safety effects of low-cost engineering measures. an observational study in a portuguese multilane...

7
Accident Analysis and Prevention 48 (2012) 346–352 Contents lists available at SciVerse ScienceDirect Accident Analysis and Prevention jo ur n al hom ep a ge: www.elsevier.com/locate/aap Safety effects of low-cost engineering measures. An observational study in a Portuguese multilane road Sandra Vieira Gomes , João Lourenc ¸ o Cardoso Laboratório Nacional de Engenharia Civil, Department of Transportation, Av. Brasil, 101, 1700-066 Lisbon, Portugal a r t i c l e i n f o Article history: Received 23 December 2009 Received in revised form 2 January 2012 Accepted 7 February 2012 Keywords: Before–after study Median Low-cost engineering measures Road safety a b s t r a c t Single carriageway multilane roads are not, in general, a very safe type of road, mainly because of the high number of seriously injured victims in head-on collisions, when compared with dual carriageway multilane roads, with a median barrier. In this paper the results of a study on the effect of the application of several low cost engineering measures, aimed at road infrastructure correction and road safety improvement on a multilane road (EN6), are presented. The study was developed by the National Laboratory of Civil Engineering (LNEC) for the Portuguese Road Administration and involved a comparison of selected aspects of motorized traffic behaviour (traffic volumes and speeds) measured in several sections of EN6, as well as monitoring of road safety developments in the same road. The applied low cost engineering measures allowed a reduction of 10% in the expected annual number of personal injury accidents and a 70% decrease in the expected annual number of head-on collisions; the expected annual frequency of accidents involving killed and seriously injured persons was reduced by 26%. © 2012 Elsevier Ltd. All rights reserved. 1. Introduction The case study described in this paper refers to the analysis of the effect of several road safety measures undertaken on a Portuguese single carriageway multilane road (EN 6), following the construc- tion of a new motorway (A5), parallel to the existing road, which links directly Lisboa to Cascais. Several low cost engineering mea- sures were implemented thereafter, on a 3.7 km long stretch of EN6 (between 3.2 km and 6.9 km) to improve its safety record. The evaluation was carried through a “before–after” study with a control group. A stretch of another single carriageway multilane road (the EN125 in Algarve) was used as control group. According to the Organization for Economic Co-operation and Development (OECD), integrated programmes on road safety aim- ing to reduce accident severity should focus on several traffic system elements, namely by adaptation of road environment char- acteristics to the intended road uses (OECD, 1984). The application of road safety measures is based on general knowledge on the road safety phenomenon, sometimes comple- mented by results from the analysis of accident data in the area to be corrected. Corresponding author. Tel.: +351 21 8443528; fax: +351 21 8443029. E-mail addresses: [email protected] (S. Vieira Gomes), [email protected] (J.L. Cardoso). Better results may be achieved with road safety interventions if procedures for their implementation include the following steps (Cardoso and Bairrão, 2001): - Network safety diagnosis to identify locations with high influence of road characteristics on accident occurrence. - Safety analysis on each selected location for potential interven- tions, to identify the main road characteristics leading to safety issues. - Selection and application of appropriate safety measures. - Safety monitoring of the corrected sites and evaluation of results. The decision process for selecting corrective measures takes into account the available budget for the programme, the safety issues identified by the diagnosis and the anticipated benefits from the potential suitable measures. Monitoring safety developments fol- lowing interventions and assessing the results obtained are very important steps for enhancing the efficiency of this type of process, since they allow the improvement of ex-ante benefit estimations, and consequently, better future investment decisions. This is espe- cially relevant when the applied safety measures are transferred from other countries’ traffic systems, due to the different response of drivers. This paper allows to increase the knowledge on the effects of safety interventions specifically applied in the Portuguese road environment, which can be different from other European countries due to particular driver behaviour. 0001-4575/$ see front matter © 2012 Elsevier Ltd. All rights reserved. doi:10.1016/j.aap.2012.02.004

Upload: joao-lourenco

Post on 23-Nov-2016

214 views

Category:

Documents


1 download

TRANSCRIPT

Page 1: Safety effects of low-cost engineering measures. An observational study in a Portuguese multilane road

SP

SL

a

ARRA

KBMLR

1

estls(

ar

Disa

kmb

(

0d

Accident Analysis and Prevention 48 (2012) 346– 352

Contents lists available at SciVerse ScienceDirect

Accident Analysis and Prevention

jo ur n al hom ep a ge: www.elsev ier .com/ locate /aap

afety effects of low-cost engineering measures. An observational study in aortuguese multilane road

andra Vieira Gomes ∗, João Lourenc o Cardosoaboratório Nacional de Engenharia Civil, Department of Transportation, Av. Brasil, 101, 1700-066 Lisbon, Portugal

r t i c l e i n f o

rticle history:eceived 23 December 2009eceived in revised form 2 January 2012ccepted 7 February 2012

eywords:efore–after study

a b s t r a c t

Single carriageway multilane roads are not, in general, a very safe type of road, mainly because of thehigh number of seriously injured victims in head-on collisions, when compared with dual carriagewaymultilane roads, with a median barrier.

In this paper the results of a study on the effect of the application of several low cost engineeringmeasures, aimed at road infrastructure correction and road safety improvement on a multilane road(EN6), are presented. The study was developed by the National Laboratory of Civil Engineering (LNEC) for

edianow-cost engineering measuresoad safety

the Portuguese Road Administration and involved a comparison of selected aspects of motorized trafficbehaviour (traffic volumes and speeds) measured in several sections of EN6, as well as monitoring of roadsafety developments in the same road. The applied low cost engineering measures allowed a reductionof 10% in the expected annual number of personal injury accidents and a 70% decrease in the expectedannual number of head-on collisions; the expected annual frequency of accidents involving killed andseriously injured persons was reduced by 26%.

. Introduction

The case study described in this paper refers to the analysis of theffect of several road safety measures undertaken on a Portugueseingle carriageway multilane road (EN 6), following the construc-ion of a new motorway (A5), parallel to the existing road, whichinks directly Lisboa to Cascais. Several low cost engineering mea-ures were implemented thereafter, on a 3.7 km long stretch of EN6between 3.2 km and 6.9 km) to improve its safety record.

The evaluation was carried through a “before–after” study with control group. A stretch of another single carriageway multilaneoad (the EN125 in Algarve) was used as control group.

According to the Organization for Economic Co-operation andevelopment (OECD), integrated programmes on road safety aim-

ng to reduce accident severity should focus on several trafficystem elements, namely by adaptation of road environment char-cteristics to the intended road uses (OECD, 1984).

The application of road safety measures is based on generalnowledge on the road safety phenomenon, sometimes comple-

ented by results from the analysis of accident data in the area to

e corrected.

∗ Corresponding author. Tel.: +351 21 8443528; fax: +351 21 8443029.E-mail addresses: [email protected] (S. Vieira Gomes), [email protected]

J.L. Cardoso).

001-4575/$ – see front matter © 2012 Elsevier Ltd. All rights reserved.oi:10.1016/j.aap.2012.02.004

© 2012 Elsevier Ltd. All rights reserved.

Better results may be achieved with road safety interventionsif procedures for their implementation include the following steps(Cardoso and Bairrão, 2001):

- Network safety diagnosis to identify locations with high influenceof road characteristics on accident occurrence.

- Safety analysis on each selected location for potential interven-tions, to identify the main road characteristics leading to safetyissues.

- Selection and application of appropriate safety measures.- Safety monitoring of the corrected sites and evaluation of results.

The decision process for selecting corrective measures takes intoaccount the available budget for the programme, the safety issuesidentified by the diagnosis and the anticipated benefits from thepotential suitable measures. Monitoring safety developments fol-lowing interventions and assessing the results obtained are veryimportant steps for enhancing the efficiency of this type of process,since they allow the improvement of ex-ante benefit estimations,and consequently, better future investment decisions. This is espe-cially relevant when the applied safety measures are transferredfrom other countries’ traffic systems, due to the different response

of drivers. This paper allows to increase the knowledge on theeffects of safety interventions specifically applied in the Portugueseroad environment, which can be different from other Europeancountries due to particular driver behaviour.
Page 2: Safety effects of low-cost engineering measures. An observational study in a Portuguese multilane road

S. Vieira Gomes, J.L. Cardoso / Accident Analysis and Prevention 48 (2012) 346– 352 347

Table 1Percentage change in the number of accidents expected as a result of several road safety measures.

Accident severity Percentage change in the number of accidents

Types of accident affected Best estimate 95% confidence interval

Median on multilane roads – rural areasa

Injury accidents All accidents −12 (−15;−8)Property damage only accidents All accidents −18 (−21;−14)

Median on multilane roads – urban areasa

Injury accidents All accidents −22 (−24;−20)Property damage only accidents All accidents 9 (+7;+11)

Small improvements of road alignmentb

Unspecified Unspecified −20 (−39; −4)Speed actuated traffic signals – minor intersections in rural areasc

Injury accidents Unspecified −11 UnspecifiedSpeed actuated traffic signals – bendsc

Injury accidents Unspecified 15 Unspecified

mftmmiaa2

aawtt

dswotbcvat

2

st“otscoaw

lsct

a Elvik and Vaa (2004).b Gitelman et al. (2001).c Barker and Baguley (2001).

From a safety point of view, road infrastructure and environ-ent interventions are especially important because they have

requently allowed casualty reductions of about 30% at interven-ioned places. Speed reduction is a major concern of all network

anagers. Whether in urban or rural areas, broadly, several specificeasures may be applied. Table 1 presents the percentage change

n the number of accidents as a result of some road safety measuresiming at reducing speed and the severity of speed related crashes,s reported internationally (Elvik and Vaa, 2004; Gitelman et al.,001; Barker and Baguley, 2001).

Low cost engineering measures (LCEM) are corrective measurespplied to the road environment or to traffic management, char-cterized by small investments and short implementation periods,hen compared with the ones for other types of road works. Fur-

hermore, LCEM can be applied within existing road limits, in bothhe construction and the operation phases.

LCEM are selected taking into account the main types of acci-ents in the site where they are applied, whose frequency oreverity they are intended to decrease; LCEM selection criteria alsoarrant that their application will not have undesirable effects

n other accident types, on traffic operation efficiency nor willhey raise serious negative environmental issues. These facts, com-ined with the low cost and quick results, justify the favourableost/benefit relations, even for those road sites where major inter-entions are planned for in the short term (3–5 years). Indeed,ccording to international experience, very good cost/benefit rela-ions may be obtained with LCEM (Gorell and Tootill, 2001).

. Safety analysis of corrective measures

The safety assessment of a road infrastructure corrective mea-ure deals with the estimation of the effect of the intervention onhe number of accidents or victims. This may be made through abefore–after” study, that includes the following steps: estimationf the safety level of a road section in a period of time previous tohe application of the corrective measure; estimation of the actualafety level in a period of time subsequent to the application of theorrective measure, and also the expected safety level that wouldccur if the measure had not been applied; the application of anlgorithm for comparing the actual development with the one thatould have occurred without intervention (Hauer, 1997).

Usually, developments in the safety level of a road section fol-

owing the application of a set of corrective measures do not reflectolely their effect. Other factors may influence the way safetyhanges over time, namely the regression-to-the-mean (RTM), longerm trends in the number of accidents and victims, changes in the

percentage of registered accidents by the responsible authoritiesand traffic volume developments (Hauer, 1997).

The RTM effect consists of a statistical trend: after a period withfrequencies very far from the average value, it is normal to observefrequencies closer to this value. Thus, after a period with very highaccident frequencies, the normal trend is that in subsequent peri-ods, lower accident frequencies will be observed. In most roadsafety interventions this is basically due to the way road author-ities select sites for improvement: interventions are implementedwhere higher accident occurrence has been observed. One way tomitigate RTM would be to randomly select the intervention sites,as described in Hutchinson (2007). This not being possible in allcases, one may use estimators developed through Bayesian infer-ence methods, where the expected number of accidents for onespecific time period and section under study is calculated multi-plying the observed number of accidents by a factor that takes intoaccount the average and the variance of the number of accidentsobserved in previous time periods (Abbess et al., 1981).

Ideally the effect due to the remaining disturbing factors couldbe estimated in two ways: using statistically derived models of theirinfluence on accident frequency, assuming all the disturbing vari-ables and their effects are known, which is seldom the case; or usingcontrol groups, assuming that the effect of the disturbing variableswill be identical in both the treated sites and the control group sites.In practice, usually control groups are used.

To control for the RTM effect the Empirical Bayes Method (EBM)was used in this study. The EBM was originally developed for thepurpose of controlling for regression-to-the-mean in before-and-after studies evaluating the effects of road safety measures (Haueret al., 2002). It states that E(m|x) – the expected number of accidentsat a site where x accidents were observed – can be estimated by:

E(m|x) = m = · E(m) + (1 − ˛) × x (1)

where x – observed number of accidents; E(m) – expected numberof accidents in a similar site, calculated using a safety performancefunction; and

˛ = E(m)E(m) + Var(m)

; Var(m) ≈ (E(m))2

k;

VAR(m) – variance of the calculated expected number of accidentsin a similar site; k – dispersion parameter of the safety performancefunction.

In this study, possible migratory effects in accident occurrence

were not considered, since the road used for comparison is notlinked to the road where the engineering measures were applied,and road sections adjacent to the study site were also subject tointervention. According to available evidence there were no major
Page 3: Safety effects of low-cost engineering measures. An observational study in a Portuguese multilane road

3 Analy

cp

ec

mdtwp

E

wo

e

3

ccssaiag7i

3

almi1nrrdcvscihwalmt

nmHs

3

pl

48 S. Vieira Gomes, J.L. Cardoso / Accident

hanges in the accident registering procedures followed by theolice forces.

Time trends associated with accident occurrence were consid-red while choosing the control zone roads, in order to achieve theorrect resemblance between roads.

The impacts on accident frequencies of traffic volume develop-ents in similar sites were taken into account using previously

eveloped safety performance functions. Earlier studies showedhat accident frequencies on multilane roads do not vary linearlyith average annual daily traffic changes: they are related to aower of 1.102 (Cardoso, 2007):

= 2.7493 × 10−4 × Ext × AADT1.102 (2)

here E = annual number of accidents in a road section; Ext = lengthf the road section (km); AADT = average annual daily traffic.

The dispersion parameter of this safety performance function isqual to 2.538 (Cardoso, 2007).

. Case study on Portuguese roads

Specific studies were carried out in Portugal to evaluate the effi-iency of engineering safety measures, since the practice in otherountries may not fit to the Portuguese traffic system. One of thesetudies has included an analysis on the efficiency of several mea-ures undertaken on a Portuguese road (EN 6): the application of

narrow median, a new pavement layer, changes on speed lim-ts and new speed-actuated traffic signals (Cardoso, 1994; Cardosond Gomes, 2005). At the time of the intervention, EN 6 was a dan-erous road, with a yearly average of 195 injury accidents, of which6 involving fatal or serious injuries and 45 frontal collisions, along

ts 18.5 km.

.1. EN6 characteristics

The EN 6 road had an almost constant transversal profile, with 12 m wide carriageway divided into four lanes, no median andateral sidewalks. Besides the usual vertical signs and horizontal

arkings, there were traffic signals at eight major intersections. Anmportant work of requalification was implemented in this road in993 between 3.2 km and 6.9 km. Several changes were introduced:ew central curbed median (0.65 m wide); top layer removal andepaving (with an antiskid bituminous material in selected bends);econstruction of the wall that protects the road from the sea; intro-uction of widening and superelevation in some horizontal curves;hanges in the safety equipment, including a complete review ofertical signs, considering the urban character of the area, lowerpeed limits and application of speed actuated signals for speedontrol, some associated with specific phases for pedestrian cross-ng (see Fig. 1). The elevated median was built by means of 0.28 migh limestone curbs. At the time of construction this solutionas criticized, the argument being that it would not conveniently

ddress the frontal collision problem, due to lack of containmentevel. The proposed concrete barrier was discarded, however, to

aintain the urban appearance of the road and its scenic integra-ion.

After this intervention, others followed, more specific, that wereot considered in the present case study due to absence of infor-ation regarding their location and exact time of application.owever, these changes do not affect the conclusions of this study,

ince they do not overlap in space with the analysed ones.

.2. Choice of the comparison group

According to Hauer (1997), a suitable comparison group shouldresent physical similarity with the road section that is being ana-

ysed and also a similar safety development in time. This last aspect

sis and Prevention 48 (2012) 346– 352

can be verified through a statistical estimator (ô) relating the num-ber of accident occurrences in the treated zones with the ones inthe comparison group in consecutive years; it is calculated usingthe following equation (Hauer, 1997):

o = (K · N)/(L · M)1 + (1/L) + (1/M)

(3)

where K – number of accidents in the treated road section in thebefore period; N – number of accidents in the comparison roadsection in the after period; L – number of accidents in the treatedroad section in the after period; M – number of accidents in thecomparison road section in the before period.

For the case study, three candidate comparison groups wereevaluated: two road sections in EN 6 (one before the treated site,from 0.0 km to 3.2 km, and the other after the treated site, from6.9 km to 18.5 km) and a multilane road section of EN125 locatedin the South of Portugal, near Faro, with similar geometric charac-teristics to the ones in EN6 before the intervention. Resulting valuesfor the estimator are presented in Table 2.

The first road section of EN6 presents an average value for the “ô”estimator considerably different from unity, meaning that it is not agood comparison group. This was expected, since this road stretchis almost straight (and, in this aspect, considerably different fromthe interventioned curvy road section); additionally several speedactuated signals were installed in this road stretch shortly beforethe intervention in the analysed stretch of EN 6. From the remainingpossible road sections, the most adjusted, according to the criteriadefined by Hauer (1997), would be the road section of EN6 between6.9 km and 18.5 km, since it presents the “ô” value closer to the unit.This was already expectable, since this road section presents verysimilar geometric characteristics and traffic volumes to the onesin the interventioned road section. However, this road section ofEN 6 was not considered for comparison, as no suitable data wereavailable for the after period. In fact, interventions were made inparts of this road stretch, for which no data register and descriptionwere possible to gather. Thus, the road section from EN125 wasconsidered as the only suitable comparison group.

3.3. Developments in traffic characteristics and behaviour

Average Annual Daily Traffic (AADT) of EN6 and EN125 for the“before” and “after” periods were obtained through several traf-fic counting stations from the National Road Administration (JAE,1985, 1990, 1995, 1996, 1997, 1998, 1999, 2000; Macedo et al.,2003). Reference must be made to the fact that both roads wereaffected by the construction in their vicinity of new roads witha higher network hierarchy, inducing traffic redistribution (seeFig. 2). In the case of EN 6, high speed drivers were diverted toa new motorway, both in interventioned and non-interventionedstretches.

Previously to the intervention, LNEC had participated in a safetystudy in EN6, which included an evaluation of the traffic behaviour(Cardoso and Castilho, 1989). Traffic volumes and average speedswere calculated through an adaptation of the method of the movingobserver. After the intervention, new measurements were made,which allowed a comparison of selected speed distribution param-eters, namely the average and the standard deviation.

A reduction in the average speed was observed in all analysedroad sections, except for a section at 13.5 km, where no median andno speed actuated traffic signals were installed (see Table 3). The

standard deviation of speed distributions are considerably smallerin the after measurements, when compared to the ones from thebefore situation; this may be explained by the use of different mea-surement methods and by the lower speeds in the after period.
Page 4: Safety effects of low-cost engineering measures. An observational study in a Portuguese multilane road

S. Vieira Gomes, J.L. Cardoso / Accident Analysis and Prevention 48 (2012) 346– 352 349

Fig. 1. View of EN 6 road, undivided (right); and with curbed median, near a speed-actuated traffic signal (left).

Table 2Selection of control group roads: calculated “ô” estimators.

Road group Road section 1987 1988 1989 1990 1991 Average

Corrected EN6 (3.2–6.9 km) Accidents 68 65 59 60 67 –

Candidate control

EN6 (0.0–3.2 km)Accidents 27 11 7 3 4 –ô – 2.50 1.50 2.27 0.49 1.69

EN6 (6.9–18.5 km)Accidents 122 123 116 123 121 –

3

c

ô

EN125 (97–103.4 km and107.4–113.85 km)

Accidents

ô

.4. Safety developments

Safety developments were analysed using three types of indi-ators: the number of expected injury accidents; the number of

57435

6000 0

445 92

5000 0

T

4081 5

3575 9

4000 0ADDT

3168 8

2583 7

3000 0

2000 0

198 5 198 6 198 7 198 8 1989 1990 1991 1992 19

Years

km 7,8 km,

Fig. 2. Impact of A5 opening in AADT development

– 1.02 1.15 0.90 0.89 0.9996 94 107 104 110 –

– 1.06 0.94 0.98 0.83 0.95

accidents involving fatal and serious injuries (KSI accidents); and

the number of head-on collisions, as the engineering measureswere intended to reduce this type of accidents and considerabledoubts on the effectiveness of the selected type of median had been

4737 6

411 36

3706 3

3498 9

4319 1

310 22

284 43

93 19 94 19 95 19 96 19 97 199 8 199 9 200 0 200 1

16,6 km 2,5, ,

on three EN6 counting sections (1985–2001).

Page 5: Safety effects of low-cost engineering measures. An observational study in a Portuguese multilane road

350 S. Vieira Gomes, J.L. Cardoso / Accident Analysis and Prevention 48 (2012) 346– 352

Table 3Traffic speeds in selected road sections of EN 6 (traffic volumes below 1400 vehicles per hour–A to C levels of service).

1988 2003

Traffic volume Speed Traffic volume Speed

3900 km Average 1197 73 719 48Standard deviation 193 11 523 8

6800 km Average 1197 77 888 63Standard deviation 193 4 508 5

7500 km Average 1059 72 790 69Standard deviation 222 9 532 5

13,500 km Average 1179 60 946 66Standard deviation 205 14 338 4

rvsbt6c

iesrdweoeabdt(aca

asoan

TOr

15,300 km Average 1179Standard deviation 205

aised. The analysis covered a period of 5 years “before” the inter-ention; and a 4 years period “after” the implementation of theafety measures. Two different periods were used in the analysisefore and after due to the fact that in 5 years after the interven-ion, several engineering measures were applied on both the EN

road and the control road (EN 125), significantly changing theirharacteristics.

Table 4 presents the observed frequency of accidents occurredn EN6, the values for the variance and the parameter, and thexpected numbers of injury accidents, of accidents with fatal orerious injuries, and of head-on collisions on the interventionedoad section of EN 6 (3.2–6.9 km). The model developed by Car-oso only applies to injury accidents; therefore, Eqs. (1) and (2)ere only applied to this type of accidents. Since no similar mod-

ls exist for KSI accidents and head-on collisions, the numbersf these accidents were calculated indirectly from the number ofxpected injury accidents, according to the corresponding percent-ge of the total observed accident frequencies. In EN6, during theefore period, head-on collisions accounted for 28% and KSI acci-ents accounted for 38% of total injury accidents; in the after period,he percentages were 4.3 and 18, respectively. In the control groupEN125), no significant differences were noted between the beforend the after period. Therefore, it was decided to use only one per-entage value for each accident type: 11% for head-on collisions,nd 36% for KSI accidents.

The expected frequency of injury accidents, head-on collisionsnd KSI accidents in the comparison group – EN 125 – are pre-

ented in Table 5. The variation of the expected annual frequencyf injury accidents as a function of AADT in EN 125 (97.0–103.4 kmnd 107.4–113.9 km) is presented in Table 6. This correction waseeded since the AADT from both roads was significantly different,

able 4bserved and expected annual frequencies of injury accidents, head-on collisions and K

oad section.

1987 1988 1989 1990 1991

Injury accidentsObserved frequency 57 62 71 69 59Var 2879.9 3323.4 3835.3 4426.1 5107˛ 0.029 0.027 0.025 0.023 0Expected frequency 58 63 72 70 60

Head on collisionsObserved frequency 16 17 24 17 16Var 230.7 266.2 307.2 354.5 409˛ 0.095 0.089 0.083 0.078 0Expected frequency 17 18 24 18 17

Fatal and seriously injury accidentsObserved frequency 18 27 31 24 21Var 417.0 481.2 555.3 640.8 739˛ 0.072 0.070 0.063 0.059 0Expected frequency 19 28 31 25 22

67 923 5711 366 2

which would lead to a biased accident estimation. These resultingvariations are considered to be the ones that would have occurredin EN6 if no safety intervention had been applied.

The variations in the expected annual frequency of injury acci-dents, of KSI accidents, and of head-on collisions as a function AADTin EN6 (3.2–6.9 km) are presented in Table 7. In order to calcu-late these values, one had to obtain the expected frequencies ofaccidents for the after period without measures applied. This wasmade using the percentages obtained from the analysis of EN125(Table 6), and correcting these for the traffic development in EN6(different from the one registered at EN125).

Comparing the expected changes with and without the mea-sures applied it was possible to verify that the corrected EN 6 roadstretch experienced a greater reduction in the number of accidentsthan the EN 125 control group road stretches; this greater reductionis especially evident as regards the number of head-on collisions.

It is possible to calculate the effect of the applied correc-tive measures, subtracting the percentage variation measured onthe comparison group road section from the percentage reduc-tion obtained in the tested road section. Benefits were obtained,especially as regards head-on collisions, for which a 50% accidentreduction was achieved (−10 head-on collisions). A 16% reduc-tion in injury accidents and a 36% decrease in KSI accidents wereobserved.

Changes in traffic speeds account for part of the reductionin the expected number of injury and serious injury accidents.Some of theses changes were the direct result of modifications in

speed limits, of reductions in lane widths and of the installationof speed actuated traffic signals; however, the opening of the newmotorway A5 in the same corridor, introduced important changesin traffic characteristics as well, since fast drivers tend to select

SI accidents, and correspondent variance and parameter for the interventioned

1992 1993 1994 1995 1996 1997

31 27 25 34.8 2295.3 2672.5 2672.5 2672.5.023 0.032 0.030 0.030 0.030

32 29 27 35

0 1 1 3.1 4.2 4.9 4.9 4.9.073 Construction 0.438 0.419 0.419 0.419

1 2 2 3

6 7 3 5.5 73.9 86.1 86.1 86.1.055 0.156 0.147 0.147 0.147

7 8 5 6

Page 6: Safety effects of low-cost engineering measures. An observational study in a Portuguese multilane road

S. Vieira Gomes, J.L. Cardoso / Accident Analysis and Prevention 48 (2012) 346– 352 351

Table 5Observed and expected annual frequencies of injury accidents, head-on collisions and KSI accidents, and correspondent variance and parameter for the comparison grouproad sections.

1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997

Injury accidentsObserved frequency 109 112 98 101 95 70 93 75 85Update to AADT* 138 156 176 198 224 178 194 210 228Var 7476.7 9528.6 12,145.3 15,482.6 19,739.4 12,540.2 14,769.2 17,400.6 20,508.3˛ 0.018 0.016 0.014 0.013 0.011 0.014 0.013 0.012 0.011Expected frequency 109 112 98 101 95 70 93 75 85

Head-on collisionsObserved frequency 15 13 11 8 10 7 11 9 14Update to AADT* 15 17 19 21 24 19 21 23 25Var 86.8 110.6 141.0 179.8 229.2 Construction 145.6 171.5 202.0 238.1˛ 0.146 0.132 0.118 0.106 0.095 0.117 0.109 0.101 0.093Expected frequency 15 13 11 8 10 7 11 9 14

Fatal and serious injury accidentsObserved frequency 46 38 46 29 35 36 20 25 31Update to AADT* 49 56 63 71 80 64 69 75 81Var 953.7 1215.4 1549.1 1974.8 2517.8 1599.5 1883.8 2219.5 2615.9˛ 0.049 0.044 0.039 0.035 0.031 0.038 0.035 0.033 0.030Expected frequency 46 38 46 29 35 36 20 25 31

* Expected accident frequency updated to the AADT changes by the application of Cardoso’s model.

Table 6Change in the expected annual frequencies of injury accidents in the comparison group road sections, adjusted for AADT development.

Average 1987–1991 Average 1994–1997 Expected change

Value Percentage

teeswo

eafnmddums

TE

Injury accidents 103

Head-on collisions 11

Fatal and serious injury accidents 39

ravelling in the new motorway. If the speed effect could bextracted from previous results, it would be possible to estimate theffect of the applied corrective measures that do not directly affectpeed choice, namely the installation of the curbed median, theidening and superelevation of horizontal curves, and the renewal

f vertical signing.Nilsson (2004) developed in Sweden the power model, which

stablishes a relationship between traffic speed and safety. Hisnalysis showed that the power model is valid for injury accidents,atal accidents and the number of injuries, but underestimates theumber of fatalities; therefore, an attempt was made to apply theseodels to calculate the change in expected frequency of injury acci-

ents, head on collisions and fatal and seriously wounded accidents

ue exclusively to differences in registered speeds. With these val-es it was possible to estimate the effect of the curbed centraledian and other non-speed related safety measures applied. Nils-

on’s model was applied to the corrected EN 6 road only, as there

able 7xpected change associated with the measures applied in EN6 and intermediate values.

Average 1994–1997

With correctivemeasures andspeed change

Without corrective measures

Adjusted fortrafficdevelopment inEN6 – with speedchange

AdjustedtrafficdevelopmEN6 – wispeed cha

Injury accidents 31 41 35

Head-on collisions 2 12 10

Fatal and serious injuryaccidents

7 16 11

81 −22 −2210 −1 −928 −11 −29

were no speed measurements on the EN 125 control group roadsections. The obtained results represent the accidents that wouldhave occurred if the speed and traffic had not changed (43 injuryaccidents, 13 head-on collisions and 14 fatal and seriously woundedaccidents).

In order to compare these values with the ones presented inTable 7 concerning the expected accident frequency with all themeasures applied and without measures applied, it was necessaryto adjust them to the traffic reduction that was experienced in EN6.The values resulting are 35 injury accidents, 10 head on collisionsand 11 fatal and seriously wounded accidents.

The expected effects of the application of a new central curbedmedian and other safety measures not related to speed were calcu-

lated subtracting the power model values from the “after” values.The results, a 10% decrease in injury accidents and a 26% reduc-tion in fatal and serious injury accidents, are also presented inTable 7. Assuming that the power model may be applied also to each

Expected changewith measuresapplied and speedchange

Expected changeassociated with speedchange

Expected changeassociated with themeasures applied

for

ent inthoutnge

Value % Value % Value %

−10 −24 −6 −15 −4 −10−10 −83 −2 −14 −8 −70−9 −56 −5 −30 −4 −26

Page 7: Safety effects of low-cost engineering measures. An observational study in a Portuguese multilane road

3 Analy

a(

msa

4

s6s

htvci

‘dssnhdadeaostmeia

LLPtsvos

Seguranc a Rodoviária. Laboratório Nacional de Engenharia Civil, Lisboa.Nilsson, G., 2004. Traffic Safety Dimensions and the Power Model to Describe the

Effect of Speed on Safety. Lund Institute of Technology and Society, Traffic Engi-

52 S. Vieira Gomes, J.L. Cardoso / Accident

ccident type, the effect on head-on collisions was also calculated−70%).

An additional analysis was made with the revised power modelade by Rune Elvik (Elvik et al., 2004). For the accident types con-

idered, only the expression for the fatal and seriously woundedccidents changed, leading to a similar reduction value (−30%).

. Conclusion

The application of corrective measures (curbed median andpeed actuated traffic signals) on EN 6 road between 3.2 km and.9 km, affected drivers’ speed choice and the safety level of thisingle carriageway multilane road stretch.

Following the opening of motorway A5, traffic volumes on EN 6ave diminished considerably, reducing the number of hours withraffic close to full capacity (with congestion in peak hours and pre-ailing low levels of service in the rest of the day). This generated theonditions for an overall increase in traffic speed and a consequentncrease in accident risk.

However, the analysis of driving speeds in the ‘before’ andafter’ periods showed that the applied measures originated a 20%ecrease in the average speed (from 77 to 63 km/h). As a result of theafety analysis described, the effects of the applied corrective mea-ures may be summarized as follows: 10% reduction in the annualumber of injury accidents; 70% decrease in the annual number ofead-on collisions; and 26% reduction in the annual number of acci-ents involving fatal and serious injuries. These numbers alreadyccount for RTM, AADT decrease and national road network safetyevelopment trends. The narrow high curb medians proved to beffective in preventing head-on collisions, provided traffic speedsre low (70 km/h or less) and narrow lanes are provided (3.00 mr less). Narrow lanes have the added advantage of promoting lowpeed choice and impeding uncontrolled vehicles form impactinghe medians with high angles. Thus, the high curb low contain-

ent median, combined with the remaining corrective measures,nsured both the preservation of the urban and scenic character-stics of the road environment, and the reduction in number ofccidents and their victims.

Systematic rational application, with technical justification, ofCEM allows a fast and significant safety improvement. SeveralCEM mentioned in this paper are currently being applied inortugal, although not always mainly for safety reasons and unfor-unately not on a methodical basis. In this context, the practice of

ystematic evaluation of safety results obtained with those inter-entions is particularly important to assess the accuracy of claimsn their cost-effective contribution to the national goal of improvedafety.

sis and Prevention 48 (2012) 346– 352

References

Abbess, C., Jarrett, D., Wright, C.C., 1981. Accidents at black spots: estimating theeffectiveness of remedial treatment, with special reference to the regression-to-the-mean effect. Traffic Engineering and Control 22 (10), 535–542.

Barker, J., Baguley, C., 2001. Road Safety Good Practice Guide. Department of theEnvironment, Transport and the Regions, London, United Kingdom.

Cardoso, J.L., Castilho, A.J., 1989. Colaborac ão no estudo da seguranc a da EN 6 –Caracterizac ão da sinistralidade e da circulac ão do tráfego rodoviário. Report n.192/89 of Núcleo de Tráfego e Seguranc a Rodoviária. Laboratório Nacional deEngenharia Civil, Lisboa.

Cardoso, J.L., 1994. Cardoso, Sinistralidade na Rede Rodoviária Nacional no ano de1990 – Report n. 152/94 of Núcleo de Tráfego e Seguranc a Rodoviária. Labo-ratório Nacional de Engenharia Civil, Lisboa.

Cardoso, J.L., Bairrão, L., 2001. Auditorias de seguranc a rodoviária ao projecto deestradas. Report n. 91/01 of Núcleo de Tráfego e Seguranc a Rodoviária. Labo-ratório Nacional de Engenharia Civil, Lisboa.

Cardoso, J.L., Gomes, S.V., 2005. Avaliac ão do impacte sobre a sinistralidade de medi-das correctivas em trecho da EN6. Report n. 217/05 of Núcleo de Tráfego eSeguranc a Rodoviária. Laboratório Nacional de Engenharia Civil, Lisboa.

Cardoso, J.L., 2007. Detecc ão de zonas de acumulac ão de acidentes na RedeRodoviária Nacional. Report n. 88/2007 of Núcleo de Tráfego e Seguranc aRodoviária. Laboratório Nacional de Engenharia Civil, Lisboa.

Elvik, R., Vaa, T., 2004. The Handbook of Road Safety Measures. Elsevier.Elvik, R., Christensen, P., Amundsen, A., 2004. Speed and road accidents. An eval-

uation of the Power Model. TOI Report 740. Institute of Transport Economics,Oslo.

Gitelman, V., Hakkert, A.S., Doveh, E., Cohen, A.,2001. A study of safety effects ofroad infrastructure improvements under Israeli conditions. In: Proc., Int. Conf.Traffic Safety on Three Continents, Moscow (CD-Rom). Swedish National Roadand Transport Research Institute, Linkoping, Sweden.

Gorell, R.S.J., Tootill, W., 2001. Monitoring local authority road safety schemes usingMOLASSES. TRL Report TRL512.

Hauer, E., 1997. Observational Before–After Studies in Road Safety. Pergamon,Oxford, UK.

Hauer, E., Harwood, D.W., Council, F.M., Griffith, M.S., 2002. Estimating safety by theempirical Bayes method. A tutorial, Transportation Research Record, vol. 1784.Transportation Research Board, National Research Council, Washington, DC, pp.126–131.

Hutchinson, T.P., 2007. Concerns about methodology used in real-world experi-ments on transport and transport safety. Journal of Transportation Engineering133 (1), 30–38 (ASCE).

JAE, 1985. Junta Autónoma de Estradas, Tráfego.JAE, 1990. Junta Autónoma de Estradas, Tráfego.JAE, 1995. Junta Autónoma de Estradas, Tráfego.JAE, 1996. Junta Autónoma de Estradas, Tráfego.JAE, 1997. Junta Autónoma de Estradas, Tráfego.JAE, 1998. Junta Autónoma de Estradas, Tráfego.JAE, 1999. Junta Autónoma de Estradas, Tráfego.JAE, 2000. Junta Autónoma de Estradas, Tráfego.Macedo, A.L., Cardoso, J.L., Cavalheiro, F.M., Marques, J., 2003. Apoio ao IEP na

realizac ão do recenseamento de tráfego na Rede Rodoviária Nacional (recensea-mento de 2001) – Final Report. Vol. II. Report n. 182/03 of Núcleo de Tráfego e

neering.OECD, 1984. Road Transport Research, Integrated Road Safety Programs. OECD, Paris.