The Corals Moved North. The Heat Followed.

During the summer and fall of 2024, sea-surface temperatures along Japan's coast ran 1 to 3°C above previous summer averages from June into November. Corals bleached from Okinawa to Honshu. The subtropical bleaching was expected. What happened at the temperate sites was not.

At Tatsukushi, a long-studied marine park at nearly 34°N, more than 40 percent of the surveyed coral was already dead by the time researchers got in the water. Another 20 percent was bleached, 26 percent visibly pale. The species hit hardest was Acropora hyacinthus, a fast-growing branching coral that does not appear in Tatsukushi's records from the 1930s. It arrived from the south over the intervening decades, established itself, and came to account for roughly 80 percent of the site's coral cover. At Numazu, on the northern Izu Peninsula at 35°N, the same species complex was completely bleached.

These corals had done exactly what a generation of coral science predicted they would do. They moved poleward into warming water, colonized new habitat, built populations. Then the water they had moved into got too hot.

The findings, published in Scientific Reports by Haruko Kurihara of the University of the Ryukyus and colleagues at four Japanese universities, document the most severe heat-driven bleaching yet reported from Japan's temperate coral communities. The refuge idea they test has been in the literature for more than a decade: warming oceans would push viable reef ecosystems toward higher latitudes, and corals migrating into those waters could establish new populations as their tropical habitats became lethally hot. The 2024 data from Japan show the destination warming alongside the source.

Two papers, fifteen years apart

Hiroya Yamano, now a professor of Earth and Planetary Science at the University of Tokyo, has spent his career inside that assumption. He helped build it. He was in the water in 2024 while it was being tested.

Yamano came into coral research through Okinawa's reefs. The 1998 global mass bleaching event let him connect field knowledge to satellite data: because he already knew where the corals were, he could point the remote sensing at those specific locations and detect bleaching from orbit. In an interview with Japan's Climate Change Adaptation Platform, he said that change in the sea is "faster and more pronounced" than people generally recognize, and that he wanted to make it visible.

In 2011, Yamano and two colleagues published a study in Geophysical Research Letters reconstructing Japanese coral distributions from roughly eighty years of records. Four of nine coral categories had expanded poleward since the 1930s, including the reef-forming A. hyacinthus and A. muricata, at rates estimated as high as 14 kilometers per year. The paper concluded, carefully, that "temperate areas may serve as refugia" for tropical corals. It also noted that fast-growing newcomers could displace the endemic temperate species already living there. The refuge language entered the literature and stayed.

Fifteen years later, Yamano is a coauthor on the Kurihara paper. He took part in the 2024 fieldwork and helped analyze the coral data. Two of the species his earlier study had flagged as successful poleward migrants, A. hyacinthus and A. muricata, were among those most severely bleached and killed. The fieldwork behind the 2011 expansion study and the fieldwork behind the 2024 bleaching study overlap at some of the same locations. The published record contains no statement from Yamano about what that was like to look at. What changed between the two sets of dives was the water.

How the heat adds up

A single hot afternoon does not bleach a reef. Thermal stress accumulates over weeks and months, and the metric coral scientists use to track it, Degree Heating Weeks, behaves less like a thermometer reading than like a sunburn: total exposure summed over time.

Each site has a baseline, the average temperature of its hottest month across decades of satellite records. When the water climbs at least 1°C above that baseline, the excess starts accumulating. A DHW of 4 means roughly two weeks at 2°C over the threshold, or four weeks at 1°C over it, enough to produce significant bleaching. A DHW of 8 is associated with severe bleaching and death.

In 2024, all eight of the study's sites, subtropical and temperate alike, exceeded 8 DHW. Three temperate sites reached values that would be extraordinary anywhere on earth. Tsushima, at 34.4°N, hit 19.1. Tagojima, at 34.8°N, reached 18.7. Numazu, at 35°N, recorded 17.2. In the 42 years of satellite record behind the study, no site had exceeded 10 before 2022.

Water temperatures at those temperate locations peaked between 29 and 30°C, 84 to 86°F, at latitudes comparable to Los Angeles or Casablanca. Nobody would call a swimming pool at that temperature refreshing. The Japan Meteorological Agency found that the annual sea-surface temperature around Japan in 2024 ran 1.44°C above the 1991–2020 mean, the highest in a record that begins in 1908.

The finding that matters most for the refuge hypothesis is not any single year's heat but the rate of change. Summer maximum temperatures rose significantly at every site over 1982–2024, and the warming rate increased with latitude. Numazu and Tateyama warmed at 0.34 and 0.39°C per decade. Ishigaki and Okinawa, the subtropical sites the corals had been leaving, warmed at 0.18 and 0.19°C per decade. The northern destinations are heating roughly twice as fast as the southern sources.

What a bleaching reef looks like

A coral colony is an animal, thousands of tiny polyps, living in metabolic partnership with photosynthetic algae embedded in its tissue. The algae supply most of the coral's energy and all of its color. When the water stays too hot for too long, the biochemistry of that partnership breaks down and the coral expels or loses its algae. What remains is a film of clear animal tissue over white calcium carbonate. You can see the skeleton through the living part.

A bleached coral is not dead. If temperatures fall soon enough, the algae recolonize and the animal recovers. But it is starving in the meantime, having lost its main food supply, and if the heat holds it dies. At Tatsukushi the survey caught the transition mid-course: 41 percent dead, 20 percent bleached, 26 percent pale. At Kushimoto, another temperate site at 33.5°N on the Kii Peninsula, the same A. hyacinthus complex bleached and paled without dying, and later observations suggested recovery.

The difference appears to have been weather. A typhoon passing through September 18 to 21 may have mixed and cooled Kushimoto's water enough to matter: eleven fewer days above the critical heat threshold than the sites that lost their coral. In the American calendar, summer ends in the first week of September, somewhere around the Labor Day cookout. The water around Japan had another six weeks of accumulation left in it, and the storm that saved one reef arrived after the season was nominally over.

Lord Howe, the Abrolhos, and the southern cases

Japan is the sharpest recent case, but high-latitude coral sites in the Southern Hemisphere have been running the same experiment for over a decade.

Lord Howe Island, roughly 600 kilometers northeast of Sydney at 31.5°S, holds the world's southernmost true coral reef. Its isolation and its mixed tropical-temperate community had attracted refuge language in the literature. In 2019, lagoon temperatures went over the local baseline, bleaching reached 83 percent at the worst-affected site, and mortality hit 40 percent in nearshore areas. Lord Howe had already bleached in 1998, 2010 and 2011, with as much as 99 percent of lagoon coral affected in 2010. Four events in two decades is a record of repeated failure to protect anything.

Along eastern Australia's subtropical coast, between 26°S and 31°S, the 2016 global bleaching event struck communities containing both tropical species arriving from the Great Barrier Reef and subtropical specialists rare in the tropics. Pocillopora aliciae, a regional endemic, bleached severely and died in quantity. Bleaching severity tracked which species were present more closely than it tracked environmental variables. What concerned the researchers was the composition of the losses: the high-latitude endemics, the corals that made these communities different from the reefs to their north, were the ones going.

At the Houtman Abrolhos Islands off Western Australia, near 29°S, the southernmost coral-reef system in the Indian Ocean, the first recorded widespread bleaching came in 2011. One monitoring site lost 99.85 percent of its coral cover. A diver there afterward would have found the reef's architecture still standing, branching skeletons intact and stripped of living tissue, already filming over with algae. Before 2011, bleaching at any Abrolhos monitoring site had stayed below 1 percent. A later study found conspicuously high thermal resistance among several common Abrolhos species during a subsequent heatwave, and its authors still concluded that the historical record shows the islands are neither permanently protected from bleaching nor indefinitely resilient.

In each case, high-latitude coral that looked stable or expanding met a marine heatwave large enough to erase whatever advantage its latitude was supposed to confer.

Where the science disagrees, and why

The literature is not really split over whether poleward expansion happens. It happens; it has been observed, measured, mapped. The disagreement is about what it can accomplish, for which species, and on what timescale, and much of it comes down to the fact that different lines of evidence measure different intervals of time.

Yamano's 2011 work showed that several Japanese corals could expand quickly, establish adult colonies, and reproduce in temperate water. A 2021 review by Abrego, Howells, Baird and colleagues worked through the proposed brakes on high-latitude establishment: winter cold, low light, unfavorable carbonate chemistry, the limits of larval dispersal, competition with kelp and other macroalgae. Experimental support for each brake varied enormously by species, which leaves room for some high-latitude establishment by broadly tolerant corals in favorable local conditions.

The fossil record takes a longer and less comforting view. Dimitrijević, Santodomingo and Kiessling found that reefs shifted poleward after four ancient hyperthermal crises, in some cases by nearly 10 degrees of latitude. Those new reefs appeared thousands to millions of years after the warming began, often after the acute stress had passed. Deep time confirms that corals relocate. It says nothing about whether they can do it inside a century.

Vogt-Vincent, Pringle, Cornwall and McManus modeled coral demography, larval dispersal, adaptation, temperature, light and pH across connected metacommunities. Their simulations produced severe losses of coral cover within 40 to 80 years and required centuries for large-scale occupation of new high-latitude habitat. Under their middle-emissions scenario, the climatic conditions corals need and the places corals actually are came apart by about 1.5°C during the second half of this century.

Huang and colleagues added a constraint that has nothing to do with heat. Modeling the Pocillopora damicornis complex around Taiwan under future conditions, they found that warming pushes the suitable temperature band poleward while the chemistry corals need moves the other way. Seawater carries dissolved mineral that corals cement into skeleton, and as the ocean absorbs more carbon dioxide that supply thins, worst at higher latitudes in colder water. A coral that finds water warm enough at 35°N may arrive somewhere it cannot build.

The open questions are specific. Can scattered colonies surviving at a new latitude reproduce the three-dimensional structure, the biodiversity and the coastal wave protection that the original reef provided? Can migration keep pace with warming this century, or only across geological time? Does temperature define the habitat, or do light, chemistry, substrate, competition and food each set limits that latitude cannot solve?

McClanahan, Darling, Beger and colleagues have argued that the word "refuge" needs breaking apart: sites that avoid heat are not the same as sites that resist it or recover quickly afterward, and the useful unit of analysis is a portfolio of sites rather than any single location. Under that framing a place can fail as a heat-avoidance refuge while still holding resistant species or recovery potential, which changes both what conservation looks for and what it spends money defending.

Joshua trees and bull trout

The refuge problem is not confined to reefs.

Climate modeling had identified Cima Dome, in California's Mojave National Preserve, as a possible future refuge for eastern Joshua trees, with favorable temperature and moisture projections running decades out. In August 2020 the Dome Fire burned nearly 44,000 acres and an estimated 1.3 million Joshua trees. The fire perimeter almost entirely overlapped the modeled refuge. The model had temperature and moisture in it. It did not have the invasive grasses that carried fire across ground that used to be too bare to burn.

In Idaho's Boise River basin, the cold, harsh, isolated mountain headwaters have long been treated as climate refuges for bull trout, a fish that needs water cold enough to exclude most of its competitors. Daniel Isaak and colleagues measured a 0.38°C increase in summer stream temperature over thirteen years and modeled a loss of 11 to 20 percent of the stream length cold enough for bull trout spawning and juvenile survival. The largest losses came in the coldest habitat, the innermost core of the refuge. It is shrinking from the center out.

A refuge is a prediction about conditions at a particular place. When the forcing is global, the place sits inside the forcing along with everything else, and the prediction holds only as long as the destination changes more slowly than the thing being escaped.

What the study establishes

One site in the Japanese study, Tateyama, did not bleach despite accumulating DHW well above the level associated with severe mortality, a result covered separately that complicates the standard metric and shows how much local conditions can bend outcomes inside a single heatwave.

The Kurihara paper does not establish that every high-latitude site will fail. Responses varied sharply by site and by species. Kushimoto's corals appear to have recovered. Lithophyllon undulatum at Tsushima and Leptastrea at Tagojima stayed healthy while colonies beside them died. What the study establishes is narrower and harder: in these waters, during this event, for these populations, the premise that the destination would stay thermally stable while the source degraded was false.

Yamano published the expansion study in a period when documenting northward coral movement still read as documenting a possible future. The species he tracked have since arrived, established, come to dominate their new sites, and bleached. Larvae are presumably still moving north from the subtropics, into water that is now warming at twice the rate of the water they left. Whatever grows back at Tatsukushi will be whatever the last two summers permitted, and the next survey will be a count of that.