CO₂ Emission Visualization from GTFS Data

Overview

A research project turned data-viz tool: simulating and visualizing how much CO₂ a city's public transit system emits, using real GTFS transit data, a physically-inspired dispersion model, and interactive Deck.GL and Chart.js visualizations. I presented the underlying research at the 31st Tri-U International Joint Seminar & Symposium at Mie University, Japan, where it won the Workshop Presentation Award.

The Problem

Public transit emissions data usually exists as aggregate numbers — total CO₂ per year, per route, per agency — which makes it hard to see where and when emissions actually concentrate across a city. I wanted something closer to a weather map: emissions as a spatial, time-based phenomenon you could actually watch move.

What I Built

I built a CO₂ emission simulation on top of GTFS transit datasets, starting with SFMTA's San Francisco data as a case study. Each simulated vehicle emits particles that disperse using a Gaussian model, get pushed around by wind, accumulate over time, and decay — closer to a real atmospheric process than a static heatmap. Deck.GL renders the spatial spread over the city map, while Chart.js tracks hourly accumulation trends alongside it.

Key Features

• A particle-based emission model with Gaussian dispersion, wind influence, and decay.
• Map-based spatial visualization built on Deck.GL, layered directly over real GTFS routes.
• Hourly accumulation charts built with Chart.js, paired with the spatial view.
• A comparison between trail-based and particle-drop visualization approaches to cut clutter in dense transit corridors.

Challenges

The hardest part wasn't the visualization — it was making the underlying model realistic enough to be worth visualizing. Naive per-vehicle emission dumps looked fine on a map but didn't behave anything like real dispersion, so I spent most of the effort tuning the wind-influenced, time-decaying particle model until it produced patterns that matched intuition (and literature) about how emissions actually spread.

What I Learned

This was my first project framed explicitly as research rather than a product, which meant writing it up formally and defending the modeling choices to an academic audience. It pushed me to be a lot more rigorous about why a visualization looks the way it does, not just whether it looks good.