Analysis on distribution of freeway accidents under various conditions in China

Abstract

This study aims to provide a current survey on the situation of freeway accidents in China. The results show that the accident rate, death toll, injury toll, and direct loss of property are 3.2, 8.4, 7.2, and 24.3 times that of the average for an ordinary highway in China. Freeway accidents occur mainly in Southern (20.77%) and Central (20.2%) China. With detailed data from Guangdong Province, the number of accidents in freeways with 80 km/h design speed (29.58/km/103 pcu) was more than three times that in freeways with design speed of 100 km/h (9.54/km/103 pcu) and 120 km/h (9.42/km/103 pcu). The total accident rate increased monotonously with the decrease in horizontal radius. The results indicate that 54.54% (/km/103 pcu) of accidents occurred on a steep slope (4%–5%), representing about 10 times that of 3%–4% slope and 20 times that of the less than 3% slope. Based on the data, the safety situation of China’s highway transportation is obviously grim, and improving freeway heavy traffic management in economically developed regions, strengthening the safeguarding of mountain freeways, applying small radius and large vertical grade with caution, and developing a monitoring system of tunnels and interchanges could be used as effective measures to prevent freeway accidents.

Introduction

By the end of 2015, freeway mileage in China reached 125.4 thousand kilometers, accounting for 2.8% of highway in the country (4.50 million kilometers). Compared to other highways, the freeway is often designed with relatively good driving environment, such as high alignment indices, good pavements, total enclosure, absence of pedestrians, no interference with low speed, and excellent traffic safety devices. Thus, the accident rate and death toll in freeways are within an average of 30%–51% and 43%–76% of ordinary highways in developed countries. ¹ However, the accident rate, death toll, injury toll, and the direct loss of property in freeways are 3.2, 8.4, 7.2, and 24.3 times that of the average for ordinary highways in China. Therefore, determining what the real laws are on the accident that occurred in freeways and how the different types of freeway environment affect the rate of accidents based on reliable databases.

Over the past several decades, historical surveys covering the features and frequency of accidents involving freeways have been a very active research area.^2,3 In the 1940s, Smeed ⁴ conducted a large-scale investigation on traffic accidents and studied the relationship between population, vehicle parc, and the accident rate based on the statistics method. Zegeer et al. ⁵ collected details on highway side accidents for 5000 miles of freeways from seven countries to study the relationship among side slope, side clearance, side obstacles, and highway side accidents. They found that vehicle accident ratio declined dramatically when the side slope changed from 1:3 to 1:7. However, the accident ratio dropped by only 2% when the side slope was changed from 1:2 to 1:3. Elvik ⁶ selected accident data of 20 typical freeways from eight countries in Europe and proposed an identification method for the danger segment. Charlton and de Pont ⁷ investigated accidents on horizontal curves and found that sharp curves led to high accident ratio (70%). Previous related studies have focused mainly on risk analysis and accident prediction for freeways using limited samples. Few attempts have been made to collect sufficient samples to analyze accident percentages concerning freeway facilities in China. These attempts have investigated mainly the features common to all types of freeway accidents, with a relatively small number of samples analyzed.^8–10 In terms of freeway accidents within China, no specialized accident databases and highway design have been created or made available, much less investigations that clarify China’s current safety situation. Thus, a study on accidents freeway environment in China is carried out.

This article is organized as follows. The section immediately following describes major sources used for collecting samples. In the next section, the distribution of accident percentages related to highway types, location, design speed, horizontal curve radius, vertical grades, and characteristics of accidents in tunnels and interchanges is presented. In the last section, some conclusions concerning the measures that need to be taken to reduce this frequency are drawn.

Data sources

In China, the National Statistics Annual Report of Road Traffic Accidents (NSARRTA) is released annually. Comprehensive data are provided in the report, including the distribution of accidents by time, territory, and road type. Generally, statistics on the national and provincial accident rates, death toll, injury toll, and the direct loss of property are gathered in the annual report. Macro data from the annual report are credible and can be used to compare accidents in the freeways with those in other highways. However, until recently, the number of reported accidents in NSARRTA was quite incomplete. Accidents in many old freeways are not reported, and minor accidents are seldom recorded because of technical or other reasons. Specifically, data on injuries due to accidents with details on the road environment have not been collected. Thus, detailed statistics on injuries due to accidents in different types in freeway environment are not available, making sample collection a time-consuming and difficult task in this study. Road design data are difficult to gather from the administration center because each freeway has its own center. Furthermore, while some old freeway centers maintain records on accidents, the design speed data are missing.

Sufficient samples to conduct a meaningful statistical analysis are acquired using four major sources, including the NSARRTA (2013), Statistical Bulletin of Transportation Industry Development (SBTID, 2013), Traffic Management Committee of Public Security Department of Guangdong Province (TMCPSDGP), different Traffic Police detachments (TPD for seven freeways, 593.099 km in total), different Freeway Administration and Maintenance Centers (FAMC for seven freeways, 593.099 km in total), and additional results provided by other scholars. Among these sources, the NSARRTA data released in 2013 were used to analyze the accidents based on accident severity, highway type, and province in China. Data from the TPD and FAMC of the various freeways were used to analyze the distribution of accidents based on design speed and distribution of accidents by alignment. Data from the TMCPSDGP were used to analyze the accident characteristics of tunnels and interchanges. As to the sample size, see Table 1.

Table 1.

Sample source and size.

Source	Observation period	Freeway length (km)	Accident amount
TMCPSDGP	2008–2012	2200.779	135,498
NSG freeway TPD and FAMC	2008–2012	72.00	1428
GZJC freeway TPD and FAMC	2008–2012	50.74	3115
JZN freeway TPD and FAMC	2006–2012	109.84	11,209
GH freeway TPD and FAMC	2008–2012	155.306	3441
KY freeway TPD and FAMC	2008–2012	125.20	2351
GZBH freeway TPD and FAMC	2007–2012	21.652	12,850
SM freeway TPD and FAMC	2008–2011	58.361	719

Research methods

Data from NSARRTA released in 2013 are national-level data and were used to analyze the distribution of accidents by severity, highway type, and province in China. Detailed information on the seven freeways in Guangdong Province was analyzed to identify the relationship between design speed, amount of accidents, and distribution of accidents by alignment. The TMCPSDGP data were used to analyze the characteristics of accidents in tunnels and interchanges. The research ideas of this article can be expressed by Figure 1.

Figure 1.

Flowchart of the article.

However, accident records in China are not so satisfied by detailed analysis. We collect accident data mainly from police office and road administration. The data from police focus on accident severity and the data from road administration focus on road facilities. Thus, they record the same accident in the different manner and also ignore the record they do not care. For this reason, we applied data mining (proximity and correlation analysis technique) to improve the accident record.

Detection of repetition by proximity technique

We set up the following proximity model to judge whether two records, respectively, from police office and road administration are the same.

If:

1. Accident data: Date.p = Date.r;

2. Accident time: |Time.p-Time.r| ≤ 30 min;

3. Accident site: |K.p-K.r| ≤ 100 m;

4. Vehicle type: Type.p = Type.r;

5. Accident severity: Severity.p = Severity.r.

Note: p represents police and r represents road administration.

Then, two records, respectively, from police office and road administration are the same.

Recover the lost data by correlation analysis technique

Some important records are missed in the data, such as accurate accident place. We use information from complete records to find the correlation of certain elements. Then, we use the relationship between the certain elements to recover the lost information.

The model should be

$$\{\begin{array}{c}\mathrm{s}\left(\mathrm{X}\to \mathrm{Y}\right)=\frac{\mathrm{\sigma }\left({\mathrm{X}}_{\mathrm{i}}{{\displaystyle \cup }}^{}{\mathrm{X}}_{\mathrm{j}}{{\displaystyle \cup }}^{}\cdots {{\displaystyle \cup }}^{}{\mathrm{X}}_{\mathrm{k}}{{\displaystyle \cup }}^{}\mathrm{Y}\right)}{\mathrm{N}}\\ \mathrm{c}\left(\mathrm{X}\to \mathrm{Y}\right)=\frac{\mathrm{\sigma }\left({\mathrm{X}}_{\mathrm{i}}{{\displaystyle \cup }}^{}{\mathrm{X}}_{\mathrm{j}}{{\displaystyle \cup }}^{}\cdots {{\displaystyle \cup }}^{}{\mathrm{X}}_{\mathrm{k}}{{\displaystyle \cup }}^{}\mathrm{Y}\right)}{\mathrm{\sigma }\left({\mathrm{X}}_{\mathrm{i}}{{\displaystyle \cup }}^{}{\mathrm{X}}_{\mathrm{j}}{{\displaystyle \cup }}^{}\cdots {{\displaystyle \cup }}^{}{\mathrm{X}}_{\mathrm{k}}\right)}\end{array}$$

(1)

where X_i, X_j, …, X_k are associated information of road traffic accident, such as accident type and vehicle involved; Y is attribute set of road traffic accident, such as accident place and direction; $\mathrm{s}(\mathrm{X}\to \mathrm{Y})$ is support degree of correlation between X and Y (%); $\mathrm{c}(\mathrm{X}\to \mathrm{Y})$ is confidence coefficient of correlation between X and Y (%); and $\sigma (\mathrm{X})$ and $\sigma (\mathrm{X}\cup \mathrm{Y})$ are support counts of X and $\mathrm{X}\cup \mathrm{Y}$ .

For example, on the same downgrade segment, we obtain 213 accident completed records but another 92 uncompleted records which only missed the accident place direction. We set up the correlation model as in Table 2.

Table 2.

Correlation model of downgrade segment.

Correlation rule {Xi, Xj}→{Y}	Support count (Xi, Xj, Y)	Support count (Xi, Xj)	Total	Support degree (%)	Confidence coefficient (%)
{truck, crash barrier}→{downgrade}	60	76	213	28.26	78.78
{truck, rollover}→{downgrade}	30	42	213	14.13	72.22
{truck, crash other facilities}→{downgrade}	7	9	213	3.26	75
{car, crash barrier}→{downgrade}	25	28	213	11.96	83.33
{vehicle collision}→{downgrade}	7	9	213	3.26	75
{rear-end collision}→{downgrade}	23	46	213	10.87	50
{truck, fall into ditch}→{downgrade}	5	5	213	2.17	100

With confidence coefficient, we can recover the 92 uncompleted records.

Results

Distribution of accidents by accident severity and highway type

Accident rate, death toll, injury toll, and direct loss of property of different types of highways have been recorded in NSARRTA 2013, otherwise called four-statistical-data. Highways in China are classified into two broad classifications: rural highways and urban roads. Urban roads are those that connect areas in the urban or to rural roads. Rural highways are those that run between cities, villages, or factories. A freeway is a type of rural highway. At the end of 2013, the total road length of China was about 4.36 million kilometers, among which the freeways accounted for 104.4 thousand kilometers or 2.4% of the total road length. However, the accident rate in freeways accounts for 7.61% of all traffic accidents, the death toll accounts for 20.20%, injury toll accounts for 17.27%, and the direct loss of property accounts for 58.19%, as shown in Table 3; these numbers are 3.2, 8.4, 7.2, and 24.3 times those of the highway’s average in China.

Table 3.

Highway types and four-statistical-data of China in 2013 highway length is cited from Ref. ¹¹

		Accident amount				Death toll
		Amount	Of total %	Length 104 km	Of length per 104 km	Amount	Of total %	Length 104 km	Of length per 104 km
Rural highway	Total	114,287	100	435.62	262.35	18,157	100	435.62	41.68
	Freeway	8693	7.61	10.44	832.66	3668	20.20	10.44	351.34
	First-class	18,198	15.92	7.95	2289.06	2356	12.98	7.95	296.35
	Second-class	36,556	31.99	34.05	1073.60	6489	35.74	34.05	190.57
	Third-class	55,936	48.94	40.7	1374.35	3420	18.84	40.7	84.03
	Forth-class	15,956	13.96	282.41	56.50	1380	7.60	282.41	4.89
	Substandard	11,948	10.45	60.07	198.90	844	4.65	60.07	14.05
		Injury toll				The direct loss of property
		Amount	Of total %	Length 104 km	Of length per 104 km	Amount (103 CNY)	Of total %	Length 104 km	Of length 103 CNY per km
Rural highway	Total	43,944	100	435.62	100.88	341,197	100	435.62	783.24
	Freeway	7587	17.27	10.44	726.72	198,552	58.19	10.44	19,018.39
	First-class	5798	13.19	7.95	729.31	245,610	7.20	7.95	30,894.34
	Second-class	15,391	35.02	34.05	452.01	674,710	19.77	34.05	19,815.27
	Third-class	8877	20.20	40.7	218.11	32,868	9.63	40.7	807.57
	Forth-class	3632	8.27	282.41	12.86	10,188	2.99	282.41	36.08
	Substandard	2659	6.05	60.07	44.27	7557	2.21	60.07	125.80

Distribution of freeway accidents by province

Mainland China can be divided into 31 provinces and 7 divisions according to their geographical zones: Eastern, Southwest, Central, Northern, Southern, Northwest, and Northeast China.

More than 40% of accidents occurred in Central and Southern China, 20.2% in the central part, and 20.77% in the south, as shown in Figure 2. Concretely speaking, the provinces with the highest percentages of accidents are Guangdong (10.74% and 0.16 accident per km), Hunan (8.33% and 0.14 accident per km), and Fujian (7.45% and 0.16 accident per km). These provinces are considered to be economically developed regions in China, with the highest average annual Gross Domestic Product (GDP) during the observation period, the highest highways density (as shown in Table 4), and the largest percentage of mountainous highways. More accidents occur in provinces with a larger number of vehicles, such as Shanghai (1.30% and 0.14 accident per km) and Zhejiang (4.61% and 0.11 accident per km), whereas Tsinghai (0.58% and 0.04 accident per km) and Neimenggu (1.15% and 0.02 accident per km) have lower accident frequencies because of smaller traffic volume.

Figure 2.

Total accident amount of freeway in different provinces in 2013.

Table 4.

Total accident amount of freeway in different provinces of China in 2013.

Province	Amount	Length	Per km	Province	Amount	Freeway length	Per km	Province	Amount	Freeway length	Per km
Beijing	117	923	0.13	Anhui	397	3521	0.11	Chongqing	–	–	–
Tianjin	63	1103	0.06	Fujian	643	3935	0.16	Sichuan	433	5046	0.09
Hebei	422	5619	0.08	Jiangxi	437	4303	0.10	Guizhou	196	3284	0.06
Shanxi	268	5011	0.05	Shandong	346	4994	0.07	Yunnan	329	3200	0.10
Neimenggu	99	4080	0.02	Henan	197	5859	0.03	Tibet	–	–	–
Liaoning	213	4023	0.05	Hubei	391	4333	0.09	Shaanxi	428	4363	0.10
Jilin	135	2299	0.06	Hunan	719	5080	0.14	Gansu	281	2953	0.10
Heilongjiang	130	4084	0.03	Guangdong	927	5703	0.16	Tsinghai	50	1228	0.04
Shanghai	112	815	0.14	Guangxi	147	3305	0.04	Ningxia	67	1344	0.05
Jiangsu	302	4443	0.07	Hainan	76	757	0.10	Xinjiang	310	2728	0.11
Zhejiang	398	3787	0.11

From the above analysis, we find Guangdong freeway has the highest percentage of accidents. By the end of 2013, private vehicles are 9.95 million and account for 84.5% of civil vehicles in Guangdong (11.78 millions). Freeway mileage is 5.703 thousand kilometers and accounts for 2.85% of highways in Guangdong province (202.9 thousand kilometers). However, traffic crashes of freeway accounts 20.54% in Guangdong, deaths toll of freeway accounts 31.98%, and injury toll accounts 18.45%.

Distribution of accidents by design speed

Because the province has the highest accident rate, Guangdong has been selected as typical province on which we conducted a detailed survey. Five-year data from TMCPSDGP and the seven different freeway TPDs and FAMCs in Guangdong were collected and analyzed in this article. According to the freeway classification by design speed, freeways in China are classified into three broad classifications: 80 km/h-freeway, 100 km/h-freeway, and 120 km/h-freeway. Traffic accidents can be classified into four broad classifications based on accident classification by personal injury and property loss:

1. Minor accident: one or two persons with slight injuries, or less than 1000 CNY property loss;

2. Ordinary accident: one or two persons seriously injured, or more than two persons with slight injuries, or 1000–30,000 CNY property loss;

3. Major accident: death of 1 or 2 persons, or 3–10 persons seriously injured, or 30,000–60,000 CNY property loss;

4. Extra serious accident: death of more than 3 persons, 11 persons seriously injured, or death of 1 person and more than 8 persons seriously injured, or death of 2 persons and more than 5 persons seriously injured, or more than 60,000 CNY property loss.

Data of accidents from the seven TPDs were classified into the four broad classifications of different design speed freeways. The number of accidents in each category can be seen in Figures 2–4 and Table 5.

Figure 3.

Accidents by design speed (/km/10³ pcu).

Figure 4.

Distribution of four kinds of accidents (%).

Table 5.

Percentage and number of accidents by accident type and freeway design speed (2008–2012 years).

Accident type		Design speed (km/h)
		80	100	120	Total a
Freeway length (km)		110.41	236.62	168.2	515.23
Minor accident	Amount*	3002	2022	1411	6436
	% of total a	46.65	31.42	21.93	100.00
	/km/103 pcu	27.19	8.55	8.39	44.13
Ordinary accident	Amount	137	180	141	458
	% of total a	29.83	39.31	30.87	100.00
	/km/103 pcu	1.24	0.76	0.84	2.84
Major accident	Amount	123	38	22	182
	% of total a	67.41	20.62	11.97	100.00
	/km/103 pcu	1.11	0.16	0.13	1.40
Extra serious accident	Amount	3	18	11	32
	% of total a	10.06	57.00	32.94	100.00
	/km/103 pcu	0.03	0.08	0.06	0.17
Total		3265	2258	1585	7108
% of total a		45.94	31.77	22.30	100.00
/km/103 pcu		29.58	9.54	9.42	48.54

a

Total in the “% of total” is the last low “Total” in the Table.

*

The total accidents number of certain design speed freeway, and the last low “Total” is sum total of 3 kinds design speed freeway.

Figure 2 and Table 5 indicate that 3268 (29.58/km/10³ pcu) accidents occurred on freeways with 80 km/h design speed, which was more than three times that of 100 km/h (9.54/km/10³ pcu) and 120 km/h (9.42/km/10³ pcu). These accidents could have occurred because 80 km/h freeways are often located in mountainous areas, which have low design speed and bad freeway environment that is not as good as that of high one, because these environments have small horizontal curve radius, large vertical grade, short sight distance, and so on.

Among the 7108 samples, 6436 were minor accidents (90.53%). The rest can be categorized as ordinary (6.45%), major (2.57%), and extra serious (0.45%) (see Figure 4).

Figure 5 shows that serious accidents (ordinary, major, and extra major accidents) have an increasing trend as the design speed increased. The higher speed limits for freeways with design speed of 120 km/h resulted in higher amounts of major accidents (57.00%) and extra serious accidents (32.94%). This increase may be because the high design speed leads to higher driving speed, which results in more serious accidents. Figure 5 shows that a large percentage of ordinary, major, and extra serious accidents occurred in100 km/h freeways. A detailed analysis of the data indicates that the higher percentage of accidents could be because most of 100 km/h freeways have special sections located in bad terrain conditions, resulting in a huge difference in operation speeds between adjacent segments, which poses considerable danger for drivers. In some conditions, tunnels or large bridges are used to address bad terrain conditions, resulting in a change of the driving freeways. Small radius is more likely to cause accidents, with the addition of small traffic volume (340 pcu/h) compared to average volume (5290 pcu/h), the accident rates/km/10³ pcu of small radius segments are quite high.

Figure 5.

Four classifications of accident percentage.

Figure 6.

Accidents by horizontal curve radius (/km/10³ pcu).

Against common rules, we find from Figure 7 that the probability of collision increases along with the decrease in radius. However, non-collision accidents do not increase monotonously with the decrease in radius. When the radius is from 800 to 1000 m, the accident rate (2.92/km/10³ pcu) becomes much higher than that of 600–800 m (0.92/km/10³ pcu). Because of small sample size for 600–800 m, the real reason is not clear and further studies need to be carried out.

Figure 7.

Distribution of different kinds of accidents by horizontal curve radius (/km/10³ pcu).

Figure 8 also indicates that collision accidents account for 81.12%, whereas non-collision accidents and rollover accidents account for 18% of the majority. Although the freeway is totally closed to pedestrians, people intruding onto the freeways often induce accidents. Fortunately, intrusions are a fairly rare occurrence and only two cases have been reported in the data.

Figure 8.

Four classifications of accident percentage of horizontal curves.

Distribution of accidents by vertical grade

According to Technical Standards of Highway Engineering and Design Specification for Highway Alignment^12,13of China, when the design speed is higher than or equal to 80 km/h, the maximum vertical grade is 5%. Absolute vertical grades have been classified into three kinds: less than 3%, 3%–4%, and 4%–5%. The number of accidents in each category based on horizontal alignment can be seen in Figures 9–11 and Table 8.

Figure 9.

Total amount of distribution of accidents by vertical grade.

Figure 10.

Different types of accident distribution by vertical grade.

Figure 11.

Four classifications of accident percentage of vertical alignment.

Figures 9 and 10 show that the total amount of accidents increases monotonously with the decrease in vertical grade. The total amount of accidents (54.54/km/103 pcu) is the highest among all other cases when the vertical grade is larger than 4%. Further analysis of the data collected shows that only freeways with 80 km/h design speed can set out such kinds of vertical curves according to Technical Standards of Highway Engineering in China (see Table 6). As explained above, most of the 80 km/h design speed freeways are located in mountainous areas, which have a more complex driving environment. Large grade segments in mountainous areas make it difficult to control vehicles, particularly heavy trucks. This kind of segment is inclined to high accidents, with the addition of small traffic volume (340 pcu/h) compared to average volume (5290 pcu/h), the accident rate/km/10³ pcu of large vertical grade segments are quite high. When the vertical grade is smaller than 4%, all kinds of accidents obviously decline (2.83/km/10³ pcu for total).

Table 6.

Percentage and number of accidents by horizontal radius.

Radius (m)	Length (m)	Accidents	Collision		Non-collision accidents		Total
			Vehicle–vehicle collision a	Vehicle–object collision b	Rollover c	Others d
Less than 600	6324.31	Amount	71	82	40	3	195.65
		% of total	102.74	119.17	57.53	4.11	283.55
		/km/103 pcu	11.21	13.00	6.28	0.45	30.94
600–800	3260.32	Amount	20	18	3	0	41.00
		% of total	31.75	28.57	4.76	0.00	65.08
		/km/103 pcu	6.13	5.52	0.92	0.00	12.58
800–1000	342.07	Amount	2	1	1		4.00
		% of total	50	25	25	0	100.00
		/km/103 pcu	5.85	2.92	2.92	0.00	11.69
1000–2000	9771.96	Amount	12	13	5	1	31.19
		% of total	7.56	7.68	3.21	0.46	18.90
		/km/103 pcu	1.28	1.30	0.54	0.08	3.19
More than 2000	6315.13	Amount	5	4	0	0	9.64
		% of total	10.38	7.41	0.74	0.37	18.90
		/km/103 pcu	0.84	0.60	0.06	0.03	1.53
Total			228/81.12%		53/18.88%		281/100%

a

Including crashing moving vehicle(s) and static vehicles(s).

b

Crashing static objects, such as safety guardrails and traffic signs.

c

Including side rollovering and rollover.

d

Including crashing pedestrian, falling into cliff, driving into ditch, being on fire, and so on.

Table 7.

Minimum radius and maximum grade of different design speed.

Design speed (km/h)		120	100	80
Horizontal radius	Minimum value (m)	1000	700	400
	Ultimate value (m)	650	400	250
Vertical grade (%)		3	4	5

Table 8.

Percentage and number of accidents by vertical grade.

Absolute vertical grade (%)	Length (m)	Accidents	Collision		Non-collision accidents		Total
			Vehicle–vehicle collision a	Vehicle–object collision b	Rollover c	Others d
Less than 3	36615.2	Amount	41	42	15	1	99
		% of total	41.68	42.26	14.91	1.15	100.00
		/km/103 pcu	1.13	1.14	0.40	0.03	2.70
3–4	2082.5	Amount	5	4	1	1	11
		% of total	42.11	40.35	12.28	5.26	100.00
		/km/103 pcu	2.18	2.09	0.64	0.27	5.17
4–5	4159.3	Amount	116	51	54	6	227
		% of total	51.25	22.50	23.75	2.50	100.00
		/km/103 pcu	27.95	12.27	12.95	1.36	54.54

a

Including crashing moving vehicle(s) and static vehicles(s).

b

Crashing static objects, such as safety guardrails and traffic signs.

c

Including side rollovering and rollover.

d

Including crashing pedestrian, falling into cliff, driving into ditch, being on fire, and so on.

Figure 11 also indicates that collision accidents (77%) accounted for a significant majority of accidents because of the vertical grade, whereas among non-collision accidents, rollover accidents (21%) accounted for the majority.

Accident characteristics of tunnels and interchanges

Characteristics of accidents in tunnels

Data on accidents in tunnels were acquired from TMCPSDGP. Accident data from 41 tunnels were surveyed and 757 accidents analyzed. Among the 41 tunnels, 14 tunnels are shorter than 500 m with a total length of 3447.65 m; 10 tunnels are between 500 and 1000 m with a total length of 6958.52 m; 13 tunnels are between 1000 and 3000 m with a total length of 22,744 m; and 4 tunnels are longer than 3000 m with a total length of 15,405.35 m. The accident type statistics in highway tunnel sections are shown in Figure 12. The statistics show that rear-end collisions, collisions onto the tunnel wall, and rollover accidents accounted for the majority of total number of accidents in tunnel sections (approximately 96.49%). The proportion of accidents caused by fire or other reasons is very small. In terms of accident by vehicle type distribution (Figure 13), heavy trucks and cars were the majority (approximately 89.85%), while the proportion of other vehicle types is relatively small.

Figure 12.

Distribution by accident type in tunnel sections (%).

Figure 13.

Vehicle distribution in tunnel sections (%).

Characteristics of accidents in interchanges

In total, 105 interchanges and 2738 accidents in Guangdong province were analyzed. Among the 105 interchanges, 75 were trumpet interchanges. Figure 14 shows that rear-end collisions, side collisions, and collision onto guard bar comprised the majority of total number of accidents within the scope of the interchange (approximately 85.07%) and that these accidents were concentrated mainly in the sections of the entrance–exit ramp. After statistical analysis, the proportion of accidents in entrance–exit ramp to total number of accidents in interchange sections was 83.33%, whereas the proportion of other accident types is 16.67%.

Figure 14.

Distribution by accident type of interchange (%).

Conclusion

Many of the characteristics associated with China’s freeway accidents are presented in this study. First, although freeway mileage is 104.4 thousand kilometers, it accounts for only 2.4% of highways in China (4.36 million kilometers), the overall accident rate in freeways accounts for 7.61% of all traffic accidents, death toll accounts for 20.20%, injury toll accounts for 17.27%, and direct loss of property accounts for 58.19%, representing 3.2, 8.4, 7.2, and 24.3 times the average for highways in China. The reasons for these accidents include high driving speed, large traffic volume, and low-level management.

Judging from the spatial location distribution, freeway accidents occur mainly in Southern (20.77%) and Central China (20.2%). The former is considered to be one of the most economically developed regions, with the highest highway density and the largest number of traffic volume in China. Guangdong (10.74% and 0.16 accident per km) and Fujian (7.45% and 0.16 accident per km), whose accident percentages are the highest in China, are located in Southern China. The latter areas have many mountain freeways, and the limited traffic infrastructure combined with inadequate driver training and management resulted in a higher probability of serious accidents occurring.

Freeways with 80 km/h design speed are usually located in mountainous areas and are areas set out with worse technical standards than those of 100 or 120 km/h design speed, such as small horizontal curve radius, large vertical grade, and short sight distance. Thus, accidents on freeways with 80 km/h design speed (29.58/km/10³ pcu) represent more than three times that of the 100 km/h (9.54/km/10³ pcu) and 120 km/h (9.42/km/10³ pcu) freeways. Minor accidents (91.08%) are the predominant crash type without serious wounds. Ordinary (5.22%) and major (3.40%) accidents account for certain percentages. However, serious accidents (ordinary, major, and extra major accidents) exhibit an increasing trend with large design speed (100 and 120 km/h) freeways, which tend to have higher speed limits. The percentage for freeways with design speed of 100 km/h is relatively higher than that for freeways with design speed of 80 km/h. The higher percentage of accidents may be because of the higher speed limits in this kind of freeway. Freeways with design speed of 100 km/h are usually located in areas with a certain percentage of distressing sections, which limit the entire freeway from 120 km/h, thereby resulting in huge differences in operation speeds between adjacent segments, which could be quite dangerous for drivers.

The total amount of accidents increases monotonously with the decrease in horizontal radius. Among the important critical radii are 2000 and 1000 m. When the radius was larger than 2000 m, the total accident rate was lesser (1.53/km/10³ pcu) than that of small ones (15.28/km/10³ pcu), and it appeared that all kinds of accidents had nothing to do with radius. Meanwhile, when the radius is smaller than 1000 m, the total accident rate is apparently higher than that of bigger ones, but the difference was not too significant. Collisions (81.12%) still accounted for a significant majority of the accidents. The probability of collision increased along with decrease in the radius. Non-collision accidents do not increase monotonously with the decrease in the radius. However, the reason for these results cannot be determined because of the small sample size in this study.

Moreover, 54.54% (/km/10³ pcu) of accidents occur on a steep slope (4%–5%), about 10 times that of the rate for the 3%–4% slope and 20 times that of the slope that was less than 3%. Driving on a large grade slope makes it difficult for drivers to control their vehicles, especially for heavy trucks. With the addition of 4%–5% grade often used in mountain freeways, drivers face a much more complex environment, resulting in a larger number of accidents. In terms of accident types, collision accidents (77%) still accounted for a great majority.

Rear-end collisions, collision onto tunnel wall, and rollover accidents comprise the majority (96.49%) of accidents in tunnels. The proportion of accidents caused by fire or other reasons is very small. Rear-end collisions, side collisions, and collisions onto guard bar represent the majority of the total number of accidents within the scope of interchange, comprising approximately 85.07%, with accidents occurring mainly in the sections of the entrance–exit ramps.

In summary, analyses on accidents in freeways indicate that China’s freeway environment has some issues in terms of safety. Improving freeway heavy traffic management in economically developed regions, strengthening the safeguarding of mountain freeways, applying small radius and large vertical grade with caution, and developing a monitoring system for the tunnels and interchanges are some solutions that could be considered as effective measures to avoid freeway accidents. However, the aforementioned results were acquired based on different samples and hence some of the conclusions were based on a small number of samples with certain limitations. Data used to analyze the distribution of accidents based on severity, highway type, and province were collected from NSARRTA (2013), which is incomplete because it ignores minor accidents and accidents in old freeways. Moreover, with the exception of the four-statistical-data, no other detailed accident records with road environment could be found. Thus, the distribution of accidents based on design speed, horizontal alignment, and vertical grade was analyzed based on the samples of TPDs and FAMCs of seven freeways. Data from TMCPSDGP were used to analyze the characteristics of accidents in tunnels and interchanges. More in-depth analyses could be conducted by combining the results from this study with the results from the subsequent years or with a larger sample size. The additional data would also enable comparisons of different alignment combinations.