Urban Heat Islands: 1 in 25 Summer Deaths in EU Cities

The core finding of this report is that the urban heat island is not a meteorological curiosity but a design outcome with a measurable death toll, and one that falls hardest on the neighbourhoods that planners historically neglected. A modelling study published in The Lancet in 2023 examined 57 million residents of 93 European cities between June and August 2015 and attributed 6,700 premature deaths to the difference between urban and rural temperatures ✓ Established Fact [1]. The same study found that 2,644 of those deaths — 39% — could have been prevented had tree cover reached 30% of each city’s area, a change that would have lowered summer temperatures by an average of 0.4°C [1]. The heat island, in other words, is both lethal and partly reversible. The question this report examines is why so few cities have reversed it, even where the physics, the mortality data and the engineering have been settled for more than a decade.

The size of the gap depends on where and when it is measured. The US Environmental Protection Agency summarises the research as follows: urban daytime air temperatures run about 1-7°F (0.6-3.9°C) above outlying areas, and night-time temperatures about 2-5°F (1.1-2.8°C) higher, with the largest differences in humid regions and in larger, denser cities ✓ Established Fact [2]. Satellite measurements of surfaces rather than air show the same pattern at planetary scale. A 2024 dataset from the Pacific Northwest National Laboratory, covering more than 10,000 cities over more than 20 years, found a positive surface heat island in more than 80% of them, averaging about 1.0°C by day and 0.8°C by night ✓ Established Fact [3]. Those averages conceal the extremes that kill: in a global study of 2003-2020, three-day surface heat island peaks averaged more than twice the warm-season median, with local exceedances of up to 10 K [4].

The trend is worsening, and it is worsening fastest precisely where it matters most. The PNNL dataset found an upward trend in heat island intensity in more than 60% of cities, rising by more than 0.1°C per decade by day ✓ Established Fact [3]. The 2022 study in Global Environmental Change went further: across urban areas worldwide, the three-day surface extremes grew faster than the seasonal median and are now 1.04 K — 31% — higher than in 2003 ✓ Established Fact [4]. The heat island is therefore not simply adding a constant increment to regional warming. It is amplifying the hottest days, the ones on which mortality curves turn steep. A city that runs 1°C above its hinterland on an average July afternoon can run several degrees above it during a heatwave, when still air and cloudless skies let stored heat accumulate night after night.

At the same time, far more people are living inside these heat islands. A study published in PNAS in 2021, covering more than 13,000 cities, found that urban exposure to extreme heat tripled between 1983 and 2016 and now reaches almost two billion urban residents ✓ Established Fact [28]. Its most striking finding concerns causation: population growth explained two-thirds of the increase, and warming only one-third [28]. Dhaka recorded the single largest rise, an additional 575 million person-days of extreme heat [28]. The implication is that the urban heat problem is driven as much by where and how cities grow as by greenhouse gases. It is a planning problem that happens to intersect with a climate problem, and treating it as purely the second has allowed the first to go largely unaddressed.

The wider mortality burden explains why this matters. The 2025 report of the Lancet Countdown on health and climate change estimated that heat-related deaths have risen 63% since the 1990s, to an average of 546,000 a year in 2012-2021 ✓ Established Fact [9]. Infants under one year were exposed to 389% more heatwave days, and adults over 65 to 304% more, than in 1986-2005 [9]. In Europe alone, researchers at ISGlobal estimated 62,775 heat-related deaths in the summer of 2024, following about 50,800 in 2023 and about 67,900 in 2022 — more than 181,000 across three summers ✓ Established Fact [8]. Most of those people died in cities, and many died in the hottest districts of those cities. The heat island does not create the heatwave, but it decides how much of the heatwave reaches a given bedroom at 3 a.m.

This report proceeds in the order the evidence demands. It sets out the physics of how cities manufacture heat, then the mortality data, then the equity gradient that determines who lives in the hottest streets. It compares how seven cities on four continents have responded, ranks the interventions that have measured effects — albedo, canopy and urban design — and then examines why, despite that evidence, the response remains fragmented, underfunded and in several jurisdictions actively reversed [35][36]. The conclusion is that the urban heat island is best understood not as weather but as a policy failure: a known, measurable, partly preventable harm that persists because its costs fall on people with little political leverage and its remedies fall across too many departments for any one of them to own.

The first mechanism is albedo — the share of incoming sunlight a surface reflects. Dark roofs and fresh asphalt reflect little and absorb most of the solar energy that falls on them, heating to temperatures far above the surrounding air. The NASA Goddard Institute for Space Studies measured the effect directly on New York rooftops through the summer of 2011: a white roof membrane reduced peak rooftop temperature by an average of 43°F (24°C) compared with a conventional black roof ✓ Established Fact [18]. That heat does not stay on the roof. It is conducted into top-floor flats and radiated into the air above the street. The EPA notes that the resulting daytime gap between city and countryside is largest where impervious, dark surfaces dominate the landscape [2].

The second mechanism is the loss of evaporative cooling. Soil and vegetation release water vapour, and evaporation consumes energy that would otherwise raise the air temperature. Paving over land removes that cooling, and removing trees also removes shade. A 2024 study in Nature Communications, using satellite data for the 500 largest cities in the world, found that urban green space lowers surface temperatures by about 3°C in the warm season ✓ Established Fact [29]. Where vegetation is absent, the energy that would have gone into evaporation goes into heating surfaces instead. The 2020 study of 108 US cities by Hoffman, Shandas and Pendleton attributed much of the heat gap between neighbourhoods to precisely this imbalance — the relative preponderance of impervious surface over tree canopy ◈ Strong Evidence [5].

The third mechanism is geometry. Tall buildings lining narrow streets form canyons that trap radiation, reflecting it back and forth between walls rather than letting it escape to the sky. By day, the canyons absorb heat into a large thermal mass of masonry and concrete; by night, that stored heat is released slowly, which is why the EPA finds the urban-rural gap is often larger at night than during the day [2]. For human health, the night-time gap is the dangerous one. A body that cannot cool down during sleep accumulates physiological strain across consecutive hot days, and the PNNL dataset confirms that night-time surface heat islands are positive in most cities worldwide [3].

The fourth mechanism is anthropogenic heat — the waste energy released by vehicles, industry and, increasingly, air conditioning. This is where the heat island becomes a feedback loop. Air conditioners do not destroy heat; they move it from inside a building to outside it, adding the energy consumed by the compressor. The Arizona State University study led by Francisco Salamanca found the effect was strongest from late afternoon to early morning, because the shallow night-time boundary layer concentrates the heat near the ground ✓ Established Fact [14]. The London study by Brousse and colleagues reached the same conclusion for a cooler, wetter city, estimating a citywide increase of 0.15°C and a rise of up to 1°C in central districts if air conditioning were widely adopted [15].

These four mechanisms interact. A dense, dark, treeless neighbourhood absorbs more sunlight, evaporates less water, traps more of the resulting heat between its walls, and then generates more waste heat as its residents turn to air conditioning to survive. Each of these factors is measurable, and each is in principle modifiable: roofs can be made reflective, trees can be planted, street orientation and building spacing can be regulated, and cooling can be supplied more efficiently. The physics therefore points directly towards the policy levers. What it does not explain is why the levers have been pulled so unevenly, which is the subject of the sections that follow.

One further point of physics deserves emphasis, because it is frequently misunderstood in public debate. The urban heat island is a local phenomenon, not a meaningful driver of global warming. The Intergovernmental Panel on Climate Change concluded in its Sixth Assessment Report that uncorrected urbanisation effects are unlikely to have inflated global land temperature trends by more than 10%, although the signal is larger in rapidly urbanising regions such as eastern China ✓ Established Fact [31]. That distinction cuts both ways. It means that cooling cities will not solve climate change, but it also means that cities do not have to wait for global emissions to fall before they can reduce the heat their own residents experience. Local heat is a local responsibility.

The mortality record of the past three decades reads as a sequence of avoidable disasters. In July 1995, a five-day heat wave killed 739 people in Chicago; the sociologist Eric Klinenberg later showed that the deaths concentrated among isolated older residents in poor neighbourhoods that the city had effectively abandoned ✓ Established Fact [33]. In August 2003, a heatwave killed about 15,000 people in France, and France’s public health agency found that living in an attic room directly under the roof raised the risk of death more than fourfold ✓ Established Fact [25]. A case-control study in the European Journal of Public Health identified the lack of thermal insulation and sleeping on the top floor as housing characteristics associated with death [39]. In each case, the built environment determined who died.

The European record has worsened since 2003 rather than improved. The Nature Medicine analysis led by Joan Ballester at ISGlobal counted 62,775 heat-related deaths in Europe between 1 June and 30 September 2024 — 23.6% above the 2023 estimate and 8.1% below the 2022 estimate ✓ Established Fact [8]. Two-thirds of the more than 181,000 deaths across those three summers occurred in southern Europe [8]. Ballester’s team argued that the figures demonstrate the need to strengthen adaptation strategies, including a new generation of continent-wide, impact-based heat-health early warning systems [8]. The Lancet study of 93 cities adds the urban dimension: the cities with the highest heat-island mortality were in Spain, Italy, Hungary, Croatia and Romania, and the lowest in Sweden, Estonia, the UK and northern France ✓ Established Fact [1].

Phoenix, in Maricopa County, Arizona, is the most thoroughly documented urban heat death record in the world, because the county publishes a detailed annual review. Deaths rose for a decade to a record 645 in 2023 ✓ Established Fact [10]. In 2024 the county confirmed 608 deaths: more than 75% occurred outdoors, 49% involved people experiencing homelessness, nearly 60% were aged 50 or over, and 57% involved substance use [11]. The county’s chief medical officer, Dr Nick Staab, framed the record plainly: heat-related deaths, he said, are especially tragic because they are preventable [11]. In 2025 the toll fell to 427, a 33% decline, after the county and the city of Phoenix expanded cooling centres and outreach ◈ Strong Evidence [12].

The Maricopa data also reveal how the heat island reaches indoors. Among indoor heat deaths, county reporting found that 88% of homes had an air-conditioning unit present, but 70% of those units were not working ◈ Strong Evidence [13]. Residents in the hottest neighbourhoods may own an air conditioner and still be unable to afford to run it, and transport barriers keep many from reaching public cooling centres [13]. The dividing line between life and death in a Phoenix heatwave is therefore not simply the presence of cooling technology but the ability to pay for electricity and repairs — a question of income that maps onto the same neighbourhoods where the heat island is most intense.

The Pacific Northwest heat dome of June 2021 showed that the danger is not confined to hot climates. British Columbia’s coroners later attributed 619 deaths to the event; 98% occurred indoors, 67% of the dead were aged 70 or older, 56% lived alone, and 93% had no air conditioning ✓ Established Fact [32]. The victims were concentrated in socially and materially deprived neighbourhoods [32]. In a region where summer air conditioning had been unnecessary for most of the twentieth century, the housing stock was designed to retain heat, and it did. As outdoor temperatures fell overnight, apartments kept warming, and residents were exposed to lethal indoor temperatures for days.

Japan offers a view of the chronic burden rather than the acute disaster. In 2025, the country’s hottest summer on record, more than 100,000 people were taken to hospital by ambulance for heatstroke between May and early October, and the June total of 17,229 was the highest for that month since records began in 2010 ✓ Established Fact [24]. The national temperature record was broken at 41.8°C in Isesaki, north of Tokyo, on 5 August 2025 [24]. These are not deaths, but they measure the load that urban heat places on emergency services every summer. Together, the Chicago, Paris, Phoenix, British Columbia and Japanese records point to the same conclusion: heat kills in the home, it kills the old and the isolated, and the building a person lives in largely decides whether they survive.

The clearest evidence that urban heat is a policy outcome comes from the history of American mortgage lending. In the 1930s, the federal Home Owners’ Loan Corporation graded neighbourhoods for lending risk, marking predominantly Black and immigrant districts in red. The 2020 study by Jeremy Hoffman, Vivek Shandas and Nicholas Pendleton compared those maps with modern satellite data for 108 US urban areas. Formerly redlined areas were, on average, about 2.6°C warmer in land surface temperature than non-redlined areas, and 94% of the cities studied showed the same pattern, with gaps of up to 7°C ✓ Established Fact [5]. The authors concluded that historical housing policies may be directly responsible for disproportionate exposure to current heat events [5].

The mechanism is not mysterious. Districts denied investment for decades received fewer street trees, fewer parks and more highways, industrial land and car parks. Southeast and western cities show the widest differences, Midwestern cities the narrowest [5]. American Forests’ Tree Equity Score quantifies the result: neighbourhoods where 90% or more of residents live in poverty have 41% less tree canopy than those where 10% or fewer do, and communities of colour have 33% less canopy than majority-white communities ◈ Strong Evidence [7]. The organisation estimates that the United States would need to plant more than half a billion trees to achieve equitable urban canopy across nearly 500 metropolitan areas [7].

The gradient is not an American peculiarity. Tirthankar Chakraborty and colleagues at Yale combined census and satellite data for 25 cities on several continents and found that poorer neighbourhoods experienced higher heat exposure in 72% of cases, because low income tends to accompany denser construction and less vegetation ✓ Established Fact [6]. The same inequality appears between cities as well as within them. The 2024 Nature Communications study of the 500 largest cities found that Global South cities have only 70% of the cooling capacity from green infrastructure available in Global North cities: the average Global South resident receives about 2.2°C of cooling from urban greenery, against 3.4°C in the North ✓ Established Fact [29].

Housing type compounds geography. Top-floor flats under dark or metal roofs are the most dangerous places to be during a heatwave, which is why the 2003 French data singled out attic rooms and uninsulated roofs as risk factors [25][39]. In Paris, about three-quarters of roofs are zinc, a metal that absorbs and conducts heat into the small attic rooms beneath it ◈ Strong Evidence [25]. In Ahmedabad and other Indian cities, the corresponding risk is the corrugated metal or asbestos roof of an informal settlement, which traps heat in a single room with little insulation — the reason reflective roof paint there lowers indoor temperatures by 5-6°C [22]. The poorest residents of hot cities often live directly beneath the hottest surfaces.

At the extreme end of the gradient are people with no housing at all. In Maricopa County, people experiencing homelessness accounted for 48% of heat deaths in 2025, even though they make up less than 0.2% of the county’s population ✓ Established Fact [12]. They sleep on the pavement and asphalt that the Phoenix cool-pavement programme was designed to address, often in the downtown districts where the heat island is strongest [17]. The British Columbia coroners’ review makes the complementary point for the housed poor: the victims were disproportionately old, alone, without air conditioning and living in deprived neighbourhoods [32]. Heat does not merely correlate with poverty; it exploits every dimension of it, from housing quality and electricity bills to social isolation and chronic illness.

The equity gradient changes the moral character of the problem. If urban heat fell randomly, it would be a public health challenge like any other. Because it falls predictably on the neighbourhoods that past policy starved of investment, and because the remedies — trees, reflective roofs, insulation, working air conditioning — are known and purchasable, the persistence of the gap is a distributional choice ◈ Strong Evidence [5][6]. This framing is contested by those who argue that heat mortality is driven mainly by age and health status rather than by neighbourhood ⚖ Contested [11]. The data suggest both are true: the individual risk factors determine who is vulnerable, and the neighbourhood determines how much heat the vulnerable must absorb.

Tokyo is the archetype of a city that has heated itself. The Japan Meteorological Agency reports that Tokyo’s annual mean temperature has risen at a rate of more than 3°C per century, markedly faster than less urbanised stations, as the heat island compounds global warming ✓ Established Fact [23]. Japan’s response has combined national heat-illness alert systems with local greening and reflective-surface programmes, but the emergency transport figures for 2025 — more than 100,000 cases — show the scale of the residual burden [24]. Phoenix represents the opposite climatic extreme. In 2021 it created the first publicly funded municipal Office of Heat Response and Mitigation in the United States, and its Tree and Shade Master Plan aims to raise canopy cover from about 12% to roughly 25% by 2030 ✓ Established Fact [27]. The city also coated streets with reflective sealant in a cool-pavement pilot measured by Arizona State University [17].

Paris has approached the problem through schools. Its OASIS programme converts asphalt schoolyards into shaded, planted, permeable spaces, prioritising schools in heat-island districts that lack green space, and opens them to local residents as cool refuges during heatwaves ✓ Established Fact [26]. Ten pilot yards were followed by 60 more funded by the city, with the goal of transforming all 700 Paris schoolyards by 2050 [26]. Medellín, in Colombia, chose linear greening. Beginning in 2017, the city planted tens of thousands of trees and plants along 18 roads and 12 waterways, forming 30 green corridors in a 20-kilometre shaded network; city hall reports that the heat island has been reduced by 2°C since 2018, and the project won a 2019 Ashden award ◈ Strong Evidence [20].

Ahmedabad shows what can be done on a small budget in a very hot, very poor city. After a 2010 heatwave, it adopted India’s first Heat Action Plan, combining early warnings, hospital preparedness and public outreach. A peer-reviewed evaluation led by Jeremy Hess estimated about 1,190 avoided deaths per year in 2014-2015 compared with the 2007-2010 baseline ◈ Strong Evidence [21]. Alongside the plan, the Mahila Housing Trust has coated more than 20,000 slum roofs in the city with solar-reflective paint, reporting indoor temperature reductions of 5-6°C [22]. Sydney and London complete the set: Western Sydney recorded 48.9°C at Penrith on 4 January 2020, the highest temperature ever measured in the Sydney basin [37], while London’s 2018 summer has become the benchmark case for modelling urban cooling interventions [15][16].

The seven cities reveal three distinct strategies. Tokyo and Ahmedabad lead with health-system responses — alerts, hospital readiness and outreach — which reduce deaths without changing the physical heat ◈ Strong Evidence [21][24]. Phoenix, Medellín and Paris lead with physical modification of the city through canopy, shade and surface changes, which reduce heat itself but take years to mature [20][26][27]. Los Angeles and Western Sydney illustrate the regulatory route through building codes, which is cheap for government but vulnerable to industry opposition [19][36]. The evidence suggests these strategies are complements rather than alternatives: warning systems save lives this summer, while trees and roofs reduce the baseline for the next thirty summers.

The comparison also reveals how much depends on measurement. The cities with the strongest claims of success — Ahmedabad and Maricopa — are those that count deaths systematically and publish them ◈ Strong Evidence [11][21]. Medellín’s 2°C claim comes from the city government and has not been matched by the kind of peer-reviewed evaluation available for Ahmedabad’s heat plan, which is why this report treats it as strong but not established evidence [20]. Paris publishes programme counts but not temperature or health outcomes for its schoolyards [26]. A city that does not measure its heat deaths cannot know whether its interventions are working, and it cannot make the budgetary case for expanding them.

Finally, the comparison shows that success is reversible. Western Sydney is the clearest example: New South Wales planning minister Rob Stokes foreshadowed a ban on dark roofs for new developments in 2021, and his successor, Anthony Roberts, scrapped it in April 2022, arguing that it placed too great a cost and regulatory burden on builders ✓ Established Fact [36]. Only a local rule in the Wilton growth area, requiring light-coloured roofs and space for trees, survived [36]. The region that recorded the hottest temperature in the history of the Sydney basin thus lost the policy most directly aimed at the physics of its heat [37]. Heat policy advances through measured evidence and retreats through lobbying, and it does both at the level of individual regulations that rarely attract public attention.

Interventions differ in what they cool: surfaces, air, indoor spaces or human bodies. The most rigorous comparison to date is the 2024 study by Oscar Brousse and colleagues in Geophysical Research Letters, which modelled nine interventions across Greater London during two hot days in the summer of 2018. Cool roofs reduced near-surface air temperature by about 1.2°C, rooftop solar panels by about 0.5°C, street-level vegetation by about 0.3°C, and green roofs by roughly zero ✓ Established Fact [15]. A companion study led by Charles Simpson estimated that widespread cool roofs could have prevented 249 of the 655-920 heat deaths in London that summer, a 32% reduction worth an estimated 615 million pounds ◈ Strong Evidence [16]. For city-wide air temperature, albedo wins.

The Australian evidence points the same way. Research from the University of New South Wales estimated that if cool roofs were applied across Sydney, peak outdoor air temperature could fall by up to 2.5°C, residential indoor temperatures by up to 4°C, and cooling energy demand by up to 40% ◈ Strong Evidence [38]. At the household scale, the Ahmedabad experience shows that reflective paint on a tin roof can lower indoor temperatures by 5-6°C at a cost affordable to slum households, which is why the Mahila Housing Trust programme won a 2019 UN Global Climate Action Award [22]. The appeal of albedo lies in its combination of speed and cost: a roof can be coated in a day, whereas a street tree needs a decade or more to reach full canopy.

Trees nonetheless deliver benefits that roofs cannot. They shade pedestrians, bus stops and playgrounds; they cool through evaporation as well as shade; and they reduce the radiant heat that a human body absorbs, which matters more for heat stress than air temperature alone. The Lancet study of 93 European cities estimated that raising tree cover to 30% would lower average city temperatures by 0.4°C and prevent 39% of heat-island deaths ✓ Established Fact [1]. The 2024 global study found green space cools surfaces by about 3°C, with substantial untapped potential in Global South cities [29]. The London model’s smaller figure for street vegetation reflects its focus on citywide air temperature rather than on the microclimate a pedestrian actually experiences under a tree [15].

Cool pavements occupy an ambiguous middle ground, and the Phoenix pilot illustrates why. Arizona State University researchers found that reflective sealant made road surfaces 10.5-12°F cooler at noon and 2.4°F cooler at sunrise, and lowered night-time air temperature at 6 feet by about 0.5°F ◈ Strong Evidence [17]. But the same surfaces reflect sunlight back onto people. The ASU researcher Jennifer Vanos warned that a person walking on reflective pavement will feel hotter, and that such coatings should not be placed in playgrounds where children are active [17]. A cooler surface is not the same as a cooler body. The lesson is that albedo belongs primarily on roofs, where reflected energy escapes to the sky, and shade belongs primarily over the places where people walk and wait.

Design and governance measures complete the list. Heat action plans with early warnings and targeted outreach have the best-documented direct effect on mortality, as the Ahmedabad evaluation shows [21]. In Maricopa County, the 33% fall in deaths in 2025 followed an expansion of cooling centres and outreach to unhoused residents, and ASU sustainability professor Patricia Solis described the result as meaningful progress while stressing that the toll remained unacceptably high ◈ Strong Evidence [12]. Building codes — the Los Angeles cool-roof ordinance, the Wilton light-roof rule — lock in gains at the moment of construction, when they cost least [19][36]. Indoor measures such as insulation and repair of existing air conditioners address the place where most heat deaths actually occur [13][32].

The evidence therefore supports a layered approach rather than a single remedy. Reflective roofs offer the largest and fastest reduction in city-wide air temperature ◈ Strong Evidence [15][38]. Trees provide the shade and radiant cooling that make streets survivable, and reduce mortality at a measurable rate [1][29]. Heat action plans save lives quickly while the physical changes mature [21]. And the most direct intervention of all — keeping the indoor temperature of the homes of the old, the ill and the isolated below lethal levels — requires working cooling and insulation rather than urban design [13][32]. The following table sets out the principal risks to this programme of action as it is currently being implemented.

The first genuine trade-off is winter. The urban heat island that kills in summer also protects in winter, and in cold climates the protective effect can dominate. A 2023 study in Nature Communications by Wan Ting Katty Huang, Gabriele Manoli and colleagues valued heat-island mortality across 85 European cities and found an average summer cost of 192 euros per adult per year, set against an average winter saving of 314 euros from avoided cold deaths ⚖ Contested [30]. Net outcomes varied widely: Geneva had a net cost of 20.7 euros per adult, Trieste 184.4 euros, while Helsinki enjoyed a net saving of 113.9 euros [30]. Manoli stressed that heat islands are not just about people feeling too hot in summer, but about increased cardiovascular and respiratory risks [30].

This finding does not undermine the case for cooling cities, but it sharpens it. Summer-only interventions — deciduous trees that shade in summer and let in sunlight in winter, operable shading, cool roofs in hot climates — can reduce heat deaths without sacrificing the winter benefit ◈ Strong Evidence [30]. Blanket albedo increases in high-latitude cities, by contrast, could increase winter heating demand and cold exposure. The Lancet study’s strongest mortality benefits were in southern and eastern Europe, where summers are long and winters mild [1]. The honest conclusion is that heat-island policy must be climate-specific: the prescription for Seville is not the prescription for Helsinki, and the evidence supports aggressive cooling in hot cities rather than everywhere.

The second trade-off is displacement. Isabelle Anguelovski and colleagues, writing in PNAS in 2019, documented how urban greening and climate-adaptation projects can raise property values and push out the lower-income residents they were meant to benefit — a process described as green gentrification ⚖ Contested [34]. The irony is sharp: the neighbourhoods most in need of trees are redlined districts that have been hot for decades precisely because they were starved of investment [5]. Planting them without protecting tenants risks transferring the cooling benefit to newcomers. The competing positions are set out below.

Beneath these genuine debates lies a governance problem that is not contested at all. Urban heat touches housing, transport, parks, building control, public health, emergency services and energy utilities, and in most cities no single department owns it. Phoenix’s creation of a dedicated heat office in 2021 was notable precisely because it was the first of its kind in the United States ✓ Established Fact [27]. At the national level, the picture is worse. The US Occupational Safety and Health Administration proposed the first federal workplace heat standard on 30 August 2024, held public hearings from 16 June to 2 July 2025, and has since given no public indication that it will finalise the rule [35]. In New South Wales, the dark-roof ban was abandoned before it took effect [36].

The political economy explains why. The costs of urban heat fall on the old, the poor, the isolated, the unhoused and outdoor workers — groups with limited political influence ◈ Strong Evidence [11][32]. The costs of remedies fall on builders, landlords and municipal budgets, whose representatives are well organised, as the New South Wales case demonstrated [36]. Heat deaths are counted months later in statistical models rather than seen on the day, which weakens the public pressure that follows floods or fires [8]. And because the benefits of a tree accrue over decades while its cost is paid this year, canopy programmes are easy to defer. Phoenix remains near 12% canopy against a 25% target for 2030 [27].

The debate over trade-offs is therefore real but secondary. The winter benefit argues for climate-specific design, not inaction ◈ Strong Evidence [30]. The gentrification risk argues for pairing greening with tenant protection, not for leaving poor neighbourhoods hot [34]. The pedestrian penalty of reflective pavement argues for putting albedo on roofs and shade over streets, not against albedo as such [17]. None of these debates explains why the hottest neighbourhoods in American cities are still, on average, 2.6°C hotter than their redlining-era neighbours nearly a century after the maps were drawn [5]. That is explained by the absence of sustained investment, and investment is a policy choice.

The first lesson is that the heat island is a design outcome. Dark roofs, wide asphalt, missing trees, street canyons and waste heat from cooling systems are the result of decisions about materials and land use, each documented and each modifiable ✓ Established Fact [2][14][18]. Cities do not have to wait for global emissions to fall to reduce the heat they themselves add, because the IPCC finds that urbanisation is a local rather than a global warming driver [31]. The Lancet estimate that more than 4% of summer deaths in European cities are attributable to heat islands places a number on the cost of that design [1]. It is a number that falls within the control of municipal governments.

The second lesson is that the burden is unequal and predictable. Formerly redlined neighbourhoods are 2.6°C hotter; the poorest neighbourhoods have 41% less canopy; poorer districts are hotter in 72% of cities studied worldwide; and the Global South receives only 70% of the green cooling available in the North ✓ Established Fact [5][6][7][29]. Within those neighbourhoods, heat kills the old, the isolated and the unhoused, mostly indoors or on the street, as the records from Chicago, British Columbia and Maricopa show [11][32][33]. Any programme that spreads resources evenly across a city will therefore underserve the people at greatest risk.

The third lesson concerns which interventions to prioritise. For lowering city-wide air temperature, reflective roofs have the strongest modelled effect, at about 1.2°C in London and up to 2.5°C in Sydney ◈ Strong Evidence [15][38]. For protecting people outdoors, trees provide shade and radiant cooling that surfaces cannot, and they carry the strongest mortality evidence across European cities [1]. For saving lives this summer, heat action plans with early warning and outreach have the most direct evidence [21]. For the indoor deaths that dominate the record, working cooling, insulation and the repair of broken air conditioners matter more than anything visible from a satellite [13][32].

The fourth lesson is that heat policy must be climate-specific and equity-protected. In cold-winter cities, heat islands can save more lives in winter than they cost in summer, which argues for seasonal measures such as deciduous shade rather than blanket albedo ⚖ Contested [30]. In hot cities, the case for aggressive cooling is overwhelming. Everywhere, greening should be paired with tenant protections to avoid displacing the residents it is meant to protect [34], and reflective materials should go on roofs rather than on the pavements where people walk [17]. These are refinements, not reasons for delay.

The fifth and final lesson is that measurement drives action. Where heat deaths are counted and published, as in Maricopa County, the numbers create pressure, and in 2025 that pressure coincided with a 33% fall in deaths ◈ Strong Evidence [12]. Where they are not, heat remains an invisible hazard, estimated in models months later [8]. With urban exposure to extreme heat having tripled since 1983 and still rising [28], the cost of inaction grows every summer. The urban heat island is a policy failure in the precise sense: a harm whose causes, victims and remedies are all known, which continues because no one has been made responsible for ending it. The first step is to assign that responsibility, give it a budget, and count every death that follows.