You are on page 1of 8

Yasuhiko Kajiya et al.

1/8

Paper No.01-3105 Factors Inducing Multi-Vehicular Collisions during Visibility Reduced by Snowstorm
Yasuhiko Kajiya*, Yasuhiro Kaneda**, Kazuhiro Tanji** *Hokkaido Development Bureau, Civil Engineering Research Institute 1-3 Hiragishi, Toyohira-ku, Sapporo, 062-8602, JAPAN Tel: +81-11-841-1111, Fax: +81-11-841-9747, E-mail: ykajiya@ceri.go.jp **Japan Weather Association, Hokkaido Head Office Kita 4 Nishi 23, Chuo-ku, Sapporo, 064-8555, JAPAN ABSTRACT Many factors are thought to relate to the occurrence of multi-vehicular collisions during snowstorm. The authors analyzed these factors based on a database and observations of multi-vehicular collisions. In addition to severe meteorological conditions and surrounding environment during snowstorm, the drivers driving attributes and the traffic factors also were found to relate closely to such accidents. Based on these results, the risk factor level was examined to assess the occurrence risk of multi-vehicular collision. When these risk factor levels were applied to collision cases, they closely expressed the occurrence risk. Although many challenges remain toward developing a quantitative evaluation method of risk assessment, the factors identified in this study enable objective assessment of the risk of multi-vehicular collisions during snowstorm by consideration of the environmental, meteorological and traffic conditions of the site. INTRODUCTION A cold snowy region is characterized by frequent traffic accidents due to slippery roads and snowstorm. These days, multi-vehicular collisions occur particularly frequently on suburban arterial roads and expressways. Because these accidents obstruct road traffic for long periods, their impact on localities is great. For this reason and in light of ever-growing road demand, effective measures are required to prevent traffic accidents and enhance road traffic safety. Figure 1 shows a radar diagram of January 7, 1998, when successive multi-vehicular collisions occurred in the Sapporo area (1). Snowfall areas are distributed in the shape of several long narrow strips parallel with the wind direction (northwest to southeast) and areas of heavy snowfall concentrate in narrow moving areas 10 to 20 km in width. A series of multi-vehicular collisions occurred in these heavy snowfall areas. This suggests that those accidents could have been prevented if drivers had been provided with preliminary information on road sections expected to become dangerous as a result of severely reduced visibility and if drivers had been encouraged to alter their route and travel schedules.

9:00 a.m. January 7, 1998

10:00 a.m. January 7, 1998

Hyvruvpyhpyyvvvr
#x

FIGURE 1 Snowfall area (radar image) and multi-vehicular collision sites.

Yasuhiko Kajiya et al.

2/8

The authors consider that the occurrence of multi-vehicular collisions during snowstorm is attributed to the complicated interaction among several factors including meteorological conditions, traffic conditions, the surrounding environment and human factors of drivers. We also assume that the risk of accident can be quantitatively assessed by integration of these factors. From such a standpoint, factor analysis of a multi-vehicular collision during snowstorm is carried out, and an accident risk assessment method is examined. FACTORS INDUCING MULTI-VEHICULAR COLLISIONS Factors inducing multi-vehicular collisions during snowstorm can be categorized as in Figure 2. Although many factors are considered to be closely associated with the occurrence of multi-vehicular collisions, they are roughly classified into environmental factors, meteorological factors, human factors and traffic factors. Except for human factors, which depend on the attributes of individual drivers, the environmental, meteorological and traffic factors will be discussed in detail.
Environmental factors Traffic factors

Open field topography Highembankment or end of a bridge Nearby structures Lack of visible guidance Curved section

Short headway Following traveling Grouped traveling Speed difference Large vehicle M ulti-vehicular collision

Reduced visibility Strong wind Cold Snowfall Icy road surface

Sudden deceleration/stop Slow driving Driver unfamiliar with the route Newly licensed driver

Accident-prone drivers (the elderly, the disabled) M eteorological factors

Human factors

FIGURE 2 Categorization of factors inducing multi-vehicular collisions during snowstorm. Materials for analysis 1. Research materials used in a factor analysis of multi-vehicular collisions during snowstorm are as follows. Multi-vehicular collision database: Data on accidents on national highways in Hokkaido (1995-1996, 23 cases) and newspaper articles (1983-1997, 56 cases) 2. Observation data from multi-vehicular collision sites: Meteorological survey and traffic survey at Kakuyama, on National Highway 275 (1997-2000) 3. Driver questionnaire survey data: "Near miss" survey of 1,574 drivers in Hokkaido

Environmental factors Figure 3 excerpts material from newspaper articles to indicate the number of multi-vehicular collisions by time. For the purpose of comparison, the mean traffic volume on national highways is included. The figure clearly shows that multi-vehicular collisions during snowstorm concentrate in daytime and lessen at dusk, when the traffic volume increases. This may be because the visibility of luminous bodies such as taillights decreases considerably during daytime snowstorm because the entirely white view at such times results in high background luminance.

Yasuhiko Kajiya et al.

3/8

12 Number of multi-vehicular collisions 10 8 6 4 2 0 10:00 12:00 14:00 16:00 18:00 20:00 22:00 0:00 2:00 4:00 6:00 8:00

1200 1000 800 600 400 200 0

Mean hourly traffic volume on national highways (1995-1996, 48 points) Number of multi-vehicular collisions during snowstorm

FIGURE 3 The occurrence time of multi-vehicular collisions during snowstorm (1983-1997, newspaper articles). The use of surrounding land and the road structure of the accident site are summarized in Table 1, based on the materials on multi-vehicular collisions that occurred on national highways in Hokkaido. 87% of the multi-vehicular collisions occurred in the non-urban district, and as much as 75% occurred on the road section of uninterrupted flow. Generally, the frequency of traffic accidents is higher in the urban district or at intersections, where acceleration and deceleration are difficult and the traffic volume is larger. However, multi-vehicular collisions during snowstorm occur along the road section of uninterrupted flow in the rural areas more frequently than in urban areas. The driver questionnaire survey, in which drivers were asked where they had accident "near misses" during snowstorm, was conducted for 1,574 drivers. An overwhelming majority of drivers reported experiencing fear of accident along the ordinary suburban road or in the open area, which strongly confirms the results of Table 1. TABLE 1 Surrounding Land Use and Road Structure at Sites of Multi-Vehicular Collisions due to Snowstorm Surrounding land use Urban area 3 (13%) Rural area 20 (87%) Road structure Intersection 1 (4%) Intersection approach 3(13%) Tunnel 0 (0%) Bridge 1 (4%) Curved section 1 (4%) Road section of uninterrupted flow 17 (75%) 1995-1996, national highways within Hokkaido

Meteorological factors Figure 4 shows the observation results recorded at Kakuyama, on National Highway 275, including the meteorological changes when multi-vehicular collision occurred. A multi-vehicular collision that involved 18 vehicles running in upward and downward lanes occurred on January 7, 1998, and 5 vehicles in the downward lane collided on March 8, 2000. Both cases occurred during severe snowstorm in daytime, under the harsh meteorological conditions of less than 100 m visibility and 10 m/s wind velocity. Although the weather was particularly severe from 3:00 to 5:00 on January 7, 1998, the January 7 accident occurred later, at 9:40, when the visible range repeatedly fell below 100 m. This suggests that the occurrence of a multi-vehicular collision does not necessarily depend only upon the meteorological conditions.

Traffic volume (vehicle)

Yasuhiko Kajiya et al.

4/8

At Kakuyama on National Highw ay 275 (Jan. 7, 1998) 16 14 12 W ind velocity (m/s) 10 8 6 4 Visibility 2 0 1:00 2:00 3:00 4:00 5:00 6:00 7:00 8:00 9:00 10:00 11:00 12:00 0 W ind veloc ity 10m
2 0

At Kakuyama on National Highw ay 275 (Mar. 8, 2000)

1000m Multi-vehic ular c ollision Visible range

16 14 12 Wind veloc ity (m/s) 10 8 6 4

1000m

Multi-vehicular collision
Visible range 100m

100m

Visibility Wind velocity

10m 2:00 3:00 4:00 5:00 6:00 7:00 8:00 9:00 10:00 11:00 12:00 13:00 14:00

FIGURE 4 Meteorological changes at the multi-vehicular collision site. Figure 5 shows the weather conditions recorded in the meteorological and traffic survey conducted at Kakuyama, when a multi-vehicular collision or a conflict occurred under reduced visibility. A conflict was defined as a state of excessively slow movement, parking and stopping, meandering, or near miss of vehicles. 15 cases were extracted from the video images, only 2 of which occurred when the visible range exceeded 100 m. When it was 50 to 100 m, 6 cases occurred within 60 minutes after the visibility fell below 100 m, which means the chance of traffic accident is high immediately after visibility becomes reduced. However an accident or a conflict had no relation with the lapse of visibility reduction when visibility was less than 50 m.
1997-2000, K akuyama 240min. Time lapse of reduced visibility (Visibility < 100m ) N ot occurred O ccurred 180min.

120min.

60min.

0min. 10 20 50 100 200 Visibility (m) 1000

FIGURE 5 Visible range when an accident or a conflict occurred during snowstorm. The relationship between the time passed after visible range fell below 100 m and the driving speed is shown in Figure 6 (1). The shorter was the time passed after the visible range fell below 100 m, the larger was the variance of driving speed, and this larger variance caused some vehicles to drive excessively slowly. In contrast, when the reduced visibility continues for a long time, the variance in driving speed among vehicles tends to become smaller and stable. The accident risk seems to increase as the variance in driving speed among vehicles increases. According to the figure, a two-hour time lapse after visibility becomes less than 100 m can be considered a threshold factor in inducing multi-vehicular collisions. 57% of these multi-vehicular collisions on national highways in Hokkaido occurred on snowy roads, and the remainder occurred on icy roads. Thus, the slipperiness of roads is also a major factor.

Yasuhiko Kajiya et al.

5/8

Vehicls S peed ( Traffic > 40 )


80 70 Vehicles Speed (km/h) 60 50 40 30 20 10 0 0hours 2hours 4hours 6hours 8hours T ime lapse of reduced visibility 10hours

Trucks and Buses Cars

FIGURE 6 Speed distribution of vehicles and time since visibility was reduced (1997-1999, Kakuyama). Traffic factors When a large vehicle is involved in an accident, a small vehicle easily becomes severely damaged or the accident tends to be otherwise severe. According to newspaper articles, although the involvement of large vehicles was unknown for 12 out of 56 multi-vehicular collisions, 40 cases involved large vehicles. The breakdown of collisions induced by other vehicles including those stopping was obtained from materials on multi-vehicular collisions on national highways in Hokkaido. Table 2 breaks down cases of such collisions. The multi-vehicular collisions were all induced by vehicle-to-vehicle collision including a case of collision with a stopped vehicle. The number of rear-end collisions is particularly large, counting for 18 cases. 17 of the rear-end collisions were with vehicles that were being driven but were stopped. This shows the mechanism of multi-vehicular collisions during snowstorm: that they take place when a following vehicle collides with a leading vehicle. According to the driver questionnaire survey, 42% of the drivers relied on the taillights of the leading vehicle to maneuver behind that vehicle while driving under reduced visibility (2). TABLE 2 Breakdown of Collisions Induced by Other Vehicles Leading to Multi-Vehicular Collisions Type of accident Number Proportion Vehicle-to-vehicle collision Head-on collision 2 9% Rear-end collision while traveling 1 4% Rear-end collision with stopped vehicle 17 74% Crossing collision 1 4% Overtaking 0 0% Sideswipe 0 0% Left turn 0 0% Between right turn and straight (opposite direction) 0 0% Other right turns 0 0% Other vehicle-to-vehicle collisions 1 4% Collision with parked vehicle Collision with parked vehicle 1 4% 1995-1996, national highways in Hokkaido

The traveling behavior of following the vehicle ahead during snowstorm is a factor that induces traffic accidents, and grouped traveling, where many vehicles run in a row, can be considered an accident multiplying factor. Figure 7 shows the relationship between the 10-minute traffic volume per lane and the proportion of vehicles driving in a group, when visibility was below 100 m (1). Grouped traveling is defined here as more than five vehicles driving in a row with less than 200 m of headway, and its proportion to the total traffic volume is illustrated in the diagram. The proportion of vehicles traveling in a group seems to increase as the traffic volume increases. However, from when the 10-minute traffic volume exceeds 40 vehicles, the proportion of grouped vehicles increases rapidly, increasing the risk of multi-vehicular collision in snowstorm.

Yasuhiko Kajiya et al.

6/8

Visibility<100m, Daytime 07:00~17:00 100% Proportion of Vehicles Driving in a Group

80%

60%

40%

20%

0% 0 20 40 60 80 10-minuite Traffic Volume per lane 100

FIGURE 7 Grouped traveling under visibility reduced by snowstorm (1997-1999, Kakuyama). Multi-vehicular collisions during snowstorm are mainly attributed to the stopping, slow driving or deceleration of the leading vehicle. Thus, when the leading vehicle drives excessively slowly or the speed variance of each vehicle is large, the accident risk is likely to increase. Figure 8 shows the relationship between the 10-minute traffic volume per lane and the driving speed of the leading vehicle in a group when visible range is less than 100 m (1). In case of higher traffic volume, excessively slow-driving vehicles hardly appear due to emergence of a vehicle stream. However, when the traffic volume is small, an excessively slow-driving vehicle that is isolated from the leading vehicle is likely to emerge. According to Figure 8, a large vehicle driving at less than 30 km/h appears when the traffic volume becomes less than 60 vehicles and a small vehicle driving at less than 30 km/h appears when the traffic volume becomes less than 80 vehicles.
Trucks and Buses Visibility<100m, Daytime 07:00 100 Speed of the leading vehicle (km/h) 90 80 70 60 50 40 30 20 10 0 0 20 40 60 80 10-minuite traffic volume per lane 100 17:00 100 Speed of the leading vehicle (km/h) 90 80 70 60 50 40 30 20 10 0 0 20 40 60 80 10-minuite traffic volume per lane 100 Cars V isibility<100m, D aytime 07:00 17:00

FIGURE 8 Speed of the leading vehicle in a group during visibility reduced by snowstorm. Figure 9 excerpts results of a survey conducted at Kakuyama, which shows the standard deviation of speed for 10 minutes and the proportion of large vehicles in accidents and conflicts under reduced visibility. The conflict in a snowstorm shows no clear correlation when the proportion of large vehicles is 0 to 20%. However, two cases that led to multi-vehicular collisions mark large vehicle rates as high as 14% and 46%, suggesting that the possibility of a conflict leading to a multi-vehicular collision is high when the proportion of the large vehicles exceeds 10%. The standard deviations of speed exceed 8km/h in case of a conflict and an accident. Higher standard deviations are thought to increase the conflict among vehicles as well as the occurrence risk of accident.

Yasuhiko Kajiya et al.

7/8

1997-2000, K akuyama 40 Standerd devision of speed (km/h) 35 30 25 20 15 10 5 0 0% 10% 20% 30% Large vehicle rate 40% 50% N either conflict nor multi-vehicular collision occurred C onflict occurred Multi-vehicular collision occurred

FIGURE 9 Standard deviation of speed and large vehicle rate at the conflict during snowstorm. RISK FACTORS OF MULTI-VEHICULAR COLLISIONS Conditions of risk factor It is thought possible to estimate the risk of multi-vehicular collisions from the environment around the route, the road structure, and the present and forecasted values of weather and traffic, by using the analysis results of factors contributing to multi-vehicular collisions during snowstorm. Using human factors of individual drivers for this estimation is difficult because it is not realistic to consider each drivers individual situation. To mitigate human factors, a system should be devised that promptly warns the driver of obstructions immediately ahead to reduce reaction delay and sudden evasive action. The items shown by factor analysis to clearly be factors in multi-vehicular collisions are summarized in Table 3. TABLE 3 Factors of Multi-Vehicular Collision during Snowstorm Factor of multiple vehicle collision Specific condition Day or night Daytime Road environment Rural road section of uninterrupted flow Visibility Less than 50 m Time elapsed since visibility became less than 100 m Within 60 minutes Road surface condition Icy, or snow-covered Traffic factors 10-minute traffic per lane 40 - 80 vehicles Standard deviation of speed 8 km/h or higher Large vehicle rate 10 % or more Classification Environmental factors Meteorological factors Risk factor level when multi-vehicular collisions occur Figure 10 shows the results of the verification of risk factor level in multi-vehicular collisions that occurred on National Highway 275 at Kakuyama. On January 7, 1998, with traffic factors (e.g., increased traffic density from the morning and a mixture of large vehicles) added to meteorological factors (intermittent snowstorm from the early dawn), the risk factor level increased. Multi-vehicular collision occurred between 9:30 and 9:40 when the number of risk factor level was the highest (risk factor level of 6). On March 8, 2000, a multi-vehicular collision occurred after the risk factor level became 8. The figure shows that the risk factor level suddenly increased also due to traffic factors (e.g., increased traffic density) together with meteorological factors (heavy snowstorm after 10:00). Thus, the risk of accidents can be expressed very well by the obtained risk factor level of multi-vehicular collisions at the time of snowfall or snowstorm.

Yasuhiko Kajiya et al.

8/8

Kak uyama on January 7, 1998 8 7 Risk factor le ve l 6 5 4 3 2 1 0 10:00 1:00 2:00 3:00 4:00 5:00 6:00 7:00 8:00 9:00 Multi-vehicular collis ion occurre d

K ak uyama on M arch 8, 2000 8 7 Ris k factor le ve l 6 5 4 3 2 1 0 10:00 11:00 12:00 13:00 4:00 5:00 6:00 7:00 8:00 9:00 M ulti-ve hicular collis ion occurre d

FIGURE 10 Verification of the risk of multi-vehicular collision. CONCLUSIONS From the meteorological analysis of multi-vehicular collisions that occurred frequently in Sapporo and its environs, this study found that accidents occurred in accordance with localized areas of heavy snowfall. In other words, meteorological factors were shown to relate strongly to the occurrence of accidents. In addition to meteorological conditions, we presumed that traffic factors such as traffic density raise the risk of accidents. We also examined the factors of the surrounding environment and the road structure, which describe the situation when accidents occur, by analyzing materials of past multi-vehicular collisions that occurred during snowstorm. We obtained factors that seemed to relate particularly closely to the occurrence of multi-vehicular collisions during snowfall or snowstorm, and made verification of these risk factor levels using accident cases. Although there is much room for examination of the level of each factor, its relative importance, etc., it was found that the risk factor level could generally represent the accident occurrence risk, which contains environmental factors, meteorological factors and traffic factors. This means that the risk of the occurrence of multi-vehicular collisions can be estimated from each factor of environmental, meteorological and traffic conditions of the site. For the future, we plan to research the relation between individual factors and their relative importance, toward establishing a quantitative estimation method of multi-vehicular collision risk. ACKNOWLEDGMENTS We would like to express our deep gratitude to Mr. Naotoshi Kanemura of Sapporo Information Center, who went to great lengths to provide radar charts of the time of the accidents in this study. REFERENCES 1. Kazuhiro Tanji, Yasuhiro Kaneda, Yasuhiko Kajiya, Feasibility Study on Weather Information-based Intelligent Winter Highway Systems -Learning from Causes of Multiple Vehicle Collisions in Blizzards-, Sixth ITS World Congress 99 Toronto, November 1999. Manabu Kaneko, Yasuhiko Kajiya, Yoshifumi Fukuzawa, Kazuhiro Tanji, Yasuhiro Kaneda, Winter Road and Human Factors - Learning from Road User Survey and Test Drive Experiments -, Fifth ITS World Congress 98 Seoul, October 1998.

2.

You might also like