Skip to main content
Prakriti Bista

Design + Engineering

DC Map: Interactive 3D Visualization

A game-like 3D map for exploring Washington D.C., with building extrusions, fly mode, landmarks, museums, and actual tree data.

DC Map: Interactive 3D Visualization

Project thesis

A city map reframed as an explorable world, using game mechanics and layered geospatial data to reward wandering.

Question

Make city exploration feel active, spatial, and a little game-like.

Decision

A browser-only build forced every map layer and interaction to earn its performance cost.

Why Google Maps wasn't enough

I live in DC and genuinely love exploring it: the museums, the monuments, the parks, the neighborhoods. But most people navigate the city by staring at Google Maps or getting lost. Neither felt like the right way to experience Washington D.C.

I wanted something that made the city feel like a world to explore rather than a set of directions to follow. Something closer to a game than a navigation app. So I built one.

A map you actually want to open

DC Map is a 3D interactive map of Washington D.C. that is meant to be fun, not just useful. What if you could explore the city from your browser the way you'd explore a game world? Discover landmarks, learn about museums, fly over the National Mall, watch the trees change color with the seasons.

The map uses Mapbox GL JS with 3D building extrusions. Real elevation data, real building geometry, rendered with a slightly stylized aesthetic that makes it feel designed rather than like a raw satellite view. You can zoom, pan, rotate, and tilt freely. The whole screen is the map.

DC Map 3D view
3D building extrusions over Washington D.C., stylized not photorealistic

The design decisions that made it feel right

Photorealistic maps signal "this is for getting somewhere." Stylized maps signal "this is a place worth spending time in."

The aesthetic choice came first and shaped everything after it. Slightly stylized buildings, muted color, clean geometry: all of it signals something different from Google Maps. This is a place you'd want to spend time in.

The information cards on landmarks and museums are deliberately minimal: a name, one sentence of context, and a search link. A detailed card creates a reading task. A brief card tells you just enough to be curious, then gets out of the way.

The layer controls give users independent toggles for landmarks, museums, trees, and parks. This was about respecting that people explore differently. Someone doing a museum tour doesn't need seasonal tree data cluttering the view. Independent layers let the user compose their own map, closer to how people actually think about a city.

Landmark and museum panels
The landmark and museum panels. Browse every place, fly straight to it, or toggle the whole layer off.

The city in layers

The city is organized into layers you can toggle independently. The tree layer marks a few dozen notable trees and groves around DC and the wider DMV, from the Tidal Basin cherries to the oaks of Rock Creek Park, while the dense canopy you see is Mapbox's own 3D tree rendering. You can change the tree colors by season: spring greens, summer deep green, autumn reds and oranges, winter grey. Parks layer in with their own seasonal coloring on top.

Landmarks and museums are the exploration layer. The map includes 10 curated landmarks and 36 museums from Open Data DC. Click on any of them and you get a small information card: what it is, a bit of context, and a search link if you want more.

The stylized 3D map of the National Mall
Get close to a landmark and its info card pops up

Points for wandering

There's a progression system built into the exploration. You discover a landmark by clicking it from within 50 meters on foot, or by flying within 40 meters of it. Discovering a landmark earns 50 XP. The XP accumulates into a level system with achievement unlocks. Progress is saved locally so your discoveries persist between sessions.

The specific mechanic matters: you earn points by getting within 50 meters of a landmark, not by clicking on it from any distance. It requires physical navigation. You have to move toward the thing to get credit. That changes the mental model from "looking at a map" to "moving through a city."

Progression · points for getting close

within range

Am I close enough to an undiscovered landmark?

Fly mode checks every 500 ms (40 m); on foot, a landmark counts when you click it from within 50 m

+50 XP

Discover the landmark

Marker turns from red → green; discovery animation fires

level up

New level + achievement unlocks

In fly mode a loop checks your position against nearby landmarks twice a second, and on foot a click only counts when you're close. You get XP for actually moving within range, not for clicking from across the map. That one rule turns looking at a map into moving through a city.

Fly mode

Fly mode lets you leave the map plane entirely and navigate DC as a first-person flyover, moving through the 3D building geometry with WASD controls and mouse look, like a free camera in a game engine.

Getting this to feel right took iteration. Camera movement with momentum and easing feels cinematic; without it, it feels like a broken drone. I went through multiple iterations of easing curves, acceleration rates, and turn sensitivity before it landed: responsive but not nervous, cinematic but controllable.

DC Map fly mode
Fly mode drops you down into the city, moving with WASD and mouse look.

The data behind the map

Most of the interesting technical work was in the data layer. DC Map uses several static GeoJSON datasets: landmark locations (manually curated), museum locations from Open Data DC, a hand-placed set of notable trees, DC's walkable road network, and a navigation graph built from road data for pathfinding.

The road network data needed preprocessing. I built a custom walk-graph generation script that converts the raw road data into a navigable node/edge graph, and Turf.js does the geospatial math in that build step. At runtime the discovery loop stays dependency-free: a small hand-written Haversine calculation handles proximity detection, so the proximity check, which runs a couple of times a second, doesn't pull in a library.

The architecture runs through a centralized StateManager that coordinates game state, landmark tracking, museum progress, XP, fly controls, and map state through custom hooks and providers. Everything that matters (discoveries, level, achievements) persists to localStorage.

System architecture

GeoJSON

Landmarks · museums

10 curated landmarks · 36 museums (Open Data DC)

GeoJSON

Trees · roads

72 hand-placed notable trees · DC walkable road lines

Mapbox GL JS · Standard style

Native 3D map

Buildings, trees, satellite toggle · zoom / pan / rotate / tilt / 360° orbit

StateManager + hooks

Game state

useGameState · useLandmarks · useMuseums · useFlyController · useExperience

localStorage

Your progress, kept

dc-game-progress · dc-experience · dc-waypoints

Turf.js runs once at build time to bake the roads into a walk graph. At runtime, proximity is a small hand-written Haversine, so the live loop stays light. A StateManager component ties the hooks together, and anything worth keeping goes to localStorage.

What it's for, and what's still missing

Navigation apps are optimized for getting you from A to B efficiently. That's useful. But exploration is a different mode. Curiosity and wandering deserve a different interface.

DC is a beautiful, walkable, historically dense city that a lot of people only experience as a sequence of tourist checkpoints. DC Map is for the other kind of visit: the one where you wander, discover something you didn't plan to find, and come away with a better sense of the place as a whole.

The open problems are the ones I hit myself: WASD fly mode only feels natural if you grew up on PC games, and the places are a static snapshot rather than a live feed from Open Data DC.

What worked

  • 3D building extrusions with Mapbox GL give the city real weight and depth
  • The fly mode camera uses momentum and easing so movement feels smooth rather than nervous, and getting that right took many iterations.
  • The XP proximity mechanic rewards moving toward a place rather than clicking it from a distance
  • Stylized map design makes it feel like an exploration tool, not a navigation clone

What didn't

  • WASD fly mode is only intuitive for users with PC gaming muscle memory
  • Landmark cards are deliberately minimal, which can leave curious visitors wanting more context
  • Data remains static. Connecting to Open Data DC API for live updates is still on the roadmap.

Built with

TypeScript · Mapbox GL JS · Next.js · React · Tailwind CSS