Black Spot Manual

GENERAL DIRECTORATE OF HIGHWAYS
ROAD IMPROVEMENT AND TRAFFIC SAFETY PROJECT
TRAFFIC SAFETY PROJECT
BLACK SPOT MANUAL
December 2001
SweRoad
ANKARA
TRAFFIC SAFETY PROJECT Traffic Safety Consultancy Services
Foreword
One of the most cost-effective road safety interventions is to eliminate so-called black spots, that is, to remedy accident-prone locations along the roads. This includes the following steps: identify the black spots, study the problems (diagnosis) at each spot, design suitable countermeasures, estimate their effects, set priorities, implement, and finally, follow up and evaluate the results. This Black Spot Manual includes all these steps. The Manual (in an earlier version) has been sent for comments to KGM. The comments have been considered in this version. It must be observed that the Manual has to be improved over time. For example, the estimated reduction factors need to be checked for Turkish conditions and the monetary values for accident and casualty reductions have to be improved. To be able to do this, it is necessary to start a systematic work to follow up the effects of different countermeasures and to compile material about different accident cost components. The main authors of this report are Mr. Kent Sjlinder and Mr. Hans Ek, SweRoads specialists on accident analysis and black spots. Ankara, December 2001
Karl-Olov Hedman Team leader
Black Spot Manual
1/81
December 2001
SweRoad
ANKARA
Contents
Foreword 1 2 Introduction Identification of black spots 2.1 Background 2.2 Used method 2.3 Using more homogeneous groups 2.4 Road safety goals and targets 2.5 Accident rate 2.6 Accident frequency 2.7 Severity index 2.8 All three criteria do not need to be satisfied 2.9 Random variation 2.10 Choice of confidence level 2.11 Using more than one-year data 2.12 Identification without damage only accidents 2.13 Calculation for road number 100 using the proposed method 2.14 Suggestions in brief 2.15 References Diagnosis 3.1 Criteria in the identification process 3.2 Location 3.3 Yes, there is a local accident concentration 3.4 Stick diagram analysis 3.5 Example from section 100 -14, kilometer 6 3.6 Diagnosis without damage only accidents 3.7 Site investigations Finding countermeasures 4.1 Introduction 4.2 Road sections 4.3 Junctions Estimating the effects of countermeasures 5.1 Introduction 5.2 Summary of proposed reduction factors 5.3 Road sections 5.4 Junctions 5.5 Improvements for pedestrians and bicyclists 5.6 More than one countermeasure Prioritizing 6.1 Introduction 6.2 Appraisal methods
2/81
Page
1 4 5 5 5 5 5 6 7 8 11 11 11 12 13 13 17 17 18 18 19 20 23 23 26 26 27 27 27 30 34 34 35 38 51 60 63 64 64 64
December 2001
Black Spot Manual
SweRoad
ANKARA 6.3 7 8 Proposed procedure for KGM
TRAFFIC SAFETY PROJECT Traffic Safety Consultancy Services 68 72 73 73 73 73 74 75 75 78 79 79 80 80 81 81
Implementation Follow-up and evaluation 8.1 Background 8.2 Planning of follow-up 8.3 Documentation of countermeasures 8.4 Target/result-oriented way of planning 8.5 Initial monitoring 8.6 Long-term evaluation 8.7 To estimate the effect 8.8 Short-term versus long-term effects 8.9 Regression-to-the-mean 8.10 Accident migration 8.11 Strange results 8.12 Change in under-coverage of accidents 8.13 Before and after periods for accident data
Black Spot Manual
3/81
December 2001
SweRoad
ANKARA
Introduction
The process of eliminating or improving accident black spots in a road network is composed of several activities, as illustrated in the following figure.
Identification of black spots Diagnosis Finding countermeasures Estimating effects
Prioritizing
Implementation Follow-up and evaluation
Identification of black spots is the procedure to locate those spots in the road network that are particularly dangerous, that is, the black spots. Diagnosis is the process to study what are the problems, the accident contributing factors and the deficiencies for each of the identified black spots. Finding countermeasures implies a methodical analysis to design suitable countermeasures for each black spot, based on actual problems and deficiencies. Estimating effects is the process to estimate the safety effects (and if necessary also other effects) and costs of suitable countermeasures. Prioritizing implies finding the best action plan (or investment program), according to some defined criteria, and based on estimated effects and costs as well as budget restrictions. Implementation is the actual realization of the prioritized measures included in the action plan (or investment program). Follow-up and evaluation is the last and very important step, which aim is to assess the actual results (effects and costs).
Black Spot Manual 4/81 December 2001
SweRoad
ANKARA
2
2.1
Identification of black spots

Background
This chapter deals with the procedure for identification of hazardous locations or black spots as they are often called. The procedure described is based on recorded accidents, data about accidents, traffic volumes and vehicle-kilometers. Other methods that can be used as compliments to accident data are not dealt with in this chapter. Examples of such methods are field investigations, conflict studies, questionnaires and interviews, etc. Identification is a first step in improving road safety at a black spot. It has to be followed by diagnosis of the selected spots, finding countermeasures, estimating effects and costs, prioritizing, implementation and at last follow-up and evaluation. These latter stages are discussed in the following chapters. In this chapter, the identification method used by KGM is scrutinized and some improvements are suggested.
2.2
Used method
The method used by KGM is called Rate Quality Control Method. It is a statistical method for identifying black spots. A statistician at the Swedish National Road and Transport Research Institute (Mats Wiklund) has scrutinized the method. The theory part of this paper is based partly on his comments. The Rate Quality Control Method consists of calculating three different parameters for each road section. In Turkey, a road section is defined as one kilometer of road. The three parameters are:

accident rate, accident frequency, severity index.
Each of these values is compared with a critical value. Thus the accident rate is compared with one critical value, the accident frequency with another critical value and the severity value with a third critical value. If a certain road section shows higher values than the critical ones for all these three parameters, the section is considered to be a black spot.
2.3
Using more homogeneous groups
The method would be better if junctions were separated from road sections and treated separately. Within junctions and sections respectively, different groups could be created, groups that are similar regarding geometry and other features. Average accident rate, average accident frequency, and average severity could then be calculated within each group. This would give the method more power to detect black spots.
2.4
Road safety goals and targets
The overall aim for the road safety work is to fulfil the safety goals and targets. The goals do not only decide what resources are needed but also influence which countermeasures
SweRoad
ANKARA
should be applied. Goals can, for instance, state that fatalities and severe injuries are to be decreased to or below a certain number. In such case, countermeasures aiming at severe casualties are most important and accidents can be allowed to happen as long as they do not result in severe or fatal injuries. On the other hand, if goals are set for accidents, then the countermeasures should aim at reducing all accidents. The process of deciding goals and targets is in itself a useful exercise. It increases the safety awareness among involved organizations. So goals and targets are necessary for the future safety work. If severe accidents are to be reduced, it is necessary to decrease accidents occurring at high speeds and pedestrian accidents, since they often lead to severe consequences. In that case, slight accidents are less interesting, for instance, accidents when a vehicle leaves the roadway where the roadside is flat and without any hazardous objects. Roundabouts, for example, do not normally decrease the number of accidents, but they drastically reduce the number of severe accidents, at least as long as safe passings are provided for pedestrians and bicyclists. The goals set should also influence the weighting factors used in the black spot identification. If goals are set for fatal and severe injuries, the weights should be adjusted accordingly.
2.5
Accident rate
In the following sections necessary parts of statistical theory are explained. In addition, comparisons with the formula used by KGM and suggestions for improvements are given. 2.5.1 Statistical theory
A j Number of accidents on section j during a certain time period. m j Number of vehicle kilometers in millions on section j during the same time period. Rj A j m j is the accident rate on section j during that time period.
Rc is the critical value for accident rate.

Section j is considered to be a black spot, from the accident rate point of view, if:
Rj Rc where Rc
n
m j
0.5 m j
Ai
i 1 n
mi
i 1
1 n
n i 1
mi Ri m
is the estimated average accident rate for sections belonging to
the same population. It is assumed that there is n such sections. -0.5 /mj is a correction for continuity when approximating with the normal distribution.
SweRoad
ANKARA
k is a constant that is chosen for the significance test. It is determined from a normal distribution and selected to give a certain significance level :
= 0,1% gives k = 5 % gives k = 10 % gives k
2.576
1.645
1.282
2.5.2 Comparison with the formula used by KGM In the English translation of the formula used by KGM, the average number of vehiclekilometers for all sections is used. This is not correct. The number of vehicle- kilometers for the tested section should be used, and not the average. It is not a quality control method if the average is used. It is just a way to get a critical value. It is, however, understandable that KGM uses the average value for vehicle-kilometers. In the US Report Safety design and operational practices for streets and highways is said average exposure of traffic during study. It should be stated more clearly that it is the average over the years for the actual road section that should be used. KGM also uses plus (+) for adding the last term in the equation. It should be minus (-) instead. 2.5.3 Suggestions for improvement
k 1.282 should be used (see below).
Each kilometer should have its individual value, mj for vehicle-kilometers. It should be
0.5 m j and not
0.5 m j in the formula.
2.6
Accident frequency
Statistical theory
Formatted: Bullets and Numbering
2.5.12.6.1
A road section is considered to be a black spot, from the accident frequency point of view, if:
Aj Ac , where Ac
Fave
Fave L j
0.5 L j
Ac is the critical value for accident frequency (= number of accidents).
Lj is the length of the road section. Here, Lj is assumed to be 1 km. Fave is the average accident frequency for all road sections. 2.5.22.6.2 Comparison with the formula used by KGM
In the English translation of the KGM method, it is stated that the average number of vehicle-kilometers should be used. Instead, it should be the length of the road section. If
SweRoad
ANKARA
vehicle-kilometers is used, it will give wrong results. Vehicle-kilometers will give a critical value that is lower as soon as the value is more than 1 million vehicle-kilometers. One million vehicle-kilometers corresponds to about 2750 vehicles per day if one year is considered. 2.5.32.6.3
k
Suggestions for improvement
1.282 should be used (see below).
It should be road length and not vehicle-kilometers in the formula. It should be

0.5 L j and not 0.5 m j in the formula.
2.7
Accident severity
Statistical theory
I f , j 9 I b , j 3 I d , j 1 , or more
2.5.12.7.1
The severity value for road section number j is S j clearly:
Severity = number of fatalities (f)*9 + number of injured persons (b)*3 + number of damaged vehicles (d)*1. This value can be divided by a suitable value. One such value could be the number of accidents. The relative severity value is then Q j S j A j , which means severity per accident. Here the assumption of Poisson-distribution cannot be used. The average value is estimated with:
n
Si Qave
i 1 n
Ai
i 1
And the variance
is estimated with:
2
1 n 1i
Qi
1
Qave .
The road section is considered to be a black spot, from the severity point of view, if:
Qj Qc , where the critical value Qc
Qave
0.5 .
2.5.22.7.2
Comparison with the formula used by KGM
In the KGM-method, it seems as if the average number of vehicle-kilometers is used also for severity. This is not correct. In addition, the last term is added in the KGM version, not subtracted.
Black Spot Manual
8/81
December 2001
SweRoad
ANKARA
In the translation available to SweRoad there seems to be a mistake in the calculations. Si is the severity value for one kilometer. It is not the severity value divided by the number of accidents for that kilometer which it should be. Also, Save is defined as average severity per accident. So, there are different dimensions for Si and Save. On road 100, for instance, the average severity per accident is 7. But the average severity per kilometer is 22, which is 3 times higher. This difference is explained by the fact that there is an average of around 3 accidents per kilometer. Another possible mistake in the formula is that the variance is estimated to be the square root of the mean. But this is only valid for Poisson distributions. And the weighting means that the severity is not Poisson-distributed. The variance is underestimated if the mean is used. 2.5.32.7.3 Comments on the weighting factors
The purpose of using weights is to put more emphasize on severe accidents than on slight ones. There are several different ways of determining such weights. One possibility is to use weights based on socio-economic costs. In the SweRoad report Methods and Values for Appraisal of Traffic Safety Improvements (May 2001), the following accident costs in million TL (1999 price level) are given: Accidents RURAL AREAS Material cost Risk value 13,973 235,959 6,741 9,432 813 0 8,716 161,889 3,796 6,865 286 0 Total 249,931 16,173 813 170,605 10,661 286
URBAN AREAS
Fatal accident Injury accident Property damage Fatal accident Injury accident Property damage
If weights are based on values for rural areas this would give the relations 300 for a fatal accident, 20 for an injury accident and 1 for property damage only. Another way to establish weights would be to base the weights on traffic safety goals if such were stated. Using weights, however, gives a higher random variation since the randomness is multiplied. The randomness is also higher when the differences in weights are higher. For this reason, SweRoad cannot recommend values that differ as much as 300 to 1. In addition, such weights will be almost equivalent to analyzing fatalities only. On the other hand, the weights cannot be too similar because then the weights would be of little use. A reasonable compromise between these two extremes is to use the factors 9:3:1 for fatal accidents, injury accidents and damage only accidents. Compared with the weights used at present by KGM (9 for fatalities, 3 for injuries and 1 for damage vehicles) this means that fatal accidents are weighted instead of fatalities,
SweRoad
ANKARA
injury accidents are given weights instead of injuries, and damage only accidents instead of damaged vehicles. These weights can be used until new weighting factors have been estimated and decided. In the Microcomputer Accident Analysis Package (MAAP) there is a function called worst. This can be used for selecting the places with the highest numbers of accidents and also with the highest number of weighted accidents. A print-out from this function is shown below. Here can be seen that weighting factors can be given to fatal accidents, serious injury accidents, slight injury accidents and damage only accidents. The area to be analyzed can also be chosen as well as the cell size.
2.5.42.7.4
Suggestions for improvement
If the translation of the KGM method is correct, the formula is wrong and should be corrected. Si and its average must correspond. Vehicle-kilometers should not be used in the formula. Number of accidents should be used instead of vehicle-kilometer when severity per accident is calculated and section length instead of vehicle-kilometer when severity per section is calculated.
k 1.282 should be used (see below).
Black Spot Manual
10/81
December 2001
SweRoad
ANKARA
The last term in the formula should be minus and not plus. Severity per accident should be used instead of total severity for each section.
2.8
All three criteria do not need to be met
As was mentioned under section 2.2, KGM defines a black spot as a road section that shows higher values than the critical ones for all three parameters (accident rate, accident frequency and severity index). SweRoad would prefer to make three additional lists to be used in the black spot identification procedure, one for each parameter. A reason for this is that sections showing high accident rates do not often have many accidents. And sections having many accidents do not often have high rates. Thus, having a section to satisfy all three criteria means that many sections will not be considered as black spots even if one or two of the parameters fulfil the criteria. There can be many cost-effective countermeasures for a section with a high rate even if the frequency is not high, and vice versa. The present black spot list based on all three parameters should be used together with the new lists. The KGM method requires all three criteria to be above its critical value. It can be said that it is a misuse of information to require that all three criteria must be met. It could instead be useful to list all spots where at least one value is above its critical value. This is useful information. When the formula discussed above are corrected, then it will also be very rare to find sections that meet all three criteria.
2.9
Random variation
Accidents normally occur at random. Since it is a random outcome, the actual number of accidents for a road section cannot be trusted to be the true value. The number of accidents differs from one year to another even if nothing has been changed. One spot can have more accidents than another spot during a certain year. But this does not necessarily mean that the first spot is more unsafe that the second one. To handle this randomness, statistical methods can be used. Accidents are normally considered to follow a Poisson-distribution. This means that the number of accidents during a year is an outcome of the statistical process and can be assessed with statistical theory. One very convenient feature with the Poisson-distribution is that the mean value and the variance are the same. Thus only one parameter has to be estimated.
2.10
Choice of confidence level
The purpose of discussing confidence levels in this kind of analysis is to decide the risk of making wrong decisions. Normally, a confidence level of 5 % is used. This means that there is a 5 % risk that a road section is considered a black spot when in fact it is not. Or said in a more understandable way, 5 out of 100 identified black spots are not really black spots. This is called type 1 error in statistical literature. These are spots where the random variations have been unfavorable during the actual period. But why do we accept to have
Black Spot Manual
11/81
December 2001
SweRoad
ANKARA
5 % risk? Why do we include false black spots at all and why do we not at least take a much smaller risk. 1 % or 0.1 % would be better than 5 %! The reason is that there is another type of error that can be made. Random variations can also be favorable for some spots. A spot can be a black spot, but due to favorable accident outcome during one year, the site is not identified as a black spot. This is a type 2 error in statistical literature. The errors are related in such a way that fewer type 1 errors give more type 2 errors. If one really wants to be absolutely sure that almost no false black spots are included (by selecting a very low type 1 error) many real black spots will be missed. The level chosen must strike a balance between these two types of errors. The choice of 5 % by KGM can be questioned and there are arguments for another confidence level, for example, 10 %. The reason is mainly that it is better to risk to include more false black spots and there by get more real black spots. The first step of the black spot analysis is the identification of the spots. This is a selection procedure that is employed by following certain steps to analyze the identified spots. Then it will be shown if the spots have potential for improvement or not. So the error made with a higher confidence level is that some unnecessary work has to be carried out for some spots. By selecting 10 % instead of 5 %, however, no serious errors are made. The only drawback is that the list to work with is extended. But also with a longer list, the work starts from the top. So having a longer list does not necessarily create more work. But, on the other hand, if a real black spot is missed, then a more serious error has been made, which cannot be corrected until new accident data is available for the next period. That is why it is better to have a 10 % confidence level than a 5 % level. Even 20 % is a level that could be considered.
2.11
Using more than one-year data
Using accidents for more than one year is favorable since the random variations to some extent tend to even out. If there are three years data and the mean value is three for each year, the mean for the sum of 3 years is 9. From a purely statistical point of view, it is favorable to have as many accidents as possible. If accidents from more than one year are added, the result would be more accidents. So why not use 3, 5 or even 10 years! There is one important reason, however, apart from the difficulties of storing many years of accident data. There should not be any changes at the spot, not in traffic flows or behavior and not in geometry or surface etc. As changes are frequent, small or big, this limits the size of the time-period. It is often considered that three years is a reasonable period for analysis. Three years is a suitable balance between having a long period for getting many accidents and a short period so that the spot is not changed too much. Spots that are known to have been changed geometrically or in other ways should be treated in a different way. Accidents before and after such changes should not be added.
Black Spot Manual
12/81
December 2001
SweRoad
ANKARA
2.12
Identification without damage only accidents
If damage only accidents are not collected, the numbers for identification will of course be smaller. But the procedures for accident rate and accident frequency will be the same, but based on casualty accidents only. Severity per accident can still be calculated. The weighting factors 9 for a fatal accident and 3 for a injury accident do not have to be changed. The possibilities to identify black spots will be somewhat reduced if only injury accidents are used compared to when all accidents can be utilized.
2.13
Calculation for road number 100 using the proposed method
Accident data for 1999 for road 100 was chosen to test the method with the suggested changes. It is the part of road 100 that passes through the Pilot Project area. The MAAP database was used to get accident data. The number of accidents was calculated for each kilometer. The number of casualties for each kilometer was calculated separately for fatalities, serious injuries and slight injuries. The number of involved damaged vehicles was calculated for each kilometer. So in all, 3 different tables were made. Data covers the following kilometers: Section 100-12 Section 100-13 Section 100-14 Km 55-82 Km 00-113 Km 00-90. Km 90 is in fact the Ilgaz junction and not a road section. It was not treated in a different way than other sections.
The data were copied into Excel sheets where the calculations were made. To this sheet, traffic volumes for 1999 were incorporated. 2.11.12.13.1 Accident rate
Section j is a black spot, from the accident rate point of view, if:
R j Rc where Rc
n
k
mi Ri m
m 0.5 m j j
Ai
i 1 n
mi
i 1
1 n
n i 1
This gives
2.0.
The critical value depends on the number of vehicle kilometers for each section.
Black Spot Manual
13/81
December 2001
SweRoad
ANKARA
Example: mj = 1 gives Rc = 3.3 and mj = 2 gives Rc = 3.0. A road section that is one kilometer long and has 2 700 passing vehicles per day gives mj = 1 for one year. 2.11.22.13.2 Accident frequency The total number of accidents is 422. Since there are 133 kilometers, the average accident frequency number will be 3,17 accidents per km. It is assumed that this is the mean value in the Poisson-distribution and the normal approximation is used. The formula gives thus
Ac
Fave
Fave L j
0.5 L j
Ac 3.17 1.282 3.17 1 0.5 1 Ac 4.956524
Thus, the critical value is 5, which means that kilometers having 5 or more accidents should be included in the black spot list. 2.13.3 Severity
The formula used is based on the severity value per accident. The weighting factors in the calculations below are assumed to be 9:3:1 for fatalities, injuries and involved damaged vehicles respectively. It is, however, for the future recommended to use 9:3:1 for fatal accidents, injury accidents and damage only accidents (see under section 2.7.3). Section j is a black spot, from a severity point of view, if:
Qj Qc and Qc
Qave k
0.5
where 2 is the estimated variance of Qj.

2 1 n 1i
n
n
Comment [k1]: Ipek. Please note that the formulas in this page are changed. These changes can not be tracked because the formulas are in a separate format
Qi
1
Qave
Si Qave
i 1 n
Ai
i 1
Qave = 7.0 severity value per accident.
and 2 = 6.4 2
Qc = 7.0+1,282*6.4-0.5=14.7
Black Spot Manual
14/81
December 2001
SweRoad
ANKARA
Thus, the critical value is 14.7, which means that kilometers having a severity value per accident higher than 14.7 should be included in the list. Please note that:

The average severity value per kilometer is = 22.1 and its variance is 2 = 27.92. The function variance in Microsoft Excel uses n in the denominator instead of (n-1). The difference is negligible. Identified potential black spots
2.13.4
The kilometers identified as potential black spots are given in tables 1, 2 and 3. The values are printed if the values are above the critical limits. This is equivalent to saying that the index is above 1. Only kilometers where at least one of the criteria is fulfilled are shown.
SECTION 100-12 Km 58 59 60 61 64 69 71 72 73 74 76 78 79 81 82 Number of accidents 6 6 9 5 6 14 6 7 8 5 8 7 Severity value per km 61 4.33 17.00 87 71 63 151 92 21.00 21.57 18.40 3.85 (16.00) 3.37 Severity value per accident Accident rate
6.73
3.37 3.85
Table 1. Kilometers identified as potential black spots on Section 100-12.
SECTION 100-13 Km 2 10 11 12 Number of accidents 8 (17) 12 23 93 102 8.00 15.32 Severity value per km Severity value per accident Accident rate 3.85
Table 2. Kilometers identified as potential black spots on Section 100-13. Black Spot Manual 15/81 December 2001
SweRoad
ANKARA
SECTION 100-14 Km 0 4 5 6 7 14 17 18 19 20 22 23 24 29 32 33 37 48 53 64 65 66 67 74 81 89 90 Number of accidents 8 5 7 17 8 7 6 Severity value per km
Severity value per accident
Accident rate 5.33 3.33 4.66 11.33 5.33 4.66
69 101 77 60 (16)
4.00 (19) 4.00 6.00 4.00 6.00 19.33 3.33 (16) (17) (17)
6 9 6 9 5
60
58
8 5 (38) (29) 16.57 16.2 15
5.20 3.25
7 5 5 8
116 81
4.55 3.25 3.25 5.20
Table 3. Kilometers identified as potential black spots on Section 100-14.
The total severity value for a section does not add any additional information. This is why it should be replaced by severity value per accident. In tables 1, 2 and 3, the severity value has been put within brackets if the severity value per accident is high but the number of accidents is low (one or two). It is important to know if the basis for the severity value per accident is only a few accidents. The calculations above only serve as an example of how to use the method. The method will be improved if junctions are separated from road sections and treated separately. Within junctions and within sections, different groups could be created, groups that are similar regarding geometry and other features. The average accident rate, the average frequency, and the average severity are calculated within each group. This gives the method more power to detect black spots.
Black Spot Manual
16/81
December 2001
SweRoad
ANKARA
2.14
Suggestions in brief
SweRoad suggests that additional lists of potential black spots should be made and include road sections as soon as one parameter is above its critical value. It is preferable to use three years data in the calculation process. This is better than calculating values for each year. The coefficients used should be based on a 10 % confidence level, that is, k 2.11.12.14.1

1.282 .
Accident rate
k 1.282 should be used. Each kilometer should have its individual value mj for vehicle-kilometers, not the average (m) in the formula. It should be 0.5 m j , not + 0.5/m in the formula.
2.11.22.14.2

Accident frequency
k 1.282 should be used. It should be road length and not vehicle-kilometers in the formula. It should be 0.5 L j , not 0.5 m j in the formula.
2.14.3

Severity value
k 1.282 should be used. Severity value per accident should be used. Vehicle-kilometers should not be used in the formula. Instead road length or the number of accidents should be used, depending on which of severity per section or severity per accident that is used. The last term in the formula should be minus and not plus.
In addition, SweRoad proposes that severity should be based on weighting of accidents (that is fatal accidents, injury accidents and property damage accidents) instead of persons and vehicles. New weighting factors should be estimated and decided.
2.15

References
Ezra Hauer, Identification of sites with promise, Transportation Research Board 75th Annual Meeting, 1996. Safety design and operational practices for streets and highways, US Department of Transportation, 1980. Mats Wiklund, Comments on Rate Quality Control Method (in Swedish). Black spot analysis documents from KGM 2000.
Black Spot Manual
17/81
December 2001
SweRoad
ANKARA
3
3.1
Diagnosis
Criteria in the identification process
The black spot identification process has identified hazardous kilometers based on three different criteria: (i) accident rate, (ii) accident frequency, and (iii) severity value. In this chapter it will be described how to study the problems, the accident contributing factors and the deficiencies for each identified black spot, the so-called diagnosis. The diagnosis is depending on which of the different criteria that has identified the site as a black spot. Depending on the values, the potential for improvement and the costeffectiveness can vary. 3.1.1 Accident rate and number of accidents The table and text below show, for two of the three criteria, how the potential can be: Number of accidents Accident Rate High Low A. B. Many A C Few B D
C.
D.
There are many accidents and the accident rate is high. This means a high potential for improvement. There are few accidents but the accident rate is high. Normally this means a potential for improvement. Only cheap countermeasures can be cost-effective, since traffic flow is low. There are many accidents but the accident rate is low. The traffic flow is probably high at this site. There can be a potential for improvement. But normally the situation is such that improving the safety situation can only be achieved with expensive countermeasures. This limits the cost-effectiveness. There are few accidents and the accident rate is low. There is little or no potential for improvement. Countermeasures should be applied only in certain cases. These kilometers are normally not selected in the black spot identification phase.
3.1.2 Identification based on number of accidents or accident rate The analysis is pretty much the same if the identification is made on the number of accidents or on accident rate. In both cases it is the accidents that have to be analyzed. Rate is a way to identify sites that can have a potential for improvement. But rate consists of accidents divided by traffic volume. The rate can be high and the number of accidents low, if the traffic volume is low. In theory, the analysis is the same, that is to look for accident patterns. But in practice, this search can be somewhat different. If the rate is based on few accidents, patterns cannot easily be found, simply because several accidents are necessary to form a pattern.
SweRoad
ANKARA
If the rate is high and the number of accidents is low a different approach has to be used. Then accidents have to be analyzed one by one. Site inspections will be more important since it can reveal deficiencies. 3.1.3 Identification based on accident severity The third criterion is accident severity. Severity is used because it is more important to find countermeasures at sites with serious accidents than at sites with not so serious accidents. First the relevance of the severity value has to be checked. Is it based on few or many accidents? The relevance is greater, of course, if the severity is based on many accidents. It must, in any case, be checked if there are single vehicle accidents or accidents with more vehicles involved. If there are single vehicle accidents, the severity can depend on dangerous roadsides and then the roadsides should be improved or guardrails erected. If there are multiple vehicles involved, it could be the road design that has to be improved. 3.1.4 Number of accidents and severity The table and text below show, for two of the three criteria, how the potential can be: Number of accidents Accident severity High Low E. F. Many E G Few F H
G.
H.
There are many accidents and the severity is high. This means a high potential for improvement. There are few accidents but the severity is high. The road design can be dangerous or there are dangerous items on the road or along the sides of the road. It can be possible to find cheap countermeasures. There are many accidents but the severity is low. This spot does not seem to be alarming if safety goals are set to decrease severe accidents. But it is necessary to look at these accidents because they can reveal dangerous situations. It is often possible to find countermeasures. There are few accidents and severity is low. There is little or no potential for improvement. Countermeasures should be applied only in very special cases.
3.2
Location
When black spot kilometers have been identified, the next step is locating the accidents within the selected kilometers. The accidents are often, but not always, concentrated to a certain part of the kilometer. It can be a junction, a sharp curve, a bridge or some other dangerous point. With MAAP this is can be done by using accident mapping. To use mapping is also a good idea in order to make sure that there are no accident clusters that cover the border between two kilometers. If that is the case, cross tables can miss these clusters, since accidents are
SweRoad
ANKARA
split up on the two kilometers. But this is more for the identification phase than for the diagnosis phase. Diagnosis could follow the structure shown in the figure below:
Local accident concentration along the road?
Yes
No
In junction
Other
Compare rate in junction accident types day night
Compare accident types day night surface conditions
The main aim with the analysis is to search for accident patterns. Patterns that can reveal deficiencies in the situation. The accidents can be split up on a number of specifications to see if there are many accidents of a certain feature, with a certain specification. It can sometimes be difficult to decide what is many or few of a certain features. In such cases, it is useful to know what the normal situation is and to compare with that. In the following section of the report percentages or averages for a number of situations are given. They have been calculated from the Pilot Project (PP) roads for 1999 and 2000. These values are for temporary use only. Revised percentages should be calculated for larger numbers of accidents.
3.3
Yes, there is a local accident concentration
If there is a local accident concentration along the road, it must be clarified if the concentration is in a junction or not. If the concentration is in a junction, there are at least three comparisons that can be made:

Accident rate can be calculated for the junction. Accident types can be compared. Day and night accidents can be compared. Accident rate for junctions
3.1.13.3.1
In the identification phase one criterion is accident rate for the kilometer. If there are safety problems at a junction, the accident rate can be calculated for that junction alone. This gives an idea of the magnitude of the problem and the potential gain with improvements. The accident rate for a junction is different from that for a section. For a section, vehicleBlack Spot Manual 20/81 December 2001
SweRoad
ANKARA
kilometers are calculated. For a junction, this corresponds to the number of vehicles entering the junction. The calculated measure will be accidents per million incoming vehicles. It could be a research project to establish normal rates for different junction types in Turkey. 3.1.23.3.2 Accident types for junctions
The composition of accident types differs depending on the geometry of the junction. The percentages for the PP roads during 1999 and 2000 are shown below. The numbers are too few to be separated into different junction types. The percentage of vehicles from same direction is higher and the percentage from adjacent directions is lower in 3-leg junctions compared with 4-leg junctions.
Accident Types Single vehicle Vehicles from same direction Vehicles from adjacent directions Vehicles from opposite directions Overtaking Pedestrian Others Total Accidents in junctions
15 % 39 % 35 % 6% 1% 1% 3% 100 %
3.1.33.3.3
Accident types for road sections
The corresponding percentages for road sections (excluding junctions) are shown below:.
Accident Types Single vehicle Vehicles from same direction Vehicles from adjacent directions Vehicles from opposite directions Overtaking Pedestrian Others Total Accidents on road sections
57 % 25 % 2% 8% 4% 2% 2% 100 %
3.3.4 Day-time night-time The accident distribution on day and night can be calculated. Normal distribution for accidents is 66 % during day-time, 30 % during night-time and 4 % for dusk and dawn (PP roads). These average percentages can be used for comparisons. A night percentage higher than 30 can indicate a special problem for night traffic. The percentages are almost the same for junctions and sections, so no separation is needed.
SweRoad
ANKARA 3.3.5 Surface conditions
The accident distribution on surface conditions can be calculated. Normal percentages are: dry surface 63 %, wet 32 %, snowy 3 % and icy 2 %. Percentages above 32 for wet conditions can indicate a special problem with wet surfaces, for instance, bad friction on wet road. Considerably more than 3 % snowy and 2 % icy accidents indicate that winter maintenance can have good potential, even though such small numbers have to be assessed with care. 3.3.6 Collision types The accident distribution on collision types can be calculated. Collision types should be used as a compliment to accident types. When the problem is many single vehicle accidents, it can, for instance, be interesting to know if they hit fixed objects or rolled over. Normal collision types for the PP roads in 1999 and 2000 are shown below:
Collision types Accidents on road sections Head on Rear end Nose to side Side to side Roll over Fixed object Pedestrian Others Total Accidents in junctions
5% 21 % 11 % 2% 24 % 13 % 2% 22 % 100 %
1% 26 % 49 % 4% 4% 7% 3% 6% 100 %
3.3.7 Vehicle types The distribution on vehicle types can be calculated. Vehicle types involved in accidents can be helpful in the diagnosis, especially if trucks are involved in more accidents than the percentages below. Normal vehicle percentages for the PP roads are shown below. Vehicle types with less than 2 % have been added to all others.
Vehicle type Automobile Minibus Pick up truck Truck Bus All others Total Percent
60 % 4% 10 % 15 % 5% 6% 100 %
Black Spot Manual
22/81
December 2001
SweRoad
ANKARA 3.3.8 Accidents
The distribution of accidents on accident severity can be calculated. Accident severity for the PP roads for 1999 and 2000 are shown below:
Fatal accident Number of accidents Percentage of accidents Serious injury Slight injury Damage accident accident only Total
159 5%
203 6%
1096 31 %
2033 58 %
3491 100 %
3.3.9
Casualties
The casualties distributed on severity can be calculated. Casualty severity for the PP roads for 1999 and 2000 are shown below:
Fatality Number of persons Percentage of persons Serious Injury Slight injury Total
255 7%
524 13 %
3083 80 %
3862 100 %
3.4
Stick diagram analysis
MAAP can be used for describing different accident details. It is called stick diagram analysis. The items can be decided from the items in the form. Below is one example using some important items.
3.5
Example from section 100-14, kilometer 6
The example below shows the situation for section 100-14, kilometer 6.
This section was identified with 17 accidents, severity value per km 101, and accident rate 11. The map shows a cluster of 13 accidents at the beginning of kilometer 6. There is another cluster of 4 accidents at about kilometer 6.6.
Black Spot Manual
23/81
December 2001
SweRoad
ANKARA
X represents damage only accidents, blue squares represents serious injury accidents and green squares represents slight injury accidents. If there had been a fatal accident, it would have been represented with a red square.
The 13 accidents can be selected from the map. This is done by using the polygon feature as illustrated in the picture to the left. The selection shows the severity and the identification for these accidents. They are saved and can be used for further analysis. These 13 accidents occurred between 100 and 370 meters. This is in fact in the same curve, which can be seen from the stick diagram analysis described below.
A stick diagram analysis for this cluster of 13 accidents on section 10014, kilometer 6 is shown in the following table.
Accident Meter Month No. of Surface number casualties condition 79 274 414 350 100 370 11 5 8 2 0 0 icy wet dry Collision Type roll over Other nose to side Accident type Single vehicle Single vehicle Vehicles from opposite directions Vehicles from opposite directions Single vehicle Single vehicle Vehicles from opposite directions Single vehicle Single vehicle Single vehicle Single vehicle Single vehicle Single vehicle Day light condition day day day
550
300
11
wet
Head on
twilight
647 648 689
200 250 220
12 12 3
0 0 3
wet wet wet
roll over roll over nose to side
day day day
829 903 1571 1585 1609 1672
150 225 120 200 250 250
6 7 8 11 11 4
0 4 1 4 0 0
dry wet wet wet wet wet
other other roll over roll over fixed object roll over
day day night day day night
All accidents occurred in a curve where the accidents are located from 100 meters to 370 meters from km 6.
Black Spot Manual
24/81
December 2001
SweRoad
ANKARA 3.5.1 Surface condition
10 accidents out of 13 occurred on wet conditions. This is more than expected and an indication that friction can be low. 3.5.2 Day-time night-time Two accidents happened during the night. This does not indicate a night-time problem. 3.5.3 Accident types The accident types are single vehicle or vehicles from opposite directions, that is, vehicles loosing control and colliding with oncoming vehicles. 3.5.4 Collision types Collision types indicate that one of the single vehicles collided with a fixed object and 6 rolled over. This shows that the sides of the road should be looked into to check the slopes and what obstacles there are. 3.5.5 Vehicle types In the table below is shown how information on each vehicle can be included, that is vehicle type and driver age. For an accident with more than one vehicle involved, accident details are reported for each vehicle, thus duplicated when there are two vehicles involved in the accident. Accidents identified as 414, 550 and 689 are accidents with more than one vehicle involved. Please observe that accident severity and number of casualties are defined for the accident and repeated for each vehicle.
Accident Number 79 274 414 414 550 550 647 648 689 689 829 903 1571 1585 1609 1672 Driver age 37 40 38 44 36 41 21 49 48 46 53 39 49 28 66 42 Vehicle type pick up truck car bus bus car truck car car minibus car car car car car car truck Accident severity slight damage only damage only damage only slight slight damage only damage only Slight Slight Damage only Slight Slight Slight Damage only Damage only No. of casualties 2 0 0 0 2 2 0 0 3 3 0 4 1 4 0 0
Black Spot Manual
25/81
December 2001
SweRoad
ANKARA
3.6
Diagnosis without damage only accidents
The possibility to make a good diagnosis is smaller if damage only accidents are not included. This is obvious since there are fewer accidents to analyze and thus more difficult to detect patterns. But the procedure for diagnosis is the same. Normal numbers and percentages shown above, have to be recalculated and based on injury accidents only.
3.7
Site investigations
A visit to the site is normally a necessary part of the diagnosis. This visit can give a lot of detailed information. The site might also have been visited earlier as part of the identification process. But it can be fruitful to make a new visit at the end of the diagnosis. The diagnosis can have shown patterns that makes a more detailed site investigation useful. It can, for example, give the investigator some new ideas about what to look for. Checklists can also be used. Appropriate parts of checklists for safety audits can be utilized. See SweRoads report Safety Audit Handbook (December 2001). The investigator should drive through the site. If it is a junction, he should enter from all directions and test all possible maneuvers. If the problem is wet accidents, he should drive when it is wet. If the problem is night-time accidents, he should drive at night. If there is a pedestrian problem, he should act as a pedestrian, etc. The investigator observes the site and the traffic situation. It can also be necessary with more objective and long-lasting measurements at the site, such as friction, sight distances, speeds, conflicts, gaps between vehicles and number of pedestrians crossing.
Black Spot Manual
26/81
December 2001
SweRoad
ANKARA
4
4.1
Finding countermeasures
Introduction
The choice of countermeasure should be based on the diagnosis (accident analysis) described in the previous chapter. When the black spot has been identified, the exact location has been determined, accidents have been analyzed and the problem has been described, this chapter can be used to find a suitable countermeasure and to get some ideas about the expected safety effects. In chapter 5, more details are given about how to estimate the effects. This chapter is divided into two main parts, Road sections (links) and Junctions (nodes). In each part a number of common problems or accident types are described. For each problem one or more countermeasures are presented together with expected effects. In some cases there are also mentioned some considerations that should be taken into account before the countermeasures are implemented.
4.2
4.2.1
Road sections
Single vehicle accidents
This accident type often implies that drivers lose control and go off the road. This kind of accident is often related with too high speed regarding the circumstances. The reason for that can be defective visual guidance, like a horizontal curve hidden behind a vertical crest curve, or unexpected situations, like a horizontal curve with a radius smaller than the minimum for the design speed. It can also be related to difference in level between driving-lane and shoulder or insufficient maintenance (potholes or damaged pavement on shoulder or driving-lane). Countermeasures Improved signing: Improved alignment: Improved skid-resistance: Improved pavement: Improved roadside area: Erect guardrails:
Warning signs, chevrons, delineators, speed-limit signs. Improve the visual guidance, enlarge the curve radius. Rehabilitation of the super-elevation, change of surface texture to increase the friction. Make sure that there is no difference in the level between driving-lanes and shoulders. Create a safety zone without rigid obstacles in order to reduce severity. If it is very difficult to improve the roadside area.
Effects In general, the effects are somewhat uncertain. For accidents in curves the effect is usually higher when there is a single curve on a generally straight road than in curvy sections. Improved alignment and improved pavement can result in higher speeds that can lead to more severe accidents.
SweRoad
ANKARA 4.2.2
Accidents with vehicles from the same direction
This accident type is usually not very common on road sections between junctions. However, the accident type occurs in hilly sections where the difference in speed between different vehicles is great. Countermeasures Add an extra lane: Divided four-lane road:
Construct climbing lanes on hilly sections. Construct a divided four-lane road if the traffic volume is high enough to justify this.
Effects Climbing lanes can be particularly useful where there is a mix of slow and faster traffic, such as on the uphill side of a steep gradient. Advance signing, for example, Climbing lane 1000 m ahead, can persuade some drivers to be patient and to wait for a safer opportunity to overtake. In that case, a positive effect can be reached even before the climbing lane. On long and steep gradients, heavy vehicles are driving slowly even downwards. Provided the traffic volume is high, a four-lane section divided by a barrier or guardrails can be effective to reduce the number of accidents related to hitting from behind. At the same time, head-on collisions will be avoided and accidents related to overtaking will be reduced. On the other hand, the number of single-accidents where vehicles hit the barrier or guardrails will be increased. Considerations With climbing lanes a clear definition of where overtaking is permitted, and where it is not, is essential. This should be done with traffic signs and with distinct lane and centerline markings. It is also essential that signs and markings are well maintained, especially at the start and end points. The start and end points of a divided section has to be chosen in such a way that the visual guidance is clear. It is also important that the end points (terminals) of the barriers and guardrails are safely designed. 4.2.3 Accidents with vehicles from opposite directions
This accident type implies that one of the drivers intersects the centerline without overtaking. This type of accident can be the result of incorrect position on the road because of deficient road markings or potholes and worn pavement due to bad maintenance. It can also depend on a driver that takes a short cut in a curve. Countermeasures Maintenance: Widening the road: Separate the directions:
Black Spot Manual
Renew the horizontal markings and/or the pavement. Construct paved shoulders, widen the driving-lanes. Install median, concrete barriers or guardrails.
28/81 December 2001
SweRoad
ANKARA
Effects The horizontal markings should generally be renewed every year. The safety effect of new horizontal markings could be a reduction of accidents if centerlines and edge-lines are renewed at the same time. The speed, however, will usually increase. Increasing the width up to normal standard has a positive effect on both number and severity of accidents. The effect of widening already existing shoulders is uncertain. If the directions are separated by a median wide enough to avoid over-running or by a barrier, head-on collisions can be reduced by almost up to 100 %. On the other hand, other types of accidents can increase, such as hitting from behind or single vehicles hitting the barrier. However, the severity is usually lower for those kinds of accidents. Considerations The start and end points of a divided section has to be chosen in such a way that the visual guidance is clear. It is also important that the terminals of the barriers and guardrails are safely designed. 4.2.4 Overtaking
Accidents related to overtaking could be between two vehicles in the same direction or between two vehicles in opposite directions. In both cases, the distance to the oncoming vehicle was too short, either because the sight distance was too short due to a curve or a crest, or because the driver that made the overtaking misjudged the distance to the oncoming vehicle. Countermeasures Increase sight distance: Add an extra lane: Divided four-lane road:
Make sure that sufficient sight distance for overtaking is provided at reasonable intervals along a road section. Construct climbing lanes on hilly sections. Construct a divided four-lane road if the traffic volume is high enough.
Effects A general improvement of the alignment along a road can reduce the number of accidents substantially, depending on the difference in alignment before and after. Climbing lanes can be particularly useful where there is a mix of slow and faster traffic such as the uphill side of a steep gradient. Advance signing, for example, Climbing lane 1000 m ahead, can persuade some drivers to be patient and to wait for a safer opportunity to overtake. In that case a positive effect can be obtained even before the climbing lane. Provided the traffic volume and the number of over-takings are high, a four-lane section divided by a barrier or guardrails can be effective to reduce the number of accidents related to overtaking. At the same time, head-on collisions will be avoided. On the other hand, the number of single-accidents, where vehicles hit the barriers or guardrails will probably be increased.
SweRoad
ANKARA
Considerations With climbing lanes a clear definition of where overtaking is permitted, and where it is not, is essential. This should be done with traffic signs and with distinctive lane and centerline markings. It is also essential that signs and markings are well maintained, especially at the start and end points. The start and end points of a divided section has to be chosen in such a way that the visual guidance is clear. It is also important that the terminals of the barriers and guardrails are safely designed.
4.3
4.3.1
Junctions
Single vehicle accidents
This accident type could occur when single vehicles continue straight on from the third leg in a T-junction or when single vehicles hit signs or traffic islands in a junction. Countermeasures Visibility: Warning signs: Speed limit: Rumble strips: Lighting:
Increase the visibility of the junction, especially from the secondary road approach. Install warning signs saying that there is a junction ahead. Change the speed limit to 70 km/h or 50 km/h through the junction. Apply rumble strips in order to increase the drivers attention and to reduce speed. If there are many accidents during dark hours, install road lighting.
Effects Traffic islands in the secondary road normally have a small safety effect in four-leg junctions. A local speed limit through the junction will reduce the number of accidents and also the severity. Lighting has a double effect. Firstly, it announces the junction in general and secondly, it makes it easier to observe traffic islands and signs as well as other vehicles etc. 4.3.2 Accidents with vehicles from same direction
This accident type could happen when one vehicle hits another from behind, for example, when the first vehicle has slowed down because of a stop or yield sign, traffic signals or turning movements. Countermeasures Visibility: Warning signs:
Black Spot Manual
Increase the visibility of the junction in order to make drivers aware of that such actions can be taken by other drivers. Install warning signs saying that there is a junction ahead.
30/81 December 2001
SweRoad
ANKARA Speed limit: Channelization:
TRAFFIC SAFETY PROJECT Traffic Safety Consultancy Services Change the speed limit to 70 km/h or 50 km/h through the junction. Provide separate lanes for left-turning and/or right-turning vehicles.
Effects A local speed limit through the junction will reduce the number of accidents and also the severity. A separate lane for left-turning vehicles has a positive safety effect, especially in 4-leg junctions. A separate lane for right-turning vehicles has normally no safety effect. Considerations When a separate lane for left-turning vehicles is used, a median, designed to give shelter for vehicles waiting in the left-turning lane, should be constructed. 4.3.3 Accidents with vehicles from adjacent directions
This type involves accidents between vehicles in the main road and vehicles entering from the secondary road. Countermeasures Speed limit: Traffic control:
Signalization: Visibility:
Lighting: Junction design: Divide a four-leg junction: Modern roundabout: Grade separation:
Reduce the speed limit to 70 km/h or 50 km/h through the junction. If there is no regulation, install yield-sign or stop-sign in the approach of the secondary road. If the junction is yieldregulated, change it to stop-regulated. If there is regulation, install traffic signals. Make sure that the junction is visible in all approaches and that there is enough sight distance. It is important that there are no billboards, advertisement signs, etc. obstructing the sight from the secondary road towards the main road. If there are many accidents during dark hours, install road lighting. Increase the angle between the intersecting roads. Change a four-leg junction into two three-leg junctions. If the traffic volume is similar on all approaching roads, consider reconstruction to a modern roundabout. If the traffic volumes are high, consider grade separation.
Effects A local speed limit through the junction will reduce the number of accidents and also the severity. A change from yield to stop regulation is effective in rural areas. Modern, traffic regulated, signals have a rather good safety effect. Time-regulated signals might increase the number of accidents.
SweRoad
ANKARA
Changing a four-leg junction into two three-leg junctions normally has a good effect on accident severity, especially if the percentage of vehicles from the secondary road is high. Considerations The angle between the main road and the secondary road should be close to 90. 4.3.4 Accidents with vehicles from opposite directions
This type involves mainly accidents with vehicles turning to the left from the main road. Countermeasures Channelization:
Separate lane for left-turning.
Effects Separate lane for left-turning vehicles has a positive safety effect, especially in four-leg junctions. Considerations When a separate lane for left-turning vehicles is used, a median, designed to give shelter for vehicles waiting in the left-turning lane, should be constructed. 4.3.5 Accidents with pedestrians
These are normally accidents in junctions between motor vehicles and pedestrians, when the pedestrians are crossing one of the junction legs. Countermeasures Marked pedestrian crossing: Pedestrian crossings marked with vertical signs and horizontal markings. Channelization: Install fences to lead the pedestrians to safe crossing locations. Secure low speed: Install speed reduction devices, such as rumble-strips, before at-grade pedestrian crossings. In urban areas speed humps can be used. Signalization: Traffic signals will separate pedestrians from motor traffic in time. Traffic signals could introduce hazards of a different kind if vehicle speeds are relatively high in the approaches to the crossing. Therefore, approaching traffic must have adequate visibility and time to stop when required. Grade separation: If the number of pedestrians and/or the traffic volume is high or if the number of children and elderly is significant, a grade-separated crossing should be considered. Effects The effect of marked pedestrian crossings is uncertain. The best effect is achieved if the marked crossing is combined with speed reducing devices. Grade separation (over- and underpasses) is very effective, if it is used by pedestrians.
SweRoad
ANKARA
Considerations When a marked pedestrian crossing is provided, the pedestrians could get a false feeling of safety when using it. Therefore, it is essential that the location is visible for the drivers and that a low speed is ensured. The use of grade-separated crossings is very much depending on the location. The location should be where it is convenient for the pedestrians.
Black Spot Manual
33/81
December 2001
SweRoad
ANKARA
5
5.1
Estimating the effects of countermeasures

Introduction
The basis for prioritizing is to have estimates of the benefits of different proposed countermeasures. Thus it is essential to make forecasts of the accident and injury outcome if a certain measure is applied. These forecasts have to be based on knowledge of the reduction factors of different countermeasures. This knowledge is best built up from research and follow-ups of the results from different places where the measures have been applied. This is a task for future safety research in Turkey. Building up knowledge will take many years and a reliable common data bank is required. Meanwhile, it is necessary to have some understanding of expected reduction factors. This can be reached by using research and development from other countries and adjust to Turkish conditions and Turkish behavior in traffic. Applying the same measure at different places can give different results. One reason is that there are not two places that are exactly the same. Another reason is that random fluctuation in the number of accidents and injuries can give different results. One special problem with the randomness is the so-called regression-to-the-mean effect. Without going into details, it is important to remember that normal selection of black spots tends to overestimate the reduction factor. This is explained in chapter 8 Follow-up and evaluation. The reduction factor estimates given in this chapter are mainly based on a Swedish handbook Effektkatalog 2000 (Ref. 1) including Swedish experiences, and the Norwegian safety handbook Trafikksikkerhetshndbok (Ref. 2). The latter summarizes knowledge from research reports from many different countries. It is, however, not certain that all these values are applicable to Turkey. There are several reasons why some of the reduction factors could be different in Turkey. However, the given estimates could form a basis for calculating Turkish estimates. For some countermeasures, where Turkish driving behavior differs significantly from European behavior, this has been specially mentioned. In the chapter: + always means increase. This means that the countermeasure is not successful. The number of accidents or casualties is larger than if the countermeasure was not implemented. Sometimes + is strengthened with the word increased, even if this is not strictly correct, increased with +10 %, for instance. - always means decrease. This means that the countermeasure is successful. The number of accidents or casualties is smaller than if the countermeasure was not implemented. Sometimes - is strengthened with the word decreased, even if this is not strictly correct, decreased with - 10 %, for instance. Reference is also made to the SweRoad report Highway design report (June 2000) and its appendices. Many of the effect estimates are given as intervals in order to show the variation or
SweRoad
ANKARA
uncertainty of the estimates. Different research projects can give different results. The results from research projects are often stated as intervals, due to randomness in accidents. The intervals can be used as a reminder of the uncertainty involved. It can also be used in such a way that if a countermeasure at one site is thought to be more favorable than at the average site, estimates in the upper end of the interval can be used and vice versa. Available estimates can refer to different variables, for instance, sometimes fatalities and sometimes fatal accidents. Often there is not detailed information available to separate a reduction factor into fatal accidents and fatalities. For this reason, the same factor can normally be used both for fatalities and fatal accidents. This can also be valid for injuries and injury accidents. Summary of proposed reduction factors
Countermeasures on sections
Road widening Climbing lanes Decrease the number of approaches Road side delineators
Estimated reduction factors Accidents -20 %

-25 % -5 %; -10 %
Comments
Fatalities -10 %
-15 % -5 %; -10 %
Injuries -15 %
-20 % -5 %; -10 %
No reduction in urban areas
Close to 0 % 0 %; -10 % -10 %; -15 %
Close to 0 % 0 %; -10 % -20 %; -30 %
Close to 0 %
In darkness on roads with bad alignment increase
Road markings General speed limits Lower speed limits during winter Local speed limits Bridge widening Side area improvement Guardrails Median barriers Vertical alignment Increased horizontal curve radius Improved signing in horizontal curves Super elevation Sight distance New surface
Black Spot Manual
-20 %
Decrease -40 % 0 % 0 % +20 %; +25 % 0 %; -20 % -5 %; -60 % -10 %; -40 % -10 %; -20 % -5 %; -15 % Around 0 %
0 %; -10 % -15 %; -20 % Depends on decrease in average speed -40 % -30 % Depends on decrease in average speed Decrease more Decrease more than accidents than accidents -20 % -30 % Not based on empirical data -20 %; -40 % -20 %; -40 % -20 %; -40 % -15 %; -20 % 0 %; -20 % -5 %; -60 % -10 %; -40 % -10 %; -20 % -5 %; -10 % Around 0 %
35/81
-20 %; -40 % -10 %; -15 % 0 %; -20 % -5 %; -60 % -10 %; -40 % -10 %; -20 % -5 %; -10 % Around 0 %
No reduction
December 2001
SweRoad
ANKARA
Increased friction -5 %; -10 % -5 %; -10 % -5 %; -10 % Decreased rutting Around 0 % Around 0 % Around 0 % Decreased 0 %; -5 % 0 %; -5 % 0 %; -5 % unevenness Prohibit overtaking -5 %; -10 % -5 %; -10 % -5 %; -10 % Variable message -15 %; -20 % -15 %; -20 % -15 %; -20 % signs Improved route Around 2 % Around 2 % Around 2 % guidance + means increase, - means decrease, when the countermeasure is applied
Table 4. Estimated reduction factors for countermeasures for road sections
Countermeasures in junctions Island on the secondary road in 3-leg junction Island on the secondary road in 4-leg junction Left-turning lane with curbs in 3-leg junction
Estimated reduction factors Accidents 0 % Fatalities 0 % Injuries 0 %
Comments
-5 %; -10 % 0 %; -10 %
-5 %; -10 % 0 %; -10 %
-5 %; -10 % 0 %; -10 % Closer to 10 % in urban areas and closer to 0 % in rural areas.
Left-turning lane, painted in 3-leg junction Left-turning lane, with curbs in 4-leg junction Left-turning lane, painted in 4-leg junction Right-turning lane Change one 4-leg junction into two 3leg junctions
0 %; -10 %
0 %; -10 %
0 %; -10 %
-10 %
-10 %
-10 %
-10 % 0 % 0 %
-10 % 0 % 0 %; -40 %
-10 % 0 % 0 %; -40 %
Roundabout
+20 %; -70 %
-50 %; -80 %
0 %; - 50 %
Can increase accidents Higher reduction factor when the percentage of vehicles from secondary road is higher Can increase accidents in central areas
Black Spot Manual
36/81
December 2001
SweRoad
ANKARA
Modern traffic regulated signals -15 %; -30 % -15 %; -30 %

-15 %; -30 % Time regulated signals increase accidents
Interchange 3-leg Interchange 4-leg Lighting in junctions Change yield to stop in rural areas Change yield to stop in urban areas Flashing yellow in signal during low traffic hours Rumble strips
-20 %; -40 % -60 %; -70 % -5 %; -10 % -10 %; -15 % 0 %; -5 % +50 %
-40 %; -60 % -60 %; -90 % -5 %; -10 % -10 %; -15 % 0 %; -5 % +50 %
-40 %; -60 % -60 %; -90 % -5 %; -10 % -10 %; -15 % 0 %; -5 % +50 %
Decrease
Counter measures -25 %; -70 % in railway junctions + means increase, - means decrease, when the countermeasure is applied
Table 5. Estimated reduction factors for countermeasures for junctions.
Decrease more Decrease more Depends on than accidents than accidents decrease in average speed -25 %; -70 % -25 %; -70 %
Estimated reduction factors Improvements for pedestrians and bicyclists Sidewalks Separate bicycle and pedestrian lanes in rural areas Separate bicycle and pedestrian lanes in urban areas Grade separated pedestrian and bicycle junctions Accidents -5 %; -10 % 0%; -5% Fatalities -5 %; -10 % 0%; -5% Injuries -5 %; -10 % 0%; -5%
Comments
Around -4 %
Around -4 %
Around -4 %
Reduction factors depend on the use of separation Marked pedestrian Can increase crossing +25 %; -20 % +25 %; -20 % +25 %; -20 % accidents Bus stop Small or no Small or no Small or no Small or no reduction factor reduction reduction factor reduction factor factor + means increase, - means decrease, when the countermeasure is applied
Table 6. Estimated reduction factors for improvements for pedestrians and bicyclists. Black Spot Manual 37/81 December 2001
Around -80 % in pedestrian accidents
Around -80 % in pedestrian fatalities
Around -80 % in pedestrian injuries
SweRoad
ANKARA
5.3
5.3.1
Road sections
Section types
2-lane roads Below are shown some average values for rural roads with speed limit 70 km/h from Ref. 1. Accidents with pedestrians and bicyclists are not included.
Road width (m) < 5.7 5.7-6.6 6.7-7.9 8-10 10.1-11.5 11.6 Motor traffic road
x)
Accident ratex) 0,456 0,416 0,376 0,360 0,336 0,320 0,248
Casualties per accident 0,52 0,52 0,52 0,52 0,52 0,52 0,53
Severe casualties per accident 0,138 0,138 0,138 0,138 0,138 0,138 0,133
Percentage of damage only accidents (%) 64 64 64 64 64 64 64
The accident rate is expressed in accidents/106 veh-km.
Table 7. Average values used in Sweden for roads with speed limit 70 km/h.
Below are shown corresponding values for rural roads with speed limit 90 km/h from Ref. 1. Accidents with pedestrians and bicyclists are not included.
Road width (m) < 5.7 5.7-6.6 6.7-7.9 8-10 10.1-11.5 11.6 Motor traffic road Accident rate 0,320 0,296 0,264 0,256 0,240 0,224 0,224 Casualties per accident 0,63 0,63 0,63 0,63 0,63 0,63 0,58 Severe casualties per accident 0,172 0,172 0,172 0,172 0,172 0,172 0,162 Percentage of damage only accidents (%) 61 61 61 61 61 61 61
Table 8. Average values used in Sweden for roads with speed limit 90 km/h.
4-lane roads Below are shown average values for 4-lane rural roads with speed limit 90 km/h and 110 km/h. Accidents with pedestrians and bicyclists are not included.
Black Spot Manual
38/81
December 2001
SweRoad
ANKARA
Speed limit (km/h) 90 110 Accident rate Casualties per accident 0,45 0,5

Severe casualties per accident 0,050 0,090 Percentage of damage only accidents (%) 61 61
0,224 0,184
Table 9. Average values used in Sweden for 4-lane roads.
Motorways Below are shown corresponding values for motorways. Accidents with pedestrians and bicyclists are not included.
Speed (km/h) 50 70 90 110 limit Accident rate 0,560 0,560 0,224 0,184 Casualties per accident 0,35 0,4 0,45 0,5 Severe casualties per accident 0,039 0,044 0,052 0,093 Percentage of damage only accidents (%) 70 65 61 61
Table 10. Average values used in Sweden for motorways.
These data (Table 7 10) are from Sweden with a different road safety situation than Turkey. The values cannot directly be used in Turkey. They can probably give a reasonable indication of relative differences between different road widths and types. A research project should be started aiming at estimating corresponding values for Turkish conditions. 5.3.2 Traffic control and equipment for sections
Road widening A widening from a narrow road to a normal 2-lane road gives according to Ref. 2 the reduction factor that injury accidents decrease with -5 %; -10 % and damage only accidents with -5 %; -25 %. Ref. 1 gives a -20 % reduction in accidents when widening from 6,5 meter to 13-meter wide rural roads. For urban roads, no reduction is estimated. The percentage reduction is the same for severe and slight accidents. A wider road decreases the number of accidents in rural roads. The same is not true for urban areas. In urban areas it is more difficult to see any effect at all since there are many junctions and a wider road implies wider junctions, which sometimes make them more dangerous. Roads with different widths have been compared with respect to their accident rates. This can be useful, but care has to be taken since road width is often correlated with other factors that can also decrease accidents. The alignment can differ, for instance. In the table below are shown estimates from Ref. 1 indicating a decrease in accident rate for rural roads with increasing width.
SweRoad
ANKARA
Road width before change (m) 6 7 9 11 Speed limit 70 km/h Decrease in accident rate with increased width to: 7m -13 % 9m -16 % -5 % 11 m -20 % -10 % -5 % 13 m -25 % -15 % -11 % -5 %

Speed limit 90 km/h Decrease in accident rate with increased width to: 7m -15 % 9m -20 % -10 % 11 m -30 % -20 % -10 % 13 m -35 % -25 % -20 % -7 %
Table 11. Estimated reduction factors for road width in Sweden.
The overall estimate for the reduction factor as a result of an average (2-lane roads) widening is -20 % for accidents. Due to increased speed, lower values are estimated for casualties, -10 % for fatalities and -15 % for injuries. Climbing lanes Climbing lanes makes it easier to overtake slower vehicles. The longer and steeper a road is, the more useful is a climbing lane. A high number of slow moving vehicles is also an indication of the need for a climbing lane. A climbing lane effects the traffic situation both before and after the lane. Before, because drivers know that they are going to have overtaking opportunities. This can decrease their frustration and limit the number of dangerous overtaking. After, because the drivers have had the possibility to overtake on a safer section. But there is also a fear that a climbing lane increases the speed and thus injury accidents after the climbing lane. The end part of a climbing lane has to be given special consideration. Drivers can speed up and make hazardous overtaking, since they know that if they are not overtaking now, they will not get a new opportunity for a long distance. Ref. 2 estimates overtaking lanes to reduce the number of injury accidents with -20 %. This includes the effect of sections both before and after the climbing lane. For overtaking lanes in both directions (short 4-lane roads) the estimated reduction factor is -40 % for injury accidents. The estimated reduction factor of climbing lanes is -25 % for accidents. Due to increased speed, it is estimated that fatalities will go down by -15 % and injuries by -20 %. Decrease the number of approaches To decrease the number of approaches should decrease the accident risk by around -5 % on high-speed roads and maybe -10 % on roads with 70 km/h speed limit according to Ref. 1. According to Ref. 2, reducing the number of approaches by 50 % will decrease the number of injury accidents by 25 %; -30 %. These factors are valid when the number of approaches are high.
Black Spot Manual
40/81
December 2001
SweRoad
ANKARA
It is estimated that reducing the number of approaches will decrease the number of injury accidents by -25 %; -30 % for sections with many approaches and where the number of approaches is reduced to 50 %. For sections with few approaches the reduction is less. Roadside delineators Research with roadside delineators shows little or no effect on accidents and injury accidents according to Ref. 2. When studying accidents during darkness only, a Finnish study shows a tendency to an increase in injury accidents when roads with bad alignment are equipped with delineators. The increase was not statistically significant. When studying accidents in darkness and during bad surface conditions only, there was a significant accident increase. Roadside delineators are appreciated by drivers because they increase comfort and make driving easier. The estimated reduction factor for roadside delineators for accidents, injury accidents and fatal accidents is 0 %. However, delineators make driving more comfortable and can lead to increased speed. Road markings Analysis performed in Ref. 2 shows that the most probable reduction factor of edge markings for injury accidents is -3 %. Estimated reduction in injury accidents for centerline markings is -1 %. Neither -3 % nor -1 % is statistically significant. In a few studies, the reduction factor has been estimated when a previously unmarked road was marked with both center and edge markings. An analysis of these studies shows a significant decrease of injury accidents by -24 %. Two research reports compare an unmarked road with a road with center and edge markings and roadside delineators. The results show a significant decrease in injury accidents by -48 %. Reduction factors largely depend on how drivers adapt to the new situation and change their speed. Road markings are often rather bad in Turkey. The situation is sometimes almost similar to not having any markings at all, that is like marking a previously unmarked road. For this reason it is estimated that road markings have a safety effect of 0 %; -10 %. It is estimated that the reduction factor for accidents of road markings is: 0 %: -10 %. Due to increased speed it is estimated that injuries and fatalities decrease less than accidents, probably by 0 %; -3 %. Change of general speed limit Speed limits are not very popular among drivers. Even in a country like Sweden, where the safety awareness is high, 50 % of vehicles in rural areas exceed the limit. Therefore, surveillance is important to keep the number of speeding vehicles down. Automatic speed surveillance using cameras could give substantial results.
SweRoad
ANKARA
One factor making drivers accept speed limits is to have limits in accordance with the geometrical design of the road. Otherwise physical measures may have to be implemented to reduce speed. There is no doubt, however, that lower speed limits decrease the number of accidents. The most severe accidents decrease more than the number of accidents. Research from many countries has shown this. The amount of the decrease depends on the change in actual speeds. The same change in a speed limit can result in different changes in actual speeds. Ref. 2 gives the following estimates based on research from different countries. The only conditions included are those where the speed limit has been changed from around 100 km/h (since the maximum speed limit is 90 outside motorways in Turkey). The theoretical calculations are based on the speed ratio model. The reduction factor of decreased speed can be estimated by using a model for the relationship between accidents and casualties, and average speeds. The model is the following: The decrease in injury accidents is supposed to decrease equal to the square of the ratio between speed before and speed after. The decrease in serious injury accidents is supposed to decrease equal to the cube of the ratio between speed before and speed after. The decrease in fatal accidents is supposed to decrease equal to the ratio between speed before and speed after raised to the power of 4. This means that the reduction factor of a measure can be estimated by first measuring or estimating the change in average speed. Example: Speed measurements at a certain site show an average speed of 97 km/h. It is estimated that a changed speed limit (e.g., from 90 to 70 km/h) at this site will decrease average speed to 88 km/h. The estimated reduction factors are:

Injury accidents (88/97)2 = 0.82, which means 1 0.82 = 18 % decrease. Serious injury accidents (88/97)3 = 0.75, which means 1 0.75 = 25 % decrease. Fatal injury accidents (88/97)4 = 0.68, which means 1 0.68 = 32 % decrease.
Change in speed limit from 100 km/h to 80 km/h. Decrease in actual speed was around 8 km/h.
Estimated reduction factor Fatal accidents Injury accidents Damage only accidents -29 % -14 % -6 % Interval (-39 %; -19 %) (-18 %; -10 %) (-40 %;+17 %) Theoretical calculations -30 % -16 %
Table 12. Estimated reduction factors for 20 km/h decrease in speed limit from 100km/h.
Change in speed limit from 90 km/h to 70 km/h and from 80 km/h to 60 km/h. Decrease in actual speed was around 5-6 km/h.
SweRoad
ANKARA
Estimated reduction factor -43 % -23 % -6 %

Interval Theoretical calculations -23 % -13 %
Fatal accidents Injury accidents Damage only accidents
(-60 %; -19 %) (-31 %; -14 %) (-40 %;+17 %)
Table 13. Estimated reduction factors for 20 km/h decrease in speed limit from 80 - 90 km/h.
Change in speed limit from 70 km/h to 60 km/h and from 60 km/h to 50 km/h. Decrease in actual speed was around 3-4 km/h.
Estimated reduction factor Fatal accidents Injury accidents -23 % -9 % Interval (-31 %; -14 %) (-10 %; -7 %) Theoretical calculations -19 % -10 %
Table 14. Estimated reduction factors for 20 km/h decrease in speed limit from 60-70 km/h.
Enforcement is important when a speed limit is changed. In Turkey massive enforcement is necessary if the above mentioned speed and accident reductions would materialize. The reduction factors of changed speed limit depend on the decrease in speed. Average reduction factors are estimated to be -10 %; -15 % for accidents, -20 %; -30 % for fatalities and -15 %; -20 % for injuries. The reduction factors are based on such changes in speed limit that the actual average speed is decreased by 5-6 km/h. Time restricted speed limits Sometimes there is a need for special speed limits during certain times when there are special conditions. Some countries have lower speed limits during the winter and others have special limits on certain bad roads during wet surface conditions. Time restricted speed limits are also used around schools where the limits are applicable to, for instance, between 7.00-17.00 on weekdays only. Special winter limits have been followed up in Finland. The speed limit was decreased from 100 km/h to 80 km/h. Reduction factors are estimated in Ref. 2 to:
Estimated reduction factor All injury accidents Fatal accidents -21 % -40 % Interval (-23 %; -16 %) (-58 %; -14 %)
Table 15. Estimated reduction factors for 20 km/h decrease in speed limit during winter in Finland.
For speed limits during wet surface conditions there are no estimates available. The above mention speed-ratio model can be used for estimating the reduction factors.
Black Spot Manual
43/81
December 2001
SweRoad
ANKARA
The estimated reduction factors for time restricted speed limits are -20 % for accidents, -40 % for fatalities and -30 % for injuries when the change in speed limit is 20 km/h. Local speed limits Lower speeds give fewer accidents. The number of severe accidents decreases more than the number of accidents. The reduction factor can be estimated using the speed-ratio model. Calculations of the reduction factor start with deciding the decrease in average speeds (not the decrease in the speed limits). Change in the average speeds is normally much lower than the change in speed limits. Bridge widening Sudden decreases in road width are always a potential hazard for accidents. Approaching vehicles can be forced to maneuvers that can be dangerous. They can come too close to oncoming traffic or they can collide with the side barrier. It is important to give the end of the side barrier of a bridge a proper design in order to avoid serious injuries if a vehicle should crash into the terminal. Pedestrians walking on the bridge could also create hazardous situations if they do not have enough space. A bridge could also be more slippery than the surrounding road. It is important when estimating the reduction factor of road widening that accidents that happen because of this are excluded from the benefits. No estimates based on real accident data have been found. It is estimated that the reduction factor is larger than for a normal road widening, maybe double that estimate. If this is assumed, it gives an estimate of -40 % for accidents, -20 % for fatalities and -30% for injuries for the length of the bridge. Side area improvement When a vehicle leaves the road it is important that it does not collide with fixed hazardous objects like trees or outcrops of rock. Steep slopes are also dangerous and will be dealt with under the guardrail section. Flattening the side to avoid vehicles from rolling over is also a safety improvement. A possible benefit with flatter sides is that a vehicle that goes off the road will have the possibility to return to the road again. The flattening, however, has become somewhat discussed during the last years. When a vehicle that has gone off tries to come back, the driver sometimes turns the steering wheel so much that when the vehicle changes direction it either rolls over or it passes over the road towards the other ditch. The latter can cause severe accidents if there are oncoming vehicles. Roadside improvements do not normally decrease the number of accidents but the severity of accidents. A maximum reduction factor (in Sweden) of side area softening is supposed to be -20 % of fatal and seriously injured persons. Ref. 2 indicates that injury accidents are decreased by -42 % when changing the side-slope from 1:3 to 1:4. A further decrease of -22 % is estimated if improving from 1:4 to 1:6.
SweRoad
ANKARA
It is estimated that the reduction factors for substantial side area improvements are 0 % for accidents and -20 %; -40 % for fatalities and injuries. Roadside guardrails Guardrails are used to prevent a vehicle from leaving the carriageway and the shoulders. It can be used at the roadside to hinder vehicles to go off the road or in the median to prevent collisions with oncoming vehicles. Guardrails on the side are erected where it is dangerous for vehicles to leave the road. The roadsides can be dangerous because the environment is hazardous with trees, rocks or stones or because of steep and high slopes. Guardrails can also be erected where there are pedestrians and bicyclists along the road. This latter purpose is not included in this chapter. Disadvantages with guardrails could be that they limit the space for pedestrians and bicyclists walking or riding along the road. The end points (terminals) of guardrails can also create severe injuries to vehicle occupants if they are incorrectly designed. Guardrails must also be so soft that a colliding vehicle is not thrown back into the traffic stream, but instead is caught by, and continues along, the guardrail. Guardrails do not normally decrease the number of accidents. They are more likely to increase the numbers. But if guardrails in a curve, for instance, is equipped with reflectors it could decrease accidents. The main benefit with guardrails is that they decrease accident severity. Ref. 1 says that a modern guardrail is as good as (if not better than) a flat roadside and a full safety zone. Ref. 2 gives the following estimates: Estimated reduction factor -43% -52% -18% Interval -48%; -41% -53%; -51% -22%; -14%
Fatal accidents Injury accidents Damage only accidents
The reduction factors for guardrails are estimated to be approximately the same as for roadside improvements: 0 % for accidents and -20 %; -40 % for fatalities and injuries. Median barriers In Ref. 1 it is estimated that median barriers reduce the number of casualties by -10 %; -15 %. Damage only accidents increase by +20 %; +25 %. Ref. 2 indicates that median barriers on multi-lane roads decrease the number of fatal accidents by -20 % and the number of injury accidents by -5 %. Damage only accidents increase by +25 %. These results are mainly from USA. Swedish tests with median barriers on wide two-lane roads show very promising results.
Black Spot Manual
45/81
December 2001
SweRoad
ANKARA
The reduction factors of median barriers are estimated to be +20 %; +25 % for accidents (increase), -15 %; -20 % for fatalities and -10 %; -15 % for injuries. Vertical alignment Ref. 2 gives the following estimates of improvements in vertical alignment:
Improved vertical alignment from over 70 to 50 -70 from 50 -70 to 30 -50 from 30 -50 to 20 -30 from 20 -30 to 10-20 from 10 -20 to under 10 Estimated decrease in accidents -20 % -10 % -10 % -7 % -2 % Interval -38 %;+1 % -20 %;0 % -15 %; -5 % -12 %; -1 % -8 %;+6 %
Table 16. Estimated reduction factors for improved vertical alignment.
It is stated that the uphill direction is safer than downhill direction. The overall reduction factor for an average improvement of vertical alignment is estimated to be 0 %; -20 % for accidents, fatalities and injuries. Increased curve radius The sharper the curve, the higher the sideways friction is needed to keep the vehicle on the road. Accidents happen when the sideways friction is too low and because of high speed or poor road surface. One way to decrease the sideways friction needed is to increase the curve radius. A drivers behavior depends on how he can foresee the curve. Accidents could happen if the driver is surprised by the sharpness of the curve. It has been shown that a single sharp curve is more dangerous than the same curve surrounded by other sharp curves. This is because the driver knows what to expect when there are many curves. For the same reason, the first sharp curve in a series of curves is more dangerous than the following ones. It is thus important that improvements in curves are made in such a way that accidents are not migrated to the next curve. Ref. 1 gives the following estimates:
Curve radius after improvement Curve radius before improvement 400 m 401-600 m 601-800 m 401- 600 m -25 % 601- 800 m -34 % -12 % 801 m -37 % -16 % -5 %
Table 17. Estimated reduction factors for improved curve radius on roads with speed limit 70 km/h in Sweden.
Black Spot Manual
46/81
December 2001
SweRoad
ANKARA

Curve radius after improvement
Curve radius before improvement 200 m 201-400 m 401-600 m 601-800 m
201-400 m
401-600 m
601-800 m
801-1000 m
1001-2000 m
-25 %
-40 % -20 %
-48 % -30 % -12 %
-52 % -37 % -20 % -10 %
-58 % -45 % -30 % -20 %
Table 18. Estimated reduction factors for improved curve radius on roads with speed limit 90 km/h in Sweden
The overall estimated reduction factor for an average improvement of curve radius is -5 %; -60 % for accidents. The highest factor concerns the case when a very sharp curve is improved to an almost straight road. Due to increased speed, the reduction factors for fatalities and injuries are less. Improved signing in curves One of the cornerstones of road safety is never to surprise a driver. If there are curves that are sharper than the drivers have reasons to expect, it is wise to have some warning for these curves. It is then advisable to put up signs that improve the drivers vision of the curve or warns for the curve. It is important that this is done in a consistent way. If other curves with the same geometry are left unattended they could be even more dangerous since drivers expectations have changed. It is also important that the signs are visible in darkness. The reduction factor for background marking signs in curves is estimated to be -20 %; 40 %. Advance warning for curves has been found to decrease the number of injury accidents by -10 %; -30 %. However, the estimates are uncertain. The overall estimated reduction factors are -10 %; -40 % for accidents, fatalities and injuries. Superelevation The technical description is that the sideways friction needed depends on vehicle speed, curve radius and superelevation (cross-fall). Unsuitable superelevation could give the result that the sideways friction is too low to keep the vehicle on the road. The behavioral aspect is that drivers do not adjust speed enough to compensate for unsuitable superelevation. Drivers that are unused to the road may not be aware of the bad superelevation and can be surprised by it. Drivers that are used to the road may know about the bad cross-fall, but tend to take a short cut making it dangerous for oncoming traffic. No reduction factors based on real accident data have been found. Increased superelevation is much like increasing the curve radius. The reduction factor is estimated to be the same as for a small curve radius increase, -10 %; -20 % for accidents, fatalities and injuries.
Black Spot Manual
47/81
December 2001
SweRoad
ANKARA
Sight distance In Sweden correction factors are used to adjust for differences in sight distances. Sight distances are classified according to classes 1 to 4, defined in the following table:
Sight class 1 2 3 4
Table 19. Sight distance classes used in Sweden.
Percentage of road length with sight distance over 300 meters 70100 % 4070 % 2040 % 020 %
These classes are the basis for correction factors on accident rates. The correction factors are shown in the table below.
70 km/h Sight distance class 1 0.9 0.94 0.99 0.99 1 1 2 0.95 0.98 1.03 1.03 1.05 1.05 3 1 1.04 1.08 1.08 1.08 1.08 4 1.05 1.09 1.14 1.14 1.14 1.14 1 0.95 0.95 0.98 0.98 1 1 90 km/h Sight distance class 2 1 1 1.03 1.03 1.05 1.05 3 1.05 1.05 1.08 1.08 1.08 1.08 4 1.05 1.05 1.08 1.08 1.08 1.08
Road width (m) < 5.7 5.7-6.6 6.7-7.9 8-10 10.1-11.5 11.6
Table 20. Estimated reduction factors for improved sight distance in Sweden.
The factors are used as follows: On a 5.7-6.6 m wide 70 km/h road, the normal accident rate is multiplied by 0.94 if the sight distance class is 1. It is multiplied by 0.98 if the sight distance class is 2, by 1.04 if the class is 3 and by 1.09 if the class is 4. Reduction factors can easily be calculated from this table. The reduction factor for a substantially improved sight distance is estimated to be -5 %; 15 % for accidents. Since improved sight distance increases speed, the estimates for fatalities and injuries are lower: -5 %; -10 %. New surface Ref. 2 has the following estimates of accidents after resurfacing compared with the old surface:
Black Spot Manual
48/81
December 2001
SweRoad
ANKARA

Estimated reduction factor +6 % -3 % Interval -12 %; +28 % -3 %; +10 %
Injury accidents Damage only accidents
Table 21. Estimated reduction factors for new surface.
A renewed asphalt surface does not seem to give a statistically significant change in the number of accidents. This holds for both injury accidents and damage only accidents. In some studies, there is a tendency for a small increase in the number of accidents. The most probable reduction factor for a new surface is 0 % for accidents, fatalities and injuries. Increased friction Increased friction has often been a countermeasure at specific accident-prone sites. The estimates given below are thus subject to regression-to-the-mean-effect and can be overestimating the true reduction factor. This fact is stated in Ref. 2 from where the estimates are taken. The estimates are about injury accidents. Damage only accidents are estimated to decrease with the same percentages.
Friction increase by Accidents on a wet around 0,1 clean surface From a friction of -40 % about 0,5 or lower (interval -55 %; -30 %) From a friction of -25 % about 0,6 or lower (interval -33 %; -17 %) From a friction of -15 % about 0,7 or lower (interval -25 %; -5 %) Accidents on a dry All accidents on a clean surface clean surface 0 % -10 % (interval -10 %; +5 %) (interval -20 %; -4 %) 0 % -6 % (interval -5 %; +5 %) (interval -12 %; -1 %) 0 % -4 % (interval -5 %; +5 %) (interval -10 %;+ 3 %)
Table 22. Estimated reduction factors for improved surface friction.
It is estimated that the reduction factor for increased friction is -5 %; -10 % for accidents, fatalities and injuries. Decreased rutting Research into the reduction factor of decreased rutting does not seem to indicate any major safety effect. It seems, however, as if higher rut depth can decrease the number of accidents. The estimated reduction factor for decreased rutting is 0 % reduction for accidents, fatalities and injuries. Decreased unevenness There seems to be a small but positive safety effect of decreased unevenness. This is valid for unevenness values that are not too low. In the research quoted, 95 % of the roads had unevenness under 5,1 mm/m.
Black Spot Manual
49/81
December 2001
SweRoad
ANKARA
The reduction factor for decreased unevenness is estimated to be 0 %; -5 % for accidents, fatalities and injuries. Prohibit overtaking There are no known research results that give the reduction factor of prohibiting overtaking. It cannot be ruled out that prohibiting overtaking can have a positive effect, for instance, in hidden depressions or junctions with bad visibility or other sites with many overtaking accidents. It is important, however, to be consistent and not to use prohibit overtaking too often. Prohibit overtaking is estimated to have a lower reduction factor than a median barrier. Since a median barrier has a reduction factor of -10 %; -15 %, it is estimated that prohibited overtaking will have a factor of -5 %; -10 %, with the same value for accidents, fatalities and injuries. Traffic regulation and information with variable message signs (VSM) VMS have been used at pedestrian crossings etc. and with speed recommendations and road surface and traffic information. Speed reductions by -10 % and up to -30 % decrease in accidents have been the results. VMS with road surface information have given somewhat lower speed, corresponding to a decreased accident risk by -15 %; -20 %. The reduction factor for effective VMSs is estimated to be -15 %; -20 % for accidents, fatalities and injuries. Improved route guidance There are no known research results that show the reduction factor of improved route guidance. It is quite possible that clear and unique route guidance has a positive effect on safety. It can avoid sudden brakes when the driver has to change lane or direction at the last second. It can also prevent drivers to take the wrong way, which leads to more vehicle kilometers traveled than necessary. It is estimated that improved route guidance will have a small reduction factor, around -2 % on accidents, fatalities and injuries.
Black Spot Manual
50/81
December 2001
SweRoad
ANKARA
5.4
Junctions
5.4.1 Junction types Simple junction. No dividing islands Dividing island on secondary road Separate left-turning lane
Modern roundabout
Signalized junction
Interchange
Comment [k2]: One of the pictures have been changed. May need some improvements in the lay out
Below are shown some reduction factor estimates from Ref. 1. The factors are based on the difference in average values for each junction design/type.
3-leg junctions in urban areas Speed limit 50 km/h Change in Change in number of number of accidents fatalities and serious injuries 0% 0% -5 % -5 % +20 %; -20 % +25 %; -10 % -30 %; - 40 % 0 %;-5 % 0 %; -5 % 0 %; -15 % +5 %; -20 % -40 %; -50 % Speed limit 70 km/h Change in Change in number of number of accidents fatalities and serious injuries 0% 0% -5 %; -10 % -5 % -10 %; -30 % +15 %; -5 % -20 %; -30 % -5 % 0 %; -5 % -20 %; -40 % -10 %; -30 % -50 %; -60 %
Junction design/type
Dividing islands on the secondary road Left-turning lane painted Left-turning lane with curbs Modern roundabout Modern signalization Interchange
Table 23. Estimated reduction factors for 3-leg junctions in urban areas in Sweden.
Black Spot Manual
51/81
December 2001
SweRoad
ANKARA

3-leg junctions in rural areas Speed limit 70 km/h Change in Change in number of number of accidents fatalities and serious injuries 0% 0% -10 %; -15 % 0% -30 %; -50 % -10 %; -30 % -20 %; -30 % -10 % 0% -40; -60 % -5 %; -20 % -40 %; -50 % Speed limit 90 km/h Change in Change in number of number of accidents fatalities and serious injuries 0% 0% -10 %; -15 % 0% -10 %; -15 % 0%
Not Not recommended recommended Not Not recommended recommended -20 %; -30 % -50 %; -60 %
Table 24. Estimated reduction factors for 3-leg junctions in rural areas in Sweden.
When the speed limit is 50 km/h in a rural area, the same reduction factor estimates are used as for 50 km/h in an urban area.
4-leg junctions in urban areas Speed limit 50 km/h Change in Change in number of number of accidents fatalities and serious injuries 0 %; -5 % 0 %; -5 % -5 %; -10 % -5 %; -10 % +5 %; -40 % +20 %; -20 % -60 %; -70 % 0 %; -10 % 0 %; -10 % -10 %; -60 % -5 %; -50 % -60 %; -70 % Speed limit 70 km/h Change in Change in number of number of accidents fatalities and serious injuries about -5 % 0 %; -5 % about -10 % about -10 % -30 %; -60 % +10 %; -20 % -60 %; -70 % about -10 % about -10 % -60 %; -80 % -40 %; -60 % - 80 %; -90 %
Table 25. Estimated reduction factors for 4-leg junctions in urban areas in Sweden.
Black Spot Manual
52/81
December 2001
SweRoad
ANKARA

4-leg junctions in rural areas Speed limit 70 km/h Change in Change in number of number of accidents fatalities and serious injuries about -5 % about -10 % about -10 % -40 %; -70 % -5% about -10 % about -10 % -75 %; -85 % Speed limit 90 km/h Change in Change in number of number of accidents fatalities and serious injuries about -5 % -5 % about -10 % about -10 % Not recommended about -10 % about -10 %
Dividing islands on the secondary road Left-turning lane painted Left-turning lane with curbs Modern roundabout
Modern signalization
-10 %; -40 %
-5 %; -35 %
Interchange
-60 %; -70 %
-80 %; -90 %
Not Recommend ed Not Not recommended Recommend ed -60 %; -70 % -80 %; -90 %
Table 26. Estimated reduction factors for 4-leg junctions in rural areas in Sweden.
When the speed limit is 50 km/h in a rural area, the same reduction factor estimates are used as for 50 km/h in an urban area. The reduction factors stated above are based on average values for each design. There could be many factors that differ between the different types and designs. This means that the whole reduction does not necessarily reflect the design standard. This is one reason why the reduction factors stated above, not necessarily are the same as the reduction factors given later in this chapter. The results from table 23-26 are used as one reference for estimates in this chapter. Dividing islands on the secondary road Dividing islands on the secondary road are, especially in 4-leg junctions, successful when the visibility of the junction needs to be improved. Otherwise, it is not regarded as a measure that decreases accidents, especially not in 3-leg junctions. One good thing with dividing islands is that it allows pedestrians and cyclists to cross the road in two steps, which increases their safety. Ref. 2 gives the following estimates:
Estimated reduction factors
Interval +5 %; +31 % -41 %; +17 % -61 %; +12 %
T junction Injury accidents X junction Injury accidents Damage only accidents
+18 % -17 % -34 %
Table 27. Estimated reduction factors for dividing islands on the secondary road. Black Spot Manual 53/81 December 2001
SweRoad
ANKARA
The reduction factors for dividing islands on the secondary road for the number of accidents, injuries and fatalities are estimated at -5 %; -10 % in 4-leg junctions and 0 % in 3-leg junctions. Separate left-turning lanes The reduction factors of left-turning lanes are uncertain. Nose-to-tail accidents, when a vehicle turns left from the primary road, is decreased to a large extent. But other accident types can be increased since the junction becomes larger and thus more difficult to overview. This could increase accidents with vehicles crossing or accidents involving other maneuvers. Installing dividing islands is normally positive since it increases the visibility of the junction for approaching vehicles. The drawback is that an obstacle is placed in the middle of the road. This could cause accidents when drivers do not see the obstacle and lose control when hitting them. This can happen when visibility is poor, in bad weather or darkness. If the junction cannot be lighted, it is important that the visibility of signs and markings is good. One way to overcome the problem with vehicles losing control is to paint the islands instead of using islands with curbs. But painted islands can also create problems. Drivers must be prevented from driving on the painted islands. It is especially important to avoid vehicles from using the left-turning lane as an overtaking opportunity. If this cannot be done by normal traffic rules, it can be necessary to physically prevent it. One solution could be to install road delineators in the islands. They have the benefit to make it difficult to drive there and if a driver does not see them, they are so soft that the driver does not lose control in case of hitting them. Painted islands are of little value if snow or mud covers the painting or if the painting is worn out. Ref. 2 gives the following reduction factors for T-junctions:
Estimated reduction factors Left-turning lanes with curbs Injury accidents Damage only accidents Painted left-turning lanes Injury accidents Damage only accidents -27 % +20 % -22 % -20 % Interval -48 %; +3 % -18 %; +75 % -45 %; +11 % -49 %; +26 %
Table 28. Estimated reduction factors for left-turning lanes in T-junctions.
It can be noted that all intervals cover 0. This means that it cannot be ruled out that reduction estimates can be 0 %. Ref. 2 gives the following reduction factors for X junctions:
Black Spot Manual
54/81
December 2001
SweRoad
ANKARA

Estimated reduction factors Interval -25 %; +22 % -49 %; +38 % -14 %; +92 % -47 %; -2 %
Left-turning lanes with curbs Injury accidents Damage only accidents Painted left-turning lanes Injury accidents Damage only accidents
-4 % -16 % +28 % -26 %
Table 29. Estimated reduction factors for left-turning lanes in X-junctions.
It can also be noted here that all intervals cover 0. This means that it cannot be ruled out that reduction estimates can be 0 %. Ref. 1 also gives the following reduction factors: Painted left-turning lane Urban Rural -10% -15 % -10% -10% Left-turning lane with curbs Urban Rural -10% 0% -10% -10%
T junction X junction
Since the intervals cover 0 % higher weight is given to the factors in Ref. 1. The reduction factor for 3-leg junctions with left-turning lanes with curbs is estimated at 0 %; -10 %, with the lower value in rural areas. The reduction factor for painted left-turning lanes is estimated at -10 %; -15 %. Since painting has probably less effect in Turkey, the reduction factor for this intervention is decreased to 0 %; -10 %. In 4-leg junctions, painted and leftturning lanes with curbs are estimated to have the same reduction factor of around -10 %. With painted left-turning lanes and road delineators erected on the islands, the reduction factors are higher than with painted lanes only. Separate right-turning lanes A separate right-turning lane does not normally improve road safety. A separate lane can increase the number of accidents, since the junction becomes wider and thus more difficult to overview. Vehicles passing the vehicle aiming at turning right can be in the shadow of the right turning vehicle. This creates dangerous situations if waiting vehicles drive into the junction without seeing the hidden vehicle. The estimated reduction factor for right-turning lanes is 0 % for accidents, injuries and fatalities. Modern roundabouts Modern roundabouts have several advantages from a safety point of view. If correctly designed they decrease vehicle speeds, which is beneficial for safety. A roundabout also creates one-way traffic, which simplifies for the drivers. This also implies that left-turning in front of oncoming vehicles is eliminated. One-way traffic also simplifies for the
Black Spot Manual
55/81
December 2001
SweRoad
ANKARA
approaching vehicle, since the driver only has vehicles coming from one direction to consider when entering the roundabout. Vehicles are conflicting at small angles in a roundabout. So, if there is a collision, the collision forces are small. This means that the risk of severe accidents is low when there is an accident. Ref. 2 gives the following estimates:
Estimated reduction factors T junction Injury accidents Damage only accidents X junction Injury accidents Damage only accidents -27 % +52 % -35 % +43 % Interval -40 %; -12 % +29 %; +78 % -46 %; -23 % +37 %; +50 %
Table 30. Estimated reduction factors for roundabouts.
The estimates are based on the situation where approaching vehicles have to yield for traffic already in the junction. Ref. 2 indicates that the number of injury accidents will decrease by -25 %; -35 %. This result comes from both junctions that previously were signalized and junctions that previously were regulated by yield signs. The reduction factors for accidents cover a large interval. It happens that the number of accidents increases after the construction of a roundabout. However, severe accidents generally decrease substantially. Therefore, modern roundabouts are one of the best countermeasures for decreasing severe accidents in junctions. It is estimated that the reduction factors for modern roundabouts are +20 %; -70 % for accidents, -50 %; -80 % for fatalities and 0 %; -50 % for injuries. Changing one 4-leg junction into two 3-leg junctions There are two reasons why transforming one 4-leg junction into two 3-leg junctions decrease the accidents even if some vehicles have to pass through two junctions:

The accident rate in a 3-leg junction is less than half of the rate in a corresponding 4leg junction. This is mainly because it is easier to get an overview of the junction. The numbers of points where vehicles can have conflicting crossing movements are 3 in a 3-leg junction but 16 in a 4-leg junction.
Ref. 2 gives the following estimates:
Black Spot Manual
56/81
December 2001
SweRoad
ANKARA

Percentage of vehicles from secondary road < 15 % +35 % +15 % 15 %; 30 % -25 % 0% > 30 % -33 % +3 % All -20 % +3 %
Injury accidents Damage only accidents
Table 31. Estimated reduction factors for changing one 4-leg junction into two 3-leg junctions.
The reduction factor of changing from one 4-leg junction to two 3-leg junctions is estimated to be around 0 % for accidents, and 0 %; -40 % for fatalities and injuries. The reduction factor increases with the percentage of vehicles coming from the secondary road. Signalization Ref. 2 gives the following estimates:
Estimated reduction factors T junction Injury accidents Damage only accidents X junction Injury accidents Damage only accidents -15 % -15 % -30 % -35 % Interval
-25 %; -5 % -40 %; +15 % -35 %; -25 % -45 %; -25 %
Table 32. Estimated reduction factors for signalization.
Estimating factors from Ref. 1 is given in tables 23-26. Installing modern traffic signals in a junction is estimated to give an average reduction factor of -30 % for accidents, injuries and fatalities in 4-leg junctions, and -15 % in 3-leg junctions. Traffic signals can, however, increase the number of accidents if installed in junctions where the percentage of vehicles coming from the secondary road is low. It is assumed that the signal is traffic regulated. Time regulated signals are not advisable. They do normally increase the number of accidents. Violating red light in signalized junctions is more common in Turkey than in the countries where the reduction factor estimates are coming from. This could imply that the reduction factors would be lower in Turkey. Grade-separated interchange Grade-separated interchanges are together with modern roundabouts the safest types of junctions. There are, however, certain things that have to be remembered. When a vehicle has left the primary road and arrives at the crossing with the secondary road, this normally at-grade junction can become dangerous. So, it is necessary to have a design that approaching vehicles are not driving too fast. It is also important to try to design approaches in such a way that vehicles do not enter the primary road in the wrong direction.
SweRoad
ANKARA
Ref. 2 estimates the reduction factor of an interchange to be -50 % (interval -57%; -46%). Ref. 1 estimates the reduction factor on accidents to be lower than -50 % in 3-leg junctions and higher than -50 % in 4-leg junctions. The decrease in severe accidents is estimated to be higher than the decrease in all accidents. The reduction factors for interchanges are estimated to be:
Interchange 3-leg Interchange 4-leg Accidents -20 %; -40 % -60 %; -70 % Fatalities -40 %; -60 % -60 %; -90 % Injuries -40 %; -60 % -60 %; -90 %
5.4.2 Traffic control and equipment in junctions Stop or yield To use stop signs in a junction is a way of simplifying for the driver. If he follows the intention with the stop, his driving task could be split up in several steps. He drives to the line and stops there. The next step is to look around and decide to start. The third step is to start. With the yield sign the driver has to do all these things simultaneously, which could be difficult. Especially older drivers sometimes have problems with junctions. One drawback with too much use of mandatory stop is that it can dilute the acceptance and thus the positive safety effect. In Ref. 1 the estimated reduction factors on accident reductions are:
Estimated reduction factor -10 %; -15 % 0 %; -5 %
Change from yield to stop in rural areas Change from yield to stop in urban areas
Table 33. Estimated reduction factors for change from yield to stop in Sweden.
In Ref. 2 the following estimates are given for injury accidents when changing from yield to stop:
Estimated reduction factor T junctions X junctions -19 % -35 % Interval (-38 %;+7 %) (-44 %; -25 %)
Table 34. Estimated reduction factors for change from yield to stop.
The estimates given by Ref. 2 may be influenced by regression-to-the-mean effects, because it is not likely that stop will have higher effect than signalization. The reduction factors for changing from yield to stop regulation are estimated to be around 10 % for accidents, injuries and fatalities in T-junctions and around 15 % in Xjunctions in rural areas. The reduction factors are less in urban areas.
Black Spot Manual
58/81
December 2001
SweRoad
ANKARA
Change from stop to yield on the other hand is estimated to increase the number of injury accidents by +39 % (interval +19 %; +62 %). Changing to four-way stop has been shown to reduce the number of accidents by -45 % in USA and Canada. Lighting in a junction Lighting in a junction decreases the number of night-time accidents and thus all accidents. The reduction factor is higher when traffic on the secondary roads is higher. Lighting is also more necessary where there are many pedestrians. The reduction factor for lighting in a junction is estimated to be -20 %; -40 % of nighttime accidents, corresponding to -5 %; -10 % of all accidents, fatal accidents and injury accidents. Where there are many vulnerable road users, the decrease of fatalities and injuries could be higher. Flashing yellow When a traffic signal is changed to flashing yellow during low-traffic hours, the number of injury accidents is estimated to increase by +50 % (interval -7 %; +165 %) according to Ref. 2. Flashing yellow is also against the Vienna Convention. It is estimated that flashing yellow during low-traffic hours will increase accidents, fatalities and injuries by about +50 %. Rumble strips Rumble strips consist of a number of painted strips across the road. The aim is to create vibrations and noise so that the driver is alerted. This is also supposed to make him decrease his speed. Rumble strips can be used before junctions, pedestrian crossings, curves or other sites where increased alertness and speed adjustments are wanted. The reduction factor of decreased speed can be evaluated by using the same model for the relationship between accidents and injuries and average speeds as was explained under section Change of general speed limit. The model is the following:

The decrease in injury accidents is supposed to decrease equal to the square of the ratio between speed before and speed after. The decrease in serious injury accidents is supposed to decrease equal to the cube of the ratio between speed before and speed after. The decrease in fatal accidents is supposed to decrease equal to the ratio between speed before and speed after raised to the power of 4.
This means that the reduction factor of a measure could be estimated by first measuring or estimating the change in average speed.
Black Spot Manual
59/81
December 2001
SweRoad
ANKARA
Example: Speed measurements at a certain site show an average speed of 55 km/h. It is estimated that applying rumble strips at this site will decrease average speed to 50 km/h. The estimated reduction factors are:

Injury accidents (50/55)2 = 0.83, which means 1-0.83 = 17 % decrease Serious injury accidents (50/55)3= 0.75, which means 1-0.75 = 25 % decrease Fatal injury accidents (50/55)4 = 0.68, which means 1-0.68 = 32 % decrease
It is estimated that rumble strips in front of a junction decrease the number of injury accidents by -30 % (interval -40 %; -25 %) in the junction, and damage only accidents by -25 % (interval -45 %; -5 %). Fatalities are reduced more, around -40 %. 5.4.3 Railway junctions Ref. 2 summarizes estimated reduction factors from different studies. Most of the studies are from USA with the designs, traffic rules and road user behavior that are prevalent in that country. Estimated reduction factors are given in the table below.
Measure Mark the junction with X-signs, when previously there were no safety devices Light and sound signal where there previously were only an X-sign Install barriers where light and sound signal already existed Install barriers where signs already existed Increase the sight distances Estimated reduction factor -25 % -50 % -45 % -67 % -44 % Interval -45 %; -5 % -55 %; -45 % -55 %; -35 % -75 %; -55 % -68 %; -5 %
Table 35. Estimated reduction factors for railway junctions.
Swedish research shows a significant increase in safety when installing half barriers in junctions where previously only light and sound systems were used. It was also seen that it is dangerous when the road and the railway are running parallel before the junction. Other measures have been to increase the visibility with the light signal in a gantry above the junction. In a junction where the sun was standing low at some times, the light signal was combined with lights on the road surface. Reduction factors for different types of measures are given in the table above.
5.5
Improvements for pedestrians and bicyclists
Sidewalks on urban roads Sidewalks on urban roads decrease the number of injury accidents for bicyclists by -30 % and pedestrians by -5 %, but the number of injury accidents with motor vehicles increases by +16 % according to Ref. 2. This means that all injury accidents will decrease by -7 % (interval: -13 %; -1 %).
Black Spot Manual
60/81
December 2001
SweRoad
ANKARA
The reduction factor for sidewalks is estimated to be -5 %; -10 % for accidents, fatalities and injuries. Separated bicycle and pedestrian lanes in rural areas Pedestrian and bicycle lanes in rural areas are normally separated from motor traffic lanes. Ref. 2 says that a separate pedestrian and bicycle lane does not necessarily decrease accidents in rural areas. The estimated reduction factor is 0 % (interval -10 %; +11 %). Ref 1refers to some research reports that indicate a positive effect. The estimated reduction factor for separate bicycle and pedestrian lanes (in rural areas) is 0 %; -5 % for accidents, fatalities and injuries. This result is a little surprising, but one explanation is that the numbers of pedestrians and bicyclists may have increased after the improvement. Another possible explanation could be that all pedestrians and bicyclists do not use the new lane and it becomes more dangerous for those who do not use the new lane. Motor vehicles could also have increased speed and do not expect vulnerable road users in their lanes. Separated bicycle and pedestrian lanes in urban areas Pedestrian and bicycle lanes in urban areas can be separated from motor vehicle traffic by curbstones. Ref. 2 shows that constructing a separate pedestrian and bicycle lane in urban areas decreases the number of injury accidents. The estimated reduction factor for separate bicycle and pedestrian lanes (in urban areas) is -4 % (interval: -7 %; -1 %) for accidents, fatalities and injuries. The lane is separated from motor vehicle traffic by curbs. Even painted pedestrian and bicycle lanes have turned out to be effective. Grade-separated pedestrian and bicycle junctions Building a grade-separated junction is estimated to decrease the number of injury accidents by -30 % (interval -44 %; -13 %). The estimated reduction factor for grade-separated pedestrian and bicycle junctions is 80 % (interval -90 %; -69 %) for pedestrian accidents. The same factors are also used for fatalities and injuries. These estimates are based on the assumption that over- and underpasses are built and planned in such a way that almost all vulnerable road users really use them. To use the over- or under-pass must not require longer time or longer distance than just crossing the street. Furthermore, there should be no major difference in level. Pedestrians in particular are very sensitive to good designs. There are many examples around the world where grade-separated junctions are not used by pedestrians to the extent anticipated. Pedestrian crossings Creating painted and signed pedestrian crossings without curbed islands does not improve safety according to Ref. 2. On the contrary, it increases the number of injury accidents. In
SweRoad
ANKARA
the table below, the estimated reduction factor of different types of pedestrian crossings are given:
Type of Pedestrian crossing Painted and signed crossing Signalized crossing on a section between junctions Signalized junction without separate phase for pedestrians Signalized junction with separate phase for pedestrians Elevated crossing Crossing with islands with curbs Estimated reduction factor on injury accidents Pedestrian Motor All injury Interval (all accidents vehicle accidents injury accidents accidents) +28 % +20 % +26 % +18 %;+35 % -12 % -2 % -7 % -12 %; -2 % +8 % -12 % -1 % -7 %;+6 %
-29 %
-18 %
-22 %
-29 %; -14 %
-49 % -18 %
-33 % -9 %
-39 % -13 %
-58 %; -10 % -21 %; -3 %
Table 36. Estimated reduction factors on accidents for pedestrian crossings.
Reduction factors for different types of pedestrian crossings are given in the table above. Bus stops There are three types of dangerous situations connected with bus stops:

The bus decreasing speed and stopping could cause dangerous situations for other vehicles. Standing passengers on their way to or from a seat can be injured when the bus is changing direction and accelerating/decelerating. The bus or other vehicles can hit pedestrians that are waiting for, or on their way to and from the bus.
The first and the last type are the most dangerous ones. To avoid the last situation it is necessary with good approach roads for pedestrians and good visibility when they go off the bus. It is especially important that they do not cross the road in the shade of the bus. In rural roads there are three types of bus stops:

The bus stops on the normal roadway. The road is widened to give a separate place for the bus (lay-by). The road is widened to give a separate place for the bus. This place is separated from the road by an island, sometimes with curbs.
In urban areas there are in principle the same alternatives, but depending on available space, parked vehicles etc. there are some more detailed designs.
Black Spot Manual
62/81
December 2001
SweRoad
ANKARA
Estimated road safety for motor traffic on rural roads Bus stop design Estimated reduction factor compared with no bus stop Bus stop on the road Small negative reduction factor Bus bay (separated with curbs) No reduction factor Lay-by No reduction factor
Table 37. Estimated reduction of accidents for different designs of bus stops (mainly on rural roads).
There are no reduction factors based on accident data available. Some indications are given in the table above.
5.6
More than one countermeasure
When more than one countermeasure is implemented at the same site, it is not possible to just add the different reduction factors in order to estimate the resulting total reduction factor for all countermeasures. Firstly, when one measure has been implemented, the next one will only effect the remaining accidents. In addition, it is possible that the first measure will also reduce the severity of the still occurring accidents and that the reduction factor for the following measure will be less than it generally is because of the effects of the first measure etc. If, for example, the speed is reduced as the first measure (e.g. by changed speed limit), then the effect of the next measure (e.g., a junction redesign) will be less, because there are fewer accidents to reduce, and the severity of the still remaining accidents and casualties is less. If, on the other hand, some other measure is implemented first, then the effect of a speed reduction could be lower than if it had been implemented first. So, the reduction factors for the different measures depends on the order of implementation and their respective effects on both accidents and severity (e.g., accident and collision types). In principle, the reduction factors could also be effected by the general safety situation (e.g., concerning seat belt use). It is, therefore, difficult to give detailed recommendations on how the reduction factors should be reduced if several countermeasures are implemented at the same site. The important thing to remember is that the above mentioned principles have to be considered in order not to overestimate the safety benefits.
Black Spot Manual
63/81
December 2001
SweRoad
ANKARA
6
6.1
Prioritizing
Introduction
The need for black spot improvements are normally much greater than what is possible to implement with available resources. Therefore, a suitable balance must be struck between the needs and what can actually be implemented. When doing this, focus should be on the most suitable safety effects, and the projects should be prioritized accordingly. Sometimes, it may be justified to deviate from this optimal order of priority, but having a list of all black spot interventions arranged in order of priority according to their estimated benefits and costs will make decision-makers aware of the reduced benefits and/or increased costs of any deviation. When planning for the implementation of black spot improvements, it is necessary to decide:

which black spots should be improved, which intervention/design (of different options) should be selected for each site, in what order and when the selected interventions should be carried out.
The process to do this is here called prioritizing. In short, prioritizing implies finding the best projects and the best action plan, according to some defined criteria, based on estimated effects and costs as well as budget restrictions. Prioritizing is normally made by so-called appraisals. Appraisals and appraisal methods are used to estimate in advance if a proposed project or plan is effective and efficient. The general principles for project appraisal and setting priorities are explained in SweRoads report Methods and values for appraisal of traffic safety improvements (May 2001). In that report, monetary values for accident and casualty reductions are also given.
6.2
Appraisal methods
For appraisal of black spot improvements, there are, in principle, two main methods (see mentioned report):

Cost-Benefit Analysis (CBA) Cost-Effectiveness Analysis (CEA) CBA

6.1.16.2.1
CBA implies, in principle, that the sum of all positive effects (benefits) of an investment is set against all negative effects (costs). In order to be able to do so, all effects, positive and negative, have to be expressed in the same kind of unit, money. This means, for example, that accident and casualty reductions, as well as travel time savings and reduced environmental impact etc., have to be given monetary values. The project yielding the best positive effects in relation to the costs should be selected first.
Black Spot Manual
64/81
December 2001
SweRoad
ANKARA
One problem with CBA is that the same amount of money one year does not have the same value another year. This has to be taken into consideration by discounting. The discounting factors depend on the number of years between the actual year and the discount year, and the discount rate. The discounting procedure should not take into account any changes or trends in current prices. It should concern real-term prices only. This implies that the values of all future benefits and costs have to be discounted and capitalized to a selected discount year and that investment costs also should be capitalized to the same year. The discount year can be different from the year when the intervention is finished and the improved road is re-opened for traffic. After that, the discounted benefits (B) and the discounted costs (C) have to be compared and analysed. As explained in the mentioned report, the most commonly used indicators are:

Net Present Values (NPV = B - C) Internal Rate of Return (IRR) Benefit/Cost-ratio (BCR = B/C) Net Benefit/Cost-ratio (NBCR = (B - C)/C)
In the above mentioned report, it is recommended that the BCR (or NBCR) should be used for setting priorities of black spot improvements. This will result in the highest total NPV in relation to investment costs. A BRC greater than 1 means that the project is profitable and a value lower than 1 indicates that the project causes a loss to society (a NBCR greater than 0 means that it is profitable and a value lower than 0 means that it causes a loss). A thorough CBA of road investment projects should theoretically include all relevant costs and benefits for society. For road projects, the most common effects and costs are:

Accident costs Travel time costs Vehicle operating costs Environmental costs Investment costs Maintenance costs
The costs are obtained by estimating the size of the effects, expressed in some suitable unit, and multiplying this effect by the monetary value. For example, first the number of reduced accidents has to be estimated (see chapter 5) and then this number must be multiplied by the monetary value per accident (see mentioned report). The principle is the same for other effects. For time savings, for instance, the number of saved hours are estimated and then this number is multiplied by the monetary value per hour. Finally, when all effects are estimated and valued, all benefits should be added and the total benefits set against the total costs.
IRR is the discount rate which equalizes the discounted benefits and the discounted costs (i.e., NPV = 0)
SweRoad
ANKARA
For black spots improvements, which mainly result in safety effects, it could be sufficient to include costs for accident and casualties, investment and changed road maintenance. For high-cost options, however, which also yield substantial other benefits than safety, it is recommended to consider all relevant effects and costs. It should be mentioned that lack of monetary values and sufficiently detailed models sometimes makes it difficult to to make a full CBA. In such cases so-called engineering estimates could be used to set values to such non-valued effects. For estimating the BCR, the following general formula can be used: BCR = (B + MC)/IC BCR B (Formula 1)
MC
IC
= Benefit-Cost Ratio = discounted value of future accident and casualty reductions (for every year of the economic life-time of the intervention, the estimated reduction in accidents and casualties should be multiplied by the relevant monetary value) = discounted value of changed road maintenance cost because of safety intervention (the value should be added to benefits if the maintenance costs are reduced, and subtracted from the benefits if the costs are increased) = investment cost of safety intervention
The discounting should consider:

the expected life-time of the intervention, the discount rate.
The tax factor should be applied to all costs (see below). 6.2.2 CEA CEA means that the positive effects, expressed in some suitable unit, for example, number of reduced accidents or casualties, are set aganist the costs expressed in terms of money (i.e., investment costs and substantial changes in road maintenance costs). The project yielding the highest ratio of positive effects compared with the costs should be selected first. If road safety projects have a special budget, which is solely intended for improving safety, there is, in principle, no need to make a full CBA in order to prioritize between safety measures and other investments. Then it is not necessary to have a monetary value of accidents and casualties. On the other hand, it is normally necessary to have a weighting scale by which different degrees of accident severity can be weighted against each other. For estimating the cost-effectiveness, the following simplified formula can be used:
Black Spot Manual
66/81
December 2001
SweRoad
ANKARA ECR = Delta AC/(AYIC + YMC) ECR Delta AC
TRAFFIC SAFETY PROJECT Traffic Safety Consultancy Services (Formula 2)
AYIC
YMC
= effectiveness/cost-ratio for the first year = estimated reduction in number of accidents and casualties the first year, weighted according to a scale, e.g., 9 for fatal accidents, 3 for injury accidents and 1 for property damage only accidents. The weighting factors should depend on the safety goals. If the focus is set on fatalities, the weight should be high for these casualties etc. = average yearly investment cost of safety intervention (total investment cost divided by the number of years corresponding to the economic lifetime of the intervention) = changed yearly road maintenance cost because of safety intervention
In principle, the tax factor should be applied to all costs (see below). Theoretically, it would be better to use discounting also for CEA. Then, the following formula could be applied: ECR* = Delta AC*/ (IC + MC) ECR* Delta AC* (Formula 3)
IC MC
= effectiveness/cost-ratio (for discounted values) = discounted number of reduced accidents and casualties for every year during the economic life-time of the intervention. The accidents etc. have to be weighted according to a suitable scale, see above = discounted value of investment cost of intervention = discounted value of changed maintenance cost because of safety intervention (the value should be added to investment costs if the maintenance costs are increased, and subtracted from the investment costs if the maintenance costs are reduced).
6.2.3 Recommended method Even if the primary goal of a safety budget is to reduce accidents and casualties, it is still of interest to society to see to it that these measures are effective not only from the safety point of view but also from all other aspects. This is especially the case for high-cost options. Therefore, SweRoad recommends that CBA (a limited version for low-cost options) should be used for black spot appraisals even if there is a special budget for road safety interventions (this is also what KGM is already doing). 6.2.4 Important parameters and values According to present KGM analysis, the number of accidents on existing roads are considered to increase directly proportional to traffic growth if no safety intervention is implemented. This can be questioned since better cars and better informed drivers will effect the accident outcome without any black spot interventions. In addition, accident statistics shows that there has not been any direct proportionality between traffic growth and increase in casualties. This is especially valid for fatalities. The annual number of fatalities has been almost constant or even reduced, while the number of vehicle-kilometers
SweRoad
ANKARA
has increased substantially. Therefore, SweRoad recommends not to use the mentioned proportionality, but either to use constant numbers of accidents over time (for the existing road) or to assume more modest increases. Concerning other effects, such as travel time and environmental impact, traffic growth has of course to be taken into account. The discount rate used for calculations of discounted values of benefits and costs is very important and has a major influence on all results. At present, KGM uses 15 percent, which is a high figure compared with many other countries. This means that short-term interventions are favoured and that long-term investments are difficult to justify. Lowering the discount rate would make substantial increases in the discounted benefits for projects with long life expectancies. SweRoad recommends that a lower discount rate should be used for road safety investments (and also for other road investments), for example, in the interval 8-12 percent. The economic life-time (1 year, 5 years, 10 years, etc.) of the interventions has to be estimated in order to make it possible to calculate the discounted values of future benefits and costs, as well as the average annual costs of investment, etc. In CBA, all taxes should be eliminated from costs. However, it is necessary to include some kind of tax factor or factors, by which all or some of the cost components should be increased (see mentioned report). SweRoad recommends that one tax factor is used and that this factor is set at 1.17, because KDV is 17 percent at present. In order to get relevant CBA-values, all costs should be multiplied by this factor. 6.2.5 Special questions Formula (1) and (3) above can give rise to some questions. For example, should in formula (1) the monetary value of accident and casualty reductions increase in real terms over time, or should the same value be applied for all years? Should in formula (3) the estimated number of reduced accidents and casualties be discounted, that is in principle, should a lower value be given today to a life saved in the future than to a similar life saved this year? These questions have to be discussed further. Awaiting such discussions, SweRoad recommends KGM to use the same monetary accident and casualty values in real terms for all years during the economic life-time, and that the number of reduced accidents and casualties for each year are discounted by the normally used discount rate.
6.3
Proposed procedure for KGM
Theoretically, to give correct answers to all the questions mentioned under introduction, prioritizing has to be carried out in one major effort in which all questions, which black spots, which intervention/design, what implementation order and when, are treated. This, however, would be rather complicated. Therefore, SweRoad proposes the following somewhat simplified procedure:
Black Spot Manual
68/81
December 2001
SweRoad
ANKARA 1. 2. 3. 4.
5.
6. 7. 8. 9. 10. 11.
12. 13. 14.
Identify the black spots (see chapter 2). Study the problems and deficiencies for each identified black spot (see chapter 3). Find suitable countermeasure(s) for each spot (see chapter 4). For each spot there should be at least one low-cost alternative and one higher-cost alternative. Estimate safety effects, costs of investment and changed maintenance costs, etc. for all potential countermeasures (see chapter 5). If the countermeasure is a high-cost option, all relevant effects and costs should be estimated. Set monetary values to accident and casualty reductions (and if necessary to other relevant effects), decide on suitable life-time of interventions, discount rate and tax factors, etc. Estimate BCRs for each alternative intervention/design for all black spots. Select the alternative with the highest BCR for each spot. Arrange in falling order of magnitude of BCR. Define the budget limit and determine which black spots could be included in a first draft action plan. Determine the BCR for the last included black spot (the one with the lowest accepted BCR) within the budget frame. Check if any high-cost alternatives for the included black spots yield higher marginal BCR than the BCR obtained for the last included spot (step 10). If this is the case, replace the low-cost alternatives for those spots with the more efficient solutions. Change the order (step 8). Check 9 to 11 again until there are no high-cost alternatives yielding higher marginal BCRs than the BCR for the last included black spot. Consider regional aspects in order to obtain a suitable balance in allocations between KGM regions. Compile the final plan for implementation of the improvements.
If CEA is used the principle will be as follows: 1-4. 5. 6. Same as above. Establish a weighting scale for fatalities, injuries and property damage only accidents. Estimate the Effectiveness/Cost Ratio (ECR) between benefits expressed in number of saved lives, injuries and accidents, and investment costs, adjusted for changed maintenance costs and tax factors. Select the alternative with the highest ECR for each spot. Arrange in falling order of magnitude of ECR. Define the budget limit and determine which black spots should be included in the first draft action plan. Determine the ERC for the last included black spot (the one with the lowest accepted ECR) within the budget frame. Check if any high-cost alternatives for the included black spots yield higher marginal ECR than the ECR obtained for the last included spot (step 10). If this is the case, replace the low-cost alternatives for those spots with the more costeffective solutions. Change the order (step 8). Check 9 to 11 again until there are no high-cost alternatives yielding higher marginal ECRs than the ECR for the last included black spot. Consider regional aspects in order to obtain a suitable balance in allocations between KGM regions.
69/81 December 2001
7. 8. 9. 10. 11.
12. 13.
Black Spot Manual
SweRoad
ANKARA
14. Compile the final plan for implementation of the improvements. Before implementing the plan (see chapter 7), a detailed program for follow-up and evaluation should be developed and implemented (see chapter 8). The above proposed procedure gives answers to which black spots should be improved and which intervention/design should be selected for each site. The described method and steps will in principle lead to an optimal allocation of funds. If it is considered to be too difficult to follow the outlined steps, it could be possible to separate the black spots into sub-groups, allocate a certain budget to each group, and then prioritize within each of these sub-groups. The groups could, for example, be based on road sections versus junctions. When there are different alternatives for each site, it should be considered that low-cost alternatives in most cases tend to give higher BCRs that higher-cost alternatives. As there are normally many black spots in need of funds for improvement, the leading principle should be to use low-cost alternatives for all sites in order to be able to eliminate as many black spots as possible. This principle will normally yield the highest total NPV within a certain budget frame. The only case when a more costly alternative should be used is when the marginal BCR for the more expensive alternative exceeds the BRC for the best alternative use of funds for other sites. To consider regional aspects is difficult. The priority list obtained after step 12 is theoretically the best. It will yield the best safety improvement in relation to the cost. On the other hand, it is perhaps necessary to consider the need to allocate some resources to every region. The principle for this should be decided in advance. One key to such a distribution could be the number of accidents on state roads in each region. Other keys could be road length and/or vehicle-kilometers traveled in the region. A simple method would be firstly to allocate a certain, limited amount of funds to all regions and then distribute the remaining (major) part according to the priority list. In some cases the listed black spots can also be included in a road rehabilitation program. In such cases it has to be checked if the black spots should be deleted from the black spot list or if they should be improved to a limited extent awaiting the more extensive improvement through the rehabilitation program. Concerning the question in what order and when the selected interventions should be carried out, the implementation should be carried out in the order indicated by the list of projects arranged in the order of falling BCRs. However, it must be said that the order of priority and the time for the implementation is normally not that important (yearly budgets and priority lists are assumed). The most important thing is that the most urgent spots are included in the list and that the selected type of intervention/design for each site is suitable to improve safety and that it is cost-effective. Simplified CB-calculation methods and values have been developed in an Excel sheet that can be used by KGM for black spot appraisals. The Excel sheet is enclosed to the previously mentioned report. SweRoad recommends that the proposed methods and values should be used in the short-term perspective.
Comment [l3]: pr
SweRoad
ANKARA
In the long-term perspective, the proposed methods and values should be checked and also changed, if necessary. This concerns, for example, monetary values for accident and casualty reductions and travel time savings, weighting factors for fatal accidents, injury accidents and property damage only accidents, and tax factors.
Black Spot Manual
71/81
December 2001
SweRoad
ANKARA
Implementation
The proposed sites should preferably be subject to a road safety audit to ensure that the proposed intervention is a suitable and effective safety measure. It should also be checked if the reconstruction can be coordinated with other construction works near the site. The work should be planned so that the construction work is carried out during timeperiods when existing traffic is disturbed as little as possible. In addition, the work zone should be arranged in such a way that existing traffic is disturbed as little as possible. In practice this may be conflicting with the safety of the road workers. To ensure their safety, it is often necessary to have harsh restrictions on passing traffic. All road works are dangerous since all road users are not familiar with the situation. It is therefore necessary to have proper warning for work zones. It is important to give good guidance to road users passing the zone, especially at night and during other conditions when there is bad visibility. Speed reducing measures must be applied to ensure low speeds for cars passing the site. Changed speed limits are often not enough. Physical measures have to be used. The safety of the workers has to be guaranteed by barriers so that inattentive drivers do not hit the workers. The accident situation must be monitored during the whole construction period to ensure that the situation is under control. When the work is finalized, another safety audit or inspection should be made before the road is re-opened for traffic.
Black Spot Manual
72/81
December 2001
SweRoad
ANKARA
8
8.1
Follow-up and evaluation

Background
It is necessary to follow-up countermeasures in order to gain knowledge about what has actually happened. The aim is to show if the investment gave good value for money and if the results in safety terms were good or bad. The purpose of this chapter is to show how follow-up and evaluation could be carried out, to discuss some important aspects and to show possibilities and limitations in the follow-up and evaluation process.
8.2
Planning of follow-up
The follow-up must be planned in advance. This is very important since once the countermeasure is applied, it is too late to make any before measurements on conflicts and speeds, etc. It often happens, that the follow-up is thought of after the site has been changed. Before - after measurements are much better than having to measure afterwards only. The follow-up should be connected to the problems that are to be solved. Before measurements could have been made as part of the diagnosis. In that case, similar measurements should be repeated in the after situation. During the after measurements, as many factors as possible should be unchanged. If before measurements cover peak traffic, after measurements should also do so, etc.
8.3
Documentation of countermeasures
The applied countermeasures must be documented in order to make a follow-up possible. The information needed is simple and limited. It is also easy to collect. But it must be observed during the implementation process, because later it will be more difficult to collect and in some cases also impossible. The documentation should contain characteristics of the road and traffic before the site was rebuilt or new equipment installed. The documentation should contain the dates when the implementation started at the site and when it was finished. This defines the end of the before period and the start of the after period. The implementation period should normally be excluded from the before and after periods. The implementation period is of interest as it is often a very complicated period from a traffic point of view. There is a danger that this complicated situation could cause accidents. It is therefore useful to analyze accidents during this period for a number of construction sites. The aim is to see if construction works are creating hazards and to study if applied procedures for signs or markings are good enough to give the right information
Black Spot Manual
73/81
December 2001
SweRoad
ANKARA
to the road users. This is a different aim than to follow-up the countermeasures. Since this manual is about the latter, the former will not be discussed further in this chapter. The exact location of the part that is improved shall be stated. From km. and meter to km and meter If the improvement means that the length of the road is to be changed, the new distances must also be given. The locations decide what part of the road should be followed up. The countermeasure should be described. The cost of the countermeasure must be specified if the real cost-benefit of the countermeasure should be calculated. All this information could preferably be stored in a file or a computerized register to be easy to retrieve. This file can also include information on where to find more detailed information, like drawings and photos on the site.
8.4
Target/result-oriented way of planning
The following figure describes an target/result-oriented way of working with road safety.
Resource goals Process and pe rformance goals/targets Common status goals /targets Overall goals/ targets
Resources
Main actors work Other actors work
Application status in the road transport system
Social and economic consequences
Feedback of experiences
Follow-up is the basis for the feedback and it should be carried out for all parts in the chain of activities. For black spot follow-up, implemented countermeasures as well as the last two boxes in the figure above are of special interest. Common status goals/targets deal with behavioral measurements. This is followed up in the initial monitoring, but can also be used in the long-term evaluation. Overall goals/targets deals with accidents and casualties. This part is most important in the longterm evaluation, but can also be used in the initial monitoring.
Black Spot Manual
74/81
December 2001
SweRoad
ANKARA
8.5
Initial monitoring
It is important to make a follow-up soon after the countermeasure has been applied. The aim is to see if the countermeasure is working as planned and that no new hazards have been created. One purpose is to find out if the road users understand what has been made and if they behave in a suitable way. Also, it is important to see if the problems that were supposed to be solved actually were eliminated or at least reduced. If there are measurements of speeds, conflicts or other measures of behavior before the countermeasure was applied, these measurements should be repeated in the after-situation. However, such measurements are often difficult to transform into safety effects. An expected change in the measured variable can tell us if the countermeasure was effective or not, but not how effective it was. Conflict studies can say more about the changes, but even conflicts are difficult to transform into accident and casualty changes. If accident data are available, they are normally of little use in this phase, since the time period is too short to reveal any significant changes. It might be advisable, however, to closely follow the reported accidents. This can be an early warning system to show if the situation has become worse than expected. This could be considered a bit pessimistic, but experience shows that road users can react in a completely different way than engineers or other safety specialists expect them to do. They are human beings and as such adjust to changes. So, therefore, one should be prepared for unexpected and bad things to occur, even though in most cases just good things happen.
8.6
Long-term evaluation
Long-term evaluation takes place during a longer period and comes after the initial monitoring period. The aim is to estimate the effect of the countermeasure. If there are before studies for speeds etc., such studies could be repeated one, two or three years after the implementation, if they have been repeated as part of the initial follow-up. Accident data, however, is more important in this stage than in the initial monitoring.
8.6.1 Changes in the environment There are always changes in traffic and traffic environment. The traffic volumes are normally increasing. The composition of the vehicle fleet is changing, cars get stronger engines and more safety equipment. Weather, traffic safety awareness among people, everything changes. All these changes influence the road safety at the sites. All of the changes from before to after cannot be attributed to the applied countermeasures.
Black Spot Manual
75/81
December 2001
SweRoad
ANKARA 8.6.2 Matched pairs
In theory, the best way of taking care of all environmental effects is to have matched pairs in a statistical experiment. This is done by matching possible sites two and two so that the two in every pair are as similar as possible. Then one site in each pair is selected at random and improved and the other one is left unchanged. This theoretically best way is difficult to achieve, since it is often hard to convince road authorities to leave sites with big problems untreated, even if the sites will be untreated for a few years only, and even if this will increase knowledge and be worthwhile in the long run. 8.6.3 Control groups
To use control groups or control sites is a common way to control for all changes in the environment. That is why it is called control. The general idea behind this is to select sites that are similar to the adjusted sites but where no countermeasures have been applied. The changes in accidents for these sites, from before to after, is said to be due to all the changes that have taken place in the environment. It is then assumed that the improved places, had they not been improved, would have had the same development. Example: The number of accidents in a control group was 200 in a before period and 180 in an after period. This 10 % decrease is therefore expected also for the improved sites if nothing had been made to them. So any decrease smaller than 10 % for the improved site is in fact not an improvement but a deterioration. Suppose instead that the number in the control group increased from 200 to 220. If the treated site had the same number of accidents before and after this means a better situation than expected, that is, an improvement. 8.6.4 Deciding sizes of control groups
The use of control groups is necessary for good estimates. It does however increase the variance of estimated effects. Higher numbers give smaller increases in the variance. The control groups should therefore be large enough so that the increase in the variance is not too large. There are different practical rules giving examples of the size. Two such rules could be mentioned:

The accident numbers for control groups should be at least 10 times the numbers at the changed site. The accident numbers for control groups should be at least 200.
The effect and reason behind such practical rules can be seen in the variance formula in the section To estimate the effect (see below). The rules are recommendations and not absolute requirements. As important, or maybe even more important, as having a low variance, is to avoid bias. There will be bias if the accident situation in control groups differs from that in the changed sites. So selecting representative control groups is more important then selecting large groups. The size of the variance can be estimated, but bias can in most cases not be estimated.
SweRoad
ANKARA
8.6.5
Test of independence
Chi-square ( 2) test is a test that is widely used to test independence. This is a test to see if the change in accidents was due to the treatment or if the change could have occurred by chance. The test variable is noted as 2. The test does not say anything about the size of the effect, just if the outcome could be due to randomness or not. It is easy to use and a way to get an initial description. It can be described by using the notation from Microsoft Excel. Excel includes a function that calculates the probability that the outcome could happen by chance. The function is called CHITEST. Example: A follow-up has given the following numbers of accidents.
Before After Total Improved sites 20 16 36 Control sites 200 220 420 Total 220 236 456
The expected numbers have to be calculated first before the CHITEST can be applied. The expected number in each cell is row total multiplied with column total divided by the total number. Thus, the expected number in the first cell is 220*36/456=17.37. All expected values are shown in the table below.
Before After Total Improved sites 17.37 18.63 36 Control sites 202.63 217.37 420 Total 220 236 456
These values are used in Microsoft Excel and the function CHITEST tells us that the probability to have an outcome like this or even more extreme is 66 %. This is much more than the 5 % needed in order to say that the countermeasure was successful. So there is no significant change due to the countermeasure. It is easy to use Microsoft Excel for calculations because it gives the probabilities directly. Other calculations give a value that has to be compared with values in a 2 -table before it is possible to get the probability and say if it is significant or not. Below is a printout that shows how the function CHITEST works.
Black Spot Manual
77/81
December 2001
SweRoad
ANKARA
8.7
To estimate the effect
When the previous test gives significant results, it is interesting to estimate the effect. Even if there is not a significant result, it is interesting to estimate the effect and to get confidence intervals for that effect. The statistical calculations will not be shown here since this is not within the scope of this paper. The notation has also been simplified to gain easier understanding. The theory can be found in different statistical textbooks, but the best references may be the works by Ezra Hauer. The formula for estimation is shown below. K, L, M and N are the number of accidents.
Before After Improved sites K L Control sites M N
The critical assumption behind this method is that the changes in the improved sites, had they not been improved, would have been the same as the changes in control sites. This assumption makes it necessary to be very careful when selecting the control sites, so that they are as similar as possible to the improved sites. We would expect K*(N/M) accidents to happen in the after period at the improved sites, if they had not been improved, or if the countermeasure had no effect. But they were improved and the actual outcome is L. Thus L/(K*(N/M)) is an index of effectiveness. Let us call this . Then 1- is the actual effect. Consider the same example as before:
Black Spot Manual
78/81
December 2001
SweRoad
ANKARA
Improved sites 20 16

Control sites 200 220
Before After
The expected number of accidents during the after period is: 220 20 * 200 = 22 This can be compared with the actual number 16. The effect is estimated to be: 1-16/22 = 27 %, that is, a 27 % decrease in accidents. Variance of the estimate is approximately estimated by: var (LM/KN)2 (1/L+1/K+1/M+1/N) (1- ) 1.96* var (1- ) gives the endpoints of a 95 % confidence interval for the effect. Since the variance for (1- ) is exactly the same as the variance for the endpoints are: 0.272727 0.497983
A 95 % confidence interval for the effect is between -0.22526 and 0.77071. Minus means an increase in accidents and plus a decrease. Since the interval covers zero, the conclusion is that this outcome could have occurred by chance. The result is that the point estimate of the effect is 0.27 (27 %) and that a 95 % confidence interval is between -0.23 and 0.77.
8.8
Short-term versus long-term effects
It is not uncommon that the initial effects are not so good as the long-term effects. One reason for this is that the road users can initially be unfamiliar with the new road site and its design. This could lead to accidents. When the road users get used to the new site, fewer accidents can happen. To avoid complicating the long-term effect estimations, the first time after the opening should be treated separately in the after period follow-up.
8.9
Regression-to-the-mean
Accidents occur at random. For example, if there are 100 sites, there are always some sites that have high numbers of accidents. The reason could be that the sites are dangerous. But there could also be high numbers because the random fluctuations were unfavorable at those sites. Then the number of accidents will decrease the next year even if nothing is made at these sites. The number of accidents tends to come closer to the mean. This is why this effect is called regression-to-the-mean. Sites selected for countermeasures normally have high accident numbers. So when the sites are improved, the effect will be overestimated during the follow-up. There are, however, ways to correct for this. One simple method is to exclude the year with the highest number
SweRoad
ANKARA
of accident in the before period when estimating the effect. So the number of accidents in the before period is estimated to be fewer than they in fact were.
8.10 Accident migration

Accident migration means that accidents are transferred to other places, normally road sections adjacent to the improved site. On these sections the accidents could increase and this would mean that the total effect of the countermeasure is decreased. Example: A straight road section is followed by three rather sharp curves. Many single accidents happen in one of the curves, the first curve for vehicles coming from one direction. Drivers have not adjusted the speed enough after a long straight road section. This curve is improved but not the other two. Then accidents could migrate to the next curve, which is now the first sharp curve, etc. Accident migration could also occur if drivers adapt to new behavior at improved sites and continue with this behavior onto old sites. Example: Speeds are often higher on new road sections. The drivers continue to have a high speed when passing onto old (unimproved) road sections, a speed that is higher than it was before. This would lead to more accidents on the unchanged road. One way to control if migration effects are present is to follow-up the accident situation on adjacent sites at the same time as on improved sites. Adjacent sites shall not be included in the control group as this could lead to an overestimation of the effect.
8.11
Strange results
Follow-ups sometimes show many strange results with accidents increasing after an improvement. When this happens, it is a strong warning signal to the engineers that something is wrong, even if it is mentally difficult to accept that the new spot or section is unsafe. But drivers and other road users are perhaps not behaving in the way road safety experts supposed they would. 8.11.1 Higher speeds
Many improvements are nice geometrical solutions that have the disadvantage, from a road safety point of view, that they make it possible to increase speed. Since speed is a very important factor behind severe accidents this is a warning signal. Increased speeds do not necessarily need to increase the number of accidents and casualties but every countermeasure that could increase speed must be examined very closely.
Black Spot Manual
80/81
December 2001
SweRoad
ANKARA 8.11.2 Subjective risk
Road users could also have wrong information. He or she could think that certain behavior is safer than is actually the case. This means that the subjective risk is lower than the objective risk. Example: Research in Sweden has shown that crossing the road at a marked pedestrian crossing, at least in Sweden, could be more dangerous than passing at places where there were no marked pedestrian crossings. The reason is probably that a pedestrian crossing outside a marked crossing is more careful than one passing at the pedestrian crossing.
8.12
Change in under-coverage of accidents
It is commonly known that all accidents are not reported (under-reporting). If the reported percentage is changed, this could of course affect the assessments made. The existence of a road safety project could change the reporting. When there is more focus on reporting from the police headquarters, reporting is likely to increase. The use of an additional reporting form makes the procedure more time consuming. This could have the effect that reporting decreases. However, changes in reporting do not matter for the results if the changes are the same at the improved sites as in the control group. This is an important factor to consider when selecting the control sites. One situation exists, however, where a change in the reporting could affect the evaluation. If the reporting is increased, it is normally increased more for accidents that have the lowest coverage. That is property damage only accidents. If a countermeasure is expected to reduce severe accidents more than slight and damage only, increased reporting could create problems. One solution would be to separate the result of the follow-up into different severity classes, and to make conclusions based on this separation.
8.13
Before and after periods for accident data
When using control groups it is necessary that these have accident data from the same time periods as the data in the improved groups. But it is not necessary that the before period has the same length as the after period. It is, for instance, possible to have three years before and one year after for a preliminary follow-up. This could later be followed by a three year after period using the same before period.
Black Spot Manual
81/81
December 2001

Black Spot Manual

Uploaded by

Copyright:

Available Formats

Black Spot Manual

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Black Spot Manual

Uploaded by

Copyright:

Available Formats

GENERAL DIRECTORATE OF HIGHWAYS

ROAD IMPROVEMENT AND TRAFFIC SAFETY PROJECT

TRAFFIC SAFETY PROJECT