Urban heat work is the data layer a city wraps around temperature — and it points two opposite ways at once: the heat a city has to shed, and the heat a city has to deliver. The instruments barely differ: a thermometer that raises a heat alert in Ahmedabad and one that throttles a heat pump in Copenhagen sit on the same procurement list. What differs is who may act on the reading.
This guide connects the urban heat island definition with 29 documented projects in 24 cities; thermal-grid generation belongs to the smart grids guide. Air quality asks the same question of a different pollutant, and the difference matters: an air index in Delhi can legally halt construction, while no heat threshold here carries automatic force.
Four instruments recur, and each appears on both sides of the split:
Our essay on what happens after a sensor fires sorts monitoring onto six rungs, from "nothing follows" to "the observation is the act". Heat splits along it by side. The supply half sits high without trying: Harbin's 156 self-operating heat-exchange stations need nobody's permission.
The adaptation half clusters at the bottom: Melbourne's readings guide where to add shade, and nothing fires. Athens climbs one rung, giving residents a risk score and directions to the nearest air-conditioned building — informed, acting alone. HeatSens reached the rung where staff decide with the data. Rio looks higher than it is: a decree gives its five levels standing, but the mayor approves each response, and even at the top tier interruption "poderá ser indicada" — may be indicated. Ahmedabad tops this half, and bought almost no sensing to get there: a forecast, a numbered band, and one nodal officer whose job is to act on it.
| City | Instrument | What the record shows |
|---|---|---|
| Ahmedabad | Forecast threshold bands, 2013 | Mortality rate ratio at 47°C fell 2.34 → 1.25 — a pilot evaluation, its authors say |
| Los Angeles | Reflective coating, ~175 lane-miles | Surfaces up to ~10°F cooler — peer review also flagged more sunlight onto pedestrians |
| Copenhagen | AI demand response on district heat | Morning peak cut up to 22% on 17 buildings; 140 buildings today — operator-reported |
| Bucharest | Live outage map, since 2021 | The map works; the network lost 37.73% of its heat in 2024 |
| Melbourne | Microclimate sensors, since 2019 | Six-plus years of open readings, no threshold attached; 16 sites deployed, six still active |
Two of the 29 entries carry peer-reviewed evidence, both on the adaptation side, both qualified. Ahmedabad's 2018 pre/post study estimated 1,190 annualised deaths avoided, but its design is ecological — it cannot separate the plan from everything else that changed. Los Angeles is stranger: Ko et al. (2022) confirmed real daytime cooling and flagged increased sunlight reflected onto people at street level. A tradeoff, not a win.
Twenty of the 29 rest on official evidence — utilities and departments reporting on themselves — and not one of the 17 supply-side entries carries a peer-reviewed number. Several headline figures are not measurements at all: Geneva's 70 GWh is projected for 2045, Oslo's 350,000 tonnes is design capacity for a plant not yet built. The only independent audit in that half checked Vancouver's rates, not its carbon; the one other independent supply figure measures a failure — Bucharest's 37.73% loss.
Nothing here failed technically. One entry is formally scaled back: Miami-Dade created what it and its funder call the world's first Chief Heat Officer and won a lower heat-advisory threshold — then Florida's HB 433 barred counties from requiring heat protection for outdoor workers, and a $400M budget gap took the post. The heat season still runs; nobody holds the title. The other losses are quieter: two-thirds of Melbourne's sensor sites have gone dark, and Zurich's Greencity pioneered a 2000-Watt label scrapped in 2024 for lack of demand. Vienna's Supergrätzl is the counterpoint: opened May 2026, evaluation still running, nothing claimed yet. The pattern is mandate, maintenance and money — never hardware.
The practical takeaway: buy the addressee, not just the thermometer. On the supply side the loop closes in a valve; on the street it closes in a person — whose authority a budget line can remove while the sensors keep reporting.
The tendency of a dense, paved city to stay hotter than the countryside around it, most stubbornly after dark (full definition). The atlas records two lines of attack on it. Surface: Los Angeles has coated roughly 175 lane-miles of residential street with a reflective sealant since 2017, prioritising its hottest neighbourhoods. Space: Paris has converted 203 schoolyards and 5 nurseries into greened, rain-absorbing "oases" that open to residents during heatwaves, with 360 targeted by 2030.
The one plan in this corpus with peer-reviewed evidence suggests yes, with caveats its own authors insist on. Ahmedabad wrote South Asia's first heat-health action plan after a 2010 heat wave with an estimated 1,344 excess deaths. A 2018 pre/post evaluation found the all-cause mortality rate ratio at 47°C fell from 2.34 before the plan to 1.25 after, estimating 1,190 annualised deaths avoided. The design is ecological — it cannot separate the plan from everything else that changed in those years, and the authors call it a pilot evaluation.
Three instruments dominate the documented record. Fixed sensor networks: Melbourne has published fifteen-minute microclimate readings as open data since November 2019, and Gründau carries seven playground sensors over a municipal LoRaWAN. Satellite and aerial survey: Mumbai's climate action plan rests on a satellite-based ward-level assessment of heat, flood and landslide risk, and Milan flies night-time thermography over its heating network to find leaks. Forecasts: Ahmedabad acts on probabilistic maximum-temperature forecasts up to seven days ahead rather than on its own sensors.
Because it is the same measurement problem pointed the other way — 17 of the 29 entries here are about heat a city must deliver rather than shed, and they use the same sensors, maps and thresholds. The difference is what the reading is permitted to do. Copenhagen's AI predictive control cut morning peak heat load by up to 22% across 17 municipal buildings because a forecast can move a valve unaided. Stockholm Exergi converts the reading into a daily price so businesses can sell waste heat back. Bucharest shows the limit: its live outage map is accurate, and the network still lost 37.73% of the heat fed into it in 2024.
Four responses to the same hazard, ordered by how far the city's own hand reaches: from a post created by philanthropy to a plan that changes what hospitals and water carriers do on a forecast day.
| Project | Since | Status | Outcome | Evidence | Run by | Cost | What the city can actually act on |
|---|---|---|---|---|---|---|---|
| Heat Action PlanAhmedabad, 🇮🇳 | 2013 | Live | Ongoing | Peer-reviewed | City government, National government, Research institution | not public | A forecast triggers an alert the municipal corporation acts on, and the peer-reviewed literature calls it South Asia's first heat-health action plan. |
| Cool Pavement ProgrammeLos Angeles, 🇺🇸 | 2017 | Live | Succeeded | Peer-reviewed | City government | ~$40,000 per lane-mile (StreetsLA); FY2023 ~$4M from the Street Damage Restoration Fee | Physical change the city controls outright: reflective coating on residential streets, prioritised in the hottest neighbourhoods. |
| Extrema Global Heat Risk ApplicationAthens, 🇬🇷 | 2018 | Live | Ongoing | Independent | Research institution, City government | not public | Information rather than infrastructure: personalised heat-risk scores and directions to the nearest cool space. |
| Miami-Dade's Chief Heat Officer and Extreme Heat Action PlanMiami, 🇺🇸 | 2021 | Completed | Scaled back | Media | Regional government | not public | A post rather than a power — created with philanthropic seed money, and abolished when the budget and state law closed in. |