道路交通事故（Road Traffic Accidents, RTA）受到交通、環境及氣候等多重因素影響，其中細懸浮微粒（PM2.5）與日內溫差（Diurnal Temperature Range, DTR）近年被認為可能增加交通事故風險。然而，既有研究對其影響方向與程度尚未獲得一致結論，且不同地理區域是否存在差異仍缺乏探討。因此，本研究以臺灣六都為研究範圍，整合 2012 年至 2024 年交通事故、空氣品質及氣象資料，建立追蹤資料（panel data）模型，分析 PM2.5、DTR 及其交互作用與交通事故之關聯；另以 2014 年至 2024 年行政區資料進行分區分析，探討不同地理位置下之空間異質性。
六都整體分析結果顯示，在未納入時間趨勢（time trend）控制時，PM2.5、DTR 及其交互作用於部分城市呈現統計顯著，其中臺南市 DTR 估計係數為 4.09（p < 0.05），高雄市 DTR 估計係數為 -2.51（p < 0.05）；高雄市 PM2.5 估計係數為 -1.09（p < 0.10），交互作用估計係數為 0.63（p < 0.10）。然而，當模型納入時間趨勢後，多數原先顯著之估計結果皆不再具有統計顯著性，僅桃園市 PM2.5 及其交互作用仍保留顯著結果，顯示長期時間結構對環境因子與交通事故關聯之估計具有重要影響。
進一步之行政區分區分析顯示，臺南市善化區、新北市林口區及高雄市鳳山區仍觀察到 PM2.5、DTR 或交互作用之顯著估計結果，而多數行政區之時間趨勢項則呈現高度統計顯著，顯示交通事故變化除受長期時間結構影響外，不同地理區域之環境暴露與交通事故關聯亦具有明顯空間異質性。
綜合而言，本研究未發現 PM2.5、DTR 及其交互作用對交通事故具有一致且穩定之整體影響；相較之下，時間趨勢控制與空間尺度會明顯改變估計結果。研究結果顯示，分析環境因子與交通事故之關聯時，若忽略長期時間趨勢或區域差異，可能導致估計結果失真或掩蓋局部效果。未來研究可進一步納入交通流量、降雨、道路型態及其他環境暴露因子，以建立更完整之交通事故風險評估模型。

Road traffic accidents (RTA) are influenced by multiple factors related to traffic, environmental conditions, and climate. Among these factors, fine particulate matter (PM2.5) and diurnal temperature range (DTR) have recently been considered potential contributors to traffic accident risk. However, previous studies have not reached consistent conclusions regarding the direction and magnitude of their associations with traffic accidents, and geographical differences in these relationships remain insufficiently explored. Therefore, this study examines Taiwan’s six major municipalities by integrating traffic accident, air quality, and meteorological data from 2012 to 2024. Panel data models are employed to analyze the associations of PM2.5, DTR, and their interaction with traffic accidents. In addition, district-level data from 2014 to 2024 are used to investigate spatial heterogeneity across different geographical areas.
The city-level results show that, without controlling for a time trend, PM2.5, DTR, and their interaction are statistically significant in several cities. In Tainan, the estimated coefficient for DTR is 4.09 (p < 0.05), while in Kaohsiung the estimated coefficient for DTR is -2.51 (p < 0.05). In Kaohsiung, the estimated coefficient for PM2.5 is -1.09 (p < 0.10), and the estimated interaction coefficient is 0.63 (p < 0.10). However, after a time trend is included in the model, most of these associations are no longer statistically significant. Only PM2.5 and its interaction term in Taoyuan remain statistically significant, indicating that long-term temporal structure plays an important role in estimating the associations between environmental factors and traffic accidents.
Further district-level analysis shows that statistically significant estimates for PM2.5, DTR, or their interaction are still observed in Shanhua District in Tainan, Linkou District in New Taipei, and Fengshan District in Kaohsiung. Meanwhile, the time-trend terms are highly statistically significant in most districts, suggesting that long-term temporal structure remains important in explaining changes in traffic accidents.
Overall, this study does not find consistent and stable associations of PM2.5, DTR, or their interaction with traffic accidents across the study areas. Instead, controlling for temporal trends and changing the spatial scale of analysis substantially alter the estimated associations. The findings suggest that ignoring long-term temporal trends or regional differences when examining environmental factors and traffic accidents may distort the estimated associations or mask localized effects. Future studies may further incorporate traffic volume, precipitation, road characteristics, and other environmental exposures to develop a more comprehensive framework for assessing traffic accident risk.

[1] Awais Piracha and Muhammad Tariq Chaudhary. Urban air pollution, urban heat island and human health: A review of the literature. Sustainability, page 19, 2022.
[2] José Ignacio Nazif-Munoz, Pablo Martínez, Augusta Williams, and John Spengler. The risks of warm nights and wet days in the context of climate change: assessing road safety outcomes in boston, usa and santo domingo, dominican republic. Injury Epidemiology, page 9, 2021.
[3] Yue Wan, Yuhang Li, Chunhong Liu, and Zhongqiu Li. Is traffic accident related to air pollution? a case report from an island of taihu lake, china. Atmospheric Pollution Research, page 6, 2020.
[4] Emtiaz Hossain Hritan and Pruek Vanna-iampikul. The impacts of air pollution on traffic accidents across the united states. Working Paper, page 13, 2021.
[5] Ling Fan and Mingqin Chen. The effects of pm2.5 concentrations on traffic violations and accident severity in guangdong china. Scientific Reports, page 18, 2025.
[6] Tian Luan, Xueliang Guo, Lijun Guo, and Tianhang Zhang. Quantifying the relationship between PM2.5 concentration, visibility and planetary boundary layer height for long-lasting haze and fog–haze mixed events in beijing. Atmospheric Chemistry and Physics, page 23, 2018.
[7] 劉遵賢, 陳昱均, 蔡孟光, and 葉富豪. 海岸氣象對空氣污染物擴散的影響. 環保簡訊, page 16, 2015.
[8] Hsuan-Min Kao, Yi-Ching Lo, Chung-Yi Li, I-Lin Hsu, Yu-Tai Lo, Hon-Ping Ma, Ping-Ling Chen, Pei-Yi Wong, Chih-Da Wu, Muhammad Atoillah Isfandiari, Yun-Kai Wen, and Ya-Hui Chang. Fine particulate matter exposure and risk of scooter collisions among older adults. Injury Epidemiology, page 9, 2025.
[9] Ta-Chien Chan, Chih-Wei Pai, Chia-Chieh Wu, Jason C. Hsu, Ray-Jade Chen, Wen-Ta Chiu, and Carlos Lam. Association of air pollution and weather factors with traffic injury severity: A study in taiwan. International Journal of Environmental Research and Public Health, page 15, 2022.
[10] Yongheng Deng, Jia He, Bingqing Li, and Xixi Shen. Does air pollution cause more car accidents? evidence from auto insurance claims. China Economic Review, page 23, 2024.
[11] Benjamin de Foy and James J. Schauer. Interpretable diurnal impacts on extreme urban pm2.5 concentrations of soil temperature, soil water content, humidity and temperature inversion. Atmospheric Research, page 107500, 2024.
[12] Madina Tursumbayeva, Aiymgul Kerimray, Ferhat Karaca, and Didin Agustian Per-madi. Planetary boundary layer and its relationship with pm2.5 concentrations in almaty, kazakhstan. Aerosol and Air Quality Research, page 210294, 2022.
[13] Limei Ke, Yao Zhang, Yingyao Fu, Xinke Shen, Yu Zhang, Xindong Ma, and Qian Di. Short-term pm2.5 exposure and cognitive function: Association and neurophysiological mechanisms. Environment International, page 11, 2022.
[14] Yuo-Hsien Shiau, Su-Fen Yang, Rishan Adha, Syamsiyatul Muzayyanah, and Giia-Sheun Peng. The exposure-response of air pollution and climate change to chronic respiratory diseases: Does residential energy efficiency matter? Urban Climate,
51:101649, 2023.
[15] Cheng-Kai Hsu. Burning gig, rewarding risk: Effects of dual exposure to incen-tive structure and heat condition on risky driving among on-demand food-delivery motorcyclists in kaohsiung, taiwan. Accident Analysis & Prevention, page 107841, 2025.
[16] Won-Kyung Lee, Hye-Ah Lee, Seung sik Hwang, Ho Kim, Youn-Hee Lim, Yun-Chul Hong, Eun-Hee Ha, and Hyesook Park. A time series study on the effects of cold temperature on road traffic injuries in seoul, korea. Environmental Research, page 7, 2014.
[17] Yuo-Hsien Shiau, Su-Fen Yang, Rishan Adha, Giia-Sheun Peng, and Syamsiyatul Muzayyanah. Dynamic and non-linear analysis of the impact of diurnal temperature range on road traffic accidents. Climate, page 20, 2023.
[18] Ministry of Transportation and Communications. 交 通 統 計 查 詢 網, 2025. Ac-cessed: 2025-07-28.
[19] Ministry of Environment. 空氣品質監測歷史資料查詢系統, 2025. Accessed:2025-07-28.
[20] Central Weather Administration. Codis 氣 候 資 料 服 務 系 統, 2025. Accessed:2025-07-28.
[21] 徐彣. 空氣品質對風險行為的影響—以交通事故與違規為例. Master’s thesis, 國立臺灣大學, 2021.
[22] Tor-Olav Nævestad, Fridulv Sagberg, Gunhild Levlin, and Torkel Bjørnskau. Competence, equipment and behavioural adaptation on norwegian winter roads: A comparison of foreign and norwegian hgv drivers. Transportation Research Part F: Traffic Psychology and Behaviour, pages 257–273, 2021