整合時空方法分析自行車與機動車輛碰撞事故風險

Integrating spatiotemporal approaches for analyzing bicycle-motorized vehicle collision risks

王華琪(Hwa-Chyi Wang)

指導教授 ： 張學孔

國立臺灣大學/工學院/土木工程學研究所/博士(2020年)

https://doi.org/10.6342/NTU202004102

摘要

本研究試圖將地理區位概念應用到現有的交通碰撞事故模型中，並將交通運輸、GIS科學與自行車安全相結合。研究結果對交通工程和地理學科領域均有意義，對於交通運輸領域，新方法考慮了城市中自行車碰撞事故的空間總體趨勢，簡化了複雜碰撞事故資訊數樣量，並證明了其在實際案例研究中的適用性。另一方面，對於地理學科領域，這些新方法也考慮了自行車碰撞事故的局部趨勢，從明確的時間結果中提供了附加的資訊，以避免高風險碰撞區域發生自行車-機動車輛潛在的碰撞衝突。
自行車現已成為城市交通運輸工具中的綠色出行選項，許多城市的自行車策略已具體提升了自行車運具使用。但是在過去的二十年中，涉及自行車的肇事次數增加幅度遠大於城市中其他的交通運具，因此，在絕大多數的城市仍然因為自行車安全的考量而無法以自行車做為主要通勤工具。
本研究提出了「整合時空資訊 (spatio-temporal information)」的方法，其能更精確的了解城市自行車的肇事型態 (collision patterns)，並能進一步減輕城市中的自行車肇事風險。首先，本研究開發了空間模型，根據涉及自行車的肇事次數、傷亡程度、以及考慮城市自行車家戶持有人口 (household ownership) 密度下，找出城市中騎乘自行車的高風險地區 (high-risk areas)。另一方面，藉由時間建模，納入微觀地點的機動車輛 (motorized-vehicle) 與自行車流量，來了解自行車肇事發生的型態。最後，本研究將這兩種方法結合起來，以便從宏觀（區域） (macro-/regional)) 到微觀（位置）(micro-/locational) 層面來評估自行車與機動車輛發生碰撞事故的風險。
在案例研究中，分別對於三年2,044筆台北-首都地區（TCR）自行車碰撞事故、以及比利時安特衛普市五年4,210筆自行車碰撞事故進行分析。通過 階段1空間工作流程，城市橋樑/主要道路被確定為事故碰撞密度最高的區域。鑑於城市橋樑上/主要道路的自行車設施由於路面空間有限，路面高程以及交通量的差異而面臨道路設計上的困難，此研究結果並不令人意外，且值得 階段2研究進一步分析，包含交通工程設施、道路環境條件、交通控制系統和駕駛行為層面及其相對應的交通工程改善策略。案例研究的成果，提供「地理」和「交通運輸」領域使用的洞察力，不僅減少城市中特定的、易發生碰撞地點的自行車碰撞風險，而且還提供與交通工程師和城市規劃師有關的資訊，以提升城市中高風險地區的自行車安全。
本研究並根據案例分析成果，探討新興（台北）和相較成熟的自行車城市（安特衛普）之間自行車碰撞風險和高風險自行車環境的異同。從研究建立兩個城市的自行車與機動車輛碰撞事故時空工作流程中，顯示了主要的風險因素及其類別，包含道路環境條件，交通工程設施和交通控制系統。經上述系統的比較分析，本研究發現，與成熟的自行車城市相比，新興的自行車城市面臨的自行車網絡連通性較差，而改善自行車設施的佈局是當務之急，以防止自行車與機動車輛碰撞。另一方面，比較分析的結果也表明，在一個成熟的自行車城市中，自行車的碰撞主要是由道路環境條件引起的，碰撞與碰撞傷害的嚴重程度呈現同樣的趨勢，並且更多地集中在交通流量較高的位置。這可能歸因於相對成熟、設計較為完善的自行車道路工程設施，使得自行車與機動車輛碰撞碰撞自然而然地集中在交通最為繁忙的位置。
此外，本研究亦發現了交通控制系統在新興和成熟的自行車城市中都扮演著重要角色。此外，無論是在新興的自行車城市還是成熟的自行車城市，交通擁堵，不適當的道路標記，較大的路口尺寸，較長的路段長度，較高的速度限制和較長的號誌周期長度都可能增加自行車-機動車輛（BMV）碰撞的風險。
整體而言，本研究成果發現自行車碰撞及其風險具有時空相關特性，在空間和時間上進行建模和分析，可以更精準了解自行車-機動車輛碰撞事故的影響因素及其風險。整合地理資訊系統（GIS）和運輸科學促進了整體研究方法的發展，因此對於道路空間的跨學科觀點，充分探索了自行車安全的複雜性，可以有助於整體道路安全的提升。

關鍵字

自行車與機動車輛碰撞 ； 城市橋樑 ； (城市)主要幹道 ； 時空建模 ； 核密度估計 ； 負二項式建模 ； 地理資訊系統

並列摘要

The research presented in this dissertation tries to apply new concepts to existing traffic collision models, and combines transportation and GIS science with bicycle safety. The research outcomes are meaningful to both fields of transportation engineering and geography: for transportation field, new approaches consider the spatial overall trend of bicycle collisions in the city, simplify the amount of complex collision information and demonstrate their applicability in real-world case studies. On the other hand, for geography field, these new approaches also consider the local trend of bicycle collisions, provide additional information from temporally explicit results, and avoid potential conflicts from bicycle collision-prone areas.
Nowadays, the bicycle has taken a prominent place as an urban mode of transportation and many bicycle strategies have significantly contributed to the increase in the cycling mode share. However, the number of collisions involving bicycles has been increasing significantly more than for other transportation modes in the last two decades and its safety concern deters people from using bicycles for their daily trips.
In this dissertation, the contribution of spatio-temporal info to a better understanding of bicycle collision patterns and to the mitigation of bicycle collision risk is achieved. On the one hand, the spatial modelling is developed to assess high-risk areas in the city in terms of bicycle safety level and to consider population density. On the other hand, the temporal modelling aims to gain insights in patterns of bicycle collision occurrences by controlling traffic and bicycle flows. Finally, the two approaches are combined in order to estimate bicycle collision risks from a macro-/regional to a micro-/locational level.
For the presented Case Studies, 2044 collisions in Taipei (Taiwan) were analysed while 4210 collisions in Antwerp (Belgium) were analysed. The data sets cover 3 and 5years respectively and includes all BMV collisions reported by the police. Through the spatial workflow, urban bridges/arterials were identified as areas with the highest density of collisions. This is unsurprising given that bicycle facilities on urban bridges/arterials face design difficulties due to limited space, discrepancy in elevation and traffic volume. Through this approach the characteristics of BMV collisions on either bridges or arterials, traffic engineering, road environment, traffic control system, and driving behaviour were then analysed in the temporal dimension. Both Case Studies conclude by not only providing an insight that can be used by both the geography and transport fields to reduce bicycle collisions risks on specific urban collision-prone locations, but also providing information relevant to traffic engineers and city planners concerning the enhancement of bicycle safety in high-risk areas in the city.
Comparison analyses have also been conducted to explore the differences and similarities of the bicycle collision risks and high-risk cycling environments between an emerging (Taipei) and a comparatively mature cycling city (Antwerp). The major contributing risk factors and the categories they belong to (i.e. road environmental conditions, traffic engineering facilities and traffic control systems) have been highlighted from the spatio-temporal workflow among bicycle-motorised vehicle collisions in these two cities.
After the above-mentioned systematic comparative analysis, we have summarized that compared to a mature cycling city, an emerging cycling city faces a poorer connectivity of the bicycle networks, and improving the layout of the bicycle engineering facilities has proved to be a priority strategy so as to prevent bicycle motorised vehicle collisions. On the other hand, the result of the comparative analysis has also demonstrated that the bicycle collisions in a mature cycling city are mostly caused by road environmental conditions, are in line with the collision injury severities and are much more clustered on the locations with a large amount of traffic. They might be attributed to their mature and well-designed road engineering facilities, which make BMV collisions naturally concentrate on the locations with high traffic exposure.
Notably, we have also found that traffic control systems play an important role across both emerging and mature cycling cities. In addition, whether in an emerging or a mature cycling city, the traffic exposure, inappropriate road marking, a larger size of road junctions, a longer length of road segments, high speed limits and a longer signal cycle length are likely to increase the BMV collision risk.
In summary, the presented research builds on the assumption that bicycle collisions and their risks are spatial and temporal. Thus, it can be modelled and analysed spatially and temporally. GIS and transportation science facilitate holistic approaches where inter-disciplinary perspectives on the road space are considered and the complexity of bicycle safety is adequately addressed.

並列關鍵字

bicycle-motorised vehicle collisions ； urban bridge ； arterial roads ； spatio-temporal modelling ； Kernel Density Estimation (KDE) ； negative binomial modelling ； GIS

參考文獻

AASHTO, American Association of State, H., Transportation, O. (1999). Guide for the development of bicycle facilities.

AASHTO, American Association of State, H., Transportation, O. (2007). AASHTO LRFD bridge design specifications.

Abdel-Aty, M., Chundi, S. S., Lee, C. (2007). Geo-spatial and log-linear analysis of pedestrian and bicyclist crashes involving school-aged children. Journal of Safety Research, 38(5), 571-579. doi:http://dx.doi.org/10.1016/j.jsr.2007.04.006

Abdel-Aty, M., Ekram, A.-A., Huang, H., Choi, K. (2011). A study on crashes related to visibility obstruction due to fog and smoke. Accident Analysis Prevention, 43(5), 1730-1737. doi:https://doi.org/10.1016/j.aap.2011.04.003

Abdel-Aty, M., Siddiqui, C., Huang, H., Wang, X. (2011). Integrating Trip