- 學位論文
應用多變量事故頻次模型分析市區車道寬度與車種組成對道路安全之研究
Using Multivariate Crash Frequency Model to Analyze Urban Lane Width and Vehicle Composition to Road Safety
摘要
我國都市因住商混合之土地使用及汽機車混合車流的環境，致使車道寬對事故的影響機制與國外文獻有所不同，且以往事故頻次的研究並未將事故依不同嚴重度或類型拆開探討，導致估計的結果有所偏誤。本研究旨在分析車道寬度與車種組成對不同類型事故的影響，並發展市區道路碰撞修正因子。研究利用101～105年台北市資料進行分析，且將事故依嚴重度、類型和涉入車種分開討論，然而準確評估車道寬度與車種組成對事故的影響存在許多挑戰，主要包括： (一)、樣本均值較低， (二)、不同事故類型和嚴重度間的相關性，為此本研究提出了多變量卜瓦松對數常態模型以避免前述議題。結果發現： (一)、不同嚴重度、類型和涉入車種的事故存在高度關聯性， (二)、外車道在二、三車道呈現愈寬事故愈多的情況， (三)、巷弄出入口、公車停靠區、左轉專用道和較高的機車佔比易導致交織或併行事件上升，提高事故風險；慢車道線提供分流行駛的環境，降低事故。
並列摘要
Taiwan’s traffic condition on urban arterials is mixed traffic, and the type of urban development is mixed-use, which caused the mechanism of lane width on crashes is different from previous literature. Besides, previous crash frequency models usually didn’t separate crashes by different severities or types, leading to biased estimates. This study seeks to explore the impact of lane width and vehicle composition on different types of crashes, and to develop urban road crash modification factors. The study data set includes crashes, roadway attributes and traffic volume of Taipei City from 2012 to 2016. However, there are many challenges to accurately evaluate the effects of lane width on the numbers of crashes. These challenges mainly include: (1) low sample mean; and (2) correlation among crash types and severities. A multivariate Poisson log-normal model is proposed in response to these challenges by simultaneously modeling crashes frequency in terms of different crash types or severities. Among findings are: (1) the correlations between crashes of different severities, types and types of vehicle involved are high; and (2) the wider of outer lane, the greater number of crashes when the number of lane is 2 or 3; and (3) access points, bus stops, left turn lanes and higher ratio of motorcycle tend to result in weaving or parallel driving events, which increase the risk of crash; while the slow lanes provide a diverging environment to reduce it.