Talstrasse 206 9 51 146 2'699'568 medium 95 low

16 Lagerstrasse/Kanonengasse 157 9 51 97 2'557'848 medium 35 low

17 Kalkbreitestrasse 125 12 54 59 2'554'387 medium 47 low

18 Dreikönigstrasse/Bleicherweg 145 10 34 101 2'489'241 medium 15 medium

19 Dörflistrasse 113 13 44 56 2'437'402 medium 80 low

20 Schaffhauserstrasse North 128 12 52 64 2'387'263 medium 17 medium

21 Stampfenbachstrasse 91 12 27 52 2'300'351 medium 105 low

22 Bahnhofquai/Bahnhofbrücke 226 14 48 164 2'278'797 medium 142 no InfraPot

23 Hardturmstrasse East/Sihlquai West 204 10 56 138 2'143'360 medium 28 medium

24 Gessnerallee 200 7 44 149 2'137'972 medium 89 low


A-28


[Acc/5a]

Acc(FSI)

[Acc/5a]

Acc(MI)

[Acc/5a]

Acc(PDO)

[Acc/5a]

avAC

[CHF/a]


25 Baslerstrasse/Bullingerstrasse 122 11 38 73 2'126'148 medium 31 medium

26 Dörflistrasse/Tramstrasse 86 10 31 45 2'011'717 medium 191 no InfraPot

27 Albisstrasse 139 17 60 62 1'984'717 medium 4 high

28 Hottingerstrasse 124 13 48 63 1'973'526 medium 44 low

29 Sihlhölzlistrasse/Selnaustrasse 157 9 39 109 1'908'920 medium 176 no InfraPot

30 Kasernenstrasse 165 4 44 117 1'874'229 medium 84 low

31 Hardstrasse 155 8 60 87 1'871'954 medium 81 low

32 Eidmattstrasse/Hegibachstrasse 102 7 37 58 1'868'632 medium 110 low

33 Herdernstrasse/Letzigraben 98 7 52 39 1'856'762 medium 107 low

34 Kornhausstrasse/Rötelstrasse 92 11 44 37 1'811'555 medium 18 medium

35 Altstetterstrasse North 84 7 37 40 1'801'841 medium 22 medium

36 Giesshübelstrasse/Bederstrasse 144 9 61 74 1'798'572 medium 29 medium

37 Feldeggstrasse/Neumünsterstrasse 96 10 33 53 1'793'672 medium 150 no InfraPot

38 Neumühlequai/Wasserwerkstrasse 238 10 68 160 1'790'557 medium 186 no InfraPot

39 Seebahnhstrasse South 206 6 63 137 1'718'877 medium 183 no InfraPot

40 Bändlistrasse/Aargauerstrasse 100 8 39 53 1'703'517 medium 108 low

41 Sihlquai East 158 11 38 109 1'690'084 medium 14 medium

42 Wasserwerkstrasse North 95 10 34 51 1'674'467 medium 57 low

43 Aemtlerstrasse 81 7 39 35 1'641'657 medium 145 no InfraPot

44 Forchstrasse 126 11 42 73 1'614'550 medium 100 low

45 General-Guisan Quai/Quabrücke 226 12 73 141 1'613'866 medium 37 low

46 Schaffhauserstrasse/Irchelstrasse 64 10 24 30 1'576'367 medium 41 low

47 Universitätsstrasse South 71 12 29 30 1'502'606 medium 9 medium

48 Limmatstrasse 223 17 62 144 4'464'480 medium 3 high



A-29

A 11 Correlation analysis


Figure 5 Correlation analysis between the section length and the number of accidents


Nodes

Figure 6 Correlation analysis between the AADT and the number of accidents


0

20

40

60

80

100

120

140

500 1'000 1'500 2'000

No o

f acc

iden

ts

length [m]

Correlation Road section length/Accidents

r = 0.27

Road section


0

20

40

60

80

100

120

140

160

0 20'000 40'000 60'000 80'000

No o

f acc

iden

ts

AADT

Correlation AADT/Accidents

r = 0.50

AADT of nodes



A-30

A 12 Network density of residential zones

Table 14 New ranking of residential zones including network densities (red = increase in priority)

Ranking Name Network length

[km]

Area

[km2]

Ratio

[km/km2]

AC

[CHF/zone] Deviation to ZH

22

[%]

Ratio Value23

[-]

Priority Primary

ranking

Primary

priority

1 University quarter 3.08 0.43 7.19 15'828'000 79.40% 0.077 high 2 high

2 Leutschenbach 4.09 1.09 3.73 6'171'000 41.23% 0.058 high 11 medium

3 Bahnhofstrasse/Rennweg 3.17 0.28 11.31 17'772'000 124.90% 0.055 high 1 high

4 Letzipark/Hohlstrasse 3.12 0.72 4.34 5'643'000 47.93% 0.046 medium 8 medium

5 Industry Herdern 0.20 0.22 0.94 1'167'000 10.40% 0.043 medium 7 medium

6 Wollishofen 23.12 1.63 14.21 17'025'000 156.92% 0.042 medium 24 low

7 Industry Altstetten/Hardhof 1.15 0.83 1.38 1'527'000 15.23% 0.039 medium 12 medium

8 Stadelhofen 1.87 0.25 7.64 8'190'000 84.31% 0.038 medium 3 medium

9 Hirzenbach 10.70 1.57 6.83 7'062'000 75.41% 0.036 medium 29 low

10 Zürichberg 23.00 2.32 9.90 8'700'000 109.31% 0.031 medium 48 low

11 Mythenquai 7.97 1.00 7.96 6'720'000 87.83% 0.030 medium 22 low

12 Triemli 8.15 1.30 6.25 4'755'000 69.00% 0.027 medium 32 low

13 Förrlibuckstrasse 0.75 0.27 2.77 1'983'000 30.60% 0.025 medium 4 medium

14 Oerlikon North 8.92 1.07 8.32 5'766'000 91.80% 0.024 medium 30 low

15 Balgrist 11.70 1.32 8.88 6'135'000 97.99% 0.024 medium 56 low

16 Station Oerlikon 4.17 0.39 10.55 7'068'000 116.47% 0.023 medium 10 medium

17 Affoltern 5.12 0.93 5.52 3‘555‘000 60.89% 0.022 medium 26 medium


22

Total value for ZH: Network length = 384.13 km, Area = 42.41 km2, Ratio = 9.06 km/km2, AC = 259’059’000 CHF, Deviation = 100%

23 𝑅𝑎𝑡𝑖𝑜 𝑉𝑎𝑙𝑢𝑒 =

1

(𝐴𝐶𝑍𝐻

𝑅𝑎𝑡𝑖𝑜𝑍𝐻)

∗𝐴𝐶𝑧𝑜𝑛𝑒

𝑅𝑎𝑡𝑖𝑜𝑧𝑜𝑛𝑒


A-31

A 13 Influence of Tram


Table 15 Ranking of traffic oriented roads without considering tram accidents (red = increase in priority, green = decline in priority)

Ranking Name Sum(ACC)

[Acc/5a]

Percentage of total acc

[%]

ACD)

[CHF/(1000*veh*km)]

Percentage of total ACD

[%]

Priority Primary

Ranking

Primary

Priority

1 Uraniastrasse24 78 100% 3‘142 100% high 5 high

2 Albisstrasse 88 89% 1‘406 92% high 4 high

3 Langstrasse North/

Kornhausbrücke24

94 100% 3‘039 100% high 7 medium

4 Badenerstrasse West 58 85% 1‘533 78% high 2 high

5 Hofwiesenstrasse/Franklinstrasse 43 96% 1‘830 99% high 11 medium

6 Badenerstrasse East 73 82% 1‘554 60% high 1 high

7 Binzmühlestrasse24 32 100% 2‘505 100% high 13 medium

8 Thurgauerstrasse/Binzmühlestrasse 43 96% 1‘792 91% medium 8 medium

9 Langstrasse South24 40 100% 2‘011 100% medium 16 medium

10 Kornhausstrasse/Rötelstrasse24 36 100% 1‘271 100% medium 18 medium

11 Schaffhauserstrasse North 56 92% 2‘467 95% medium 17 medium

12 Militärstrasse 44 98% 1‘070 99% medium 20 medium

13 Manessestrasse/Steinstrasse24 38 100% 1‘395 100% medium 21 medium

14 Sihlstrasse/Bahnhofstrasse 65 93% 2‘310 67% medium 6 high

15 Sihlquai East 89 96% 1‘392 92% medium 14 medium

16 Altstetterstrasse North24 29 100% 1‘239 100% medium 22 medium

17 Rämistrasse 56 77% 1‘739 83% medium 10 medium

18 Weinbergstrasse 26 96% 902 98% medium 23 medium

19 Hohlstrasse West24 46 100% 1‘547 100% medium 24 medium

20 Dreikönigstrasse/Bleicherweg 54 93% 1‘372 84% medium 15 medium

24

No tram accidents recorded


A-32


[Acc/5a]


[%]

ACD)

[CHF/(1000*veh*km)]


[%]

Priority Primary

Ranking

Primary

Priority

21 Brandschenkenstrasse24 35 100% 906 100% medium 27 medium

22 Altstetterstrasse North 72 99% 2‘007 99% medium 28 medium

23 Hardturmstrasse West 35 83% 783 74% medium 12 medium

24 Stauffacherstrasse 28 85% 1‘547 95% medium 25 medium

25 Universitätsstrasse South 27 66% 1‘794 74% medium 9 medium

26 Baslerstrasse/Bullingerstrasse2424 41 100% 711 100% medium 31 medium

27 Lagerstrasse/Kanonengasse24 52 100% 1‘137 100% medium 35 low

28 Seilergraben/Heimstrasse 52 96% 2‘534 98% medium 32 low

29 Krähbühlstrasse 28 93% 412 97% medium 33 low



A-33

Nodes

Table 16 Ranking of nodes without considering tram accidents (red = increase in priority, green = decline in priority)


[Acc/5a]


[%]

ACD)

[CHF/(1000*veh*km)]


[%]

Priority Primary

Ranking

Primary

Priority

1 Bellevue 124 92% 9'660 85% high 1 high

2 Bürkliplatz 56 92% 8'272 95% high 6 high

3 Heimplatz 133 95% 8'172 97% high 3 high


Dörflistrasse25

35 100% 7'324 100% high 10 high

5 Bucheggplatz 94 95% 6'794 91% high 2 high

6 Bahnhofstrasse/Museumstrasse 56 98% 6'224 99% high 11 medium

7 Pfingstweidstrasse/Hardstrasse25 33 100% 5'781 100% high 9 high

8 Römerhof 23 85% 5'372 95% high 14 medium

9 Limmatplatz 41 80% 4'956 72% high 5 high

10 Langstrasse/Lagerstrasse25 50 100% 4'934 100% high 8 high

11 Schaffhauserstrasse/Glattalstrasse25 32 100% 4'836 100% high 22 medium

12 Escherwyssplatz 76 95% 4'829 96% high 7 high

13 Utoquai/Kreuzstrasse25 29 100% 4'812 100% high 26 medium

14 Schimmelstrasse/Manessestrasse25 55 100% 4'636 100% high 29 medium

15 Walchebrücke/Neumühlequai25 63 100% 4'488 100% medium 17 medium

16 Central 55 86% 4'375 69% medium 4 high

17 Weststrasse/Manessestrasse25 35 100% 4'200 100% medium 40 medium

18 Sihlhölzlistrasse/Manessestrasse25 33 100% 4'132 100% medium 38 medium

19 Talstrasse/Bleicherweg 32 94% 4'072 97% medium 30 medium

20 Thurgauerstrasse/Dörflistrasse25 18 100% 4'048 100% medium 34 medium


Aemtlerstrasse

25 96% 4'016 99% medium 32 medium

22 Zwinglihaus25 21 100% 3'939 100% medium 19 medium

25

No tram accidents recorded


A-34


[Acc/5a]


[%]

ACD)

[CHF/(1000*veh*km)]


[%]

Priority Primary

Ranking

Primary

Priority

23 Tobelhofstrasse/

Dreiwiesenstrasse25

16 100% 3'928 100% medium 37 medium

24 Letzigrund25 26 100% 3'903 100% medium 21 medium


Seebacherstrasse

17 71% 3'884 72% medium 15 medium

26 Bahnhofplatz/Löwenstrasse 39 98% 3'832 81% medium 23 medium

27 Farbhof25 26 100% 3'816 100% medium 42 medium

28 Utoquai/Falkenstrasse25 32 100% 3'812 100% medium 45 medium

29 Kasernenstrasse/Lagerstrasse 17 89% 3'780 96% medium 35 medium

30 Kornhausbrücke/Rousseaustrasse25 27 100% 3'636 100% medium 28 medium

31 Wipkingerplatz 34 92% 3'619 96% medium 12 medium

32 Schmiede Wiedikon 21 88% 3'568 76% medium 24 medium


nsm - advancing the model for safety improvement of the road

Documents