基于交通流量控制下的交通气象风险等级研究OA
Research on the classification of traffic-meteorological risk levels controlling for traffic flow
城区交通流量是交通气象风险研究中的关键混杂变量,其日际波动大且难以被精确统计,导致气象因素的独立风险难以被有效量化.本文基于2019-2023 年温州市区气象与交通事故资料,通过因子分析消除各气象要素的多重共线性,利用具有稳定通勤特性的工作日时段样本构建逻辑回归模型(Logistic Regression Model,LR),将模型输出的事故高发概率转换为低、中、高 3 个交通气象风险等级,并推广至全年时段.为验证该方法的有效性,同步建立了未控制交通流量的全时段(工作日+非工作日)模型进行对比.研究表明:气象因素对温州城区交通事故的影响存在季节性差异,汛期的温度因子和降水因子包含了可与事故机理直接关联的因素,而非汛期的气象因子与事故之间则以更为复杂的间接影响关系为主;工作日模型在拟合优度及检验结果上均优于未控交通流量的全时段模型,其中汛期和非汛期的检验准确率较全时段模型分别提升 1.6%和9.6%.本文通过筛选工作日样本削弱了交通流量带来的混杂效应,在提升模型预测能力的同时有效量化了交通事故中气象因素的独立风险,这有助于将交通安全保障的侧重点聚焦于气象因素本身,为公众出行和交通管理提供了一种更加科学、精准的决策依据.
Urban traffic flow is a key confounding variable in research on traffic meteorological risks.Its significant daily fluctuations and difficulties in precise estimation obscure the independent risk contributions of meteorological factors.This study utilized meteorological and traffic accident data from the urban area of Wenzhou from 2019 to 2023.Factor analysis was employed to eliminate multicollinearity among various meteorological elements.A Logistic Regression model(LR)was constructed using samples from weekdays,which exhibit stable commuting patterns,to control for variability in traffic flow.The model's output of high accident probability was converted into three traffic meteorological risk levels(low,medium,and high),and this framework was applied to the entire year.To validate the effectiveness of this approach,a comparative full-time model(incorporating both weekdays and non-weekdays)without traffic flow control was simultaneously developed.Results indicate seasonal differences in the impact of meteorological factors on traffic accidents in Wenzhou's urban area.During the flood season,temperature and precipitation factors contained elements directly linked to accident mechanisms,whereas in the non-flood season,the relationship between meteorological factors and accidents was predominantly more complex and indirect.In addition,the weekday model demonstrated superior goodness-of-fit and test performance compared to the full-time model without traffic flow control.Specifically,the test accuracy improved by 1.6%during the flood season and 9.6%during the non-flood season.By selecting weekday samples to control for the confounding effect of traffic flow,this study enhances the model's predictive capability while effectively quantifying the independent risk of meteorological factors in traffic accidents.This approach helps focus traffic safety assurance efforts on the causal influence of meteorological factors themselves and provides a more scientific and precise decision-making basis for public travel and traffic management.
谷铁城;郑峰;吴贤笃;杨烨坤;郑裕;范亮
温州市气象局,浙江 温州 325000||东海国家气候观象台,浙江 温州 325000温州市气象局,浙江 温州 325000||东海国家气候观象台,浙江 温州 325000温州市气象局,浙江 温州 325000||东海国家气候观象台,浙江 温州 325000温州市公安局交通管理局,浙江 温州 325000温州市公安局交通管理局,浙江 温州 325000温州市气象局,浙江 温州 325000||东海国家气候观象台,浙江 温州 325000
天文与地球科学
气象风险交通流量混杂变量控制逻辑回归模型
meteorological risktraffic flowconfounding effect controllogistic regression model
《气象科学》 2026 (4)
376-384,9
浙江省气象科技计划项目重点项目(2021ZD26)
评论