On this page
Some problems are difficult to reason about when they are presented only as numbers, tables or separate technical layers. Flood risk is one of them. An elevation value, a water-level scenario and the shape of a built environment may each be understandable in isolation, while their combined consequences are much easier to examine spatially.
Between January 2024 and March 2025, I worked on a research-and-development proof of concept exploring flood and sea-level scenarios using a three-dimensional digital twin environment. The work brought together geospatial data, spatial modelling, visual presentation and an experimental natural-language interaction route.
The value of this project, for me, lay as much in bringing those concerns into one understandable experience as in any individual technical component.
The problem with separated representations
A spatial problem can be described by many specialised datasets. The people interpreting the outcome, however, may need to answer a much simpler question: what would a particular scenario mean for this place?
A map can locate features. A model can calculate or represent conditions. A three-dimensional environment can show how spatial relationships appear. But if those representations remain disconnected, the user still has to perform much of the integration mentally.
I wanted to investigate whether a digital twin could bring those perspectives together in a way that allowed scenarios to be explored rather than merely described. It was an experimental question about representation and interaction, not a claim that a prototype could replace professional flood-risk assessment.
Working from data towards an understandable scene
The publicly described proof of concept combined GIS datasets, spatial logic and a three-dimensional environment. It allowed flood and sea-level scenarios to be represented as visual overlays, alongside an experimental way to express questions in ordinary language and connect them to structured scenario outputs.
A great deal of the design work concerned the boundary between the data and the scene. Spatial information needs to retain its meaning while being transformed into something that a person can inspect. An attractive visual representation has limited value if it obscures the assumptions used to construct it.
The interaction challenge was related. A question phrased conversationally is useful as a starting point, but the system still needs a structured interpretation of what is being requested. That distinction became especially important in my later research into governed analytics, although the two projects have different analytical responsibilities and levels of assurance.
A prototype is a means of investigation
One of the benefits of a working proof of concept is that it exposes questions a written proposal can leave unresolved. How should scenarios be selected? What does the viewer need to know about the underlying data? Which visual changes help understanding, and which merely look impressive? How can an interaction remain clear when several forms of information are brought together?
These are user-experience questions as much as technical ones. A three-dimensional environment can increase the apparent sophistication of a system without necessarily improving comprehension. The objective should be to make the spatial implications easier to explore, with appropriate regard for uncertainty and evidence.
I developed the prototype while working in a professional simulation environment. The public account here intentionally stays at the level of research purpose, general capabilities and design observations. It does not describe protected internal interfaces, proprietary modelling details or unpublished customer material.
Connections with current work
The most direct connection is with my continuing interest in the relationship between analytical results and their presentation. In determinIQ, Geographic Intelligence is conceived as a controlled presentation route for governed analytical results, rather than an independent authority for generating those results.
That is not the same architecture as this earlier prototype, and the connection should not be mistaken for an identical implementation. It is a retrospective observation that the responsibility boundary between source information, computation and explanation keeps reappearing in different forms of software.
The project also reinforced a broader point about research: building an interface can be a way of asking a technical question. A functioning prototype provides something concrete to inspect, challenge and refine. Its usefulness as research does not depend on presenting it as a completed operational system.
What can and cannot be concluded
My public professional project record describes this work as an R&D proof of concept, with a date range and a high-level account of its GIS, three-dimensional and natural-language elements. It supports a description of the investigation undertaken. It does not establish independent scientific validation, external assurance of the underlying flood model or a proven production deployment.
Those distinctions are worth retaining. This was a working investigation into how spatial information might be made more understandable. It is that process of connecting complex information to a useful human interaction that makes the project relevant to the Foundations collection.
Historical context
Research programme: Making data analysis understandable, reproducible and accountable
Topic: Foundations
Explore further
Supporting material is listed below.
Further reading is listed below.