Gold Coast • Queensland road crash data, 2020–2024
Where should road-safety investigation start?
The city is split into 500 m squares. Each square is ranked by serious crashes in 2021–2023, and the ranking is then checked against 2024.
Loading the data…
The shortlist
Ranked on serious crashes in 2021–2023. Checked against 2024.
Each square is 500 × 500 m. Optional points show 2021–2023 serious crashes; crashes at the same spot overlap.
Shortlist table and CSV
| Rank | Roads and suburbs | Road manager | Serious crashes 2021–2023 | Serious crashes 2024 |
|---|
Cluster counts and extent
Cluster numbers only apply to this fit. The diagonal measures each cluster's bounding box. A 250 m neighbourhood can link chains of crashes that stretch much further.
| Cluster | Serious crashes | Bounding-box diagonal (m) |
|---|
Who manages these roads
The Department of Transport and Main Roads (TMR) manages state-controlled roads such as the Pacific Motorway. The City manages local roads. A list the City can act on needs to separate the two.
| Road manager | Serious crashes 2024 | Caught by original top 20 | Caught by its own top 20 |
|---|
This breakdown was added after the 2024 check. It describes the original shortlist and does not change it. Choose "Council roads only" above to see the City's own list on the map.
Counts versus clustering
Both methods use the same grid and the same number of squares. Settings were chosen on 2023, so 2024 stayed unseen.
DBSCAN settings were chosen using 2023 results, then both methods were refitted on 2021–2023. The small gap in 2024 does not show that either method is better.
Crash counts by year and severity
These filters change the counts below only, not the shortlist.
Download monthly counts (CSV)Method and data quality
Data
The data covers police-reported casualty crashes. The snapshot has only six months of 2025, so 2025 is left out. The source is revised over time, so older years may differ from what was published back then.
The source's Gold Coast label can disagree with the current boundary. Every serious crash the source places in the Gold Coast stays in the totals.
An uncertainty model that didn't work
Smoothing every square with the same prior also leaves the ranking unchanged.
Other tools
A small read-only MCP (Model Context Protocol) server exposes these results to AI tools. A Power BI report and BigQuery spatial queries were checked against the same data snapshot.
Method, reproducibility and sources
Coordinates are converted from GDA2020 to MGA zone 56 so distances are in metres. The 500 m grid has a fixed origin that doesn't depend on where crashes are. DBSCAN uses a 250 m neighbourhood and at least five crashes, chosen on 2023.
Road names and road managers come from the same crash records: the streets most often recorded in each square, and the source's controlling-authority field.
Without traffic volumes the counts cannot measure risk per trip. Choosing a treatment would need an engineering assessment; this study does not estimate causes or crash reductions.
TMR crash data · Queensland boundary data · Source manifest (JSON) · Full results (JSON)
Crash and boundary data © State of Queensland, used under Creative Commons Attribution 4.0. This independent study is not endorsed by the City of Gold Coast or TMR.