Forecasting Cities: From Static Maps to Algorithmic Urban Models
- Anish Khare
- May 7
- 6 min read
The world of urban planning has been undergoing a silent revolution. For the past century, city planners have relied on three primary tools: the paper map, the physical survey, and the census conducted once every decade. But the data supply these produce has not kept up with the pace of cities, especially in the global south. In many emerging cities, urbanization has outpaced both tech and government infrastructure creating what I call “data bottlenecks” that prevent planners from solving many urban challenges.
Chief among them is something we in India call the “delayed census report.” Put simply, census data is too slow. In India, census data released can be 5-7 years out of date by the time it reaches your desk. Planners have long had to contend with archaic demographic and housing stock data when making critical decisions like where health services should be located, where new schools should be built, or where to expand public assistance programs. This problem forced me to ask myself two questions. First, why is planning based on a 10-year-old snapshot of the city we call “Snapshot Planning?” and second, how do we graduate to “Dynamic Urbanism?” As it turns out, the answer to both questions is AI.
Experiencing Two Worlds
When I lived in United States, I used to marvel at how comprehensive US Census Bureau’s datasets were. Planners there enjoyed regular (annual), high-resolution data about everything from commute routes to the age of every home on a single street block. It gave planners like me great confidence that we had a reliable “data backbone” to make evidence-based decisions that impacted communities.
Imagine my surprise when I came back to India and realized that our cities lacked even the most basic datasets about our cities. India is the world’s fastest-growing major economy. We have a higher GDP growth rate than the EU, US, Russia, Japan and China combined! We also have the largest technocracy in the world with over 6 million+ people working in the IT and AI sector alone. Despite this, we currently find ourselves in the middle of the backlash from our last census, which was supposed to be conducted in 2021, but got delayed to who-knows-when. Where we in America ask “why hasn’t that building been demolished?”
While in India we ask “why was that entire building not recorded in the census?”
That cognitive dissonance set me off on a thought journey: if we have so many people working in tech & IT how can we use them to solve world’s biggest democracy’s largest planning pain points? How can we use AI tackle issues in a country that’s changing at the speed of a sprinter, instead of a marathon runner?
AI for Faster Censuses: Meet your Synthetic Neighbors
In many fast-growing heritage cities in India, waiting a decade to plan isn’t an option. So how do you plan without a census? The answer: Synthetic Population Modeling. Synthetic population modeling is essentially creating digital twins of your city’s residents. Instead of physically counting everybody’s heads, sampling information about small sets of people (through detailed sample surveys), and then using AI to statistically “inflate” that information to the entire population.
Using these digital clones, we can model what-if scenarios. How will housing needs change if we build a new metro corridor today? How many new schools will we need in 5 years? By knowing the future attributes of a city, we can better plan for what a city needs tomorrow instead of what it needed ten years ago. Synthetic population modeling is basically a new world solution for an old problem.
“Eyes in the Sky”: Earth Monitoring Satellites
If censuses tell us who lives in a city, satellites can tell us what lives in a city. When I was in school, city expansions were mapped by having technicians painstakingly trace the outline of every building onto a computer screen. In the age of Computer Vision (CV), that same building footprint can be programmatically identified using satellite imagery and models like Mask R-CNN. Roads can be measured in width, trees can be counted, public squares can be located, all this with nearly military grade accuracy.
So what does this mean for improving property taxes in Indian cities? For one, cities like Mumbai and Bengaluru have vast networks of unauthorized buildings, many of which escape taxation because the municipal body responsible for collecting those taxes isn’t aware of their existence. By using satellite imagery to count the actual number of floors in a building and using AI to match that information with property tax registries, these civic bodies have uncovered invisible floors. Doing so has helped these cities grow their property tax revenues by up to 40%.
And when official data is scarce, AI can use proxy variables to estimate where people live and how many there are. One of the best proxies for economic development and wealth is nighttime lights as seen from space. Satellites can now estimate how many people have moved to a city within months of them moving by looking at how much a city’s “glow” has changed over time.
Predictive Urban Modeling: Goodbye Maps, Hello Magic Crystals
If you’re an urban planning undergrad or professional like me, you’ll find the transition from “Descriptive GIS” to “Predictive GIS” most exciting.
Map Cleaning: Gone are the days of spending days manually fixing “sliver polygons” or “dangling nodes” on your digital map. AI can now automatically “heal” these maps in a matter of seconds.
Urban Growth Modeling: Based on past trends and patterns of development, we can now feed an AI where exactly our cities will grow (to 2031) years before construction even begins. In Jaipur, cities have started using this model to understand where open space will be devoured by urbanization so they can buy that land, develop parks, or increase transit options before it’s too late.
Digital Twins: Every city needs a digital twin. Imagine if we could monitor our cities in real-time by taking the sensors on every smartphone (and IoT device) and merging it with a live, digital model of our cities. By overlaying this sensor data on a “live map” of our cities, we could now watch traffic, pollution, and even wait-times for elevator banks at metro stations update before our eyes.
Case Studies: AI at Work in India
India’s Smart City Mission was launched in 2015 with the goal to use technology to make 100 Indian cities “smart.” Here are some solutions already being deployed:
Pune: AI-based traffic applying Intelligent Traffic Management System (ITMS) decreased congestion on arterial roads by up to 19%.
Bangalore: Drone surveillance and AI are being used to map slums and triage which slums need water and sanitation services most urgently.
Indore: AI-powered waste management enabled route optimization for garbage collection trucks. Indore is now considered the cleanest city in India.
Conclusion:
The transition from a static map to a dynamic algorithmic model does not replace the urban planner; rather, it evolves their role. But what they can do is empower urban planners with more relevant, high-resolution data than they’ve ever had access to. Urban planners are the data orchestrators of their cities. In an era of rapid urbanization, we can’t afford to make decisions based on stale datasets. Let’s put our billion plus population and tech talent to work creating AI-backed solutions to problems that have plagued cities for decades.
What features do you find useful in the context of your city?
0%Map Cleaning
0%Urban Growth Modeling
0%Digital Growth
0%All of the above
Sources:
The Alan Turing Institute. (n.d.). Synthetic population estimation and scenario projection (SPENSER). https://www.turing.ac.uk/research/research-projects/synthetic-population-estimation-and-scenario-projection
Bihar Shodh Samaagam. (2025). The impact of India's delayed 2021 census on urban development and policy. Bihar Shodh Samaagam, 3(2), 120–121. https://www.biharshodhsamaagam.com/header_banner/sub_menu_banner_1747026296_223.pdf
Dadhich, G., & Hanaoka, S. (2025). Urban growth prediction in Jaipur city using cellular automata and machine learning. Revista Latinoamericana de la Papa, 29(1), 1–15. https://papaslatinas.org/index.php/rev-alap/article/download/56/53/103
Inakhiya, G. (2025). Artificial intelligence and smart cities in India: A conceptual and policy review. Town and Regional Planning, 87, 148–160. https://journals.ufs.ac.za/index.php/trp/article/view/9623
Indian Institute of Public Administration (IIPA). (2024). AI for smart cities and infrastructure management. GyanKOSH. https://www.iipa.org.in/GyanKOSH/posts/ai-for-smart-cities-and-infrastructure-management
Press Information Bureau (PIB). (2024). Cabinet approves IndiaAI Mission to strengthen the AI ecosystem. Government of India. https://www.pib.gov.in/PressReleasePage.aspx?PRID=2178092
ResearchGate. (2024). Leveraging artificial intelligence for slum mapping and upgrading programs using drones: The case of Bangalore city, India. https://www.researchgate.net/publication/379055595_Leveraging_Artificial_Intelligence_for_Slum_Mapping_and_Upgrading_Programs_Using_Drones_The_Case_of_Bangalore_City_India
Comments