高速公路交通事故与气象条件的关系研究

选取2013—2017年河北高速公路逐日发生的一般等级以上交通事故起数、伤亡人数、经济损失及同期气象要素资料，基于序关系分析法将事故量及等级量化为事故指数，分析了对高速通行高影响的雾天、雨天和雪天事故的年际、月际及日变化特征，并构建日均交通事故指数与气象要素的非线性回归模型。研究表明：（1）交通事故发生时不利天气条件下交通事故总起数、死亡事故起数及人数、经济损失为雾天>雨天>雪天，而伤人事故及人数为雨天>雾天>雪天。（2）雨天、雪天交通事故峰值与当年雨雪极端天气事件有关，雾天单起交通事故的直接经济损失呈上升趋势；事故月、日分布有显著性差异，表现为雾天事故高发月份在9月至翌年2月、雨天事故多发于在5—10月、雪天事故多发于1—2月、12月；05—08时和00时为事故高发时段。（3）分别针对不同等级日雨量和能见度与日均交通事故指数建立了一般等级以上交通事故模型，非线性回归效果显著。

The Relationship Between Road Accident Severity and Recorded Weather

Based on the data of traffic accident number, casualty number, economic loss and meteorological elements in the Hebei Express-way from 2013 to 2017, the number and grade of accidents were quantified into accident index. The interannual, interlunar and diurnal varia-tion characteristics of the fog, rain and snow days with high impact on highway traffic are analyzed. The daily average traffic accident index and the nonlinear regression model of meteorological factors are constructed. The results show that:(1) The total number of traffic accidents, fatalities and economic losses under bad weather conditions at the time of traffic accidents are in the order of foggy, rainy, and snowy, while the number of injuries and casualties are in the order of rainy, foggy, and snowy. (2) The peak value of traffic accidents in rainy and snowy days is related to the extreme weather events in that year, and the direct economic loss of single traffic accident in foggy days is on the rise. There are significant differences in the month and day distribution of accidents. The month with the highest incidence of foggy accidents is from September to February of the next year. The month with the highest incidence of accidents in rainy days is from May to October. The snow days are mostly in January-December, and the time between 05:00-08:00 BT and 00:00 BT is the high incident time period. (3) Us-ing precipitation, visibility and traffic accident index, higher-than-normal traffic accident prediction models are established, and their effect as judged by non-linear regression is very significant.