What this map is
992,225 casualty road accidents published by Taiwan's National Police Agency, drawn as a heatmap. It opens on the whole country; pan and zoom directly, with no county to pick first and no radius to set. Redder areas are where accidents concentrate.
What you can look up
- Area All of Taiwan including the outlying islands. At street level a cell is about 13 m
- Period 2024/01 to 2026/08, 32 months
- Scale 992,225 accidents, 4,210 deaths, 1,320,732 injured
- Time range Last 3 months, 6 months, a year, or drag the start and end month freely
- Fatal only Shows only accidents where somebody died. That map looks nothing like the all-accidents one
- Locate me The crosshair at the top right jumps to where you are
The panel on the left updates as you move, showing accidents, deaths and injuries in view.
Filter by vehicle type
Below the panel you can show only accidents involving one type of road user. This matters because different people care about completely different subsets:
- A parent checking the walk to school wants accidents that hit a pedestrian; what the other vehicle was does not matter
- Planning a commuting route by bike, the relevant layer is bicycles and slow vehicles
- Riding a motorcycle every day, the motorcycle layer is clearer than the total
- Living on a road used by gravel trucks, trucks can be viewed on their own
The heat scale is recomputed from the filtered counts, so even though pedestrian accidents are only about a seventh of the total, the map still shows where they cluster.
Two things to know. One accident can count under two types — a car hitting a motorcycle appears under both, because both are involved; that is the definition, not double counting. Injuries are hidden while filtering, because the raw data records injuries per person and never attributes them to a vehicle type; the death figure then shows the number of fatal accidents.
Prefer numbers to a map
Every county and district has its own statistics page: accident counts, casualties, the roads that appear most often in accident addresses, the hours when crashes cluster, primary causes and the vehicle types involved. Those pages are in Chinese for now. Start from the area index.
What the colours and dots mean
Heatmap colour is relative, not absolute. The same place looks different at national and city zoom because the scale is recomputed from what is currently on screen. Nationally only the six metros run red; zoom in and the differences within a city open up.
At street level dots appear. Each is an aggregate for that spot, sized by count, and clicking one shows the actual number of accidents and casualties there.
Source and updates
From the National Police Agency casualty road accident datasets (types A1 and A2) on Taiwan's open data platform, under the Open Government Data License version 1. The agency refreshes the rolling data at 17:00 on the 1st and 15th of each month. The datasets used, their IDs and the field list are on the data source page.
Limitations, please read
This data cannot rank "the most dangerous intersections". Taichung records the most accidents in the country, yet holds only 19 of the national top 200 hotspots while New Taipei holds 121. That gap comes from how each city geocodes addresses, not from how dangerous the roads are.
26% of addresses are written as a road-segment reference point rather than the exact spot, so at maximum zoom a position may be tens of metres out. More traffic also means more accidents, and this data carries no traffic volume, so an accident rate per vehicle-kilometre cannot be calculated.
The full write-up is on limitations, the cleaning steps are on method, and the build is described in how it is built.
Disclaimer
This site re-presents public government data and makes no warranty as to its accuracy, completeness or timeliness. It is not a basis for property, insurance or any other transaction decision. Refer to the competent authority's own published figures.
Other tools on this site
- Old Photo Motion — Turn an old photo into a 6-second clip that blinks and smiles
- age-me — See the same person older or younger
- Campsite Legality Check — Which of 1,807 campgrounds comply, violate, or are not listed
- Flood Potential Map — WRA flood potential under 10 rainfall scenarios, by township
All run on one machine at home; the data tools use Taiwan government open data and state their limits.
The numbers behind this map
All figures below come from the same cleaned dataset the map draws from: 992,225 accidents, 4,210 deaths and 1,320,732 injured, covering 2024/01 to 2026/08.
By county and city
| County / city | Accidents | Deaths | |
|---|---|---|---|
| 臺中市 | 181,783 | 381 | |
| 新北市 | 122,592 | 390 | |
| 桃園市 | 112,432 | 414 | |
| 臺南市 | 112,148 | 446 | |
| 高雄市 | 86,810 | 378 | |
| 彰化縣 | 61,061 | 322 | |
| 臺北市 | 57,009 | 164 | |
| 屏東縣 | 40,544 | 320 | |
| 新竹縣 | 29,585 | 122 | |
| 雲林縣 | 29,183 | 265 |
Accident counts follow population and traffic volume rather than danger. Taichung records the most, which mainly reflects how many vehicles are on its roads.
By party type
| Type | Parties involved | Share | |
|---|---|---|---|
| 機車 | 1,226,784 | 57.5% | |
| 小客車(含客、貨兩用) | 562,431 | 26.3% | |
| 人 | 185,959 | 8.7% | |
| 慢車 | 79,177 | 3.7% | |
| 小貨車 | 48,238 | 2.3% | |
| 大貨車 | 11,951 | 0.6% |
Counted per party rather than per accident, because one collision involves two or more. Motorcycles dominate, which matches how people travel here.
By hour of day
The peak is 17:00 to 18:00 with 91,171 accidents, the evening commute.
| Hour | Accidents | |
|---|---|---|
| 00:00 | 9,192 | |
| 01:00 | 5,562 | |
| 02:00 | 4,031 | |
| 03:00 | 3,401 | |
| 04:00 | 3,903 | |
| 05:00 | 7,713 | |
| 06:00 | 22,056 | |
| 07:00 | 87,882 | |
| 08:00 | 81,440 | |
| 09:00 | 60,693 | |
| 10:00 | 58,979 | |
| 11:00 | 62,766 | |
| 12:00 | 58,129 | |
| 13:00 | 47,144 | |
| 14:00 | 48,867 | |
| 15:00 | 53,513 | |
| 16:00 | 66,568 | |
| 17:00 | 91,171 | |
| 18:00 | 67,981 | |
| 19:00 | 47,561 | |
| 20:00 | 35,613 | |
| 21:00 | 30,682 | |
| 22:00 | 22,730 | |
| 23:00 | 14,648 |
Most common primary cause
- 111,136 未保持行車安全距離
- 101,994 其他不當駕車行為
- 77,253 恍神、緊張、心不在焉分心駕駛
- 56,543 尚未發現肇事因素
- 51,919 有號誌路口,轉彎車未讓直行車先行
These are the attending officer's assessment recorded in the data, not a court finding.