The more you cool your living room, the more you heat up your street. This isn’t a metaphor: it’s atmospheric physics, measured with coupled energy and urban climate models. A team from Arizona State University simulated ten summer days in Phoenix with and without the waste heat expelled by air conditioning units, and found that this waste heat raises nighttime air temperature by up to 1°C in some parts of the city [1]. During the day the effect is negligible because the atmosphere disperses it easily; at night, with the atmospheric boundary layer more compressed, that heat lingers between buildings, warming the air right when your body most needs to cool down to sleep. Part of the solution to urban heat is actively generating more urban heat.
But air conditioning is just the entry point to a much bigger problem, with two distinct fronts: the heat that’s already killing people in Europe this very summer, and the water that’s already erasing entire countries from the map. Let’s look at both, with updated 2025 and 2026 data, without limiting ourselves to the United States again.
The island that manufactures its own heat
The underlying phenomenon is 200 years old, and it was discovered, almost by accident, by the man who named the clouds. Luke Howard, a London pharmacist and father of cumulus and cirrus nomenclature, compared thermometers inside and outside London in the early 19th century and found the city was consistently warmer than the surrounding countryside, with the difference most pronounced at night [2]. It’s the first documented description of the urban heat island.
The mechanism combines several mutually reinforcing effects [3]: asphalt and concrete absorb solar radiation during the day and release it slowly at night; buildings block the wind that would disperse that heat; vegetation, which would cool the air through evaporation, has been replaced by impermeable surfaces; and waste heat from cars, industry, and air conditioners adds to the mix. The temperature difference between a city and its rural surroundings can exceed 5°C on average and reach peaks of 7°C on clear nights, according to the synthesis from the U.S. Environmental Protection Agency [4].
Europe can no longer pretend this is someone else’s problem
Here’s the piece of data that was missing, and it shifts the center of gravity of this whole issue: Europe is, per capita, the continent with the most heat-related deaths on the planet, and the trend is getting worse, not better. The summer of 2022 caused 61,672 heat-attributable deaths across 35 European countries, with Italy, Spain, and Germany leading in absolute numbers [5]. The 2026 European Lancet Countdown report on climate and health puts the 2024 figure at around 62,000 heat-related deaths, and confirms that 99.6% of monitored European regions have seen heat-related mortality rise over the last decade [6]. Official extreme heat warnings have more than quadrupled since the 1990s [6].
And 2025 and 2026 haven’t been exceptions — they’ve been confirmation. Preliminary estimates put excess mortality from European heatwaves at around 16,500 people in 2025, and roughly 20,390 in 2026, though these numbers are preliminary and subject to revision as the full epidemiological analysis progresses [7]. In other words: you don’t need to go to Phoenix to find the problem. It’s in Madrid, in Rome, in Berlin, right now.
That said, Phoenix remains the most extreme laboratory for studying solutions, precisely because it has spent more years treating heat as a managed emergency. In 2024 it strung together 113 consecutive days above 37.8°C, and Maricopa County confirmed 608 heat-related deaths; in 2025 that figure dropped to 430, the second consecutive annual decline [8]. The county itself acknowledges there are several factors behind the improvement and that the causal link to municipal measures hasn’t been statistically isolated, but the timing lines up with when the city created, in 2021, the first Office of Heat Response and Mitigation in the United States, with public funding and a career climatologist at its head [9].
Shade isn’t distributed randomly — it’s distributed by income, and this has been measured across different countries, not just the US. An analysis of 5,723 US municipalities published in PLOS ONE found that in 92% of the urbanized areas studied, low-income neighborhoods have less tree cover than high-income ones — 15.2% less on average — and are 1.5°C hotter in summer; in some northeastern US cities the gap reaches 30% less tree cover and 4°C more heat [10]. It has a name in the literature — the “luxury effect” — because urban tree cover behaves almost like a premium consumer good. Shade, which should be the cheapest and most democratic adaptation resource there is, also has a zip code.
Rotterdam: less wall, more room for water
On the night of January 31, 1953, a storm surge broke the dikes in the southwest of the Netherlands and killed more than 1,800 people [11]. The response was the Delta Plan, and its most spectacular piece is the Maeslantkering: a barrier with two floating arms — each as long as the Eiffel Tower is tall — that closes itself when a computer system detects a risk of marine flooding [11]. Since 1953, the country hasn’t suffered another deadly coastal flood of comparable magnitude.
But the interesting shift over the last two decades isn’t more wall — it’s less wall and more room. The Room for the River program, launched after the near-catastrophic river floods of the 1990s, lowered some dikes and widened river channels, returning floodplains that had been reclaimed a century earlier [12]. A wider river needs less dike height to move the same amount of water during a flood surge. At the city scale, that same spirit shows up in water squares like Benthemplein, opened in 2013: on dry days it’s a sports court, and during heavy storms it becomes a reservoir capable of holding 1.7 million liters of water [13].
Venice: the barrier that’s already showing its expiration date
MOSE, the system of 78 steel gates protecting the Venice lagoon, cost €6 billion and became operational in 2020. Between 2020 and 2025 it closed 108 times to prevent flooding. In the first two months of 2026 alone it had already been activated 30 times [14]. That acceleration isn’t a system failure — it’s exactly what physics predicts: as the sea rises and the city itself keeps sinking due to subsidence, the same barrier that works today will need to close more and more often and for longer stretches, to the point of potentially having to stay shut for weeks each year, blocking the port, disrupting the lagoon ecosystem, and requiring permanent pumping systems just to maintain the interior water level [14].
A study published in Scientific Reports in 2026 modeled four possible futures for Venice — keeping the lagoon open with more defenses, walling off the historic center, sealing the lagoon entirely, or retreating and relocating selected monuments inland — and its conclusion is uncomfortable but clear: no strategy preserves the city as we know it today indefinitely [14]. Costs range from €500 million for additional dikes to more than €30 billion to seal the lagoon, and up to €100 billion for a partial relocation [14].
And here’s the data point that answers the question of whether sea-level rise is “very long-term”: the researchers themselves calculate that, even under the most favorable emissions scenario, the major decisions and the start of construction need to happen within the next few decades [14]. This isn’t a threat for the next century. It’s a decision Venice is already discussing right now.
How much time do we really have?
This question only has an estimated answer. The IPCC projects a global average sea-level rise of between 0.28 and 1.01 meters by 2100 depending on the emissions scenario, with a low-confidence range that doesn’t rule out approaching 2 meters if the destabilization processes of Antarctica’s ice sheets turn out to be faster than expected [16]. The real source of uncertainty lies in Greenland and, above all, Antarctica: the IPCC itself speaks of “deep uncertainty” in the dynamics of those ice sheets, because they involve collapse mechanisms that still aren’t well modeled [16]. A 2026 Oxford University study adds another uncomfortable nuance: even if the world managed to stabilize global temperature tomorrow, sea-level rise and permafrost thaw would keep going for centuries, because these are processes with an enormous time lag relative to the temperature that drives them [18]. This means that much of the sea-level rise expected between now and 2100 and beyond is, in a sense, already locked in by past emissions.
For most of the world’s coastal cities — Rotterdam, Venice, New York, Barcelona — it’s a decades-long process, manageable with the kind of infrastructure planning we’ve already seen, though with construction timelines that require decisions now, not fifty years from now. But for the low-lying Pacific island nations, the question already has an answer, and it isn’t reassuring. Tuvalu, with an average elevation of barely two meters and a highest point of 4.5 meters, has seen sea levels in its region rise 50% faster than the global average over the last three decades [17]. The Tuvaluan government estimates that by 2050 half the capital will experience regular tidal flooding, and that between 90 and 95% of the country could be permanently submerged by 2100 [17]. This is no longer an abstract projection: in 2023 Tuvalu signed the Falepili Union treaty with Australia, which recognizes the country’s sovereignty even if its physical territory disappears underwater, and offers 280 climate visas per year; in the first round, more than a third of the country’s population applied [17]. In parallel, Tuvalu is building a complete digital replica of the country to preserve its identity and government even if the physical land ceases to exist [17]. In other words: for the lowest-lying islands on the planet, this isn’t long-term — it’s a process already underway on a one-generation timeline; for other coastal cities, it’s a decades-long race against engineering deadlines that are also measured in decades, which leaves much less margin than it appears at first glance.
Singapore: the city that invented its own water
When Singapore became independent in 1965, it had no major rivers, barely any aquifers, and depended on water-supply agreements with Malaysia that had expiration dates. The response, managed by the state agency PUB, is the “four national taps”: rainwater captured in a reservoir network covering two-thirds of the island, imported water, desalinated water, and NEWater — wastewater treated with microfiltration, reverse osmosis, and ultraviolet disinfection until it exceeds WHO drinking-water standards [20]. Today it covers around 40% of national demand, with a target of reaching 55% by 2060 [20]. A significant share of the water that five and a half million people drink passed through a toilet at some point; chemistry doesn’t care where it came from.
On heat, Singapore plays the same obsession with measuring before building. The Cooling Singapore consortium, with participation from ETH Zurich, is developing a digital climate twin of the city to simulate the effect of changing materials or vegetation before spending a single dollar on actual construction [21], and the LUSH regulations require developers to replace, within the building itself, the greenery that construction removes at ground level [22]. Hence that aesthetic of skyscrapers covered in vegetation, which is, at bottom, fairly boring green-square-meter accounting.
What the evidence actually supports, without embellishment
The urban heat island kills tens of thousands of people every summer in Europe, not just in American desert climates, and the trend over the last decade is getting worse, not better: this is about as well established as anything in urban climatology can be [2][3][4][5][6]. The solutions that work are, for the most part, unspectacular: trees, shade, room for water, water recycling, well-maintained barriers; Rotterdam and Singapore have spent decades proving it [11][12][13][20]. But sea-level rise introduces something heat doesn’t have: engineering deadlines that are already running. Venice needs to decide its future within this decade to have any margin to build in time, and Tuvalu is no longer deciding whether to adapt but how to preserve its identity without its territory [14][17].
The honest conclusion is that there’s no single deadline for the whole planet: there’s a different timeline for each city, and some of those timelines, contrary to what you’d intuitively expect, have already started running. Countries with strong economies will be able to build protective measures with time to spare, but the most affected — Pacific islands and densely populated deltas — will practically cease to exist, and paradoxically, these are the countries that have contributed the least to emissions. That said, apocalyptic or alarmist scenarios should currently be ruled out, at least regarding temperature rise and sea level, since there’s still time to get this under control if things are done right.
REFERENCES
[1] Salamanca, F., Georgescu, M., Mahalov, A., Moustaoui, M. and Wang, M. (2014). Anthropogenic heating of the urban environment due to air conditioning. Journal of Geophysical Research: Atmospheres, 119(10). — Reliable
[2] Howard, L. (1833). The Climate of London, Deduced from Meteorological Observations. Harvey and Darton, London. — Historical/observational source
[3] Oke, T. R. (1982). The energetic basis of the urban heat island. Quarterly Journal of the Royal Meteorological Society, 108(455), 1–24. — Reliable
[4] United States Environmental Protection Agency (EPA). Heat Island Effect: research synthesis on urban temperature differentials. — Official statistics
[5] Ballester, J., et al. (2023). Heat-related mortality in Europe during the summer of 2022. Nature Medicine, 29. — Reliable
[6] Lancet Countdown Europe (2026). 2026 European report on health and climate change. — Reliable
[7] Preliminary excess-mortality estimates for the 2025 and 2026 European heatwaves (Indiana University; Poznań University of Medical Sciences), as compiled on Wikipedia, “2026 European heatwaves.” — Under review
[8] Maricopa County Department of Public Health (2024–2026). Heat-Related Deaths Reports, 2024 and 2025. — Official statistics
[9] City of Phoenix (2021). Office of Heat Response and Mitigation. — Institutional document
[10] McDonald, R. I., et al. (2021). The tree cover and temperature disparity in US urbanized areas. PLOS ONE, 16(4). — Reliable
[11] Rijkswaterstaat / Dutch Delta Programme. Technical documentation on the 1953 flood and the Maeslantkering barrier. — Institutional document
[12] Government of the Netherlands. Room for the River Programme (Ruimte voor de Rivier). — Institutional document
[13] De Urbanisten / Municipality of Rotterdam (2013). Water Square Benthemplein. — Institutional document
[14] Lionello, P., Giupponi, C., Di Fant, V., Pasquier, U., Nicholls, R. J. and Vafeidis, A. T. (2026). Long-term adaptation pathways for Venice and its lagoon under sea-level rise. Scientific Reports, 16, 9438. — Reliable
[15] (combined with reference 14 in the final text)
[16] IPCC (2021). Sixth Assessment Report (AR6), Working Group I, Chapter 9: Ocean, Cryosphere and Sea Level Change. — Reliable
[17] NASA / Government of Tuvalu / Government of Australia. Sea-level rise data for Tuvalu and the Falepili Union treaty (2023). — Official statistics / institutional document
[18] Environmental Change Institute, University of Oxford (2026). Study on the long-term irreversibility of sea-level rise and permafrost thaw after temperature stabilization. — Reliable
[19] IPCC (2021). AR6, assessment on the likelihood of abrupt methane release from clathrates and methane hydrates during the 21st century. — Reliable
[20] PUB, Singapore’s National Water Agency. NEWater programme and “Four National Taps” strategy. — Official statistics
[21] Cooling Singapore / Singapore-ETH Centre. Urban digital climate twin. — Institutional document
[22] Urban Redevelopment Authority, Singapore. LUSH programme. — Institutional document
[23] White, G. F. (1945). Human Adjustment to Floods. University of Chicago; verified in Di Baldassarre, G., et al. (2015), Water Resources Research. — Reliable
Note on nuanced sources: reference [7] is labeled “Under review” because these are preliminary excess-mortality estimates, explicitly flagged as subject to change as the full epidemiological analysis progresses; they’re used here only as a trend indicator, not as a final figure. Reference [8] is official statistics, not peer-reviewed: the criteria for classifying a “heat-related death” have varied slightly between years. The cost and timeline figures for Venice’s adaptation strategies in reference [14] are projections from a single recent modeling study, not executed figures; they’re presented as such in the text.