Back
Report numberRA-2015-003
TitleNetwork Safety Management: A ranking of dangerous road segments on the TEN-T network in Flanders
Subtitle
AuthorsKurt Van Hout
Caroline Ariën
Stijn Daniels
Published byPolicy Research Centre for Traffic Safety 2012-2015
Number of pages111
Date09/04/2015
ISBN
Document languageDutch
Partner(s)Universiteit Hasselt
Work packageWP5: Ranking and evaluation of the measures
Summary

The European directive 2008/96/EC on road infrastructure safety management involves the implementation of procedures relating to road safety impact assessments, road safety audits, the management of road network safety and safety inspections by the member states. This Directive was implemented into Flemish legislation by the decree of June 17, 2011 (published in the Belgian Official Journal on July 4, 2011) and the consecutive Decision of the Flemish Government of Feb, 3 2012 on the execution (published in the Belgian Official Journal on April 19, 2012).

 

This legislation imposes, among others, a classification of road segments with a high number of accidents and a classification of the road safety of the road network. As the first classification is targeted at the number of crashes, the second is targeted at the potential to improve road safety. The application area of both the EU-directive and the Flemish Decree is limited to those road segments that make up the trans-European road network (TEN-T), which in Flanders largely resembles the motorway network.

 

In this report a method is described and applied to screen the TEN-T road network in the Flanders region. The goal of this screening is the selection of a relatively small group of road segments that can be subjected afterwards to an in-depth investigation that will allow proposing appropriate measures. The network screening as such is therefore not an ultimate goal, but merely the first step to improve the road safety on dangerous locations.

 

The Empirical Bayes (EB) method is in the scientific literature considered to be the state-of-the-art approach for identifying dangerous road segments and is therefore adopted in the present study. The EB approach offers a solution to two problems when estimating the number of crashes: it corrects for regression to the mean and it improves the accuracy of the estimation by calculating it as a weighted average of the actual number of crashes and the normal number of crashes where the latter is based on the outcomes of a risk model.

 

For the ranking purpose six risk models were built for Flemish highways based on the accident and traffic intensity data for the period 2008-2010. On the one hand there was some variation in the model form of the risk model by including more or less explanatory variables in the model or by changing the function form. On the other hand the lengts of the road segments was varied. Based on these six different models we tried to get more insight into the optimal length of a road segment, the model form en the number of explanatory variables.

 

The model describes the number of crashes as a function of a number of explanatory variables. Two models are built: one simple model only including the length of the segment (L) and traffic intensity (I) as explanatory variables and a more elaborate one built on the variables included in the subsegment database of the Flemish Traffic Centre. Apart from length (L) and traffic intensity (I) the daily number of trucks (Z), the year (jaar), pavement (verh), number of driving lanes (brs), the presence of an emergency lane (pech) and bus lane (bus) and a porch for driving lane signalisation and finally road quality (sprd) are included in the final model. The normal number of crashes μ can be calculated by:

 

Risk models for segments

μ = e-17,0652 * L0,9532 * I1,0266  (model 1)

μ = e-9,7586 * L0,9705 * I0,2285 * e0,0266*10^-3  (model 2)

μ = e12,086 * L0,9500 * I-2,3606 * e(0,2680*10^-3*I + -0,3007*10^-8*I² + 0,0312*10^-12*I³)  (model 3)

 

Risk models for subsegments

μ = e-7,9658 * L0,7856 * e(1,1680*10^-4*I - 0,1589*10^-8*I² + 0,0079*10^-12*I³)  (model 4b)

μ = e-8,6804 * L0,8772 * Z0,1003 * e(1,0456*10^-4*I - 0,1598*10^-8*I² + 0,0084*10^-12*I³) * Cjaar * Cverh * Cbrs * Cpech * Cbus * Cportiek * Csprd  (model 5)

 

The number of crashes on Flemish highways thus increases with both segment length and traffic intensity. The factors Cx are safety performance factors which influence the global values, apart from L and I. The number of accidents in  2008 e.g. is significantly higher compared to the number of accidents in 2009 and 2010. Road sections with a concrete pavement show 20% less accidents compared to road sections with a bituminous pavement. Even when corrected for traffic intensity, the number of accidents increases when the number of driving lanes increases. Furthermore road sections with an emergency lane are characterised by more accidents while road sections with a bus lane have less accidents. Also road sections containing porch signalisation have more accidents. It appears that no clear relation exists between the road surface quality and the number of crashes.

 

Road safety can be expressed in different manners. Therefore four indicators were used to construct different rankings: number of crashes, crash density (crashes/km), crash risk (crashes/vehicle-km) and the safety potential (expected minus normal number of crashes, expressed per km road length). As mentioned before, segment length and traffic intensity are major influential factors. Because of the important influence length has on the number of crashes we only compare the rankings based on the normalised indicators per km and ignore the ranking based on crash numbers after model 1.

 

The used indicators each illustrate a different aspect of road safety and the results therefore show considerable differences in the outcomes of the different rankings. The ranking based on crash density gives the longer (901 m on average for ranking based on segments 167 m on average for sugsegments) and busiest (77.000 vehicles per day on average for segments and slightly over 86.000 vehicles per day on average for subsegments) road sections in the tops with the highest ranked road segments. A ranking based on crash risk selects the least traffic intense sections (57.000 vehicles per day on segments and almost 63.000 vehicles per day on average for subsegments), that are often short (761 m on average for segments and 145 m on average for subsegments). The ranking based on safety potential ends up with sections that have a traffic intensity that is in between the other two (69.000 vehicles per day on average for both segments and subsegments). The average length of the top 25 highest ranked segments is 965 m. For subsegments this average length is 144 m.

 

In total, the 3 rankings (based on density, risk and potential) fo segments (top 25’s) (model 1) contain 37 different segments. Rankings based on subsegments (model 5) end up with 75 different subsegments in the three top 50’s. Nevertheless, 15 segments are present in each of the three top 25’s and 30 subsegments are present in each of the top 50 of subsegments. These segments have thus a high score for crash density, risk and safety potential and are listed in Tabel 3 and Tabel 4. Although the length of the road segments was segments and subsegments respectively, comparable road segments can be found in both tables. These pairs are visualized by means of the colors. Remarquable, the Antwerp’s Ring is well represented in these summary lists.

 

Tabel 3: 15 segments which are represented in each of the top 25 of segments with highest ranking (model 1)

R1 (richting 2), Borgerhout tot Antwerpen-Oost

A3 (E40 richting 1), complex Bertem

R1 (richting 2), Berchem tot Borgerhout

A3 (E40 richting 1), Bertem tot Heverlee

R1 (richting 1), Borgerhout tot Berchem

A10 (E40 richting 2), Gent-St.-Pieters tot Zwijnaarde

R1 (richting 1), Berchem tot Antwerpen-Zuid

A10 (E40 richting 2), complex Erpe-Mere

R1 (richting 1), knoop Antwerpen-Oost (Ring 1)

A13 (E313 richting 2), complex Massenhoven

R1, (richting 2) knoop Antwerpen-Oost (Ring 2)

A13 (E34 richting 1), Antwerpen-Oost tot Wommelgem

R1 (richting 2), Linkeroever tot Antwerpen-Centrum

A1 (E19 richting 2), complex Mechelen-Noord

A14 (E17 richting 1), Gent-Centrum tot Gentbrugge

 

 

Tabel 4: 30 segments which are represented in each of the top 50 of segments with highest ranking (model 5)

Nr, richting

Van KM

Tot KM

Deel van

R1, richting 1

7,166

7,216

Antwerpen-Oost tot Borgerhout

8,158

8,531

Borgerhout tot Berchem

9,286

9,406

9,477

9,59

Berchem tot Antwerpen-Zuid

12,072

12,121

Le Grellelaan tot Antwerpen-Centrum

13,393

13,471

Antwerpen-Centrum tot Linkeroever

13,471

13,523

14,486

14,514

R1, richting 2

4,894

4,85

Deurne tot Merksem

6,569

6,397

Borgerhout tot Antwerpen-Oost

6,722

6,569

7,034

6,722

13,524

13,472

Linkeroever tot Antwerpen-Centrum

14,384

14,232

14,405

14,384

14,929

14,792

15,044

14,948

Antwerpen-West tot Linkeroever

A1 (E19), richting 1

27,553

27,635

Kontich tot UZA

A1 (E19), richting 2

34,457

34,413

Antwerpen-Zuid tot Wilrijk

34,502

34,457

A10 (E40), richting 1

33,892

34,074

Erpe-Mere to parking Wetteren

43,307

43,347

Merelbeke tot Zwijnaarde

49,798

50,001

St-Denijs-Westrem tot parking Drongen

A10 (E40), richting 2

44,828

44,699

Gent-St. Pieters tot Zwijnaarde

46,059

45,986

St-Denijs-Westrem tot Gent-St. Pieters

A13 (E34), richting 1

0,687

0,800

Antwerpen-Oost tot Wommelgem

0,800

1,265

A13 (E34), richting 2

0,754

0,718

Wommelgem tot Antwerpen-Oost

A14 (E17), richting 1

51,355

51,722

Gent-Centrum tot Gentbrugge

99,500

99,701

Zwijndrecht tot Antwerpen-West

 

The road segments mentioned above are therefore suggested to be the first to be subjected to an in-depth analysis and inspection to clarify the accident causes that will help to define remedial measures.

DownloadPDF icon RA-2015-003.pdf
Lijn

Mission

The Policy Research Centre for Traffic Safety carries out policy relevant scientific research under the authority of the Flemish Government. The Centre is the result of a

cooperation between Hasselt University, KU Leuven and VITO, the Flemish Institute for Technological Research.

Partners

Leuven vito