Traffic accident is an event that cannot be in the predictions and involve unintentional vehicle with or without other road users, resulting in human casualties (suffered minor injuries, serious injuries, and death) and property loss. According to Tanariboon and Satienanam (2005) intended to increase road safety requires a comprehensive approach, involving all components of the accident: human, vehicle and road environment. According to Jasa Marga Indonesia Criteria traffic accident victims are:
1. Injury is the state of the victims suffered minor injuries which did not endanger life and / or do not require further help or treatment in hospital. For example, a small wound with little bleeding and the victim are conscious, burns, sprains from mild limb without complication, the patient were conscious not unconscious or vomiting.
2. Injury is the state of the victims suffered severe injuries that can life-threatening and require help / further treatment at the hospital immediately. For example, injuries that cause the state of the patient decreases, usually wound on the head and shaft of the head, limb fractures with complications accompanied by great pain and heavy bleeding, collision or injury of patient body causing damage to equipment inside.
3. Death is a state where there are people with signs of physical death. The victim died was an accident victim who died at the scene, died during the trip to the hospital, or died while hospitalized.
Otherwise, Austroads(2000), mentioned that some models are proposed to be the cause of accidents in road safety literature. This model generally describes crashes as a result of driver behavior that is not true according to the demands of the environment or to the characteristics of road vehicles, or to both. Accidents can happen when a driver's performance level is insufficient to meet demand for the environmental performance of road. Most of the time, the ability of drivers exceed the performance demands. If not, the three factors that can contribute to accidents in traffic operations is the human factor, the road environment and vehicle factors.
2.2 Exposure Analysis
Nichols and Lester (2002) mentioned to exposure rate in such cases are ââ‚¬Å“event ââ‚¬"based,ââ‚¬Â using total volume passing through the point. Exposure to vehicular and other conflicts that are susceptible to accident occurrence varies with many factors, including volume levels, road side activity, intersection frequency, degree of access control, alignment, and many others.
An accident on the highway can be mentioned as the incidence of negligence by the party concerned (in the aspects of road user behavior, vehicle maintenance and road conditions) or due to environmental factors (including natural disasters) that could result in collisions (including "out of control" case and the collision or victim in a vehicle against an object inside or outside the vehicle such as passenger buses.) involving at least between two moving vehicles where the damage or injury that occurs in every person, property, vehicles, structures or animals and recorded by the police.
Kao (1991) mentions the comparison between the crash site, the number of opportunities for accidents to occur, must be taken into account. Whereas if it remains the case then the observed differences in accident rates may be due to the occurrence of errors of differences in road design. for example, when the volume of traffic at one location may be higher than the other. The same type of analysis that is applied when compared to the experience of accident in two different time periods in one location. The accident exposure was of intent with the proportional number of opportunities for accidents. Accident exposure was expressed by the number of accidents; this figure is calculated by length of vehicle trips, (Million vehicle kilometers) as a reference.
2.3 Identification of black-spot locations
The definition and identification of hazardous road locations are closely related activities and are in some cases impossible to distinguish. From a theoretical point of view, however, it is possible to make a distinction between the definitions of a hazardous road location and the identification of it. A theoretical definition of a hazardous road location is any location that has a higher expected number of accidents than other similar locations as a result of local risk factors present at the location (Elvik, 2008)
(Nicholas J.Garber et al, 1999) mentioned hazardous locations are sites where the accident frequencies, calculated on the basis of the same exposure data, are higher than the expected value for other similar locations or conditions. Any of the accident rates or accident summaries described earlier may be used to identify hazardous locations. A common method of analysis involves the determination of accident rate based on the same exposure data for the study site with apparent high accident rates and several other sites with similar traffic and geometric characteristics.
A rather popular type of accident research is accident black spot analysis. It is indeed one of the most promising areas of traffic safety improvement, because it is directed to a very important aspect of safety: the architecture of traffic environment. Accident black spot analysis is also the best example to show the limitation of accident analysis. This regards detection and selection, problem analysis and evaluation (Dang, 2004)
An identification of black spot is the first step to improving road safety. It has followed by diagnosis of the selected spot, finding countermeasures, estimating effects, prioritizing, and implementation and last follow up and evaluation (SweRoad, 2001)
(McGuigan, 1981, 1982) proposed the use of potential for accident reduction, as the difference between the observed and expected number of crashes at a site given exposure. (Hakkert and Mahalel, 1978) proposed that black spots should be defined as those sites whose accident frequency is significantly higher than expected at some prescribed level of significance. (Mahalel et al., 1982) proposed the road sites selected for treatment should maximize the expected total accident reduction by treatment.
2.3.1 Accident Frequency
Using frequency as an index is the simplest method. Hazardous locations can be ranked according to a number of accident occurrences. This simple range by number of accidents is easy to explain. This method is practically to begin which requires small data fields. It serves the purpose indefinitely in many jurisdictions. Extensions of this method can be adopted by using number of victims instead of accident occurrences (Yaowaret J, 2006)
According to (Hauer, 1996) some researchers rank locations by crash rate (crashes per vehicle-kilometers or per entering vehicles), some use crash frequency (crashes per km-year or crashes per year), and some use a combination of the two. Furthermore, there is a wide range of methodologies available, ranging from simple models based on actual crash counts to advanced statistical models based on estimates.
2.3.2 Accident Rate
European Union Road Federation (2002): Crashes frequency (in this method only the number of crashes are considered); Hazard potential ratio (here, the number of crashes and the traffic volume are considered together); Joint method with crash frequency and crash risk ratio (this method aims at developing a selection of road sections according to their levels of crash frequency, to prioritize the measures to be taken depending on the road sectionââ‚¬â„¢s crash risk ratio); Confidence interval method (this method is based on the application of a statistical test with the aim to determine if the particular crash risk ratio of a road section is significantly higher than the mean value of this ratio); Method of the crash severity ratio (road sections are classified according to their gravity rate, which is calculated using weighted coefficients for different types of casualties); Risk rate method (this method defines for each of the different characteristics influencing the crash risk ratio of road sections a risk function: road sections that have a risk function value higher than a predetermined value are considered dangerous); and Inventory of the crash risk elements in the road (the purpose of this method is to identify road sections that present some high potential crash risk characteristics without having a particular high crash rate).
2.4.3 Accident Severity
The Accident severity methods are used to identify hazardous location based on the severity of accidents at each location. Each location could be transformed to a certain unique unit by multiplying an equivalency factor. The popular unique unit is equivalent property damage only. So, this technique is then called the equivalent property damage only (EPDO). The equivalency factors vary by area or assumption as describe in the following.
Zegeer (1982) state the formula as shown in Equation 2.1 was used to evaluate highway locations. The accident severity was classified into five levels according to definition of the national council. The accident severity levels are (1) fatal accident-one or more fatal injuries, (2) a type injury (incapacitating) accident bleeding wound, distorted member, or person carried from scene, (3) B-type injury (no incapacitating) accident bruises, abrasions, swelling, limping, (4) C-type injury( probable injury) accident no visible injuries but complain of plain, and (5) PDO accident property damage only.
EPDO = 9.5(F+A) +3.5(B+C) +PDO Eq. 2.1
F = number of fatal accidents,
A = number of A-type injury accidents,
B = number of B-type injury accidents,
C = number of C-type injury accidents,
PDO = number of property damage only accidents.
Kao (1991) assessed equivalency values for weighting severity of the traffic accidents in Bangkok, Thailand by using the results of economic study by JICA (1987). The EPDO is denoted in equation 2.2.
EPDO = 45F+4.5J+PDO Eq. 2.2
F = number of fatal accidents,
J = number of injury accidents, and
C = number of A-type injury accidents,
PDO = number of property damage only accidents.
Prince of Songkhla University and Asian Center for transportation Studies (2003) studied the traffic accident in Songkhla Province, south of Thailand, during the period of 1993 to 1994. The employed PDO is shown in equation 2.3.
EPDO = 4F+3Js+ (Jm+PDO) Eq. 2.3
F = number of fatal accidents,
Js = number of severe injury accidents,
Jm = number of minor injury accidents, and
PDO = number of property damage only accidents.
An alternative approach of severity index methods is called Equivalent Total accident Number (ETAN). This method includes all accident cases and causalities. The ETAN can be denoted as in Equation 2.4
ETAN = aF+bJ+TAN Eq. 2.4
ETAN = Equivalent total accident number on the site,
F = number of person who died on the site,
J = number of person who injured on the site
TAN = Total accident number on the site, and
A and b = calibration factors
However, this method assumed that the severity is equivalent unit values that multiply the victims and accident occurrence with equivalent factors. The severity value for section j (sj) is calculated as shown in equation 2.5
Sj = Wj(Fj) + Ws(Ij) + (PDOj) Eq. 2.5
Fj,Ij,PDOj = Number of fatal accidents, injury accidents and poverty damage only accidents at section j, respectively
Wf,Ws,Wd= Calibration factor
2.4.4 Rate Quality Control Method
Michael D. Pawlovich (2007) the rate quality control method identifies those sites where crash rate is greater or significantly greater than the average crash rate for similar sites across the state or similar region. Similar to the crash rate method, the rate quality control method adds some statistical control for determining the critical crash rate. The rate quality control method applies a statistical test to determine the significance of a site's crash rate when compared to the mean crash rate for similar sites. The statistical test applied is based on the Poisson distribution, the commonly accepted distribution for crashes. Use of the rate quality control method effectively eliminates sites with high crash rates but low exposures. Inputs for the rate quality control method, for identification of hazardous sites, include: average crash rate (per 100 million vehicle miles) for site category, crash rate at the site, and level of statistical significance. Determination of each site category's average crash rates must be done with care, considering the nature of the sites and their surrounding environment. Site categorizations must be carefully designated and each site then assigned to a particular category. Site categories can be developed using a variety of features, including: rurality, number of lanes, surrounding land use, road types, etc. The purpose of the categories is to facilitate comparison of site crash rates with like sites, to the degree possible. However, this categorization of sites can be taken to unreasonable limits. Therefore, limiting the number of categories to a number which is tenable (that is, neither too large to be unmanageable or that would reduce sample size below statistical reliability nor too small to adequately describe sites) is strongly advised. One suggested breakdown utilizes a combination of rurality of the roadway (urban or rural) and the number of lanes. The categorization utilized should reflect the question being addressed. After categories have been established, computation of the average rates for each category ensues. Many state transportation agencies calculate statewide averages for many categorizations. To compute the critical crash rate for a site, use the following equation:
Rc = Ra + k (Ra/M)1/2 + 1 / 2M Eq.2.6
Rc = the critical crash rate
Crashes per Million Vehicle Miles (MVM) or Million Vehicle Kilometers (MVK)m used
for Sections Crashes per Million Vehicles (MV) used for spots
Ra = average crash rate for the entire population of sites within the category
k = a probability constant, where the higher the value of k, the higher the value of the critical crash rate. Some common k values are:
k = 3.090 for a 99.9% level of confidence
k = 2.576 for a 99.5% level of confidence
k = 1.645 for a 95% level of confidence
k = 1.282 for a 90% level of confidence
M = millions of vehicle miles (or kilometers) for sections or millions of vehicles for spots Use of a high k value will result in a shorter list of critical sites but confidence that those sites are hazardous is increased. Critical crash rates for low ADT highways are higher because fewer crashes occur within low exposure sites. Also, the use of multiple years of crash data lowers critical crash rates due to the variability of crashes at a site over time.
Using the above equation, develop for each categorization a list of critical sites and order them by a Safety Index, which is simply the actual rate divided by the critical rate. The steps involved in using the rate quality control method are:
If not already done, locate all crashes in accordance with accepted coding practices.
Compute system wide average number of crashes per MV or MVM for each category of highway, based on total data for all sites of each category.
For each site, determine the vehicle exposure, M, during the study period.
Compute the critical crash rate, Rc, for each site within each category using the equation above.
Compute the actual observed crash rate at each site for the same time period.
Compare the actual crash rate with the critical rate for each site and prepare a list of all sites within each category with rates exceeding the critical value.
Compute the Safety Index for each site and rank the list for each category by the Safety Index.
As mentioned, the quality control methods utilize a statistical test to refine the decision-making
process involved in determining a site's hazardousness. Also, these methods allow agencies to
determine priorities by grouping locations according to their functional classification and rank within these classifications. Also, sites having higher crash frequencies than average for their category can be quickly singled out for special attention. Though this improves over the previous methods, it still has notable deficiencies. First, the statistical test utilized is somewhat ambiguous and suspect. The addition of the Poisson distribution probability constant adjusts the critical rate equation in order to limit the number of sites judged critical. However, the reasoning behind the use of this probability constant in the equation is somewhat unclear. Adjusting the critical rate by a standard deviation or two fits with standard statistical practice, but the third element in the equation (1 / 2M) has a less clear meaning. Additionally, the entire premise of crashes being distributed as per the Poisson distribution has been questioned in recent literature. The Negative Binomial distribution has, recently, been judged a better representation. This may not matter due to the simplicity of this equation and its intended use, but it might introduce some bias due to over dispersion. Finally, the choice of which k-factor value to pick is highly subjective, giving rise to possible ambiguity in results from year to year. Second, the method is quite data intensive, if simply because it needs to be in order to achieve the gains. For each site and site category the user must track several different types of data that wouldn't be needed under the spot map, crash frequency/density, crash rate, and frequency-rate methods. The categorization development process involves the subjective determination of categories through examination of site characteristics throughout the jurisdictional region. Many site characteristics are now in computerized databases but not all, thus requiring some data collection. Then, once the site categorizations have been developed, each site must be categorized and the method steps listed above must be run for each site within each category. Third, only crashes and volumes are included in the equation. While the categorizations address other types of data, as the categorizations become more refined, more data must be collected. Again, this might be the price of better refinement in list generation. Thus far, none of the methods have addressed the idea of including crash or injury severity into the determination of hazardous site ranking lists.
2.4.5 Conclusion regarding applications methods
After reviewing the significant literature of previous research, one well-known technique called RQCM is the most widely used to identify the hazardous locations, regarding to the available accident data, this research applies the accident rate, accident frequency, and severity index in RQCM to identify black spot locations in the study area.