Unlocking urban intelligence: Integrating LLMs and InSAR for Scalable City Risk Management

Cities are constantly shifting—whether it’s aging infrastructure, heavy rainfall, underground construction, or groundwater extraction, all these forces quietly reshape the urban landscape. And now, urban deformation is measurable—thanks to satellite radar technology (InSAR).

Today,InSAR generates hundreds of thousands of high-resolution deformation time series over every city, capturing millimetric changes across buildings, roads, slopes, and other critical infrastructure. It’s an incredible source of insight—but also a massive volume of data that is almost impossible to navigate and interpret without technical expertise.

This is the motivation behind our latest project within the SPACE4Cities initiative: making advanced satellite deformation analytics accessible to any city manager, engineer, or planner—without needing to master the complexities of InSAR, time series modeling, or geospatial clustering.

Over the past five years, Detektia has specialized in bringing satellite radar (InSAR) technology into the real world of infrastructure monitoring. At the core of our platform is a robust API designed for scalability, integration, and advanced on-demand processing. This API serves deformation time series, calculates metrics, and connects seamlessly with GIS systems, digital twins, and engineering dashboards.

But access to deformation data alone is not enough. That’s why we’ve built an extensive set of post-processing algorithms that allow users to:

1. Decomposes deformation into vertical and horizontal components for deeper insight.
2. Detects trend shifts and acceleration events early
3. Clusters similar deformation behaviors across large urban areas using 2D time series analysis.
4. Correlates deformation with environmental and human-driven factors like rainfall, temperature, or construction.
5. Generates alerts and early warnings based on deviations from expected behavior.
6. Monitors the structural health of thousands of buildings and infrastructures over time.
   

These capabilities are powerful—and critical for effective urban risk management—but they are often too complex for non-specialists to use directly.

In the Space4Cities project, we are taking a major step forward: integrating Large Language Models (LLMs) and AI agents into our system to act as intelligent interpreters of geospatial deformation data.These AI agents don’t just fetch information—they reason over it. They use function calling and model context protocols to orchestrate API interactions, execute clustering or forecasting routines, cross-check results with weather or construction datasets, and generate explanations in plain, human language.

Imagine asking:

Which neighborhoods in the city have experienced significant subsidence since January?

Has the recent rainfall increased the deformation rate near the tunnel entrance?

Are any slopes in the city moving faster than before, and when did that change begin?

How has this underground construction affected the structural health of nearby buildings?

The LLM processes the question, queries the appropriate time series, runs the clustering or breakpoint detection algorithms, combines the output with contextual variables, and returns a clear, focused answer—backed by data, but free from technical jargon.

The real breakthrough lies in the integration of advanced APIs with LLMs. The core processing power—deep analytics, high-resolution deformation data, and complex geospatial insights—lives in the API. What LLMs unlock is intuitive access to that power. Now, users can ask critical questions in plain language and instantly tap into tools that were once reserved for specialists. This fusion not only simplifies access—it transforms it, enabling faster decisions, better coordination, and broader engagement across city departments. Urban planners, engineers, emergency teams, and environmental analysts can all interact with the same powerful intelligence layer, tailored to their needs—without needing to be data experts.

This work is made possible by the maturity and stability of our existing infrastructure. Detektia’s API is already operational and used in real projects across Europe and Latin America. We process Copernicus Sentinel-1 data continuously, and all our models are designed for reproducibility, scalability, and explainability. We are now combining this foundation with cutting-edge AI to bring the next generation of urban monitoring to life through this SPACE4cities project..

With SPACE4Cities, we’ll be piloting this integration across real city environments, testing how LLMs respond to realistic queries, how they prioritize risks, and how they justify decisions using geospatial reasoning.

The result?

A more intelligent, transparent, and responsive way to manage the geometric evolution of our cities.

We’re not just monitoring cities. We’re helping them think, and giving city managers the answers to the questions that truly matter.

---

This is part of the SPACE4Cities project that has received funding from the European Union’s Horizon Europe Research and Innovation Programme via EU Agency for the Space Programme (EUSPA).