The simulation needs a larger screen — open this on a desktop. The ten analyses below work here.

Loading transit schedule…

Speed 30×

Transit Pulse

Boston rapid transit + traffic

MBTA subway/light rail (real GTFS times) + modeled interstate/freeway traffic

Ask the map

Welcome! Ask me about anything you can see on the map.

All the answers will be based on the data from the simulation.

Example prompts

Date & time

—trains
in motion

—cars on the
highway network

—air passengers
at Logan

Trips in service by hour

Year 2026

Month September

Day 22

Time of day —

People

grouped totals — zoom in for individuals

Click a dot for that resident's card; click a transit line for its route and headway.

The People layer is the Boston agent simulation: 1,182,015 simulated residents across the simulation's extent — a box wider than the city itself, so trips crossing the boundary keep both ends. Each is given a real home building and matched to a real time-use diary, so what they are doing at any minute reflects survey data rather than invention. Each dot is one simulated person.

The population streams by viewport: zoom in and the tiles for that area load everyone who is there, so you are looking at the real, whole population of that patch and can click any individual for their id, age, sex, employment and what they are doing right now. Zoomed out you get a representative sample instead — a screen has nowhere near enough pixels for 1.18M dots, so drawing them all would cost a lot and show you less.

Nobody is routed yet: people travel in straight lines between home, work and school, so they cross the river rather than using bridges. Activities that happen neither at home nor at work are placed at a real destination wherever a source says one exists: errands at a shop near home, lunch at an eating place near work, and evening eating and drinking at a venue the City licenses to be open at that hour. Everything else still defaults to home, because nothing in the data says where it happened.

Transit

Each transit pulse is one scheduled trip; click a line for its route, direction, and headway (minutes since the previous train on the same line/direction).

Shadows & sun

real solar position for the chosen date and time

Automobile traffic

free flow — short dim tail
near capacity — longer tail
jammed — bright crawling queue
one dot ≈ 100 cars

Development pipeline

under review
board approved
under construction

Bluebikes

real dock volumes, modelled destinations

Logan flights

real domestic flights for this date

Logan flights: domestic flights are real BTS on-time records — every reported flight on its actual wheels time, climbing or descending on the true bearing of its origin or destination, width by estimated passengers (the tail number's aircraft × a ~85% load factor).

International flights are modeled from DOT monthly route totals: real routes, carriers, frequencies and per-flight passenger averages, on typical bank times (Europe departs in the evening, arrives mid-afternoon) rather than actual times — no open per-flight international source exists. Aircraft cast shadows only below ~500 ft.

Air passengers are non-residents, separate from the People layer, dwelling at the terminal per the model stated above — watch them pool before the evening Europe bank.

Air passengers

non-residents inside the real terminal buildings — arrive 40–120 min before departure, disperse within 45 min of landing (a stated model for reading ground-traffic demand)

How this is modelled

Traffic pulses are modeled, not measured: MassDOT's annual-average traffic count (AADT) for each interstate/ freeway segment, spread across the day using a real hourly shape measured at Massachusetts's own FHWA continuous-count stations, then colored/slowed by volume vs. a rough estimated road capacity (the standard BPR congestion function) — a statistical estimate, not live GPS or sensor data. Only interstates/freeways are shown; arterials and local streets aren't modeled yet.

The sun's path (and the shadows it casts) follows the time slider, plus the month/day sliders above it — move those to see how shadows lengthen and shift direction with the seasons (longest around the December solstice, shortest around June). The year slider picks the calendar — which dates are weekdays, Election Day, and how much of the development pipeline exists yet; the day stays a representative one rather than a real calendar date, so the sun isn't tracking today's actual date either.

Sources

Transit and traffic. MBTA GTFS via Mobility Database, under the MassDOT Developers License Agreement · MassDOT Traffic Inventory + FHWA TMAS.

Buildings. Footprints and heights from Overture Maps Foundation, buildings theme — © OpenStreetMap contributors, available under the ODbL. Height where Overture has it, else estimated from floor count. The extract covers central Boston, so pulses running past its edge (the Red Line's Braintree branch, say) show without buildings.

Bikes. Bluebikes System Data, under the Bluebikes Data License Agreement. Published here only as hourly departures and arrivals per dock, never as trip records.

Water. Census TIGER/Line AREAWATER (Suffolk, Middlesex & Norfolk counties).

Massachusetts comprehensive planning · M.G.L. c. 41 § 81D

The ten analyses a Boston plan needs, built from the same people

Every resident moving on the map above carries an age, an income, a household and a job. Those same people are the evidence base for the ten analyses Massachusetts master plans, Community Planning grants and One Stop applications all turn on — each one below computed from the source the Commonwealth names, with the requirement it answers stated first and every gap said out loud.

✓

Does the simulation match reality?

Before any of the analyses above can be trusted, the population on the map has to behave like Boston. This is that test, run against traffic counts the synthetic residents were never shown.

Validation: simulated drivers vs counted traffic

Screenline test: every MassDOT road segment with a real traffic count in the map's extent (20,957 segments) against the 62M simulated driver crossings from the diaries' own travel modes. No calibration — the synthetic residents were never shown a traffic count.

road class	segments	rank correlation (ρ)	who drives here
freeway (1,2)	1,324	-0.01	regional through-traffic dominates — the stated negative control
arterial (3)	4,543	+0.06	mixed local + regional
collector/local (5,6)	15,084	+0.31	resident-generated traffic dominates — the test

The correlation rises exactly as the road class localizes — the signature a residents-only model must produce: it ranks the streets its own residents load (ρ = +0.31 across 15,084 local segments) and rightly cannot explain freeways carrying the whole region. The hourly rhythm found its missing population. Boston has 16,466 residents whose job is driving — truck, delivery, taxi and bus operators — and the simulation had them parked at a workplace dot all day. Putting them on real road alignments for their shifts lifted the midday road population from 15% to 28% of the simulated peak (counted: 80%) and the hourly correlation from 0.58 to 0.70. What remains is the rest of the non-commute day: errands, appointments and service trips still glide straight instead of routing on streets, and regional through-traffic is outside a residents-only model by construction. Mode share checks out exactly: 86% of simulated travel minutes are in a car, the national diary share.

Who is actually in the counts. Census LEHD origin–destination records where the worker in every job sleeps: of the 1,049,449 jobs inside this map's extent, 60.6% are held by people who live outside it — 565,974 from the rest of Massachusetts and 70,318 from other states, arriving every weekday and absent from a Boston-only population by construction. (175,623 residents commute the other way.) MassDOT's own truck volumes add the freight: 2.66% of counted vehicle-miles, 5.9% on freeways, peaking on I-93, Route 1 and Chelsea's Broad Street beside the petroleum terminals. Through traffic is the one genuine residual — and the freeway-versus-local gradient above is its fingerprint. Roughly two thirds of work-related car trips here are made by people the simulation does not contain; that is a scope limit of a city-sized population, not a behavioural error, and it closes when the population goes regional.

hourly profile, each normalized to its own peak — simulated driver-crossings · counted vehicles on the network (AADT × TMAS)