Data Reporter

Data ReporterData Reporter》

2026年5月甫正式通車的淡江大橋，因實體機車道寬度過窄引發爭議。不過，爭議並非只在淡江大橋，台灣其他橋梁機車道長期也都有實體分隔過窄與事故頻傳的問題，橋梁上的傷亡事故有9成以上都是機車事故。在民間開始注重交通安全與機車騎士權益後，新北市政府近年來開始陸續拆除橋梁的機車道實體分隔島、拓寬機車道，取得初步成效。《報導者》使用獨家開發的事故地圖進一步分析新北市橋梁，發現曾拆除實體分隔的路段事故下降7、8成，而交通量相近、採汽機車混流的橋梁更比實體機車道分隔的橋梁，事故下降2到4倍。

The Tamsui River Bridge, which formally opened to traffic only in May 2026, has stirred controversy because its physical motorcycle lane is too narrow. Yet the controversy is not confined to the Tamsui River Bridge: for years, motorcycle lanes on other bridges in Taiwan have also suffered from narrow physical dividers and frequent crashes, and more than 90 percent of bridge casualties involve motorcycles. As civic groups have begun paying closer attention to traffic safety and motorcyclists’ rights, the New Taipei City Government has in recent years started removing the physical divider islands that separate bridge motorcycle lanes and widening those lanes, with preliminary results. Using an accident map it developed exclusively, The Reporter further analyzed bridges in New Taipei and found that crashes fell by 70 to 80 percent on sections where physical separation had been removed, while on bridges with comparable traffic volumes, crashes were two to four times lower on mixed car-motorcycle bridges than on bridges with physically separated motorcycle lanes.

改革需要共識，共識又需要實證數據才能凝聚。新北市交通局將開啟機車道與安全的相關研究，學者專家也認為拋棄過往「實體分隔才能保護機車」的車種分流舊思維，改用國際上常見的車速分流，「不強制機車走機車道」或許是台灣的交通改革新方向。

Reform requires consensus, and consensus requires empirical data if it is to cohere. The New Taipei City Transportation Department will begin research on motorcycle lanes and safety, and scholars and experts also argue that Taiwan may need to abandon the old vehicle-segregation thinking that “only physical separation can protect motorcycles” and instead adopt the speed-based separation common internationally. “Not forcing motorcycles to use motorcycle lanes” may become a new direction for Taiwan’s traffic reform.

天氣舒爽的平日午夜，大漢橋新莊端的機車道上有數台工程車、壓路機和10多名工人，他們正在進行劃設機車道標線的工程。前幾天，工人們已經將分隔快車道和機車專用道的實體水泥分隔島拆除，今夜等柏油鋪完後即可劃上熱拌漆標線，因為橋梁都是直線畫起來相對快，一個晚上就能完工。

On a weekday midnight in pleasant weather, several engineering vehicles, a road roller, and more than a dozen workers were on the motorcycle lane at the Xinzhuang end of Dahan Bridge, carrying out work to mark motorcycle-lane lines. A few days earlier, the workers had removed the concrete divider island separating the fast lane from the motorcycle-only lane. Once the asphalt was finished that night, they could paint the hot-mix road markings; because bridges are straight, the lines are relatively quick to draw, and the work can be completed in a single night.

工人領班說，民眾、議員這幾年愈來愈注重機車的權益，3、4年前他們開始打通橋梁的分隔島，每一段距離就把分隔島敲掉做出缺口，讓機車道發生事故而塞車時，後方機車可以提前從缺口駛進汽車道離開。今年（2026）他們則是同時施作大漢橋雙向全段和浮洲橋部分的分隔島拆除工程。

The crew foreman said that in recent years the public and city councilors have paid increasing attention to motorcyclists’ rights. Three or four years ago, they began opening gaps in bridge divider islands, breaking away sections at intervals so that when a crash blocked the motorcycle lane, motorcycles behind it could leave early through the gap and enter the car lane. This year, 2026, they are simultaneously removing divider islands along the full length of Dahan Bridge in both directions and on part of Fuzhou Bridge.

大漢橋另一端，板橋往新莊方向的機車道實體分隔於今年4月就已全改成標線，差異很明顯。領班說，因為板橋端匝道複雜，以前幾乎每週都會發生汽車誤闖機車道的情形，「汽車還是可以開機車道，但因為很窄，有些駕駛怕輪胎擦到分隔島，就開超級慢，後面機車全部都塞住。」現在改劃標線後，還是不少汽車會誤闖機車道，但上橋後可以直接切回快車道，不再有塞住機車道的問題。

At the other end of Dahan Bridge, the physical separation of the motorcycle lane from Banqiao toward Xinzhuang had already been fully converted into lane markings in April this year, and the difference is obvious. The foreman said that because the ramps at the Banqiao end are complex, cars used to stray into the motorcycle lane almost every week. “Cars can still drive in the motorcycle lane, but because it is very narrow, some drivers are afraid their tires will scrape the divider island, so they drive extremely slowly, and all the motorcycles behind them get stuck.” Now that the lane has been changed to markings, many cars still mistakenly enter the motorcycle lane, but after getting onto the bridge they can shift directly back into the fast lane, so the motorcycle lane no longer gets blocked.

凌晨，大漢橋的機車道標線終於完工。隔天午夜，這批工人又開始到隔壁連結板橋和樹林的浮洲橋，繼續拆除機車分隔島的工程。

In the early hours, the motorcycle-lane markings on Dahan Bridge were finally finished. At midnight the next day, the same workers began work on the neighboring Fuzhou Bridge, which connects Banqiao and Shulin, continuing the removal of motorcycle divider islands.

2026年5月，連接板橋和新莊的大漢橋兩側機車道實體分隔島全部拆除，改以槽化帶跟汽車道區隔。機車道寬度由原本2.65公尺拓寬至3公尺。浮洲橋也拆除南側（樹林端往板橋方向）上橋約200公尺範圍的機車道實體分隔島。

In May 2026, all physical divider islands for the motorcycle lanes on both sides of Dahan Bridge, which connects Banqiao and Xinzhuang, were removed and replaced with channelized markings separating them from the car lanes. The motorcycle lanes were widened from 2.65 meters to 3 meters. On Fuzhou Bridge, the physical motorcycle-lane divider island was also removed on the south side, in the roughly 200-meter section leading onto the bridge from the Shulin end toward Banqiao.

為什麼要拆機車道實體分隔島？新北市交通局專門委員林昭賢接受《報導者》訪問時指出，因為橋梁直線段車速比較快，早期橋梁設計都採分流設計，認為可以保護機車。但台灣機車數量每年都在增加，過去的機車道寬度與容量早已不足，再加上這幾年很多民間團體跟媒體開始重視機車權益，才開始進行機車道實體分隔拆除跟拓寬車道。

Why remove the physical divider islands for motorcycle lanes? Lin Chao-hsien, a senior specialist at the New Taipei City Transportation Department, told The Reporter that because vehicles travel faster on straight bridge sections, early bridge designs used segregation, believing it could protect motorcycles. But the number of motorcycles in Taiwan has risen every year, and the width and capacity of motorcycle lanes from the past have long been insufficient. Added to that, many civic groups and media outlets in recent years have begun emphasizing motorcyclists’ rights, which is why the removal of physical dividers and the widening of motorcycle lanes have begun.

《報導者》比較兩種橋梁，第一種為同一橋梁拆除機車道分隔島路段的前後事故差異，包括福和橋與重新橋；第二種為無機車道設計、汽機車混流的橋梁。比較的橋梁中，實體分隔機車道橋梁的機車道橫斷面寬度皆為約2.5公尺，而混流的車道寬約為3.5～4公尺。

The Reporter compared two kinds of bridges. The first involved before-and-after differences in crashes on sections of the same bridge where motorcycle-lane divider islands were removed, including Fuhe Bridge and Chongxin Bridge. The second involved bridges designed without motorcycle lanes, where cars and motorcycles mix. Among the bridges compared, the cross-sectional width of motorcycle lanes on bridges with physically separated motorcycle lanes was about 2.5 meters, while mixed-traffic lanes were about 3.5 to 4 meters wide.

本文標定的這6座橋的傷亡事故幾乎都是機車事故：從2020年1月起至2026年4月，總計2,621起傷亡事故中機車涉入有2,548起，高達97%，僅有3％的傷亡事故與機車無關。如何降低機車騎士傷亡，成為橋梁設計的重要課題。

Almost all casualty crashes on the six bridges marked in this article were motorcycle crashes. From January 2020 through April 2026, motorcycles were involved in 2,548 of the 2,621 casualty crashes, as high as 97 percent; only 3 percent of casualty crashes had nothing to do with motorcycles. How to reduce deaths and injuries among motorcyclists has therefore become a key issue in bridge design.

連結台北市中正區、文山區與新北市永和區的福和橋，在2023年7月拆除北側機車道實體分隔島。為排除剛施工後駕駛習慣改變的數據差異影響，分別採取前兩年（2021年1月到2022年12月）和後兩年（2024年1月到2025年12月）的數據比較。

Fuhe Bridge, which links Taipei’s Zhongzheng and Wenshan districts with Yonghe District in New Taipei, removed the physical divider island on its north-side motorcycle lane in July 2023. To exclude distortions caused by changes in driving habits immediately after construction, the comparison uses data from the two years before, January 2021 to December 2022, and the two years after, January 2024 to December 2025.

福和橋拆除分隔島後，車禍數據皆有減少，後兩年共有122起傷亡事故，其中機車涉入的事故為117起，追撞類事故86起為最大宗，自撞31起次之。記錄在地圖範圍內共有61起事故點位，在台北端上橋處卻僅有3起事故，同一處約減少8成事故。

After the divider island on Fuhe Bridge was removed, crash figures declined across the board. In the following two years there were 122 casualty crashes, of which 117 involved motorcycles; rear-end-type crashes were the largest category, at 86, followed by 31 single-vehicle crashes. Within the mapped area, 61 crash locations were recorded, yet at the Taipei-end bridge entrance there were only 3 crashes, a reduction of about 80 percent at the same location.

拆除分隔島前兩年，重新橋共有317起傷亡事故，其中機車涉入的事故為312起，追撞類事故224起為最大宗，自撞66起其次。超過兩台機車以上的連環追撞車禍就有120起，其中有19起是5台機車以上的事故。記錄在地圖範圍內共有278事故點位，其中新莊端上橋起始點約50公尺範圍內有9起，也就是南側機車道實體分隔島的起點。

In the two years before the divider island was removed, Chongxin Bridge had 317 casualty crashes, of which 312 involved motorcycles. Rear-end-type crashes were the largest category, at 224, followed by 66 single-vehicle crashes. There were 120 chain rear-end crashes involving more than two motorcycles, including 19 crashes involving five or more motorcycles. Within the mapped area, 278 crash locations were recorded, among them 9 within roughly 50 meters of the starting point onto the bridge at the Xinzhuang end, the beginning of the physical divider island on the south-side motorcycle lane.

拆除分隔島後兩年，重新橋共有188起傷亡事故，其中機車涉入的事故為185起，追撞類事故123起為最大宗，自撞52起其次。超過兩台機車上的連環追撞車禍就有68起，其中有10起是5台機車以上的事故。記錄在地圖範圍內共有164事故點位，在新莊端上橋處50公尺範圍內卻僅有3起事故，同一處約減少近7成事故。

In the two years after the divider island was removed, Chongxin Bridge had 188 casualty crashes, of which 185 involved motorcycles. Rear-end-type crashes were the largest category, at 123, followed by 52 single-vehicle crashes. There were 68 chain rear-end crashes involving more than two motorcycles, including 10 crashes involving five or more motorcycles. Within the mapped area, 164 crash locations were recorded, yet there were only 3 crashes within 50 meters of the bridge entrance at the Xinzhuang end, a reduction of nearly 70 percent at the same location.

為何拆除實體分隔島可以減少機車事故？台灣機車路權促進會理事、企鵝交通手札創辦人吳祥瑀分析，當機車從較寬的平面汽機車混流車道匯入到橋梁上較窄的機車道時需要整理車流成一或二排，此時車與車之間容易產生擦碰。若把機車道入口的實體分隔島拆除，除了減少視覺壓力外，也能延長車流整理的緩衝跟容錯空間，因而減少自撞、追撞、擦撞事故。

Why can removing physical divider islands reduce motorcycle crashes? Wu Hsiang-yu, a board member of the Taiwan Motorcycle Rights Promotion Association and founder of Penguin Traffic Notes, analyzes that when motorcycles move from wider surface-level lanes where cars and motorcycles mix into narrower motorcycle lanes on a bridge, the traffic flow must be sorted into one or two rows, and vehicles are then prone to brushing or colliding with one another. If the physical divider island at the entrance to the motorcycle lane is removed, it not only reduces visual pressure but also lengthens the buffer and margin for error as the traffic flow sorts itself out, thereby reducing single-vehicle crashes, rear-end crashes, and sideswipes.

如果僅是拆除機車道實體分隔改為標線型機車道就能減少事故，那汽機車完全混流的橋梁的事故風險又有何差異？

If merely removing the physical divider of a motorcycle lane and replacing it with a lane marked by paint can reduce crashes, then how does the crash risk differ on bridges where cars and motorcycles are fully mixed?

《報導者》選定大漢溪流域三座有交通量數據的橋梁，大漢橋和浮洲橋在今年3月和5月開始拆除機車道實體分隔島前，都屬於雙向有機車道實體分隔的橋梁；新海橋則沒有機車道，汽機車完全混流。

The Reporter selected three bridges in the Dahan River basin for which traffic-volume data are available. Before Dahan Bridge and Fuzhou Bridge began removing physical motorcycle-lane divider islands in March and May of this year, both were bridges with physically separated motorcycle lanes in both directions. Xinhai Bridge, by contrast, has no motorcycle lane, and cars and motorcycles are fully mixed.

- 大漢橋有467起事故，其中446起為機車事故；因採汽機車車道實體分隔，機車事故中，和汽車碰撞的僅有25起。機車事故中的追撞類事故有346起，其次為自撞55起。

- Dahan Bridge had 467 crashes, of which 446 were motorcycle crashes. Because cars and motorcycles were physically separated by lane, only 25 motorcycle crashes involved collisions with cars. Among motorcycle crashes, rear-end-type crashes numbered 346, followed by 55 single-vehicle crashes.

- 浮洲橋則有556起事故，其中545起為機車事故，機車事故中和汽車碰撞的有32起；機車事故中的追撞類事故有359起，其次為自撞116起。

- Fuzhou Bridge had 556 crashes, of which 545 were motorcycle crashes; among motorcycle crashes, 32 involved collisions with cars. Rear-end-type crashes accounted for 359 motorcycle crashes, followed by 116 single-vehicle crashes.

- 新海橋有133起事故，其中129起為機車事故，機車事故中的追撞類事故有68起，其次為自撞40起；機車涉入的事故中，和汽車碰撞的有36起。

- Xinhai Bridge had 133 crashes, of which 129 were motorcycle crashes. Among motorcycle crashes, rear-end-type crashes numbered 68, followed by 40 single-vehicle crashes. Of the crashes involving motorcycles, 36 involved collisions with cars.

以新海橋為基準對比交通量和事故數，大漢橋和浮洲橋的機車通行量約為新海橋的1.13倍和0.95倍，但機車傷亡事故卻為新海橋的3.46倍和4.22倍。事故類型比例也有所不同，大漢橋和浮洲橋的追撞類事故占78％和66％為最多，新海橋的追撞類事故雖也占最多比例，但僅有53％。

Using Xinhai Bridge as the baseline to compare traffic volume and crash counts, motorcycle traffic on Dahan Bridge and Fuzhou Bridge was about 1.13 times and 0.95 times that of Xinhai Bridge, respectively, yet motorcycle casualty crashes were 3.46 times and 4.22 times those on Xinhai Bridge. The proportions of crash types also differed. Rear-end-type crashes made up the largest shares on Dahan Bridge and Fuzhou Bridge, at 78 percent and 66 percent. On Xinhai Bridge, rear-end-type crashes were also the largest category, but only at 53 percent.

此外，《報導者》亦選定新店溪流域兩座有交通量數據且鄰近的橋梁：福和橋在2023年7月拆除北側機車道實體分隔島前屬於雙向都有機車道實體分隔的橋梁，永福橋則沒有機車道，汽機車完全混流。兩橋比較的數據皆為2020年起至2023年6月的區間，根據新北市交通局資料，福和橋與永福橋平日尖峰平均每小時雙向通行6,016輛和4,731輛機車。

The Reporter also selected two nearby bridges in the Xindian River basin for which traffic-volume data are available. Before Fuhe Bridge removed the physical divider island from its north-side motorcycle lane in July 2023, it had physically separated motorcycle lanes in both directions; Yongfu Bridge, by contrast, has no motorcycle lane, and cars and motorcycles are fully mixed. The comparison data for both bridges cover the period from 2020 through June 2023. According to New Taipei City Transportation Department data, Fuhe Bridge and Yongfu Bridge carried average weekday peak-hour two-way motorcycle volumes of 6,016 and 4,731 motorcycles, respectively.

- 福和橋拆除前有279起事故，其中266起為機車事故。因採汽機車車道實體分隔，機車事故中，和汽車碰撞的僅有6起；機車事故中的追撞類事故有178起，其次為自撞66起。

- Before the removal on Fuhe Bridge, there were 279 crashes, of which 266 were motorcycle crashes. Because cars and motorcycles were physically separated by lane, only 6 motorcycle crashes involved collisions with cars. Rear-end-type crashes accounted for 178 motorcycle crashes, followed by 66 single-vehicle crashes.

- 永福橋則僅有91起事故，其中86起為機車事故。機車事故中自撞有36起，其次依序為追撞類29起和21起擦撞類事故；機車涉入的事故中，和汽車碰撞的有30起，追撞類和擦撞類事故各占約一半。

- Yongfu Bridge had only 91 crashes, of which 86 were motorcycle crashes. Among motorcycle crashes, 36 were single-vehicle crashes, followed in order by 29 rear-end-type crashes and 21 sideswipe-type crashes. Of the crashes involving motorcycles, 30 involved collisions with cars, and rear-end-type and sideswipe-type crashes each accounted for about half.

用交通量和事故數對比，福和橋的機車通行量約為永福橋的1.27倍，但機車傷亡事故卻為永福橋的2.92倍。事故類型也有所不同，福和橋的追撞類事故占67％為最多，永福橋的追撞類事故僅有34％。

Comparing traffic volume and crash counts, motorcycle traffic on Fuhe Bridge was about 1.27 times that of Yongfu Bridge, yet motorcycle casualty crashes were 2.92 times those on Yongfu Bridge. Crash types also differed: on Fuhe Bridge, rear-end-type crashes were the largest category, at 67 percent, while on Yongfu Bridge they accounted for only 34 percent.

數據是否代表機車道實體分隔對機車騎士造成的安全風險高於汽機車混流設計？逢甲大學中區區域運輸發展研究中心副主任鍾慧諭說，若數據差異不大，需要更大量數據分析，但這兩組的數據差異相當明顯，機車道實體分隔橋梁的機車事故風險明顯大於汽機車混流的橋梁。她分析，當機車數量超過一定密度，又被實體分隔限制機車在較窄的車道，碰撞機率大，事故數自然增加。

Do the data mean that physical separation of motorcycle lanes creates greater safety risks for motorcyclists than a mixed car-motorcycle design? Chung Hui-yu, deputy director of Feng Chia University’s Central Taiwan Regional Transportation Development Research Center, says that if the differences in the data were small, a much larger dataset would be needed for analysis; but in these two sets, the differences are quite clear, and the motorcycle crash risk on bridges with physically separated motorcycle lanes is plainly greater than on bridges with mixed car-motorcycle traffic. Her analysis is that once the number of motorcycles exceeds a certain density, and physical separation confines them to narrower lanes, the probability of collision rises and the number of crashes naturally increases.

除了橋梁之外，平面道路也不乏有實體分隔機車道的設計。往來台北市區和新北淡水的大度路，原本約有3.6公里筆直的實體分隔島，但在今年2月，台北市拆除全段的分隔島並取消機車專用道，將原本3線禁行機車的快車道和2線機車道配置，改為4線車道，內側2車道禁行機車，外側2車道供汽機車混流行駛。

Beyond bridges, physical separation for motorcycle lanes is also common on surface roads. Dadu Road, which runs between central Taipei and Tamsui in New Taipei, originally had a straight physical divider island of about 3.6 kilometers. In February of this year, however, Taipei City removed the divider island along the entire stretch and abolished the motorcycle-only lane, changing the road from a configuration of three fast lanes closed to motorcycles and two motorcycle lanes into four lanes, with the two inner lanes closed to motorcycles and the two outer lanes open to mixed car-motorcycle traffic.

吳祥瑀認為，台北市政府拆除機車道分隔島並非為了提升機車騎士安全，而是因應淡北道路工程增加，大車增多而調整路型。有些人會覺得汽車加入變危險，但對機車騎士而言，可騎乘的空間從原本僅有約2.5公尺寬，變成兩個車道約7公尺寬，超車空間、駕駛容錯度都提高，安全的效益應該更多。也因為拆除分隔島後，右轉汽車可以提前進入最外側車道右轉，未來應該會減少許多右轉汽車撞擊直行汽車的「向右側撞」的事故。

Wu Hsiang-yu believes that Taipei City Government did not remove the motorcycle-lane divider island in order to improve motorcyclist safety, but to adjust the road layout in response to increased construction on the Danbei Road project and a rise in large vehicles. Some people may feel that adding cars makes things more dangerous, but for motorcyclists, the space they can ride in has expanded from only about 2.5 meters wide to two lanes, about 7 meters wide. Passing room and the margin for driver error both increase, so the safety benefits should be greater. And because, after the divider island is removed, right-turning cars can enter the outermost lane earlier to turn right, many “right-side impact” crashes in which right-turning cars hit cars going straight should be reduced in the future.

從事故地圖數據可知，實體分隔機車道改用標線分隔車道後，機車事故會減少；汽機車混流橋梁的機車事故，也比實體分隔機車道來得少。下一個問題是：車種分流真的不安全嗎？現行分流的道路設計又要怎麼調整？

The accident-map data show that after physically separated motorcycle lanes are changed to lanes separated by markings, motorcycle crashes decline; motorcycle crashes on mixed car-motorcycle bridges are also fewer than on bridges with physically separated motorcycle lanes. The next question is: is vehicle-type segregation really unsafe? And how should existing segregated road designs be adjusted?

吳祥瑀認為，不管是混流還是分流，事故的關鍵在於速差，特別看到的是機車道內「機車與機車之間的速差」。台灣因為駕照取得門檻極低，騎士的駕車技術、速度差異大，很多人把機車當腳踏車在騎，就容易被超車，過窄的車道加上速差大就容易發生事故。

Wu Hsiang-yu believes that whether traffic is mixed or segregated, the key to crashes lies in differences in speed, especially the “speed differences among motorcycles” inside motorcycle lanes. Because Taiwan’s licensing threshold is extremely low, riders vary widely in driving skill and speed. Many people ride motorcycles as if they were bicycles, making them easy to overtake; narrow lanes combined with large speed differences make crashes likely.

此外，橋梁上的分流設計在遇到快車道有匝道匯出時，也會迫使單純要直行下橋的機車道，必須要往上或下繞一個彎才有機會回到平面道路主線上。吳祥瑀指出，除了彎道容易產生事故外，騰空且過窄的機車道也因為救護車難抵達產生救援困難問題。

In addition, when a fast lane on a bridge has an exit ramp, the segregated design on the bridge also forces motorcycle lanes that are simply meant to carry riders straight off the bridge to bend upward or downward before they have a chance to return to the main surface road. Wu Hsiang-yu notes that, besides the fact that curves can easily generate crashes, elevated and overly narrow motorcycle lanes also create rescue difficulties because ambulances have trouble reaching them.

鍾慧諭指出，台灣大部分平面車道都是汽機車混流，在橋梁上汽機車混流也不會是問題，但一定要減少機車任意變換車道、鑽車行為，這部分可加強教育跟執法。

Chung Hui-yu points out that most surface lanes in Taiwan already mix cars and motorcycles, and mixing cars and motorcycles on bridges would not be a problem either, but arbitrary motorcycle lane-changing and weaving must be reduced; this can be addressed through stronger education and enforcement.

吳祥瑀跟鍾慧諭都建議，橋梁機車道也可以比照不強制待轉的邏輯，改為「不強制走機車道」，讓跟得上汽車車流速度的機車就走快車道，不敢騎很快的機車騎士就繼續騎機車道，用車速分流思維取代車種分流，可望大幅減少速差、降低事故。

Wu Hsiang-yu and Chung Hui-yu both suggest that bridge motorcycle lanes could follow the logic of not mandating two-stage left turns, and be changed to “not mandating use of the motorcycle lane.” Motorcycles that can keep up with the speed of car traffic could use the fast lanes, while riders who do not dare ride very fast could continue using the motorcycle lane. Replacing vehicle-type segregation with speed-based separation could greatly reduce speed differentials and lower crashes.

只是，前提要件除了社會需要高度共識外，也需要有扎實的交通安全實證研究結論支持。從機車道實體分隔、維持機車道但改以標線分隔，到汽機車混流，哪種條件更安全更有效率？鍾慧諭說，「台灣幾乎沒有這樣的研究，是『零』，我們極度欠缺。」

The prerequisites, however, include not only a high degree of social consensus, but also solid empirical research findings on traffic safety. From physically separated motorcycle lanes, to retaining motorcycle lanes but separating them with markings, to mixed car-motorcycle traffic, which conditions are safer and more efficient? Chung Hui-yu says, “Taiwan has almost no such research. It is ‘zero.’ We are extremely lacking.”

吳祥瑀說，從初步數據來看，開放汽機車混流的安全性應該高於原本的實體分隔機車道，但若採更寬鬆的解釋，「只要證明汽機車混流沒有比較危險，就應該讓機車駕駛多一個選擇，畢竟台灣是民主社會，人民本就有選擇的自由。」

Wu Hsiang-yu says that, judging from preliminary data, the safety of opening lanes to mixed car-motorcycle traffic should be higher than that of the original physically separated motorcycle lanes. But under a more permissive interpretation, “as long as mixed car-motorcycle traffic is shown not to be more dangerous, motorcycle riders should be given one more choice. After all, Taiwan is a democratic society, and people inherently have the freedom to choose.”

本報導資料取自警政署交通事故（A1、A2類）開放資料，如112年傷亡道路交通事故資料等，資料區間為2019年至2026年4月底，共2,404,341筆不重複事故資料。

The data for this report are drawn from the National Police Agency’s open data on traffic crashes, categories A1 and A2, such as the 2023 casualty road-traffic crash dataset. The data period runs from 2019 to the end of April 2026, comprising 2,404,341 non-duplicate crash records.

Data Cleaning and Statistics I. Bridge List and Crash Data Compilation資料清洗與統計 一、橋梁列表與事故資料彙整

我們首先透過台灣機車路權促進會整理的全台灣橋梁列表，接著在台北市、新北市交通局公開報告中，取得各橋梁尖峰時間交通量數據。最後根據機車道設計、交通量與專家建議選出6座橋梁作為本文的討論對象，分別為大漢橋、新海橋、浮洲橋、永福橋、福和橋與重新橋。

We first used the nationwide bridge list compiled by the Taiwan Motorcycle Rights Promotion Association. We then obtained peak-hour traffic-volume data for each bridge from public reports issued by the Taipei City and New Taipei City transportation departments. Finally, based on motorcycle-lane design, traffic volume, and expert recommendations, we selected six bridges as the subjects of this article: Dahan Bridge, Xinhai Bridge, Fuzhou Bridge, Yongfu Bridge, Fuhe Bridge, and Chongxin Bridge.

We then obtained data on crashes occurring on bridges in the following two ways:接著透過以下兩種方式取得發生在橋梁上的事故資料：

- 橋梁GPS座標標記：由於發生於橋梁上的事故，在資料中「道路型態子類別名稱」欄位中並不會全部被標記為「橋梁」，可能會被標記為直路等類別。為了盡可能準確蒐集發生於橋梁中的事故資料，我們在地圖上標記每個橋梁的空間範圍（標準為至少涵蓋橋梁全部範圍、最遠畫至橋下路口為止），篩選座標位於橋梁範圍內的A1、A2事故點位。

- Bridge GPS coordinate marking: Because crashes occurring on bridges are not all marked as “bridge” in the “road-form subtype name” field of the data, and may instead be marked as straight road or other categories, we marked the spatial range of each bridge on a map in order to collect bridge-crash data as accurately as possible. The standard was to cover at least the full extent of the bridge, extending at most to the intersection under the bridge. We then filtered A1 and A2 crash locations whose coordinates fell within the bridge range.

- 橋名比對：我們將上述6座橋的橋名，與事故資料中的「事故地點」欄位比對，篩選出地點中包含橋名的。

- Bridge-name matching: We matched the names of the six bridges above against the “crash location” field in the crash data, filtering for records whose location contained the bridge name.

因為現場事故登記的方式與座標精度不同，有可能有事故地點符合，但登記座標並非在實際事故發生地點（而是在附近的地標或警察局）。

Because the way on-site crashes are registered and the precision of coordinates vary, it is possible for a crash location to match while the registered coordinates are not the actual crash site, but instead a nearby landmark or police station.

II. Crash-Type Marking二、事故型態標記

- 為了更精細呈現車禍事故類型分布，我們將原始資料中「事故類型及型態子類別名稱」欄位，根據「行動狀態」與「碰撞部位」等資訊，重新分類事故類型，分類依據參考交通研究中碰撞構圖的型態。 例如：原始分類為「其他」的事故，若前二當事人的行動狀態皆為「向前直行中」，且碰撞不為分別為「前車頭」與「後車尾」，我們則將此資料標記為「直行追撞」，詳細對照表請見此連結。

- To present the distribution of crash types in greater detail, we reclassified the original “crash category and subtype name” field according to information such as “movement status” and “point of impact,” using as reference the collision-diagram typologies found in transportation research. For example, if a crash originally classified as “other” involved the first two parties both moving “straight ahead,” and the points of impact were not respectively “front of front vehicle” and “rear of rear vehicle,” we marked the record as “straight-line rear-end crash.” For the detailed comparison table, please see this link.

- 其他無法判斷的點位將保留原分類，或標記為「無法判斷」。

- Other points that could not be determined retained their original classification or were marked “undeterminable.”

依據以上資料篩選規則與人工校對排除不合理的路名（例如行政區錯置等）後，我們進一步篩選容易發生在橋梁上的事故類型，包含同向直行擦撞、同向擦撞、直行追撞、停等追撞、臨停追撞、追撞與自撞等7種類型，作為本文分析對象。

Based on the data-filtering rules above, and after manual checking to exclude unreasonable road names, such as mismatched administrative districts, we further filtered for crash types likely to occur on bridges. These included same-direction straight sideswipes, same-direction sideswipes, straight-line rear-end crashes, stopped-and-waiting rear-end crashes, temporarily stopped rear-end crashes, rear-end crashes, and single-vehicle crashes: seven types in all, which serve as the objects of analysis in this article.

共在6座橋梁中篩選出2,621件不重複A1、A2事故，其中1,249件位於橋梁標記範圍內、1,372件在範圍外但事故地點與橋名相符。文中的統計皆使用上述共2,621件事故分析，唯地圖僅顯示GPS座標位於橋梁範圍內的1,249件事故，呈現橋梁上事故熱點分布。

Across the six bridges, we filtered out 2,621 non-duplicate A1 and A2 crashes, of which 1,249 were within the marked bridge ranges and 1,372 were outside those ranges but had crash locations matching bridge names. The statistics in this article all use the above 2,621 crashes for analysis, while the map displays only the 1,249 crashes whose GPS coordinates fall within the bridge ranges, showing the distribution of crash hot spots on the bridges.

Statistical Limitations統計限制

第一，根據現場事故記錄的方式與標準不同，地圖中的少數事故GPS點位可能有所飄移，或是記錄為事故發生後移車等待員警到現場等位置，並非所有點位皆是碰撞發生當下地點。

First, because the methods and standards for recording crashes on site differ, a small number of GPS points on the map may have drifted, or may have been recorded as the location where vehicles were moved after the crash to wait for police to arrive, rather than the location of the collision itself. Not every point is therefore the place where impact occurred.

第二，每個橋梁空間內的事故點位可能有填報時座標錯誤的狀況，雖經我們人工篩選、排除不合理的路名，但仍可能有少量缺漏。

Second, crash points within each bridge’s spatial area may include coordinate errors made during reporting. Although we manually filtered the data and excluded unreasonable road names, there may still be a small number of omissions.

陽明交大運輸與物流管理學系教授吳昆峯表示，員警在處理橋梁事故時的最優先任務往往是恢復車流通行而不是處理事故，出現大量隨意填寫GPS點位的情形並不令人意外。但員警填寫的文字資訊相對可靠，可初步作為事故位置的依據。

Wu Kun-feng, a professor in the Department of Transportation and Logistics Management at National Yang Ming Chiao Tung University, says that when police handle bridge crashes, their top priority is often restoring traffic flow rather than processing the crash, so it is not surprising that GPS points are often filled in arbitrarily. But the textual information entered by police is relatively reliable and can serve as an initial basis for locating crashes.

Seeking Truth in Depth, Walking with Many Voices深度求真 眾聲同行

獨立的精神，是自由思想的條件。獨立的媒體，才能守護公共領域，讓自由的討論和真相浮現。

An independent spirit is the condition of free thought. Only independent media can safeguard the public sphere, allowing free discussion and truth to emerge.

在艱困的媒體環境，《報導者》堅持以非營利組織的模式投入公共領域的調查與深度報導。我們透過讀者的贊助支持來營運，不仰賴商業廣告置入，在獨立自主的前提下，穿梭在各項重要公共議題中。

In a difficult media environment, The Reporter persists in using a nonprofit model to devote itself to investigative and in-depth reporting in the public sphere. We operate through reader sponsorship and support, do not rely on commercial product placement, and, on a foundation of independence and autonomy, move among the major public issues of our time.

今年是《報導者》成立十週年，請支持我們持續追蹤國內外新聞事件的真相，度過下一個十年的挑戰。

This year marks the tenth anniversary of The Reporter. Please support us as we continue tracking the truth of news events at home and abroad and meet the challenges of the next decade.