OroraTech is using space-based thermal imaging to help identify, track and predict wildfires across the globe. Based in Munich, the company provides data tools designed to support fire services, infrastructure operators and government agencies with earlier alerts and clearer decision-making.
At the centre of its operations is Dr Martin Langer, CEO and CTO, who brings an extensive background in aerospace engineering and satellite development. In this interview with FSJA Editor Iain Hoey, Dr Langer explains what OroraTech does, how its technology works and what’s driving its expansion into new regions and markets.
Like many startups, we stumbled into it. We spun out of the Technical University of Munich with a strong technology idea and saw thermal data from space was under-used. The fire case found us after the 2018 Los Angeles fires shocked me.
Soon after, people in Australia and South America asked if we could process thermal data when government systems were too complex or incomplete.
We launched a minimum viable product in 2019 and quickly gained interest. New versions followed, alongside satellite launches to close data gaps. Coming from outside fire helps us approach problems with a fresh mind.
The biggest difference is that they’re small. I call them shoebox-size satellites – literally about the size of a shoebox – around 10 kilograms. Instead of building large, expensive satellites that try to cover many missions, we built cost-efficient spacecraft that pinpoint the use case and focus on fire.
That’s also different from many government satellites, which serve a broad variety of use cases. We focused on a niche – back then it was still a niche – with the intention to build many of them.
Government satellites are excellent for scientific applications, but their revisit rates can be low. You may get one excellent measurement per day – but how do you close the gaps in between? Our answer was to launch many small satellites to fill those gaps, while continuing to leverage all existing satellites.
When we started OroraTech, our first years were spent figuring out how to utilise the existing data and we still do that. We process data from dozens of existing satellites and from our own.
Dr Martin Langer
We aggregate all of it into a homogeneous information set, because most users don’t care about satellites at all – they want to see information. Our job is to make that information better.
We do something different from many government approaches: we fly GPUs on our small satellites and run fire processing on the satellite. We don’t downlink all the raw data; we identify the fires in orbit. That saves time and cuts the delay from identification to information for the customer.
A satellite detects a fire, translates that into coordinates, size and related attributes and ships it within minutes to a ground station.
Dr Martin Langer
It’s then integrated into the platform – not as a standalone image, but as abstracted information that merges with everything else we know about that fire. If it’s a new fire, it gets a new ID; as more information arrives, it accumulates across space and time.
Most users upload a shapefile or define an area. As soon as something happens, an alarm triggers. They get a WhatsApp message or a push notification. They don’t have to log in – they receive a notification and, when they click, they see the event and the context.
That was a major learning for us. We began with identification – the real-time aspect. Then we realised you also need the past and the future. Our Fire Spread model projects the next six to twelve hours; it’s a fire progression model built a bit differently.
A fully physical model has many parameters – twenty or more – that you will never know perfectly. There’s always imperfection.
We tuned our model to use those parameters but also to run fast and to update all the time. When we learn something new from satellites or other measurements, it corrects itself. It’s a feedback loop – essentially continuous correction.
Because we collect information from fires globally, the model can also learn from those events. That helps build a better picture of how fires progress.
Does it work 100% perfectly all the time? No – and it doesn’t need to, because it constantly corrects. Real fires involve human interventions; conditions change. If you can compensate with speed and rapid updates, you can deliver the best practical results.
That feature came directly from customer feedback. Users love the spread model and run it in the field on their phones. They asked for a way to implement fire breaks – to put a break on the map and see how that would change progression.
It functions as a what-if scenario. You can enter real fire breaks or planned lines and understand the effect immediately. That turns the model into a planning tool you can use in the field and it also lets crews send information back from the ground.
For us, satellite information is one part of the picture. There are other components that stream into the platform. We merge them to give a more holistic view of what’s happening.
Dr Martin Langer
We began with the idea of doing proper post-fire damage analysis – understanding what has burned down. From there we learned two things. First, knowing what has already burned is critical for spread modelling; it won’t burn twice. Second, it’s important to track those areas over the next years.
So, we started with one thing in mind and learned, through users and customers, that it’s important for other tasks too. Having burnt areas in a common database that’s maintained over time becomes a key component, both for immediate assessment and for planning and modelling in the seasons that follow.
Over the years we’ve discovered many more users and use cases than we expected. We do a lot of B2B. Early on and still today, that includes commercial forestry companies, owners of large infrastructure on the ground and insurers – anyone affected by fire. Recently, we even had a chain of hotels come in.
Governments are big users too, in many forms: counties, provinces, states and full national deployments. We have country-scale contracts in places like Greece and Canada, where the focus is a holistic view of the entire country.
With all of them, you’re learning all the time. The questions are similar, asked in different ways. It’s about building resilience before a fire and coping when it’s active.
The pressure on our clients is growing as situations become more extreme and the number of fires increases. You need a system that shows the whole picture so you can decide what to do next.
We’ve also learned that the scope is year-round. At the start we emphasised identification and early detection. Now it’s about the full fire cycle – before, during and after. And we’re working on how to use AI: building better models, but also breaking down barriers to data access.
There are still barriers. We’re learning how to reduce them – through notifications, through interfaces that give simple explanations, even connecting to large language models so people can get a quick read of the situation.
Dr Martin Langer
Many people affected by fires and many who are fighting them, don’t have time. They need reduced barriers and the most important information delivered in a way they can understand and act on – without jumping through hoops every time.
This question always comes up: what’s the best technology – towers, drones, satellites? The answer is that you need all of them for different operations. For me, satellites are the backbone – the safety net and base layer of the data you create.
Think of the fire “digital twin” we’ve built. It runs all the time, 24/7, giving you global coverage. We process around five terabytes of satellite data per day.
A user can log in, see their area with no special preparation and do it in a minute. If you have high-value assets or local sensors – towers, drones, anything else – please use them and fuse them with our data to create even better results.
Satellites provide the ground layer that feeds everything else. That base layer supports decisions at pace and scale.
