Kamalasudhan A, Mitra S, Bo Huang and Chin H C
Department of Civil Engineering,
National University of Singapore, Singapore
Accident data, collected for many years, serve as the ground base for programs designed to reduce the number of traffic accidents. These accident databases are usually in the form of linear record file system, which enable an extensive amount of research to be undertaken using statistical methods. But both the databases as well as the analyzed information lacked visibility, which is essential for better understanding and good decision-making (1). Further many of the required analyses have a strong locational element, and as such they may suggest some form of geographic computer based data management system. GIS has been identified as an excellent system for storing and managing these types of data and also as a potential tool for improving accident analysis process. One of the reasons is that it provides an efficient system of linking a large number of disparate databases and also provides a spatial referencing system for reporting output at different levels of aggregation. Even though traffic safety seems to have many easy and logical connections to GIS, and its applications have even been proposed earlier (2), the development of a useful GIS database proved to be more difficult than anticipated. Works are limited to just visual interpretations of displayed aggregate accident data (3).
Traffic safety programs can provide benefits only when effective countermeasures are applied to the locations and areas that really need safety treatment. Hence the practice in analyzing accident data has always been preceded with identification of high accident locations followed by identification of factors and selection of countermeasures. Several techniques for identification of accident locations have been established (4), but GIS has only been applied to display such locations (5) and to analyze problematic cases (7).
The fundamental objective is to determine the factors that contribute to accidents at those spots and to take actions that will reduce crash frequency or severity. Determining such factors like roadway features and road user characteristics is usually done through a macroscopic study involving a large accident database (6). However making inferences from a single site, solely based on historical data does not always yield consistent results, because accidents do not usually occur at the same specific location. They may be distributed over an area although they may be caused by factors in a specific location. Hence it is more appropriate to identify accident-prone areas, with a subset of locations having high number of accidents instead of accident spots. Identification of accident-prone areas is best done on a GIS platform, which can facilitate further investigation of accident causation.
The goal of this study is to identify accident-prone areas using GIS tools. Specifically this involves plotting individual accident locations, identifying accident-prone areas using GIS’s spatial analysis tools, and presenting a diagnosis of accident characteristics by different types at these high accident locations.
The accident data used in the study are taken from the National Road Accident Database, while the expressway data have been collected using GPS. A brief description of the expressway and accident data is provided below.
As a digital map database of Singapore is not available the expressway profiles are obtained through GPS mapping. There are eight expressways in Singapore covering a total distance of 148km. These expressways provide uninterrupted high-speed travel (Max 90km/hr) for motorists with an average speed of around 50km/hr even during rush hours.
This spreadsheet file included all the accident records for the years (1992, 1994, 1996, 1998 and 2000). The street ID in this tabular record is used to filter the accident records that occurred on expressways alone. This file included information such as the location where accident occurred in local grids, type of accident, type and severity of injuries, date and time of occurrence, type of road surface, the number of vehicles involved etc. The general expressway accident statistics account for nearly 25% of the total number of accidents. Moreover the number of accidents has doubled from the year 1992 to 2000 from 635 to 1380 respectively.
Fig. 1. Location of all crashes: Singapore Expressways in 2000
The accident data in local grid coordinates are converted to latitudes and longitudes, and exported as X, Y coordinates and overlaid on the expressway layer. One problem in accident analysis is the quality of data. The records are based on the police-reported accidents and these may be wrongly reported. However these can be easily seen on the GIS maps. Some modifications may be applied to correct wrong entries. An example of an accident distribution is shown in Figure 1.
In this study, the analysis has not been confined to any particular class or type of accidents and rather towards utilizing the open nature of GIS in data manipulation. Hence it was decided to consider all the crashes initially and determine accident-prone locations by developing density maps.
Density maps are produced for all the years that are considered from 1992 to 2000 with the input data as individual accident points. Density maps show the highest concentration of a particular type of accidents. They are useful for looking at patterns rather than at locations of individual features. It creates a continuous raster surface from a set of input point features representing a magnitude per unit area, in our case the number of accidents per square kilometer. Hence by density maps the places where accidents are clustered can be easily distinguished.
Fig 2. Identified Accident-Prone Areas: Singapore Expressways
Identification of accident-prone areas
From the density maps produced for different years, the accident-prone areas can be easily distinguished. In order to get the real locations wherein the accidents are highly clustered all through these years, the raster layers are added up together. Thus the summed up layer gives us a density map showing accident density distribution.
For better localization of the potential areas the resulting layer is reclassified i.e. producing a different raster layer by modifying classification range, color coding etc. Reclassification makes it easy to understand the distributions and also to make decisions from the displaying map. Figure 2, shows areas on Singapore expressways where accidents are highly clustered.
As these areas have been determined based on 5 years of data, it can be argued that these areas may have certain factors which contribute to accidents recurrence there. The accidents may be clustered over an area but these can be caused by factors in a specific location. Hence the next step after identification of high accident locations is to perform a detailed safety analysis on these areas. Usually much less thought has been given to the safety analysis stage that follows.
However it will help to determine abnormal accident pattern and would lead to selection of sites for appropriate remedial measures for the improvement of safety. This analysis of collision trends of specific accident types can be flexibly done and visually displayed using GIS.
Based on the available attributes of the accident data it has been observed that the distribution of accidents varies depending on the time of a day, the type of road surface at the time of accident, the type and number of vehicles involved in the collision and so on. The following section describes how GIS software may be used to highlight specific accident types.
Fig 3: Accident-Prone Areas: Spatial distribution of accidents by road surface
Accident based on type of road surface
Pavement surface characteristic is one of the important factors, which determines the safety of a vehicle when negotiating a curve or at the time of sudden application of brakes. Present study reveals that major percentage of accidents occurred at both accident-prone and rest of areas when the road surface condition was dry. But comparing the percentage of accidents which occurred when the surface was wet it has been found that the accident-prone areas contribute more, leaving us the clue for viewing such accident location distributions to look for the factors and take countermeasures.
Accident based on vehicle involvement
The accidents are also grouped based on the vehicle involvement and it has been observed that around 58% of the expressway accidents were multiple vehicle accidents and the rest were single vehicle accidents. Further the collision of a vehicle with stationary objects has been more in accident-prone areas. On the other hand even though most of the multiple vehicle accidents are caused as a result of head to rear type of collision, the percentage of sideswipe collision is found to be more in accident-prone areas, which needs to be focused. Hence analysis of specific types of accidents may reveal important geometric and traffic control and regulatory factors which are responsible for their causation.
Further while investigating the distribution of accidents over different times of the day, it is found to be almost uniform over the evening peak and night time but slightly higher during the morning peaks. Fatality percentage is clearly higher during night. When considering the speed limit there is a clear demarcation between the accident percentage contribution and the other lower speed limit zones. These types can also be visually studied for their distribution patterns. Framing various combinations of above accident types can still improve the analysis. They help to localize accidents and thus arrive at some more useful conclusions. Further the determination of accident-prone areas can well be approached using only accidents that result in personal injury or death, as the cause of such severe accidents cannot be reasonably attributed to randomness. Hence by proceeding in this way and having the information about the spatial distribution of accidents it is possible to model the accidents with different geometric, traffic control/ regulatory factors and environmental factors that might be responsible for a particular type of accident.
Fig 4: Accident-Prone Areas: Spatial distribution of accidents by vehicle involvement
Conclusion and Recommendations
The GIS is an effective tool to display different types of spatial accident distribution on digital road network. The use of GIS enables relevant accident data to be quickly processed and displayed on a map. GIS has also been used as a tool to identify hazardous locations along the expressways depending on the historical road accident data. These in turn help to improve the safety of road by advanced planning and maintenance of the so-called accident-prone areas. The expressway database can be manipulated by adding roadway geometrics, information about traffic control and regulatory factors, flow data at different stretches of those areas. Information thus attained will consist of the data that can be easily accessed or referred to at the time of making crucial decisions in relation to the type of improvement to be introduced in a particular area.
- Alind Saxena, Ganesh Babu., R. K. Bajpai, S. M. Sarin. GIS as an Aid to Identify Accident Patterns, Map India 2002.
- Mwatelah, K.Z. Applications of GIS to Analyze Causes of Road Traffic Accidents- Case Study Kenya. International Conference on Spatial Information for Sustainable Development, 2001.
- Florido, L., Mussone, L. and Rinelli, S. An Analysis of Urban Vehicular Accidents through A GIS System in order to Monitor and to Increase Safety. Proceedings of the 6th International Conference on Computers in Urban Planning and Management, 1999.
- Hauer, E. Identification of Sites with Promise. Transportation Research Record 1542, TRB, National Research Council, Washington,D.C., 1996, pp. 54-60.
- Chin, H.C., and Quddus, M.A. The Use of Geographic Information Systems in Traffic Accident Analysis. Proceedings of the 8th International Conference on Transport and Logistics, 2001.
- Peled, A., Haj-Yehia, B. & Hakkert, A.S. ArcInfo- Based Geographical Information System for Road Safety Analyses and Improvement. ESRI User Conference Proceedings, 1996.
- Miller, J.S. Geographic Information Systems-Unique Analytical Capabilities for the Traffic Safety Community. Transportation Research Record 1734, TRB, National Research Council, Washington,D.C., 2000, pp. 21-28.