11.–12.06.2025 #polismobility

EN Icon Pfeil Icon Pfeil
EN Element 13300 Element 12300 DE
From all perspectives

Making roads safer with data from the community

Share page
PrintPrint page Read duration ca. 0 minutes

In conversation with Michaela Grahl, project manager at the Initiative für sichere Straßen GmbH, about the background to the gefahrenstellen.de website, data in road safety work and safe routes to school.

Michaela Grahl | © Initiative für sichere Straßen GmbH

Michaela Grahl | © Initiative für sichere Straßen GmbH

What is the motivation for the Hazardous Sites website as well as for the Safe Streets Initiative?

The motivation came from a personal negative experience. One of the three founders (Arno, Henrik and Jörn Wolter) had himself had a negative experience at a dangerous spot - a typical spot that basically everyone in the neighbourhood knew about and said "I won't let my child drive past there because I know that the tyres always squeal there". So the idea came up that you had to digitise this knowledge and make it available to others.

The company behind it is basically a small IT specialist. From this perspective came the idea: let's create something digital for road safety - that was in 2014. The whole thing was then set up as part of the research project "FeGiS+" (early detection of danger spots in road traffic through smart data).

Who is the project aimed at? The map is very visual and can be used without further expertise. But is it also intended as a tool for planners?

The idea was to build a holistic concept that offers added value for all areas. The danger score, as it can now be seen on the map, is primarily intended for users, for road users, because it is easy to interpret. For the municipalities, for traffic planners and also for the police, we have an extra access area, the Pro-Portal, where you can analyse the data in more detail.

The hazard score is then calculated primarily from police accident data?

We currently use the public Destatis accident atlas - this contains the accidents in which people were injured or killed, which are best documented. This is the very valid basis and currently still makes up the main part of the score. But we are also concerned with early detection. So what if not so many accidents have happened yet, but this place is developing? That's where the user data comes into play, together with the vehicle data. We have safety-critical braking manoeuvres - critical braking data - from telematics tariffs. We do this very selectively, put this information on road sections and evaluate the road sections according to the methodology of RWTH Aachen University (a FeGiS+ project partner, editor's note). This then recognises when a spot deviates from the standard value and generates a colouring. So there is a basic evaluation via the accident data and this additional data, which increases the danger level when defined limit values are exceeded.

There are two different levels on the map: the user reports and the score; in other words, what is obviously dangerous from the data, and then the places where people feel uncomfortable. How is that connected?

That mixes a bit. We have studied this connection in cities where we have done a lot of advertising - Bonn and Aachen, for example. With our scientific partners, we made over 200 site visits as part of the FeGiS+ research project. In the process, we found that many user reports had been received for locations that were already coloured red in the accident data in the score. These are often already known accident blackspots, which we then confirm again from the user's point of view. Other places, however, had not yet attracted such attention - we also assessed many of them as dangerous on the spot. But not every report has an immediate influence on the score.

What does the process look like?

We evaluate the reports based on the comments, as an input control, whether this is a report in our sense. Then we automatically rate it according to how much interaction there is at this danger spot. A pin is already a summary of several user interactions: One reports, another makes a comment, the next supports. This results in a level of activity, and above a certain level, this is included in the score.

Do you then still check the reported jobs manually or is there a process for this?

In the research project, we found that the messages are already very valid. However, with some messages, where not so much information was available because the activity level was relatively low, it was sometimes simply not clear what was meant.

And from that they then worked out this threshold value, so to speak.

Exactly, simply from experience in combination with these on-site inspections.

There are also other citizen science projects that work in a similar way - for example SimRa or the GehCheck app of FUSS e.V. Could these data be merged somehow?

We regularly exchange information with other projects and always check whether it is feasible. The problem is often that the projects request data differently. With us, they are collected in a very structured way, so we can also automate the assessment. The other portals all tend to have a different direction - they don't only ask about danger spots, but also about other things. That's why it's difficult.

Actually, it would be nice to have one big project at the end, with information from all possible data sources and for all types of traffic.

Absolutely. Actually, that is also our approach. We record all types of road users via the users, but we also always think about what data could be added. It has to fit in terms of content and data quality. But you can't just integrate it like that, you have to see how the data source is to be evaluated. That's a bit of a pipe dream.

At the moment, you are working on the project "HarMobi" (Harmonising Mobility), which aims to include as many modes of transport as possible. What is that about?

The background is that in FeGiS+ we identified danger spots and many of them were intersection areas. As we found out, it is difficult to find out what is really the conflict there when there is only a user message and maybe the perspective from the vehicle data. In "HarMobi" we want to look at the situation together by trying to capture all types of road users.

Even with site visits, the danger is not always completely clear. We have seen that ourselves: Sometimes it depends on very specific constellations - times of day or situations where there is perhaps particularly heavy lorry traffic and it changes the visibility conditions. Sometimes it's so diffuse that you can't always see the danger when you're on site. And then it's a matter of being able to deduce from the data how safe this intersection is. In the end, it should provide assistance in planning so that you can evaluate in advance whether a planning alternative is safe or not.

But how can data help in the planning phase?

The aim of the project is to find out about typical conflicts at intersections of a certain type. Based on the data we collect, we want to find out for these types of intersections where the typical conflict situations are or at which intersections things go better. In this way, we can already say in the planning phase, with these standardised data collected in large quantities, where there will probably be a conflict. But we are still at the very beginning here, I can probably tell you more about this in a year's time.

There are already standardised designs for intersections, for example in the Guidelines for the Design of Urban Roads (RASt), on which experts have thought about what the safest intersection looks like and how different concerns can be weighed against each other. How do you hope to improve this with your data?

Over time, you notice that things that were once planned in a certain way no longer work well today because, for example, traffic flows are changing. For example, cycling traffic is steadily increasing, and some intersections are not well designed for it. In some cases, we are not yet sure what safe traffic routing would look like. We definitely want to take data for this. And the e-scooters complicate the whole thing even more.

How do you get the e-scooter data?

We have cooperations. The project is initially only related to Aachen, and there to certain intersection areas that we define in advance because we have a very good data situation there. As a cooperation partner, we also have a pedelec rental company - the pedelecs have a completely different driving behaviour, completely different speeds than normal bicycles - especially in intersection areas. We also evaluate truck telematics, car telematics, bus and e-scooter data. Pedestrians are also included in principle, but probably only via camera data, because it is difficult to capture them.

We try to really pick up everyone who is on the road at this intersection: Either they have already installed sensors on their vehicles themselves or we are still developing an app that can then be used to record this situation.

On your website gefahrenstellen.de you also have a school route function. When I select a start, a destination and a tolerated danger score, I get a suggested route to school on foot or by bike. When I tried it out, I sometimes got unusual routes as results.

The School Route Planner is still a beta version - the routing is one of the next developments. We developed the routing together with HeiGIT (Heidelberg Institute for Geoinformation Technology, operator of the routing service openrouteservice.org, editor's note). We are once again working closely with HeiGIT to see how we can better integrate information such as crossing aids, stairs or cycle paths into the map base. Unfortunately, the problem is that these things are not always as well documented as one would like. Some things are already entered in openstreetmap - traffic lights, some crossing islands - but this is completely location-independent, sometimes very well maintained, sometimes not.

© gefahrenstellen.de

© gefahrenstellen.de

What was the idea behind the school route calculation?

We want to encourage parents to consider how their child could cycle. For primary school children there are still the school route plans, but older children also cycle. Parents should look at the route alternatives and then really decide for themselves: Is this better than the route you might always take by car, or not? Just like the police always say: make your children aware of dangerous places. That's what we want to do with our offer.

This is also a process for parents when they are out with their children and put themselves in their position. I could imagine that one suddenly perceives the traffic space quite differently.

Yes, and the parents of course have much more experience of road use than their children. We simply notice that this is a topic that interests many people. Schools also approach us and find it good to talk about such topics, such as parent taxis. Parents are insecure, which is why they don't let their children drive or walk to school alone.

Do they then work directly with the schools?

There are schools that we work with. Some of them like what we have to offer and link to the school route planner or run a campaign in which parents report danger spots. We did something like this in Königs Wusterhausen near Berlin, for example. Parents' initiatives in the region have called for danger spots to be reported. The municipality is also currently working with our Pro-Portal and is trying to eliminate or warn of the dangers. The city is very active with its prevention council. I think that's great.

So is the work of accident commissions also an area that you would like to improve?

We want to make it a little easier. We have noticed that the commissions naturally have a lot to do, depending on the region. And we simply want to automate certain analyses - hence this danger score. We have simply found that the construction of intersections or the elimination of accident blackspots are not as simple as one might imagine. Sometimes there is a lack of data to assess the whole thing properly and to be sure that it will work better afterwards.

Author

Jan Klein