Lea Schollerer and Christian Geiß, The German Aerospace Center (DLR)
There is a fundamental challenge: how do we protect people from disasters that haven’t happened yet, in places where they don’t yet live? The answer requires us to understand not just where hazards might occur, but where vulnerable populations might be located in the future.
The growing challenge
There has been an increase in the number of recorded natural hazard events in the last decades, including earthquakes, fires, floods, and even compounding disasters. Such events can cause huge losses, especially in settlements with high population densities. It can be expected that this situation will intensify in the future as the world’s population grows and climate change increases the number of both single and multi-hazard disasters. As a result, more people will be exposed to natural hazards than ever before. In order to develop mitigation strategies for possible future damage events, detailed information on the projected spatial distribution of the population and further exposed elements e.g. buildings are required. Ultimately, quantitative models that describe future risk situations are essential.
Learning from the past – a novel approach: Earth Observation (EO) and Artificial Intelligence (AI)
Our work within the PARATUS project addresses this challenge by using satellite observations and artificial intelligence to anticipate where people will be living in the future. This approach combines long-term earth observation timeseries data from the last two decades with advanced modelling techniques to project exposure spatiotemporally. Therefore, patterns of urban growth and land use change from the past are analysed with AI methods to predict future development of population and urban areas.
Linking exposure to hazard models
The generated information on future risk-related exposure can be linked to models of natural hazards to show how many people will be affected by earthquake, fire, or flood events in the future. This has been done before: for the earthquake- and tsunami-prone areas of Lima and Callao, Peru. Here, the development of the population was predicted up until the year 2035. When combined with earthquake and tsunami simulation data, the results showed that there is a population growth in high peak ground acceleration areas and also in tsunami inundation areas. These results can be incorporated in information systems for (multi-)risk assessment, so stakeholders can explore possible future scenarios.
Expanding the approach to Istanbul
The current work aims to follow the described pathway in a more exhaustive way, for instance for the highly dynamic megacity of Istanbul, which faces significant exposure to earthquakes and landslides. Istanbul presents a particularly complex case due to its rapid urban transformation and multi-hazard environment. Within PARATUS the aim is to innovatively provide a complete picture of risk by modelling and anticipating the exposure of the built environment in addition to population forecasts.
Looking forward
The resulting exposure assessment datasets demonstrate how earth observation and AI techniques can support disaster risk reduction. Early and sustainable urban planning can incorporate these projections and derived future risk scenarios to guide development decisions and apply strategies to reduce the risk, e.g. by proactively designate safer zones for residential development. This forward-looking approach allows cities to build resilience into their growth trajectory rather than relying solely on current conditions.This information enables the evaluation of the systemic risk and vulnerability of settlements for proactive disaster preparedness and mitigation strategies.
