How Much of a City Can You Rebuild From Its Open Data?

Vienna keeps a record of every tree in the city. The registry covers around 200,000 trees on public land: streets, school grounds, city squares, parks. For each one it notes its exact position, species, age, height, and width.

What an amazing dataset, I thought to myself, while developing two side projects to get more familiar with the city I moved to five years ago: Hof, named after the German word for courtyard, and Bim, after the Viennese slang for tram. Both use the city's public data. Hof is a map of buildings and people. Bim is a live map of transit.

So I kept pulling in more data: which era each building was built in, how the population is distributed by living space, how the city voted, who owns the property, where people were born, where the city is growing and where it is shrinking. Vienna's open data portal has over 300 datasets, and the more I combined them the more it opened up.

So far so good. Keep adding data and the picture keeps getting better.¹

But is rebuilding a city just a matter of adding more data? A matter of scale? Assume, as a thought experiment, that we had everything: every footstep, every transaction, every conversation, infinite sensors, unlimited compute. Could you reconstruct the whole city? Is it possible to simulate the places we inhabit without running out of memory?

What the data can tell you

Probably more than you'd think.

Hof maps 572,781 buildings coloured by the era they were built, with 2,040,914 residents distributed across them by living space. Ten layers of data you can turn on and off. You can see which districts have the least tree cover, then overlay it with how people voted. You can see where Vienna is growing and where it is shrinking, and combine that with what people earn, or how many people don't have a vote at all.

Bim is a live map of the city in motion. Vienna doesn't publish the coordinates of its transit fleet, only countdowns at stops. So every moving vehicle is an estimate, calculated from arrival times and known routes. You can see where trams slow down in rush hour, how they thin out at night, and where the schedule doesn't match reality.

I caught myself looking out the window every time the No. 12 tram passed, checking whether its real-world position matched the 3D vehicle on my screen. I even spent some time at Praterstern, one of the city's busiest transit hubs, just watching trams and buses pull in and out. It was somewhat satisfying seeing the physical city sync up with the map I had built.

Neither project is a simulation. They are reconstructions built strictly from what the city records and publishes.

If we push our thought experiment to its limit, though, we end up with something much bigger: a complete digital twin of the city, a real-time model that can predict what happens next. But that immediately raises a mess of bigger questions. Who owns all this data? Who controls the system, and who gets to decide how it's used? Map every detail, and you quickly realize data isn't just information but power itself.

Where the reconstruction fails

Every social system — a city, a country, a company — is some sort of a computational engine. It takes inputs, produces outputs, and moves through states in between. People, resources, and energy go in. Services, infrastructure, and decisions come out. The city itself constantly changes states: people arrive and leave, businesses open and close, institutions form and dissolve. It is an ongoing algorithm of political decisions, incremental improvements, and legacy code.

In Why Information Grows, César Hidalgo views every complex system exactly as vast computational networks that process information. Taken together, it's a system that computes products and physical order that no single individual could create alone.² A system’s ability to compute depends on three distinct layers: information, which can be encoded into data and stored on disk; knowledge, the cognitive frameworks humans use to interpret that data; and knowhow, the tacit, experiential mastery required to perform complex tasks. An open data portal provides an extraordinary snapshot of the first layer. But the city's true computation happens in the non-codifiable interaction between the second and third — the unwritten norms, institutional memory, and relational trust that sit entirely outside any database.

And those cannot be pulled from any portal.

The tree registry has 200,000 entries. It’s an extraordinary testament to civic record-keeping. But drawing so many points across a screen only reveals the limit of what open data can do. You can map the coordinates, but the city's real life happens in the unquantifiable space between the people living beside them.