Subway Riders Are Getting Hotter, and 85,000 Social Posts Prove It

The London Underground has hit 47 degrees Celsius. That is 116 degrees Fahrenheit, hotter than the highest air temperature ever recorded in the city. And it happened below ground, in tunnels originally built for horse-drawn carriages, where millions of commuters spend portions of every weekday.

A study published March 10 in Nature Cities has now put numbers on what subway riders have been saying for years: it is too hot down there, it is getting worse, and the pattern tracks closely with rising surface temperatures. What makes this study unusual is its data source. Instead of deploying sensors or conducting surveys, Northwestern University researchers Giorgia Chinazzo and Alessandro Rotta Loria scraped 16 years of social media posts and online reviews to build one of the largest datasets ever assembled on underground thermal discomfort.

22,000 complaints, three cities, 16 years

The team collected more than 85,000 posts from X (formerly Twitter) and Google Reviews mentioning subway conditions in London, Boston, and New York between 2008 and 2024. Using natural language processing, they filtered for thermal complaints, keeping posts that used words like "hot," "warm," "scorching," and "boiling" in a temperature context while excluding irrelevant uses like "hot dogs" or "warm smiles."

After filtering, 22,000 posts remained as confident thermal complaints spanning all seasons, times of day, and years. The researchers then correlated these complaints with recorded surface temperatures.

The relationship was direct and consistent. For every 1 degree Celsius increase in surface temperature above 10 degrees Celsius, thermal complaints rose by 10% in Boston, 12% in New York, and 27% in London. Complaints peaked in summer, especially July, and spiked during major heatwave years like 2018 and 2019.

Why underground spaces trap heat

Unlike streets and buildings, underground environments do not cool down quickly. Soil and rock act as thermal insulators, absorbing heat from trains, passengers, lighting systems, and the surface above, then holding it. In some metro systems, below-ground temperatures exceed surface records because the heat has nowhere to go.

The problem compounds over time. Rotta Loria quantified rising subsurface temperatures in a 2023 study, documenting warming trends in transportation tunnels, parking garages, and basement facilities. But long-term data on how commuters actually experience that warmth has been scarce. The crowdsourcing approach fills that gap.

Winter coats and rush-hour misery

One finding that initially surprised the researchers: commuters complained about heat even in winter. The likely explanation is straightforward. People dress for freezing outdoor temperatures, then descend into a subway system that is 15 or 20 degrees warmer than the street. Bundled in heavy coats, they feel overdressed and uncomfortable almost immediately.

Timing patterns also emerged. In London, complaints spiked in the late afternoon. In Boston and New York, they peaked around noon and again at 9 p.m. On weekends, complaints dropped across all three cities, even though underground temperatures were not necessarily lower. Chinazzo suggested that lower crowding, more casual clothing, and a different state of mind, such as being on leisure time rather than commuting to work, may account for the difference.

Targeted cooling instead of constant cooling

The practical value of complaint mapping lies in its granularity. Running ventilation fans continuously throughout every station all day is expensive and often unnecessary. But if transit agencies know that complaints spike between 4 and 6 p.m. on weekdays at specific stations, they can target cooling interventions to those windows.

Rotta Loria put it simply: enhancing cooling during specific times when people are uncomfortable means spending less energy and less money. The approach treats social media data as a real-time proxy for thermal monitoring, cheaper and more responsive than installing dedicated sensor networks across aging infrastructure.

Beyond discomfort: infrastructure and health

Extreme underground heat is not just an inconvenience. It deforms train rails, accelerates wear on mechanical components, affects groundwater quality, disturbs foundations of civil infrastructure above, and disrupts subterranean ecosystems. Heat exposure in crowded, poorly ventilated underground spaces also poses genuine health risks, particularly for older adults and people with cardiovascular conditions.

Extreme heat already kills more people annually than any other natural hazard. Extending that risk into enclosed underground environments where millions of people spend time daily adds a dimension that urban heat planning has largely overlooked.

What this study cannot tell us

Crowdsourced data has inherent biases. People who post about subway conditions skew toward those with social media accounts and the inclination to complain publicly. The dataset does not capture the experiences of riders who are uncomfortable but silent, or those without internet access. The three cities studied, London, Boston, and New York, all have old, deep metro systems. Results may differ in newer, shallower, or better-ventilated systems in other parts of the world.

The study also measured perceived discomfort rather than physiological heat stress. A complaint about being hot is not the same as a measured core body temperature. And natural language processing, while effective at scale, introduces classification errors that manual coding would catch.

Still, the researchers emphasize that the technique could be applied to metro systems anywhere. It requires no new hardware, no survey budgets, and no cooperation from transit agencies. The data already exists, posted daily by riders who do not realize they are contributing to climate science.

The study was supported by the National Science Foundation (award number 2046586).