ArcInfo - BASED GEOGRAPHICAL INFORMATION
SYSTEM FOR ROAD SAFETY ANALYSES & IMPROVEMENT
A. Peled(1), B. Haj-Yehia(2), A.S. Hakkert(3)
ABSTRACT
Road safety involves three major components: the road system, the human factor, and the vehicle element. Those three elements interlaced, and linked through georeferencing traffic events, are the basis of road safety analyses and improvement. The location perspective seems to be the most suitable methodology by which to analyze different traffic events. Geographic Information Systems offer an advanced engine to drive, both area-wide and location-oriented investigations. The possibility to raise and solve, easily, problems related to street segments, streets, intersections, and neighborhoods, may ease much of the labour-intensive production effort. Thus, more emphasis may be given to complex analyses and in-depth investigations. The paper describes an Arc\Info-based (GIS) road safety analysis system. This newly developed software package was designed for the Haifa Municipality in Israel, and may be adapted very easily to any other city. The package was tested successfully with accident data during a three-year period, and was adopted as the basic tool for road safety management, analyses and improvement. Work is being carried out now to add a "Before\After" analyses module. The modular design of the system enables to add, easily, additional modules too.
_________________________________________________________________
(1) Dr. Ammatzia Peled, Senior Lecturer, University of Haifa, Dept. of Geography, Mt. Carmel Campus, Haifa 31905, Israel. Tel: +(972)4-824-0148. Fax: +(972)4-8246-814
(2) Haj Yehia Basheer, M.Sc., Senior GIS Engineer, Pelled-GIS Mapping Ltd., 67, Albert Schweitzer St., Haifa 34995, Israel. Tel: +(972)48343-591. Fax: +(972)4-8343-763.
(3) Prof. A.S. Hakkert, Associate Professor, Transportation Research Institute, Technion City, Haifa 32000, Israel. Tel: +(972)4-8292373. Fax: +(972)48225716. E-Mail: Hakkert @ tx.technion.ac.il.
INTRODUCTION
Road safety is an important issue in most countries. It involves large numbers of people, vehicles, and road sections, and as such, is a complex issue to understand and manage. The interaction of the various safety system components dictates an interdisciplinary approach that almost always needs a factual basis. Because of the size and scope of the problem, the required data bases are generally large, require constant updating and changing, and need to be accessible to a wide variety of authorities.
Many road safety issues involve the road infrastructure and its associated activities and land-uses. Road safety relates also to other fields of activity, such as education, driver training, publicity campaigns, police enforcement, the court system, public health, and vehicle engineering. The environment can be both urban and rural. Some of the issues mentioned dictate a socio-economic and demographic approach, some an epidemiological approach, and for some, a locational approach is most suitable. The location-associated approach can either be general, being applied to a geographic area (macro), or be site-specific (micro).
Although the total number of road accidents generally continues to rise, in terms of accident rates, road safety is improving in many countries. Considering the complexity of the issue, and the ever-increasing difficulties in providing effective safety measures, large data bases and efficient means of analysis and management are essential [Lubkin and Maleck, 1989].
Although safety comprises a large array of activities, this paper will mainly concern itself with the road infrastructure, which is one of the large and complex components of the system.
Safety analyses are performed by various authorities -- both central and local, by research institutes, by companies, and by individuals. Almost invariably, data are needed for the analyses, and data bases are necessary to contain the information. Data handling has been carried out on computers since the 1960s, and since the 1970s-1980s, more and more of the data handling and management are performed on increasingly more powerful PCs and workstations. Many of the required analyses have a strong locational element, and as such, suggest some form of geographic computer-based data management system. The types of analysis conducted include periodic reports on a variety of aspects and specific queries. These analyses can be general, such as the severity of accidents by type of accident, by age of drivers, etc., or they can be location-oriented, such as the types of accidents at various locations on the network. On a more detailed level -- for engineering purposes -- in many cases, a collision diagram is constructed, presenting the types of collision at a certain location.
Accident data alone are not sufficient. To be interpreted, they must be placed in a context. They can be compared to another time period (trends); likened to a similar area or type of location (locational analysis); they can be related to the level of activity (rates). Only when placed in these contexts do accident data become meaningful and useful.
BACKGROUND
Until recently, most locational analysis was map-based and was conducted manually. Maps provide a clear and immediate visual impression of the accident distribution in a town or rural area, identifying those locations that have accident concentrations. The advantage of maps over the use of orderly accident lists lies in the fact that they present the geographic aspect of the accident distribution required for an understanding of the problem leading to useful solutions. Neighboring streets, sections, or intersections immediately become obvious, and can be associated with specific features, land-uses and activities.
The main disadvantages of the use of accident maps are the inability to duplicate maps easily, the lack of easy transferability, difficulties in indicating many accident classifications simultaneously, and the difficulty in updating the maps from time to time. In order to overcome these drawbacks, use is generally made of a computerized geographic data base [Peled, 1982; Peled and Hakkert, 1982; Prastacos, 1991].
Road safety analyses and improvements involve large diversified georeferenced data bases. The two major data bases are, of course, the road system layout and accident information. Many other data bases can be correlated to road safety, among which are: positional and physical road characteristics; traffic counts; intersection-related data, such as signals and signal programs; traffic signing; public transportation services and routes, etc.
Because of the difficulties associated with the merging of various data files, the sophistication of analysis can easily suffer. Studies such as the accident analysis of junctions with different kinds of traffic control require merging of at least four data files: accidents, physical data, type of traffic control, and traffic volumes. A study of the levels of safety on streets with public transport again requires various data files. Such studies are difficult to conduct on a comprehensive scale. Geographical information systems as described in the following sections become almost essential.
SYSTEM DESCRIPTION
Data in a GIS system for road safety improvement are generally stored in separate thematic layers. Each layer represents one theme of the overall system and comprises spatial information and non-spatial information stored in the system database, and sometimes in auxiliary files. All these are linked together to establish the georeferenced database in which each feature is represented by its location and attributes.
The separation into the specific thematic layers was established as a result of the extensive system analysis phase of the research effort. In addition to road safety analysis, the broader needs of a municipality also influenced the conclusions. For example, educational institutions and facilities were separated from other municipal institutes and landmarks. This was done to serve specific needs of the Department of Education, whereas for the Road Safety Analysis System, all these institutions could have been stored in one general thematic layer, with the proper identification as to the type of institution.
The second phase of the system analysis focused on the use, reconstruction, and accessibility to various data elements incorporated in the Road Safety GIS relating to the projected output, query handling, and data manipulation needs. For example, it was concluded that less than 5% of the queries relates to information on vehicles involved in the accidents. Less than 10% of the queries are connected to the specific information regarding accident injuries, and usually data analysis relates to the accident information. This leads to the establishment of different auxiliary files of vehicles and casualties, which in turn are linked to the major database accident files that are attached to the spatial information (positional) file. This methodology made it possible to implement most of the queries in a simple and rapid manner, where only in a "micro type" analysis does the system need to select the information contained in the auxiliary files. Although this type of query is more time-consuming, the additional execution time (to access the auxiliary file) is negligible.
The total implementation of this methodology resulted in an efficient system where the response time is brief compared to the complexity of the analyses and queries.
DATABASE STRUCTURE
The road safety database includes 13 thematic layers and 6 auxiliary files. These layers are divided into two major groups: feature-based layers and "graphic" layers. There are 6 feature-based layers, as follows:
Street Layer. The street layer comprises all the links (segments between intersections) of the entire municipal road network, including trails, steps, pedestrian streets, and non-paved streets. These were digitized from 1:5000 and 1:10,000-scale maps.
Non-spatial data of the street's links are grouped into five major categories:
1. Identification attributes -- includes: ID numbers for each link; street addresses at the start and end nodes; ID numbers for the start and end nodes;
2 Physical information -- includes: width of the pavement, sidewalks, separation, parking lanes, and road "right-of-way" (the formal cadastre road width);
3. Class definition -- includes: classification of pavement, street hierarchy, usage, traffic direction;
4. Traffic volume -- includes: Average Annual Daily Traffic (AADT) and year of observation;
5. Accident data -- includes: total number of accidents for the past five years; total number of accidents in current year; ratio of total number of accidents to million kilometers travel of passing vehicles.
Junction Layer. The junction layer is composed of all the intersections of all the segments stored in the street layer. This includes also "pseudo nodes" that occur when a street is split (topologically), due to a physical phenomenon such an intersecting pedestrian path or a pedestrian-operated traffic light.
Non-spatial data of the junction layer are also grouped into five major categories:
1. Identification attributes -- includes: internal node ID number; official intersection ID number as declared by the Ministry of Interior, ID number of two of the streets intersected at the junction;
2. Type -- includes: classification by physical type of intersection and type of intersecting streets;
3. Traffic volume -- includes: entering AADT and year of observation;
4. Physical and operational information -- includes: number of approach legs; type of control; operation of control;
5. Accident data -- includes: total number of accidents in the past five years; ratio of this number to the AADT; total number of accidents in current year.
Accident Layer. The accident layer includes all accidents observed within the city boundary. The accidents are correlated to the street network in order to calculate and define their position in a world coordinate system that complements the given street address.
The non-spatial accident data are divided into six major groups:
1. Identification data -- includes: traffic police ID number; date; time of day;
2. Location information -- includes: type of address and address. These correlated items enable the system to import and manipulate all types of location definition used in various road safety databases and analysis packages;
3. type of accident;
4. Road factor -- includes: engineering, traffic and environmental aspects that are recorded at the time of the accident, and may assist in-depth study of the event;
5. updating status -- serves to follow the status of data-gathering and the processing status of the accident information;
6. Auxiliary data -- includes: pointers to the relevant records in the vehicle and injured persons files that are involved in the accident.
Educational Institutes Layer. The educational institute layer includes all the institutes from primary schools up to university campuses. Non-spatial data for this layer include the institute type, major studies' branches, number of students, and street address.
Activity Centers' Layer. The activity centers' layer includes both centers that attract various segments of the population at given times, and also some known landmarks in the area. This combination is a result of the system analysis phase. The idea was to establish enough known landmarks that will serve as control and guide points to position accidents and other events where the observer failed to record the full street address. In the newly developed GIS system, an event may be positioned by verbal and textual remarks, such as "200 m. from the Metropolitan Museum, along 5th Avenue in New York City, NY."
Neighborhoods' Layer. The neighborhoods' layer includes the internal municipal borders of neighborhoods. In the first stage of the research, this layer was graphic-based only. It was upgraded from graphic to feature-based, to support the "neighborhood-competition" analysis module (see System Application chapter). The non-spatial data are: neighborhood ID, name, population, total street length, and traffic volume.
Graphic Based Layers. In addition to the feature-based layers that are extensive in terms of non-spatial attributes, and are much used in terms of queries and analyses, there are additional data and features that are used only as term references and identification framework. These layers are totally active, and may be approached as any other (feature-based) layer. Yet, due to their characteristics and usage, they were separated and defined as graphic-based only.
1. Street names -- includes textual graphic representation of the street names to be used as a reference in visual observation on the screen or for rapid map compilation. This is implemented once for every municipality, and should not be mistaken for the georeferencing capabilities of the system that may reproduce the names for each given or selected street;
2. Railway system layer. This layer may be regarded as graphic only, but is correlated to the street layer by the junctions only;
3. Shore layer: includes the shore line;
4. Boundary layer: includes the municipal boundary;
5. Frame layer: a typical graphic layer that holds the framework of any visual representation;
6. Coordinate system grid. Serves as a reference grid for the study. The grid's resolution may differ for each municipality;
7. Photologging Inventory layer. This layer serves for visual representation of all the locations that were recorded as digital images. Each of the icons presented on the screen represents an image that may be shown by pointing to the relevant icon.
Auxiliary Files. In addition to the eleven thematic layers, there are six auxiliary files that are correlated and complement various layers. As mentioned above, the methodology was to separate rarely used data from extensively used data. In the case of streets, the idea was to separate the general information of the streets, thus preventing an unnecessary duplication for each street segment (link), such as street name, for example.
Street File. The street data file stores information related to the streets as one unit. It is linked to the street segment layer through the link-ID number of the street layer. Beside the ID information that also includes the street name, there are accident-related data calculated for the entire street in the same fashion as described above regarding the street segment layer. This type of segmentation enables rapid response to queries related to street segments and whole streets, using only one geographical representation of the street network system.
Injury File. This file stores information on all casualties. It is linked directly to the relevant accident by a special pointer in the accident layer non-spatial attributes, and also by the accident ID number stored in the injured persons' files. This enables data retrieval in both directions.
Vehicle File. This file stores information on all vehicles and drivers involved in accidents. It is linked to the accident layer as described above in the injured persons' file section.
Traffic Volume Files. These two files store traffic volume counts in street segments and intersections separately. They serve to calculate annually the needed AADT values for each street segment and intersection. At present, it is envisaged to hold only current traffic volume data in this file, together with year of data collection. A historic traffic volume file, including previous traffic counts and dates, could be added as a separate layer.
Project Files. These files include information about road improvement projects, such as: project ID, name, "start" date, "finish" date, type of improvement, the links and intersections within its region and affected influence area.
Image Files. The digital images generated and represented in the photo-log layer are stored in a special directory in various raster formats as: SVF, Lattice, ERDAS, SUN Raster, and others. The system is able to identify the format and represent it when requested.
Development of Algorithms and Programs
The system comprises two main packages: an updating/editing module and an analysis/query module. The program's routines were written in "C" programming language and ArcInfo macro language (AML) in a UNIX environment. These two modules enable to conduct swift and easy analyses, aided by user-friendly menus.
Updating and Editing Module
At this stage, the updating/editing module is available only for the accident coverage. This application has the possibility to add new accidents from a file, add a single accident interactively, and update single accident parameters. Moreover, an option was built to remove a single accident. It is possible to locate accidents at an intersection or at a street link by converting all the accepted forms of addressing. It is also possible to locate accidents by using the remarks and other parameters, or by a temporary position if the address is incomplete. A quality control and zero check processing are implemented while entering data, such that the system prevents feeding incorrect data to the database.
Location Module. A location module has been developed which can greatly improve the locational coding of police-reported accidents. Once the road network has been digitized, a large number of land-use and landmark features were added, including schools, institutions, shopping malls, cinemas, hospitals, fuel stations, etc. The police officer coding the accident can call up the section of road in question, which then appears before him with all associated features. He can then locate the accident accurately, although the section has no house numbers. Locating accident by house number or intersection are the type of coding generally used in Haifa.
Special tools are given to the user in the "location module" for such cases. By marking the relevant street, a two-coloured ruler is shown along its path. The length and number of the ruler's segments are controlled by the user. The start point may be defined at an intersection or an arbitrary point, such as in the proximity of a landmark. When digitizing the street network, a special landmark layer was also generated. This layer comprises schools, known institutions, shopping centres, malls, hospitals, fuel stations, theatres, cinemas, libraries, etc. The user may call up this layer, and by marking the relevant landmark, the system will automatically calculate the closest point on the relevant street, and the ruler start point will be positioned on that very location.
Queries and Analysis Module
The analysis/query module provides a wide range of outputs, such as: data retrieval; standard periodic reports; hazardous location analysis; Before and After Analysis; "neighborhoods' competition" analysis; area-wide analysis; display of data upon request and creating maps with required detail and scale. The analyses are performed according to three levels of complexity, demands and users: (1) An easy query and analysis by using default parameters; (2) A fixed analysis with options to choose thresholds, criteria, etc., and (3) producing flexible reports and maps.
This module was developed in a modular manner, to enable joining other modules of analysis in the future. The "Before and After" Analysis and the "neighborhoods' competition" analysis modules were added in the second phase of this research. This was made possible by the "modular strategy" under which the research effort was conducted.
SYSTEM APPLICATIONS
Hazardous Location Analysis
In this option, the road network is presented in five options: streets, street segments, intersections, and a combination of the first two with the third, respectively. The criteria are available in three options: total number of accidents, ratio of this number to the AADT value, and both. After choosing the targeted network segments and criteria options, the user is then obliged to specify the thresholding values for each category of criteria. The results are presented on the screen, complemented with a textual report that may be addressed during the interactive session. If the results are satisfactory, both the map and the report may be stored for off-line production. At present, junctions are plotted when exceeding a threshold number of accidents and sections in excess of a threshold number of accidents per kilometer. When appropriate traffic volumes are entered into the system, thresholds could be defined as number of accidents per vehicle-km travelled, or per vehicle entering the junction.
AREA-WIDE ANALYSIS
in this option, a special temporary buffer layer is crossed with the accident layer. The user is given a choice to specify the distance and center points among all the educational institutes and centers of activity. In this option, the user may choose all the institutes or any variation by type. The cross-correlated accidents are chosen by three options: all accidents or pedestrian accidents only; injured person's group in terms of age (again, all variations of age groups are optional), and the final alternative is a time-span in two separate groups (optional). The results are presented as an overlay of the buffer, or without the buffer layer. By default, the road network is presented, but it is optional to activate any desired layer before the actual map storage.
Before and After Analysis
One of the important assignments of road researchers and engineers is to investigate how road improvements affect road accidents. The main objectives of this investigation are: (a) to determine whether the number of accidents decreased or increased after the road project, and (b) to resolve whether the number of accidents increased during the road improvement project. In this analysis, there are three options:
Updating the Project Database
The newly developed system provides tools to add new projects, and to remove and update an existing project. A project is determined by: project number, name, "start" date, "finish" date, type of improvement and region (a set of street network elements: street links and intersections). The project region is divided into two types: the set of improved street network elements and the set of affected elements.
Textual Reports
The system provides a series of reports that describe the accidents. Prior to the "start" date, during or after the "finish" date of the project, and also during any specified period. This is given either for the project region, or affected region and for the whole municipality, as a controlling set.
Maps
The system provides maps for each project, for periods as described above. These maps depict the accidents in the project region. Intersections and links, within the region and the affected region are shown by different symbols. Intersections are labeled by the total number of accidents. Also, accidents on links are shown by a special symbol on the accurate position. Links are labeled by the total number of accidents. These are surrounded by a box in the middle of the link.
Neighborhoods' Competition Analysis
In order to fight road accidents, the Haifa municipality decided to commend neighborhoods that achieve reductions in accident numbers and severity. For this purpose, the city was divided into 10 administrative neighborhoods. Major streets were joined together as a separate neighborhood. The user may compute, for each neighborhood, an index of road safety, or, alternatively, the rate of accident change during a specific time period. Neighborhoods are classified and sorted by the index or by the road safety "rate." There are eight options to compute the index and rates:
(a) Total number of accidents; (b) total number of accidents per street length; (c) total number of accidents per population; (d) total number of accidents per traffic volume; (e) weighted accidents; (f) weighted accidents per street length; (g) weighted accidents per population; and (h) weighted accidents per traffic volume.
Weighted accidents are calculated by accident severity (light, severe, and fatal).
CONCLUSIONS
The Road Safety GIS system was developed with the intention of rendering it useful and accessible to the many different city departments. The methodology was to build it as a composition of separate modules and sub-modules allowing future development and upgrading. At the present state of development, much effort was put into developing a comprehensive and flexible GIS system. It enables to supply it with the accident data under strict quality control checks, including a check on the accuracy of the accident location. Due to its modular set-up, in future information on traffic volumes, bus routes and time-tables, further land-use information can be easily added, enhancing the system's benefit.
Applications developed so far include the identification and analysis of hazardous locations, area-wide safety analysis, the conduct of "before-after" studies in connection with road improvements, and the evaluation of a neighborhood safety competition. These applications have convinced the Haifa municipality of the system's use and potential.
REFERENCES
1. Lubkin, J.L. & Maleck, T.L. "ARES: An Accident Report Entry System for Local Agencies." In Micro-Computer Applications in Transportation III, ASCE, New York, 1989.
2. Peled, A. "Geographic, Geodetic and Cartographic Data Base for Mapping Road and Traffic Characteristics in Urban Areas." TRI Interim Paper presented at the Symposium on Road Accident Mapping and Location Analysis, Haifa, Israel, 1982.
3. Peled, A. & Hakkert, A.S. "Geodetic Data Base for Mapping Traffic Characteristics and Road Accidents." TRI Report 82-1004, Transportation Research Institute, Technion, Haifa, Israel, 1982.
4. Prastacos, P. "Integrating GIS Technology in Urban Transportation Planning & Modeling." 70th Annual Transportation Research Board Meeting, Washington, D.C., U.S.A., 1991.
is‹‚.ŒÆAeas." TRI Interim Paper presented at the Symposium on Road Accident Mapping and Location Analysis, Haifa, Israel, 1982.
3. Peled, A. & Hakkert, A.S. "Geodetic Data Base for Mapping Traffic Characteristics and Road Accidents." TRI Report 82-1004, Transportation Research Institute, Technion, Haifa, Israel, 1982.
4. Prastacos, P. "Integrating GIS Technology in Urban Transportation Planning & Modeling." 70th Annual Transportation Research Board Meeting, Washington, D.C., U.S.A., 1991.