A data platform for road safety
Critical
Road Segments
— Czechia
On Czech roads, an average of 500 people die every year. We know where.
The platform combines open data from the Czech Police (Policie ČR), the Road and Motorway Directorate (ŘSD), the Prague road administration (TSK hl. m. Prahy), and the methodology of the Transport Research Centre (CDV) into a single interactive map. 14 regions. 8,877 ŘSD traffic census segments, plus 954 segments and 704 junctions of the TSK network in Prague. Accidents from 2015 to the present.
Accidents and casualties: Czech Police, Jan 1, 2015 – Oct 5, 2026 (last day of available data). Societal losses: sum of the annual CDV estimates for 2015–2024, valid as of 2024 — CDV has not yet published a newer estimate.
The problem
Since 2015, more than 5,700 people have died on Czech roads. More than 23,000 others were seriously injured. Total societal losses estimated by the Transport Research Centre (CDV) for 2015–2024 alone come to almost CZK 1 trillion.
Most of these accidents happened on repeatedly dangerous segments — and until the next one occurs, no one talks about them.
Losses: sum of the annual estimates of socio-economic losses from road accidents published by the Transport Research Centre (CDV) for 2015–2024 (CZK 993.1 bn), in current prices of each year and including accidents not recorded by the Czech Police. Source: CDV press release for 2024 (in Czech) and earlier CDV press releases.
What it does
Nine views of the same data — from the map to places for inspection.
Instead of laborious manual data work, you get answers in minutes: which segments show repeated accident rates, under what conditions, with what impact on social cost, and where investment has the highest priority. Fully responsive UI — from phone to large monitor.
A map that shows the segments that are actually risky
After picking a region, you see green, yellow, and red segments — the safety class per the implementing decree to Section 18m of Act No. 13/1997 Coll. (§18m zákona č. 13/1997 Sb.) The same map for a mayor, a regional councillor, a police officer, and a journalist — one shared baseline that removes disputes about whose numbers are right.
Switch the metric to RSI (Relative Safety Index), H (Integral indicator of Total Societal Losses), accident count, or TOPSIS (the composite index) to see the same segments from another angle. The slide-out Top 30 segments panel respects the active filter and gives you a ranking ready to feed straight into a repair plan.
Segment detail and individual accidents
Click a point and a popup shows the date, accident type, cause, damage amount, and the number of injured. The ŘSD segment ID is a link to the segment detail, including traffic intensity and the year of the traffic count.
Insurers map claim events to a specific segment in full context. Regions get factual input for negotiations with municipalities — a concrete list of accidents by date, cause, and severity instead of aggregated averages.
Weather as a risk multiplier
Some segments show a low overall accident rate yet markedly elevated risk under specific conditions. A 12-cell matrix (temperature × precipitation × day/night) shows, for each segment, when risk multipliers diverge significantly from the segment's long-term average. Cells with low observation counts are explicitly flagged so they don't lead to conclusions drawn from a small sample.
The historical matrix is complemented by a 48-hour risk forecast based on the current weather forecast.
Time trends with a clear distinction for projection
Views By year, Seasonal, By region, Correlation, and Social cost — five angles on the same data. The current year is computed as a seasonally-weighted projection and is clearly distinguished in the chart by hatching, so it's immediately obvious where measured data ends and an estimate begins.
A solid foundation for press releases, annual reports, and council briefings. For deeper analytical work there is a Spearman rank-correlation matrix between segment metrics — covered in detail in the Methodology section.
Table and CSV export to plug into your systems
Filter segments, sort them by any metric, and download the result as CSV. Each row contains the ŘSD segment ID, which enables direct import into GIS, repair-tracking systems, and data warehouses — without any further transformation.
For initial use a browser is enough. For regular data exchange the same outputs are available via a REST API and, through an MCP server, directly in the Claude assistant — ask in plain language (“Which D1 segments in the Vysočina region are the most dangerous?”) and Claude fetches the numbers from the same calculations the application uses — including accident clusters, places for inspection, weather effects, and the risk forecast.
Prague: segments and junctions of the TSK network
The ŘSD network has only 24 traffic census segments in Prague. For the capital the application therefore also uses traffic counts from the Prague road administration (TSK) — RAR, RSI, H, accident count, and TOPSIS are computed for 954 segments and 704 junctions of the monitored network. The map colours them by the selected indicator just like ŘSD segments and applies the same accident filter.
TSK segments and junctions have their own rankings. Accidents at nodes belong to junctions and a junction indicator is relative to the number of entering vehicles, so they are not mixed with ŘSD segments into one order. The Top 30 panel on the map therefore switches between ŘSD segments, TSK segments, and TSK junctions. Details in the Methodology section.
Segment analysis works for Prague segments and junctions too — accident clusters, weather effects, and the 48-hour risk forecast. For Prague the weather effect is estimated from the Prague network, not from the mostly rural ŘSD roads. Search finds a segment by street name as well.
48-hour risk forecast for a segment
How risk on a selected segment changes over the next 48 hours given the weather forecast, time of day, and traffic. The chart shows, hour by hour, the change in risk for an average vehicle versus the segment average — it is not an accident forecast. The weather effect comes from the segment's accident history and, for segments with few accidents, leans on roads of the same class.
The forecast comes with a segment map and an elevation profile from ČÚZK with the freezing level — you can see where on the segment it may freeze or snow. Useful for winter maintenance, fleet dispatch, and preventive driver information.
Accident clusters within a segment (KDE+)
The indicators tell you which segment is dangerous. Segment analysis shows where on it: the KDE+ method (Bíl, Andrášik, Janoška, 2013) finds places where accidents concentrate more than chance would explain and ranks them by strength. Accidents appear on the segment map, and a strip of accidents and their density runs below the elevation profile.
By default only stable clusters are shown. Clusters follow the same accident filter as the map, so you can narrow them down to injury accidents or winter months, for example. The analysis works for ŘSD segments and for segments of the TSK network in Prague (pictured: Kbelská street). The procedure is described in the Methodology section.
Places for inspection: where to send a safety inspection
An accident cluster alone does not tell you whether there is something to fix on site. The Places for inspection view links accident clusters across the whole network with how the road is laid out at that spot and with the indicator of the whole segment. The matrix shows where risky geometry meets the worst quartile of segments — that is where an on-site inspection makes the most sense.
The Pareto 20 % option highlights the worst fifth of places by severity-weighted harm or by socio-economic losses; nothing is filtered out. Outside built-up areas across Czechia, the worst fifth of places carries roughly half of the severity-weighted harm. Clicking a place opens Segment analysis with its clusters — input for a safety inspection plan.
Who it's for
Four audiences. Four angles on the same data.
Regions
Where to invest in infrastructure first. A repair plan based on the Relative Safety Index (RSI) and total societal losses (H). Segment export with ŘSD IDs.
Insurers
Per-segment risk score for rate setting, claims analysis, and prevention. Segment-level granularity instead of regional averages.
Professionals
Reproducible foundation for studies, expert reports, and strategic documents. Published methodology, open formulas, monthly updates.
Public and journalists
Public data sources, transparent formulas. A foundation for analyses and articles on road safety, free of marketing numbers.
Methodology
Three transparent indicators. Every formula shown in full.
Unit: accidents per million vehicle kilometres per year.
Numerator is Reinhold's weighted sum Z. Unit: accidents per million vehicle kilometres per year.
CDV unit cost values (AF, AS, AM) for 2015–2024 are on the Methodology page.
Where the rankings from the individual indicators diverge, the application also offers a TOPSIS composite index — a single order that combines the accident rate relative to traffic intensity, the severity of outcomes, and the number of injury accidents. Severity counts once, even though two indicators describe it (RSI and H).
Demo
Request a demo
Send a few words about your organization and your interest. I'll get back to you as soon as possible.
Or email directly: info@zbirovsky.com