When Does Heat Become Forest Failure?

On the hottest afternoons, a satellite sees patches of tropical canopy cross 46.7 degrees Celsius, the temperature at which heat begins to destabilize a protein complex essential to photosynthesis. A 2026 study estimates that this now happens across 57 million hectares, an area larger than France. Two decades earlier, the same calculation yielded 43 million hectares.

The estimate brings together measurements made at two very different scales. In the laboratory, researchers established the benchmark by heating individual leaves. In orbit, a thermal sensor averages thousands of leaves in sun and shade into one temperature for each patch of canopy. The study treats those measurements as comparable, counting the patches whose average crosses the laboratory benchmark. But can a result from one heated leaf tell us when an entire forest is in trouble? And if more canopy is crossing that benchmark, how close does that bring us to saying a forest is beginning to fail? The answers are not in the estimate itself. They are in the studies it was assembled from.

Three temperatures, three outcomes

The studies behind these satellite maps use 46.7 degrees Celsius because that is where photosystem II, the part of the chloroplast that splits water and begins the chemistry of photosynthesis, starts to destabilize. A survey of 147 tropical tree species produced that average. A broader survey of 218 species found that the critical temperature varies with latitude, from the low 40s at high latitudes to just over 50 degrees in the lowland tropics.

In 2023, Christopher Doughty and colleagues compared that leaf threshold with canopy temperatures measured by ECOSTRESS, a thermal sensor on the International Space Station. Across the Amazon, Central Africa and Southeast Asia, midday canopy peaks in dry periods averaged about 34 degrees, with a high tail above 40. Thermocouples clipped to upper-canopy leaves showed that those leaves already spend 0.01 percent of their time above 46.7 degrees. The team then warmed upper-canopy leaves by 2, 3 and 4 degrees at sites in Brazil, Puerto Rico and Australia, and the hottest leaves did not warm in step with the air: their time above 46.7 degrees rose from 0.01 to 1.3 percent, a hundredfold for a few degrees. Carried forward in a model, that relationship puts a potential tipping point in metabolic function at 3.9 degrees Celsius, plus or minus half a degree, of further warming in the air over tropical forest, counted from the temperatures those forests live in now rather than from the preindustrial baseline used in climate targets. That is the warming that would hold the hottest leaves above 46.7 degrees often enough to break the forest's metabolism. A rise of that size over the tropics sits inside the worst-case emissions pathway of the climate projections, RCP 8.5, which is why the authors framed their result as a choice still open.

Nina van Tiel and colleagues extended the comparison in 2026. Using a longer satellite record, they mapped how far canopy temperature stayed below the critical threshold for 208 tree species across South America, Southeast Asia and Central Africa from 2001 to 2020. The area where canopy temperatures exceeded the average threshold grew from 43 to 57 million hectares. Van Tiel and colleagues project that it will reach 83 million hectares by 2050 and 160 million by the end of the century.

A second threshold sits far lower, where what suffers is carbon uptake rather than the cells themselves. At Doughty's tropical study site, carbon uptake began to fall when leaves reached about 35 degrees. Across the tropical forests examined by Pau and colleagues in 2018, gross primary productivity, the total carbon fixed through photosynthesis, climbed with canopy temperature up to about 28 degrees; above that it kept climbing, but more slowly. German Vargas and Roberto Cordero measured how fast two Costa Rican lowland rainforest species fixed carbon at different leaf temperatures in 2013, and in one of them the rate began to fall at 37 degrees.

Long-term forest plots reveal a third threshold, based on stored carbon. Across the tropical plots analyzed by Martin Sullivan and colleagues in 2020, stored carbon remained stable until the mean daily maximum temperature of the warmest month reached about 32 degrees. Above that temperature, carbon stocks declined. Standing carbon is a balance: a stand takes carbon in as new wood and gives it back as trees die and decay. In the hotter plots the new wood came on more slowly while trees went on dying at the usual rate, which left less wood standing at any given moment. In their model, two degrees of global warming would expose three-quarters of tropical forest to temperatures above that long-term carbon threshold.

The mark the maps use is also the one most often read as a death line. As noted above, 46.7 degrees is where photosystem II begins to destabilize. Laboratory tests catch that moment by watching a leaf glow. Chlorophyll always re-emits a little of the light it absorbs but cannot push through photosynthesis, and as photosystem II destabilizes, more of the absorbed light comes back out as that faint glow. The jump in fluorescence is what marks the critical temperature in the assay. That mark was long read as a lethal limit. Klaus Winter argued in 2024 that reading it that way underestimates the heat tolerance of tropical leaves. In a 2025 experiment on a tropical tree in the genus Calophyllum, his team heated leaves to the critical temperature and two degrees beyond it. After fourteen days, the leaves had recovered with almost no dead tissue. Across broader datasets, severe irreversible leaf damage generally begins several degrees above the critical temperature.

Leaves do adjust when they grow in warmer conditions. Across the 218-species dataset, the critical temperature rose 0.3 degrees for every 1-degree warmer growth environment, climbing from about 41.5 degrees at an Alaskan tundra site to 50.8 degrees in lowland Amazon rainforest, the highest site average the survey recorded. Seasonal measurements at two sites found adjustment of the same limited size. Leaf tolerance therefore climbs at roughly a third the rate of the temperature it has to withstand.

These temperatures describe different outcomes: forests losing stored carbon, individual trees losing photosynthetic capacity and leaves showing acute cellular strain. The maps by Doughty and van Tiel compare canopy temperatures with the threshold associated with the last outcome, acute strain in leaves, which is the outcome an orbiting sensor is least equipped to see.

What the satellite can see

Doughty's team and van Tiel's both needed satellite observations to map heat across millions of hectares. The thermal sensors mentioned earlier read the infrared radiation leaving the top of the canopy and convert it into land-surface temperature, the average radiometric temperature of the surface inside a pixel. Air temperature and individual-leaf temperature are separate measurements.

Doughty used ECOSTRESS because its pixels cover 38 by 70 meters and the International Space Station carries it over forests at different hours of the day. Van Tiel needed a twenty-year record, so her team used MODIS. Its one-kilometer pixels are much coarser, but the long MODIS record spans the study's 2001-to-2020 window.

Cloud-contaminated observations are masked from both products. Tropical and subtropical forests are among the cloudiest non-polar biomes, so cloudy conditions are underrepresented in the resulting maps. The clear-sky observations do not represent a random sample of the conditions a wet forest experiences.

A pixel also combines sunlit crowns, shaded gaps and cooler leaves beneath the canopy surface. The hottest leaves occupy only a small part of that area, so their temperature disappears into the average.

In 2024, Olivier Manzi and colleagues measured individual leaf temperatures alongside air temperature and the canopy average. Sun-exposed tropical leaves were 8 to 10 degrees hotter than the air, and leaves on the most exposed crowns were up to 20 degrees hotter than the air. Some approached 50 degrees at midday while the canopy average remained well below 50 degrees.

Sunlight heats a leaf, while evaporation and moving air carry heat away. A leaf's size and the readiness with which it opens its stomata (the pores that release water vapor) help determine its temperature. A satellite pixel is too coarse to resolve those differences among leaves.

The satellite average therefore conceals some sunlit leaves that cross 46.7 degrees. But measuring only the hottest leaf would misrepresent the canopy, because one exposed leaf cannot reveal the photosynthetic response of all its leaves. Repeated satellite passes can show the trend, a growing area of canopy crossing the threshold, while the count of irreversibly damaged leaves stays out of reach. Whether a hot leaf becomes a dead leaf depends on something no pixel records, the amount of water the tree still has to spend on cooling.

How water protects a tree from heat

A leaf sheds heat by evaporating water. Water drawn from the roots leaves through the stomata and carries heat with it. In dry air, water vapor leaves the leaf faster, so the tree must draw more water through its roots. In 2020, Marielle Smith, Scott Saleska and colleagues found that declining tropical photosynthesis on hot days tracked air dryness more closely than air temperature. Photosynthetic productivity remained stable at air temperatures up to 38 degrees.

A tree short of water usually closes its stomata to slow further loss, which also stops evaporative cooling. Renée Marchin and colleagues combined heat and drought in a 2022 glasshouse experiment on twenty tree and shrub species. Watered plants came through the heatwave with leaf temperatures a mean of 3.5 degrees below the point at which their photosystem II would destabilize. In the droughted plants that margin closed to zero, the leaves reaching the critical temperature itself. Some species, however, responded to extreme heat by opening their stomata, cooling their leaves at the cost of using up the water left in their tissues. Current vegetation models omit that response.

Trees move water from their roots to their leaves through continuous internal columns. A long-running rainfall-exclusion experiment in the eastern Amazon followed drought-stressed trees from water loss to death. In 2015, Lucy Rowland and colleagues reported what had failed in the trees that died. Air bubbles had broken those water columns, so the roots could no longer supply the crown. The stored sugars, meanwhile, were largely intact, which ruled out the competing explanation that drought kills by starving a tree of the food photosynthesis makes.

Large trees must lift water farther, which makes them especially vulnerable to hydraulic failure during drought. They also hold more carbon. During the 2015 and 2016 El Niño, mortality rose among the largest trees in South American plots. Across the continent, tropical forests stopped gaining carbon.

By contrast, intact African tropical forests monitored through a network of 100 long-term plots continued gaining live biomass during the same El Niño.

The heat around the Osa

Regional differences like that one are the reason a pantropical figure cannot be read down to a single peninsula. Doughty's maps cover the Amazon, Central Africa and Southeast Asia, while van Tiel's cover South America, Southeast Asia and Central Africa. Because neither study includes Central America, their estimates cannot be applied directly to Costa Rica or the Osa Peninsula.

Still, the Osa's wet, hot lowland forests grow in a climate with a mean annual temperature of 25 to 26 degrees and 3,000 to 7,000 millimeters of rain a year. A 2015 airborne survey measured canopy heights up to 67 meters and aboveground carbon stocks exceeding 225 tonnes per hectare. Those carbon stocks were among the highest the survey had recorded in the Neotropics.

Keeping the canopy standing prevents the additional heat caused by clearing. An intact canopy shades the soil, while sunlight evaporates water from its leaves and carries heat away. Once the trees are removed, that cooling ends. Luke Zeppetello and colleagues found that cleared tropical land becomes hotter at the surface, with the greatest warming in the largest clearings.

In the Amazon, Edward Butt and colleagues found that regional forest loss warms intact forest up to 100 kilometers away, producing more than four times the warming attributed to nearby clearing alone. Other tropical studies have found that trees remain hotter as far as tens of meters inside a new forest edge because shade and humidity decline there. Juliano Prevedello and colleagues measured one degree of local surface cooling after reforestation, showing that restored tree cover can reverse some local warming.

Set beside each other, the studies mark costs rather than a moment of failure. A leaf crossing 46.7 degrees is a leaf under strain, and Winter's recovered from it in fourteen days. Carbon uptake starts falling more than ten degrees lower. Standing carbon turns down where the warmest month's daily maximum averages 32. What turns heat into death is water: the trees that died in the Amazon exclusion experiment died with their sugars intact and their water columns broken.

Two kinds of heat reach those leaves. One comes with the climate, on a schedule no landowner sets. The other arrives when the canopy next door comes down, and that one is decided property by property, in permits that can be refused, on the Osa as anywhere else in the tropics. Every permit refused is heat that never reaches the leaves the satellites are watching.