Mexico City's heat-related posts rise exponentially once the air passes 25 degrees C

Professor Hannes Taubenböck stated the problem the study was designed around: "Satellites and weather stations can tell us where it is hot and how hot it is. But they cannot reveal whether people cannot sleep, struggle to work, feel unsafe or suffer during their daily commute." ^[9] The design answer was to use what people wrote as a second data source. A geolocated post has both words and coordinates, so a complaint about a sleepless night can sit on the same map as the temperature and building density around it ^[2]. The work comes from Geolingual Studies, a field being developed at JMU Würzburg that joins remote sensing, linguistics, digital humanities and geography ^[11].

The threshold is the cleanest result. Once the air passed 25 degrees C, lead author Dr. Richard Lemoine-Rodríguez said, "heat-related social media posts from Mexico City increased exponentially, showing that heat becomes a public concern above this threshold." ^[3] Rising posts show rising talk about heat. The step from talk to "public concern" is the authors' interpretation. The press release does not give the number of posts, say whether counts were scaled to each district's population, or report health records such as hospital admissions.

The second of those gaps matters most for the map. Densely built districts hold more people, and more people produce more posts on every subject. A raw count would rise there even if each resident minded the heat equally. Lemoine-Rodríguez reported the hotter and the denser areas together ^[4], so the release alone cannot separate temperature from crowding. The study did use census data ^[2], the obvious source for a per-resident rate. The paper, "From Measured to Experienced Heat," in IEEE JSTARS, is where to check whether that correction was applied ^[13].

The sample has a boundary too. A geolocated post exists only when someone has a phone, an account and location sharing turned on. A resident without them is absent from the data however hot the flat gets. The symptoms users described ^[5] are genuine accounts of distress, but they come from volunteers with no sampling frame. That puts them closer to a self-selected survey than to a health registry.

The method is strongest where it finds things no sensor could. Professor Carolin Biewer said users described having to choose between heat and air pollution: "If they were to open a window to cool down, they would at the same time be forced to inhale polluted air." ^[8] A temperature map would record that home as hot and nothing more. Other posts linked the heat to social inequality and climate change, or worried about birds, pets and other animals needing water ^[12].

In my view the paper makes a good case for adding residents' own accounts to heat maps. It makes a weaker case, so far, for using post counts to rank where heat does the most damage. Taubenböck put the goal this way: "To adapt cities to increasing heat, it is not enough to know how hot they are. We also need to understand what that heat means for people." ^[10]