18
TITLE OF PAPER: Identifying Secondary Crash Characteristics for the California Highway System
KEY WORDS: Secondary crash, Crash analysis, California
ABSTRACT:
This paper uses the crash records, which are part of the Highway Safety Information System (HSIS), for California highway system from 1999 and 2000 to identify the prevailing characteristics of secondary crashes for the aforementioned system. The crash records are aggregated to the California Department of Transportation (Caltrans) District level for analysis. The study compares differences in the characteristics of secondary crashes and primary crashes with respect to time of day, roadway classification, primary collision factors, severity level and type of accident. This study defines a secondary crash, which is any crash that results from the non-recurring congestion or emergency response associated with a primary crash, as any crash that occurs in the same direction within sixty minutes of a primary crash and no more than two miles upstream.
For all comparisons, a proportional test is used to assess the presence of significant differences between secondary crashes and primary crashes. One such comparison is a comparison between rural and urban Caltrans’ Districts. As expected, a higher proportion of crashes in urban Districts are classified as secondary crashes than in rural Districts. The typical secondary crash on the State of California Highway System is a rear-end, property damage only crash on a greater than a four lane urban freeway that occurs during one of the peak periods and is caused by excessive speed.
Identifying Secondary Crash Characteristics for the California Highway System
1. INTRODUTION
A crash is a non-recurring event that not only causes a reduction of roadway capacity but may lead to another crash, known as “a secondary crash”. Secondary crashes can increase congestion, incident duration, fuel consumption and emissions. Many states have proposed various programs to reduce secondary crashes and estimate their benefits in crash reduction. Therefore, understanding the characteristics of secondary crashes can help decision makers select better traffic safety programs.
The paper analyzes secondary crash characteristics for the State of California highway system for urban and rural districts. The study involves statistical analyses to compare differences in the characteristics of secondary crashes and primary crashes with respect to time of day, roadway classification, primary collision factors and type of accidents as well as severity level (fatality, PDO, injures). Also, the paper estimates crash rates for secondary crashes. The study uses 1999 and 2000 datasets from the Highway Safety Information System (HSIS) to estimate the likelihood of secondary crashes. The paper defines secondary crashes as those crashes that occur in close proximity both temporally and spatially to other crashes.
2. LITERATURE REVIEW
Some previous research attempted to find a relationship between primary crashes and secondary crashes. Raub (1997) proposed a straightforward method to link primary crashes to secondary crashes. A crash was classified as a secondary crash if it occurred less than fifteen minutes after an initial crash and no more than one mile distant. Later, Karlaftis et al. (1998) used Raub’s method to develop a logistic regression model for secondary crashes. The studies suggested that more than fifteen percent of all crashes might have resulted from an earlier incident.
Moore (2004) established boundary criteria of two hours and two miles and excluded secondary crashes in the opposite direction of initial crashes. Moore’s study showed a lower frequency of secondary crashes (secondary crashes per primary crash range between 0.015 and 0.030) on Los Angeles freeways than previous investigations.
Previous research showed inconsistent secondary crash rates which might be related to differences in the definition and measurement of secondary crashes. Since the data collection related to secondary crashes, such as queue length and duration, seemed to be seldom observed and difficult to measure, the resulting discrepancy was expected. However, Lindley and Tignor (1979) suggested that the time it takes for the queue to clear ranges from five to more than sixty minutes, depending on the incident type, number of vehicles involved and time of day. Al-Deek (1995) analyzed incident related queue generation on I – 880 in northern California and suggested that the maximum queue length ranges from two miles to five miles and lasts from forty minutes to more than two hours. For this paper, the time to clear the queue is assumed to last sixty minutes and a queue length is defined as two miles.
3. DATA
The project uses 1999 and 2000 crash data from the Highway Safety Information System (HSIS). This database is maintained by the Caltrans’ Traffic Operations Office. The California Highway Patrol has the responsibility to collect and investigate crashes. The variables that this project uses include the date, district, route number, post mile, hours, side highway, road classification, severity, cause and type of accident. A total of 170,866 and 183,988 crashes occurred in 1999 and 2000, respectively. The project considers all of the crashes in the analysis of the frequency distribution of secondary crashes on the California Highway System.
The California Highway System consists of twenty-three Interstate Highways, seven US highways and 218 State Routes (the odd-even directional relationship is the same as US highways). Each route in the California Highway System is given a unique number. In other words, route numbers used on one system are not duplicated on another system. The milepost system runs from south to north and west to east and the milepost values start over again at each county line.
The State of California is divided into twelve highway districts as shown in Table 1. Table 1 presents the population, area and population density (population/area) by district. Based on population density, District 4 (Alameda County, Contra Costa County, Marin County, Napa County, and San Francisco County, San Mateo County, Santa Clara County, Solano County, Sonoma County), District 7 (Los Angeles County, Ventura County), District 11 (Imperial County, San Diego County) and District 12 (Orange County) are defined as urban districts.
Table 1: Population, Area and Population Density by District in 1999 and 2000
District
Population (a)
1990
Population (a)
2000
Area in 2000 (b)
(square miles)
Pop/Area in 2000
(1) Humboldt
273,554
298,599
9,349
32
(2) Shasta
310,270
340,719
27,259
12
(3) Sacramento
1,908,766
2,280,022
12,344
185
(4) Santa Clara
6,023,577
6,783,760
5,347
1,269
(5) Santa Barbara
1,208,861
1,356,626
11,200
121
(6) Fresno
1,712,447
2,081,643
22,457
93
(7) Los Angeles
9,532,180
10,272,535
5,906
1,739
(8) San Bernardino
2,588,793
3,254,821
27,270
119
(9) Inyo
28,237
30,798
13,237
2
(10) San Joaquin
1,155,461
1,369,642
10,861
126
(11) San Diego
2,607,319
2,956,194
8,380
353
(12) Orange
2,410,556
2,846,289
790
3,604
Total
29,760,021
33,871,648
154,398
219
(a) http://www.dof.ca.gov/HTML/DEMOGRAP/table1.xls
(b) http://quickfacts.census.gov/qfd/states/06000.html
4. METHODOLOGY
The overall database must be analyzed to assess its suitability. First, any data missing either a time or post mile is excluded from the analysis. Since the crash database is enormous, the researchers develop a program to filter and analyze the data as a text-based function. The filter logic is based on the time-space and direction of traffic flow. All crash data are input into the program and then the program compares each data pair and produces a data set of secondary crashes.
Second, a proportional test is used to find differences in the secondary crashes for a pair of districts for secondary crashes and the difference between secondary crashes and primary crashes. The proportional test of two populations requires estimates of P1 and P2 which are the estimators of the population proportions. The statistic test is given as:
The null hypothesis H0 : P1 = P2
H1 : P1 ≠ P2 if Z > Zα/2 or Z < Z-α/2
(1)
Where
P1 and P2 = two proportions to be compared
p = pooled estimate
n1 and n2 = population sample size
x1 and x2 = the number of successes for populations 1 and 2
A significance level of ninety-five percent is used for all of the hypotheses tests. The paper tests the following null hypotheses.
The proportion of secondary crashes in urban districts is not significantly different from the proportion of secondary crashes in rural districts.
The proportion of primary crashes by time of day is not significantly different from the proportion of secondary crashes of the same classification.
The proportion of primary crashes by crash severity is not significantly different from the proportion of secondary crashes of the same classification.
The proportion of primary crashes by collision type is not significantly different from the proportion of secondary crashes of the same classification.
The proportion of primary crashes by primary collision factor is not significantly different from the proportion of secondary crashes of the same classification.
The proportion of primary crashes by road classification is not significantly different from the proportion of secondary crashes of the same classification.
Finally, an exposure-based rate for secondary crashes and primary crashes is estimated to compare overall risk. The exposure rate is based on vehicle-miles of travel (VMT). Estimating crash-related VMT is calculated using the following equation.
Crash-related VMT = L x Q x C x T x N (2)
Where
L = number of lanes
Q = queue length
C = traffic count inside a traffic queue
T = time to clear a queue
N = number of crashes
5. CHARACTERISTIC OF SECONDARY CRASHES
5.1 Urban Districts and Rural Districts
There are seven criteria comparisons between a pair of crashes that need to be met before classification as a primary-secondary crash pair.
Crash comparisons include:
Matching district county, route number, occurrence date, and side of highway
Within sixty-minutes of each other
Within a two-mile distance of each other
The number of primary and secondary crashes for each district in 1999 and 2000 is shown in Table 2.
Table 2: Number of Primary Crashes and Secondary Crashes by District in 1999 and 2000
District
Year 1999
Percent
Year 2000
Percent
Primary Crashes
Secondary Crashes
Primary Crashes
Secondary
Crashes
(1) Humboldt
2477
26
1.05
2427
27
1.11
(2) Shasta
2597
32
1.23
2780
38
1.37
(3) Sacramento
10653
416
3.91
11403
405
3.55
(4) Santa Clara
39191
1908
4.87
44037
2258
5.13
(5) Santa Barbara
7352
214
2.91
8127
238
2.93
(6) Fresno
8089
188
2.32
8898
248
2.79
(7) Los Angeles
46849
2439
5.21
48795
2523
5.17
(8) San Bernardino
18083
703
3.89
19613
773
3.94
(9) Inyo
409
5
1.22
432
4
0.93
(10) San Joaquin
7872
197
2.50
8333
238
2.86
(11) San Diego
11930
532
4.46
12849
564
4.39
(12) Orange
15018
678
4.51
15947
788
4.94
Total
170520
7338
4.30
183641
8104
4.41
Urban districts (District 4, 7, 11 and 12) have higher percentages of secondary crashes that range from 4.46 to 5.21 in 1999 and from 4.39 to 5.17 in 2000. The high rate of secondary crashes is expected to occur in regions with high traffic volumes. Specifically, Districts 4, 7, 11 and 12 account for 75.7% of total statewide secondary crashes and 66.2% of statewide primary crashes.
In 1999, the Districts can be divided into six different levels of secondary crash proportions as shown in Figure 1. The proportional tests indicate that:
At level 6, District 7’s proportion of secondary crashes is significantly higher than all.
At level 5, Districts 4, 11 and 12 are not significantly different from each other, but they are different from other levels, and only less than District 7.
At level 4, Districts 3 and 8 are not significantly different from each other, but they are different from other levels.
At level 3, Districts 5 and 10 are not significantly different from each other, but they are different from other levels.
At level 2, Districts 6 and 9 are not significantly different from each other, but they are different from other levels, and only greater than District 1 and 2.
At level 1, Districts 1 and 2 are not significantly different for each other, but their proportions are less than other levels.
1999
Level
District
Highest Proportion
6
7
5
4, 11, 12
4
3, 8
3
5, 10
2
6, 9
Lowest Proportion
1
1, 2
Figure 1: Ranking the proportional test for secondary crashes in 1999
The 2000 data, which is similarly described by Figure 2, indicates that the Districts can be divided into five different levels of secondary crash proportions. Although the ranking among urban districts differs for years 1999 and 2000, the frequency of secondary crashes in urban districts appears to be higher than in rural districts. Also, the proportional tests for secondary crashes suggest that the frequency of secondary crashes in Districts 3 and 8 appear less than the urban districts, but more than all other rural districts. The lowest frequencies of secondary crashes appear to occur in Districts 1 and 2.
2000
Level
District
Highest Proportion
5
4, 7, 12
4
11
3
3, 8
2
5, 6, 10
Lowest Proportion
1
1, 2, 9
Figure 2: Ranking the proportional test for secondary crashes in 2000
5.2 Time of Day
Secondary crashes occur more often during rush hour traffic in the morning and evening. Table 3 shows that for 13 of 24 hours during day the proportion of primary crashes with respect to overall crashes are different from secondary crashes for year 1999, but during the peak hour, the proportion of secondary crashes is higher than that of primary crashes. The number of hours with significantly different proportions occurs less frequently in 2000 than 1999. The critical time of day for secondary crashes is 17:00 pm – 18:00 pm, where the secondary crashes account for (9.4% and 8.8%) of overall crashes on year 1999 and 2000, respectively. In the morning, the highest probability of secondary crashes occurs between 7:00 am and 8:00 am. When comparisons are made to the mean percentage of secondary crashes, the result indicates that am and pm peak hour proportions of secondary crashes are significant higher than mean secondary crashes.
From 0:01 to 5:00, the proportion of primary crashes is equal or higher than the proportion of secondary crashes; this may be the result of low traffic volumes and short queue lengths impacted by the primary crash. However, when a crash occurs after 5:00 am, traffic may be more likely to queue due to higher commute volumes. Under these circumstances, the proportion of secondary crashes may be higher than the proportion of primary crashes. In the evening, the proportion of secondary crashes is higher than the proportion of secondary crashes in the morning. The higher proportion in the evening may result from the higher proportion of primary crashes.
Table 3: Proportional Test Between Primary and Secondary Crashes by Time of Day
Time of day
Primary Crash Percentage
Secondary Crash Percentage
Proportional Test Result
1999
2000
1999
2000
1999
2000
0:01-1:00
1.8
1.90
1.5
1.52
Reject H0 (P)
Reject H0 (P)
1:01-2:00
1.5
1.58
0.8
1.43
Reject H0 (P)
Fail to reject H0
2:01-3:00
1.5
1.56
1.2
1.60
Reject H0 (P)
Fail to reject H0
3:01-4:00
1.1
1.12
1.0
0.81
Fail to reject H0
Reject H0 (P)
4:01-5:00
1.2
1.25
0.9
1.21
Reject H0 (P)
Fail to reject H0
5:01-6:00
2.1
2.22
2.6
2.90
Reject H0 (S)
Reject H0 (S)
6:01-7:00
3.9
3.82
4.6
3.76
Reject H0 (S)
Fail to reject H0
7:01-8:00
6.0
6.07
6.7
6.45
Reject H0 (S)
Fail to reject H0
8:01-9:00
5.8
5.55
6.5
6.29
Reject H0 (S)
Reject H0 (S)
9:01-10:00
4.5
4.39
4.3
4.08
Fail to reject H0
Fail to reject H0
10:01-11:00
4.4
4.32
4.0
3.89
Fail to reject H0
Fail to reject H0
11:01-12:00
4.9
4.84
4.7
4.59
Fail to reject H0
Fail to reject H0
12:01-13:00
5.7
5.48
4.7
4.63
Reject H0 (P)
Reject H0 (P)
13:01-14:00
5.4
5.46
5.2
5.02
Fail to reject H0
Fail to reject H0
14:01-15:00
6.5
6.52
5.5
6.00
Reject H0 (P)
Fail to reject H0
15:01-16:00
7.8
7.75
7.2
7.72
Fail to reject H0
Fail to reject H0
16:01-17:00
7.7
7.60
7.7
7.66
Fail to reject H0
Fail to reject H0
17:01-18:00
8.4
8.36
9.4
8.82
Reject H0 (S)
Fail to reject H0
18:01-19:00
6.1
6.13
7.2
6.73
Reject H0 (S)
Reject H0 (S)
19:01-20:00
3.7
3.78
3.8
3.96
Fail to reject H0
Fail to reject H0
20:01-21:00
2.8
2.90
2.5
2.80
Fail to reject H0
Fail to reject H0
21:01-22:00
2.7
2.87
2.7
2.76
Fail to reject H0
Fail to reject H0
22:01-23:00
2.6
2.56
3.3
3.08
Reject H0 (S)
Reject H0 (S)
23:01-24:00
1.9
1.96
1.9
2.26
Fail to reject H0
Fail to reject H0
Note: Reject H0 (S) represents the proportion of secondary crashes is higher than primary crashes at significance level = 0.95.
Reject H0 (P) represents the proportion of primary crashes is higher than secondary crashes at significance level = 0.95.
5.3 Collision Type
The rear end collision is the predominant secondary collision type for year 1999 and 2000 as shown in Table 4. While rear end is a frequent collision type for primary and secondary crashes, the proportional test indicates that secondary crashes belong to the rear end collision type more often than primary crashes. Other test results show that the proportion of total primary crashes is higher than the proportion of total secondary crashes for all other collision types because rear-end type crashes account for about two thirds of all secondary crashes. Also, each district in California indicates that the rear end proportion of secondary crashes is higher than that of primary crashes.
Table 4: Primary and Secondary Crash Distribution by Collision Type and Proportional Test
Collision Type
Primary Crash Percentage
Secondary Crash Percentage
Proportional Test Result
1999
2000
1999
2000
1999
2000
REAR END
44.1
44.81
66.99
67.66
Reject H0 (S)
Reject H0 (S)
HIT OBJECT
21.37
21.4
11.68
11.78
Reject H0 (P)
Reject H0 (P)
SIDESWIPE
15.83
16.06
13.93
12.75
Reject H0 (P)
Reject H0 (P)
BROADSIDE
9.39
8.92
3.8
3.9
Reject H0 (P)
Reject H0 (P)
OTHER
4.16
3.84
1.64
1.96
Reject H0 (P)
Reject H0 (P)
OVERTURNED
3
2.91
0.8
0.86
Reject H0 (P)
Reject H0 (P)
HEAD-ON
1.41
1.39
0.76
0.67
Reject H0 (P)
Reject H0 (P)
AUTO-PEDESTRIAN
0.72
0.66
0.38
0.41
Reject H0 (P)
Reject H0 (P)
NOT STATED
0.01
0.01
0.01
0.01
Fail to reject H0
Fail to reject H0
Note: Reject H0 (S) represents the proportion of secondary crashes is higher than primary crashes at significance level = 0.95.
Reject H0 (P) represents the proportion of primary crashes is higher than secondary crashes at significance level = 0.95.
Based on his study of crash contributing factors, Campbell (2003) suggests that rear end collisions occur when the leading vehicle decelerates, stops or moving at a lower constant speed. The conditions that exist in a traffic queue after a primary crash support the high frequency of rear end type secondary crashes.
5.4 Primary Collision Factor
Table 5 shows the distribution of collision factors in 1999 and 2000. Speeding is the main factor for primary crashes and secondary crashes. The hypothesis test results indicate that speeding accounts for a higher proportion of secondary crashes than primary crashes. Moreover, other test results by district indicate that speeding is a significantly higher cause for secondary crashes than primary crashes. The result of collision type analysis shows that rear end collision type for secondary crashes results from fluctuating speeds in a traffic queue. At the end of a queue, discontinuity in traffic speeds occurs; as a result, excessive speed may have a more significant impact on secondary crash causes. Therefore, speeding is the primary collision factor for secondary crashes and at a higher proportion than that of primary crashes.
5.5 Road Classification
Most crashes occur on urban freeways with greater than four lanes, and a higher proportion of secondary crashes with respect to overall crashes occur on these facilities (urban freeways with greater than four lanes) as shown in Table 6. The proportion of secondary crashes for urban freeway with greater than four lanes is higher than the proportion of primary crashes. As previously discussed in the earlier proportional tests, urban areas have a higher probability of secondary crashes. Moreover, a higher likelihood of secondary crashes is associated with speeding collision factor. Therefore, a primary crash occurring in an urban area on a high speed facility, such as a freeway should have high probability of secondary crashes.
Table 5: Primary and Secondary Crash Distribution by Primary Collision Factor and Proportional Test
Primary Collision Factor
Primary Crash Percentage
Secondary Crash Percentage
Proportional Test Result
1999
2000
1999
2000
1999
2000
SPEEDING
42.27
43.23
66.42
67.46
Reject H0 (S)
Reject H0 (S)
OTH VIOL HAZOUS
22.05
21.57
13.91
12.86
Reject H0 (P)
Reject H0 (P)
IMPROPER TURN
13.24
13.71
5.49
5.31
Reject H0 (P)
Reject H0 (P)
ALCOHOL
5.95
5.87
2.60
2.58
Reject H0 (P)
Reject H0 (P)
FAILURE TO YIELD
5.15
4.88
0.97
0.99
Reject H0 (P)
Reject H0 (P)
OTH THAN DRIVING
4.23
3.94
3.72
4.04
Reject H0 (P)
Fail to reject H0
FOLLOW TOO CLOSE
3.46
3.53
4.99
4.86
Reject H0 (S)
Reject H0 (S)
UNKNOWN
1.77
1.62
0.97
1.07
Reject H0 (P)
Reject H0 (P)
FELL ASLEEP
0.85
0.76
0.11
0.14
Reject H0 (P)
Reject H0 (P)
OTH IMPR DRIVING
0.66
0.52
0.63
0.56
Fail to reject H0
Fail to reject H0
INVALID CODE
0.37
0.36
0.19
0.15
Reject H0 (P)
Reject H0 (P)
NOT STATED
0.00
0.00
0.00
0.00
Fail to reject H0
Fail to reject H0
Note: Reject H0 (S) represents the proportion of secondary crashes is higher than primary crashes at significance level = 0.95.
Reject H0 (P) represents the proportion of primary crashes is higher than secondary crashes at significance level = 0.95.
Table 6: Primary and Secondary Crash Distribution by Road Classification and Proportional Test
Road Classification
Primary Crash Percentage
Secondary Crash Percentage
Proportional Test Results
1999
2000
1999
2000
1999
2000
URB FRWY >= 4 LN
63.78
64.98
84.08
84.43
Reject H0 (S)
Reject H0 (S)
URB ML DV N-FRE
11.18
10.67
4.05
3.71
Reject H0 (P)
Reject H0 (P)
RUR 2 LN ROAD
8.46
8.13
1.89
1.90
Reject H0 (P)
Reject H0 (P)
RUR FRWY >= 4 LN
6.08
6.11
4.35
4.42
Reject H0 (P)
Reject H0 (P)
URB 2 LN ROAD
3.24
3.17
0.59
0.90
Reject H0 (P)
Reject H0 (P)
URB ML UND N-FRE
2.65
2.50
0.74
0.88
Reject H0 (P)
Reject H0 (P)
RUR ML DV N-FRE
1.96
1.87
1.95
1.65
Fail to reject H0
Fail to reject H0
OTHER
1.46
1.45
1.73
1.57
Fail to reject H0
Fail to reject H0
RUR ML UND N-FRE
0.71
0.68
0.22
0.16
Reject H0 (P)
Reject H0 (P)
URB FRWY < 4 LN
0.28
0.29
0.22
0.37
Fail to reject H0
Fail to reject H0
RUR FRWY < 4 LN
0.19
0.15
0.19
0.01
Fail to reject H0
Reject H0 (P)
Note: Reject H0 (S) represents the proportion of secondary crashes is higher than primary crashes at significance level = 0.95.
Reject H0 (P) represents the proportion of primary crashes is higher than secondary crashes at significance level = 0.95.
5.6 Crash Severity
Crashes severity distributions are shown in Table 7, which indicate that property damage only (PDO) crashes occur most frequently for both primary and secondary crashes. In addition, the proportional test for PDO shows that thery account for a higher proportion of secondary crashes than primary crashes. Other proportional tests indicate that the probability of crash severities is greater for primary crashes than secondary crashes.
Table 7: Primary and Secondary Crash Distribution by Crashes Severity and Proportional Test
Severity
Primary Crash Percentage
Secondary Crash Percentage
Proportional Test Results
1999
2000
1999
2000
1999
2000
PDO
65.39
64.98
72.25
72.14
Reject H0 (S)
Reject H0 (S)
Fatal
0.82
10.67
0.40
0.43
Reject H0 (P)
Reject H0 (P)
Severe Injure
1.97
8.13
0.98
1.10
Reject H0 (P)
Reject H0 (P)
Other Visible Injury
12.47
6.11
8.01
7.69
Reject H0 (P)
Reject H0 (P)
Complaint of Pain
19.36
3.17
18.36
18.64
Reject H0 (P)
Fail to reject H0
Note: Reject H0 (S) represents the proportion of secondary crashes is higher than primary crashes at significance level = 0.95.
Reject H0 (P) represents the proportion of primary crashes is higher than secondary crashes at significance level = 0.95.
6.0 EXPOSURE BASED RATES
6.1 Crash Rates
For this step, estimating crash-related VMT is required to consider the exposure to traffic or the rate of secondary crash occurrence. Queue length and time to clear a queue is based on the earlier assumptions of two miles and one hour. The number of lanes is a classification variable. However, traffic counts inside a traffic queue associated with previous crashes are most often unobserved. This work assumes capacity: 2300 vph for urban and 1800 vph for rural.
Figure 3 shows crashes per VMT (in millions) by Caltrans district in 1999. In urban districts, secondary crash rates are higher than primary crash rates, but in rural districts, the results are not consistent because Districts 1 and 2 have higher crash rate for primary crashes in contrast to other rural districts. Figure 4 shows crashes per VMT (in millions) by Caltrans district in 2000. In the urban districts, secondary crash rates are higher than primary crash rates; however, in rural districts, the results are not consistent because District 1 still has a greater primary crash rate than secondary crash rate in contrast to the other rural districts. In both study years, District 1 has the highest primary crash rate per VMT and District 7 has the highest secondary crash rate per crash related VMT while District 9 has the lowest rates for both primary and secondary crash rates.
Figure 3: Crash Rates in 1999
Figure 4: Crash Rates in 2000
6.2 Injury Rates
Injury rates help to quantify the magnitude of the risk associated with both class of crash. Figures 5 and 6 shows Districts related to injuries per VMT (in millions) by Caltrans district in 1999 and 2000. In urban districts, secondary crash rates are higher than primary crash rates, but in rural districts, the results are still not consistent due to Districts 1 and 2. Both districts have different primary crash injury rates greater than secondary crash injury rates. In 1999, District 1 has the highest injury rates for primary crashes and District 12 has the highest injury rates for secondary crashes while District 9 has the lowest injury rate for primary crashes and District 1 has the lowest injury rate for secondary crashes. Table 21 shows injuries per VMT related to primary crashes and secondary crashes in 2000. In 2000, District 1 has the highest injury rates for primary crashes and District 11 has the highest injury rates for secondary crashes while District 11 has the lowest injury rate for primary crashes and District 2 has the lowest injury rate for secondary crashes. The low frequency of secondary crashes and lack of urban environments may help explain the inconsistent results for Districts 1 and 2.
Figure 5: Injury Rates in 1999
Figure 6: Injury Rates in 2000
6.3 Fatality Rates
Typically, the fatality rates for secondary crashes are not very consistent because of the low incidence of secondary crash fatalities, and consistent results will require longer temporal aggregation periods. Figures 7 and 8, years 1999 and 2000 respectively, show total fatalities per VMT (in millions) by Caltrans districts. Surprisingly, the results are consistent where secondary crash fatality rates are higher than primary crash fatality rates in urban districts. In rural districts, due to the high overall fatality rate of primary crashes and the lack of secondary crash fatalities, primary crash fatality rates are higher than secondary crash fatality rates. Once again in 2000, the urban districts’ secondary crash fatalities rates are higher than primary crash fatalities rates except for District 11, where the secondary fatality rate was unusually low. In the rural districts, the primary crash fatalities rates are higher than secondary crash fatalities rates, except District 3, where the secondary fatality rate was unusually high.
In 1999, District 1 has the highest fatality rates for primary crashes and District 6 has the highest fatality rates for secondary crashes. In 2000, District 1 has the highest fatality rates for primary crashes and District 3 has the highest fatality rates for secondary crashes. In both years, District 12 has the lowest fatality rate for primary crashes.
Figure 7: Fatality Rates in 1999
Figure 8: Fatality Rates in 2000
7. CONCLUSIONS
The frequency of secondary crashes depends on the presence of high traffic flows. The highest rate for secondary crashes is District 7. Overall, the probabilities of secondary crashes in urban districts are higher than rural districts. The overall secondary crash rate is about 4.4% of the total crashes and the rate increases slightly from year 1999 to 2000. The frequency distributions of secondary crashes with respect to time of day, crash severity, collision type, primary collision factor, and road classification is different than that of primary crashes. An initial crash that occurs during peak period is more likely to be followed by secondary crashes than during other periods. The highest collision factor is speeding (defined as exceeding the posted speed limit) for primary and secondary crashes. In addition, urban freeways with more than 4-lane is the type of road that has the highest number of secondary and primary crashes. A simple solution to reduce secondary crashes is to enforce drivers’ compliance with posted speed limit at the location of high traffic flow especially in urban area with freeway more than 4-lane. Moreover, drivers upstream of a primary crash should be informed to reduce their speeds to decrease the likelihood of secondary crashes.
The exposure based crash rate for secondary crashes indicates that the risk of a secondary crash is higher than that of a primary crash. This rate is also higher when driving in urban districts. One may infer that the risk for a secondary crash is higher in urban areas than rural districts. However, some districts that are located in very rural areas such as Districts 1 and 2 show that their exposure rates for crash rates, and fatality rates for primary crashes are higher than the rates for secondary crashes.
Identifying crashes occurring inside a traffic queue associated with a previous crash is used to define a secondary crash. Such circumstances need to have a better data collection process. Further study may consider the data collection process for secondary crashes. The process may involve the integration of police records, emergency response team, and ITS technologies such as traffic surveillance, roadway loop detectors, and GIS/GPS location information. A better understanding of secondary crashes allows traffic engineer to address the problem with a more appropriate solution.
REFERENCES:
Al – Deek, H., Garib, A., Radwan, A.E., 1995. New method for estimating freeway incident congestion. Transportation Research Record 1494 (1), 30-39.
Campbell, B.N., Smith, J.D., Najm, W.G., 2003. Examination of Crash Contributing Factors Using National Crash Databases, DOT HS 809 664. National Highway Traffic Safety Administration, Washington, D.C.
Karlaftis, M.G., Latoski, S.P., Richards, N.J., Sinhs, K.C., 1999. ITS impacts on safety and traffic management: an investigation of secondary crash causes.” ITS Journal 5 (1), 39-52.
Lindley, J.A., Tignor, S.C., 1979. Getaway flow rates for freeway incident and geometric bottlenecks. Public Road 43 (1), 1-7.
Moore , J.E., Giuliano, G., S., Cho 2004. Secondary accident rates on Los Angeles freeways. Journal of Transportation Engineering 130 (3), 280-285.
Ruab, R.A., 1997. Occurrence of secondary crashes on urban arterial roadways. Transportation Research Record 1581 (1), 53-58.
Skabardonis, A., Noeimi, H., Petty, K.F., Rydyzewski, D., Varaiya, P., Al-Deek, H., 1995. Freeway Service Patrol Evaluation, UCB-ITS-PRR-95-5, Institute of Transportation Studies, University of California, Berkeley, CA.
The post 18 TITLE OF PAPER: Identifying Secondary Crash Characteristics for the California Highway appeared first on essayfab.