Comparative analysis of accident severity causation and heterogeneity on low-grade highways in mountainous areas

doi:10.16265/j.cnki.issn1003-3033.2025.07.1744

Abstract

Abstract:

In order to explore the heterogeneity sources of the influencing factors of traffic accident severity on low-grade highway and analyze the causal mechanisms of accident severity, based on traffic accident records of a low-grade highway in Chongqing city in the past 7 years, combined with the classification of collision vehicle types, the accident severity inducement analysis model was constructed using the fixed parameter Logit, mixed Logit, and random parameter Logit models considering heterogeneity sources. The differences in goodness-of-fit and heterogeneity of different models were analyzed, and the influence of significant variables on accident severity was quantified using average elasticity coefficient. The results show that the Logit model with random parameters considering heterogeneous sources has the highest goodness-of-fit under the same type of accident conditions. In the comprehensive accident model, the non-motor vehicle of the involved party has the greatest influence on the accident severity. The frontal collision is a heterogeneous variable that obeys a normal distribution of (-0.668, 0.749²). The mean and variance of the parameters are positively influenced by summer and large accident-causing vehicles. The season and involved vehicle types exhibit the largest effect intensity in the vehicle-to-vehicle accident model. The parameters associated with summer, mid-sized accident-causing vehicles, and side collisions follow a one-sided triangular distribution with mean (variance) of 0.586, 0.948, and 0.631, and have mean heterogeneity with the factors of night and horizontal curve radius greater than 1 600 m. The severity of vehicle-to-non-motor vehicle accidents is significantly affected by the variables of non-motor vehicle, night, and frontal collision. The parameters corresponding to the night and involved non-motor vehicle variables obey the unilateral triangular distribution of the mean (variance) 2.040 and 1.330, the mean value of parameters is significantly correlated with the weekend and age over 59 years old. Reasonably segmenting the accident dataset of low-grade highways in mountainous areas helps to reduce the heterogeneity effect in the factors affecting accident severity and improve the accuracy and reliability of accident severity cause analysis results.

Key words: mountainous low-grade highways, accident severity, causality analysis, heterogeneity, Logit model

CLC Number:

X928.02事故处理

YAN Jie, XING Xiaoliang, SUN Peng, ZHANG Yu, PENG Hanjie, SUN Tongbo. Comparative analysis of accident severity causation and heterogeneity on low-grade highways in mountainous areas[J]. China Safety Science Journal, 2025, 35(7): 201-208.

Figures/Tables 5

Table 1

Table 2

Table 3

Table 4

Table 5

References 15

[1]	国家统计局. 数据查询[EB/OL]. (2023-06-20). https://data.stats.gov.cn/easyquery.htm?cn=C01.
[2]	胡立伟, 贺雨, 侯智, 等. 山区高速公路交通事故风险多维度耦合研究[J]. 中国安全科学学报, 2024, 34(5): 17-27. doi: 10.16265/j.cnki.issn1003-3033.2024.05.1497
	HU Liwei, HE Yu, HOU Zhi, et al. Multi-dimensional coupling study on traffic accident risk of highway in mountainous areas[J]. China Safety Science Journal, 2024, 34(5): 17-27. doi: 10.16265/j.cnki.issn1003-3033.2024.05.1497
[3]	王芳菲, 彭立飞, 张诗, 等. 山区高速公路隧道交通事故特征分析:以G65包茂高速6条隧道为例[J]. 隧道建设:中英文, 2023, 43(5):816-825.
	WANG Fangfei, PENG Lifei, ZHANG Shi, et al. Characteristics analysis of traffic accidents in six tunnels of G65 Baotou-Maoming highway in mountainous regions[J]. Tunnel Construction, 2023, 43(5): 816-825.
[4]	张洁, 张萌萌, 李虹燕. 基于二元Logistic模型的城市道路交通事故严重程度分析[J]. 交通信息与安全, 2022, 40(5): 70-79.
	ZHANG Jie, ZHANG Mengmeng, LI Hongyan. An analysis of severity of traffic accidents on urban roadways based on binary Logistic models[J]. Journal of Transport Information and Safety, 2022, 40(5): 70-79.
[5]	MUSSONE L, BASSANI M, MASCI P. Analysis of factors affecting the severity of crashes in urban road intersections[J]. Accident Analysis & Prevention, 2017, 103: DOI: 10.1016/j.aap.2017.04.007.
[6]	梁乃兴, 闫杰, 杨文臣, 等. 基于ARIMA-LSTM的高速公路交通安全组合预测模型研究[J]. 重庆交通大学学报:自然科学版, 2023, 42(4): 131-138.
	LIANG Naixing, YAN Jie, YANG Wenchen, et al. A combined prediction model for highway traffic safety based on ARIMA-LSTM[J]. Journal of Chongqing Jiaotong University: Natural Science, 2023, 42(4): 131-138.
[7]	IRANITALAB A, KHATTAK A. Comparison of four statistical and machine learning methods for crash severity prediction[J]. Accident Analysis & Prevention, 2017, 108: DOI: 10.1016/j.aap.2017.08.008.
[8]	KARLAFTIS M G, VLAHOGIANNI E I. Statistical methods versus neural networks in transportation research: differences, similarities and some insights[J]. Transportation Research Part C: Emerging Technologies, 2011, 19: DOI: 10.1016/j.trc.2010.10.004.
[9]	SHANGGUAN Qiangqiang, WANG Junhua, FU Ting, et al. An empirical investigation of driver car-following risk evolution using naturistic driving data and random parameters multinomial logit model with heterogeneity in means and variances[J]. Analytic Methods in Accident Research, 2023, 38: DOI: 10.1016/j.amar.2022.100265.
[10]	ALNAWMASI N, MANNERING F. The impact of higher speed limits on the frequency and severity of freeway crashes: accounting for temporal shifts and unobserved heterogeneity[J]. Analytic Methods in Accident Research, 2022, 34: DOI: 10.1016/j.amar.2021.100205.
[11]	宋栋栋, 杨小宝, 祖兴水, 等. 基于均值异质性随机参数Logit模型的城市道路事故驾驶员受伤严重程度研究[J]. 交通运输系统工程与信息, 2021, 21(3): 214-220.
	SONG Dongdong, YANG Xiaobao, ZU Xingshui, et al. Examination of driver injury severity in urban crashes: a random parameters Logit model with heterogeneity in means approach[J]. Journal of Transportation Systems Engineering and Information Technology, 2021, 21(3): 214-220.
[12]	潘义勇, 吴静婷, 施颖, 等. 建成环境对老年人交通事故严重程度异质性影响[J]. 交通运输系统工程与信息, 2022, 22(3): 224-230.
	PAN Yiyong, WU Jingting, SHI Ying, et al. Heterogeneous effects of built environment on elderly traffic accident injury severities[J]. Journal of Transportation Systems Engineering and Information Technology, 2022, 22(3): 224-230.
[13]	杨文臣, 谢碧珊, 房锐, 等. 山区双车道公路机动车碰撞事故严重度致因比较分析与预测[J]. 交通运输系统工程与信息, 2021, 21(1): 190-195.
	YANG Wenchen, XIE Bishan, FANG Rui, et al. Comparative analysis and prediction of motor vehicle crash severity on mountainous two-lane highways[J]. Journal of Transportation Systems Engineering and Information Technology, 2021, 21(1): 190-195.
[14]	SUN Ming, SUN Xiaoduan, SHAN Donghui. Pedestrian crash analysis with latent class clustering method[J]. Accident Analysis & Prevention, 2019, 124: DOI: 10.1016/j.aap.2018.12.016.
[15]	COOK J, KIMUYU P, BLUM A G, et al. A simple stated preference tool for estimating the value of travel time in rural Africa[J]. Journal of Benefit-Cost Analysis, 2016, 7: DOI: 10.1017/bca.2016.13.

变量属性	变量名称	变量类型(占比)
因变量	事故严重度	仅财产损失事故(61.5%)、伤亡事故(38.5%)
驾驶人	性别	女(12.3%)、男(87.7%)
驾驶人	年龄/岁	<25(17.3%)、≥25且<39(32.5%)、≥39且<59(32.1%)、≥59(18.1%)
车辆	肇事车型/涉事车型	小型车(55.3%/48.8%)、中型车(24.4%/17.6%)、大型车(9.8%/9.6%)、非机动车(10.5%/24.0%)
车辆	事故类型	追尾事故(34.0%)、撞固定物(12.6%)、侧面碰撞(13.3%)、正面碰撞(40.1%)
道路	平曲线半径/m	<800(39.4%)、≥800且<1 600(36.3%)、≥1 600(24.3%)
	坡长/m	<1 000(67.5%)、≥1 000(32.5%)
	坡度/%	<2(46.2%)、≥2且<5(37.4%)、≥5(16.4%)
	路面状况	干燥(86.3%)、潮湿(13.7%)
环境	时段	日间(79.3%)、夜间(20.7%)
	季节	春季(29.7%)、夏季(19.5%)、秋季(17.2%)、冬季(33.6%)
	是否周末	工作日(68.3%)、周末(31.7%)
	天气	晴天(82.1%)、非晴天(17.9%)

变量	固定参数Logit		混合Logit		考虑均值异质性的随机参数Logit		考虑均值和方差异质性的随机参数Logit
变量	参数估计	P值	参数估计	P值	参数估计	P值	参数估计	P值
截距	-1.885	0.000	-1.960	0.000	-1.969	0.000	-1.930	0.000
侧面碰撞	0.593	0.002	0.612	0.001	0.613	0.001	0.618	0.001
肇事车型:中型车	0.861	0.000	0.906	0.000	0.919	0.000	0.900	0.000
肇事车型:大型车	0.748	0.000	0.787	0.000	0.772	0.000	0.586	0.020
肇事车型:非机动车	2.447	0.000	2.556	0.000	2.596	0.000	2.569	0.000
涉事车型:非机动车	2.912	0.000	3.041	0.000	3.065	0.000	3.038	0.000
夜间	0.512	0.001	0.639 (0.639)	0.000 (0.000)	0.639 (0.639)	0.000 (0.000)	0.636 (0.636)	0.001 (0.001)
正面碰撞	-0.399	0.006	-0.617 (0.861)	0.003 (0.039)	-0.617 (0.861)	0.003 (0.039)	-0.668 (0.749)	0.001 (0.049)
均值异质性	—
夜间,年龄≥59岁	—	—	—	—	-0.678	0.046	-0.663	0.048
正面碰撞,夏季	—	—	—	—	0.495	0.053	0.540	0.061
方差异质性	—
正面碰撞,肇事车型:大型车	—	—	—	—	—	—	1.454	0.058
模型评估	—
对数似然值	-787.022		-786.363		-783.234		-781.427
AIC	1 590.000		1 590.700		1 588.500		1 586.900
伪R²	0.338		0.338		0.341		0.342

变量	固定参数Logit		混合Logit		考虑均值异质性的随机参数Logit
变量	参数估计	P值	参数估计	P值	参数估计	P值
截距	-2.545	0.000	-2.548	0.000	-2.469	0.000
夜间	0.379	0.048	0.377	0.052	—	—
秋季	0.630	0.012	0.637	0.012	0.594	0.021
冬季	0.515	0.020	0.517	0.020	0.491	0.030
肇事车型:大型车	0.771	0.000	0.778	0.000	0.852	0.000
夏季	0.803	0.001	0.776 (0.776)	0.001 (0.001)	0.586 (0.586)	0.024 (0.024)
肇事车型:中型车	0.899	0.000	0.876 (0.876)	0.000 (0.000)	0.948 (0.948)	0.000 (0.000)
侧面碰撞	0.825	0.000	0.806 (0.806)	0.000 (0.000)	0.631 (0.631)	0.019 (0.019)
均值异质性	—
夏季,平曲线半径≥1 600 m	—	—	—	—	0.702	0.069
肇事车型:中型车,夜间	—	—	—	—	0.981	0.042
肇事车型:中型车,平曲线半径≥1 600 m	—	—	—	—	-1.139	0.033
侧面碰撞,夜间	—	—	—	—	1.347	0.004
侧面碰撞,平曲线半径≥1 600 m	—	—	—	—	-1.137	0.084
模型评估	—
对数似然值	-495.684		-495.804		-485.770
AIC	1 007.368		1 007.608		995.541
伪R²	0.196		0.196		0.212

变量	固定参数Logit		混合Logit		考虑均值异质性的随机参数Logit
变量	参数估计	P值	参数估计	P值	参数估计	P值
截距	-2.545	0.000	-2.548	0.000	-2.469	0.000
平曲线半径:≥800且<1 600 m	-0.437	0.040	-0.463	0.045	-0.459	0.049
正面碰撞	-0.814	0.001	-0.867	0.001	-0.897	0.001
肇事车型:中型车	1.065	0.018	1.112	0.017	1.042	0.026
夜间	0.811	0.006	0.899 (0.899)	0.009 (0.009)	2.040 (2.040)	0.004 (0.004)
肇事车型:非机动车	1.410	0.000	1.562 (1.562)	0.000 (0.000)	1.569 (1.569)	0.000 (0.000)
涉事车型:非机动车	1.370	0.000	1.474 (1.474)	0.000 (0.000)	1.330 (1.330)	0.000 (0.000)
均值异质性	—
夜间,周末	—	—	—	—	-1.483	0.044
夜间,年龄≥59岁	—	—	—	—	-1.447	0.035
涉事车型:非机动车,周末	—	—	—	—	0.583	0.048
模型评估	—
对数似然值	-284.173		-284.453		-279.350
AIC	580.347		580.907		576.700
伪R²	0.230		0.229		0.243

事故类型	自变量	弹性分析
事故类型	自变量	仅财产损失事故	伤亡事故
综合事故	夜间	-0.043 8	0.053 8
	侧面碰撞	-0.046 3	0.035 6
	正面碰撞	0.038 5	-0.058 4
	肇事车型:中型车	-0.051 0	0.062 4
	肇事车型:大型车	-0.014 7	0.029 7
	肇事车型:非机动车	-0.153 6	0.019 6
	涉事车型:非机动车	-0.780 4	0.187 2
机-机事故	夏季	-0.032 9	0.113 9
	秋季	-0.020 2	0.082 5
	冬季	-0.028 3	0.136 1
	侧面碰撞	-0.036 8	0.047 5
	肇事车型:中型车	-0.042 6	0.081 6
	肇事车型:大型车	-0.028 3	0.074 8
事故类型	自变量	弹性分析
事故类型	自变量	仅财产损失事故	伤亡事故
机-非事故	夜间	-0.151 0	0.008 1
	平曲线半径:≥800 且<1 600 m	0.1034	-0.034 8
	正面碰撞	0.119 4	-0.051 1
	肇事车型:中型车	-0.093 4	0.009 9
	肇事车型:非机动车	-0.209 9	0.028 7
	涉事车型:非机动车	-0.951 9	0.232 3