The Most Detailed Map of Auto Emissions in America

Category: Best visualization (small and large newsrooms)

Country/area: United States

Organisation: The New York Times

Organisation size: Big

Publication date: 10 Oct 2019

Credit: Denise Lu, Nadja Popovich

Project description:

Coal plant smokestacks have long been the poster-children of climate change. But transportation is the largest source of planet-warming greenhouse gases in the United States today – and the bulk of those emissions come from driving in cities and suburbs. For this project, we analyzed data from Boston University’s Database of Road Transportation Emissions to show readers how driving-related emissions in their hometown have changed over time and how they compare to emissions from other metro areas across the country.

Impact reached:

Dozens of local media outlets from around the country — from Pittsburgh to Jacksonville — picked up the piece and used the dataset to report their own local stories on driving-related emissions. (See list below.)

Additionally, the map from the piece was requested and used by a representative of the Minnesota State House in a committee meeting on transportation-related emissions.


Des Moines

Boston (interview with researcher)

New Haven


Rochester, N.Y.





Techniques/technologies used:

We used GDAL and rasterstats, a Python library, to process raw raster images from the Boston University research group. Mapshaper was useful in helping us trim the large rasters by various Census shapes until we found the right geography for our story. (We went through a few different analyses, first looking at metro-level, t We used QGIS to see the geospatial data on a map and test out various color scales. To analyze the numerical and per capita values, we used Python to parse raw population data and Javascript to sketch out small multiple charts and maps of different cities to find any interesting trends over time. For the final product, we made a custom slippy map, mainly using the D3.js library, sandwiching together static layers of the data and map labels along with an interactive SVG layer for users to interact with to find out more details about specific metro areas. This allowed us to keep the original map projection of the raw data so none of the pixels would be resampled. We could also add a cartographic touch with custom road shield labels to emphasize the roads where there were high emissions. We geolocated readers as they landed on the page so readers would instantly see their own metro area on the map. The article page was powered by ArchieML, a tool that links a Google Document to the article page, which allowed for easy editing and flexibility in moving components around, very useful for us during the storyboarding process. Finally, for the print product, which included a doubletruck of the map, we used QGIS to bucket our data and export individual data layers to Photoshop. This method allowed us to control the exact CYMK values we wanted to use to make sure the data was clear in

What was the hardest part of this project?

Climate change is often a subject that feels abstract to readers. Greenhouse gas emissions are not something we encounter with our senses every day. We do not see them with the naked eye, or even smell them by nose, the way we experience other pollution. Yet, most of our actions result in these emissions.

The hardest part of this project was finding a way to tell the visual story of driving-related GHGs in a way that would connect with readers.

The high level of detail offered by the road-level emissions maps from Boston University was a big draw because it helped make the abstract idea of driving-related emissions feel less abstract.  Through the maps, readers could see emissions coming from the very roads they drive on every day, making a direct linkage back to the source: Cars. (The research team mapped emissions down to the road-segment level.)

However, relying on the detailed, road-level map data for the top visual was limiting for storytelling purposes, as these maps only show total emissions and cannot be easily translated to per-capita measurements (not as a technical issue, but rather an issue of emissions allocation). The per-capita data, however, is much more useful for comparing city by city greenhouse gas releases, and for readers who wanted to understand their own impact.

We reconciled these competing needs by using the detailed map as the base for our lead graphic, in order to provide readers with visual context, but overlaying it with information that is not directly reflected in the map in the form of a tooltip. 

This top graphic provided a personalized “hook” to get people to understand the story more personally. Then we were able to use the rest of our data analysis to compare trends over time and across cities lower down.

What can others learn from this project?

A single dataviz doesn’t have to do all the work, but can serve as a visual “nut graf” that summarizes data graphics further down in the piece. We used our top visual as a compelling hook into the rest of the piece, but were able to tell a more nuanced story further down the page once we captured a viewers attention and established them in a sense of place.

Others can also learn to be bold in their use of maps to help locate readers within the story. We built our own large slippy map geolocated to a users’ location for this piece in order to help readers understand the story through connection to their own city/town. Readers were able to intimately explore the dataset at the “new view” (or zoomed in scale) because it offered a more interesting story and a more personal connection to the data.

Project links: