A decades-long heating experiment reveals that Earth’s most stable soil carbon isn’t as permanent as climate models assumed—and microbes may be awakening it just as the planet warms.
Soil scientists have long treated the deep, stable pool of carbon locked in forest soils as a kind of climate savings account — chemically bound, microbially uninteresting, and safe for centuries. A 37-year warming experiment in the woods of central Massachusetts has just torn a hole in that assumption.
After decades of steady heating, the most stubborn fraction of soil organic matter at Harvard Forest began to break down, releasing carbon dioxide that models had assumed would stay in the ground. The world’s longest-running soil warming experiment now suggests a feedback loop between microbes and climate that is stronger, and slower to arrive, than mainstream projections have captured.
The finding matters because forest soils store more carbon than the atmosphere and all living plants combined. If even a modest slice of that pool becomes volatile under sustained warming, the arithmetic of every emissions pathway shifts.
A plot heated for nearly four decades
The experiment began in 1991. Buried heating cables kept a set of forest plots at Harvard Forest 5°C warmer than the surrounding ground, year-round, without interruption. That temperature offset was chosen to bracket the upper end of global warming projections available at the time.
For context, global average temperatures have risen roughly 1.1 to 1.4°C since the Industrial Revolution. The Harvard plots are a look into a future the planet has not yet reached — but is heading toward.
Jerry Melillo has run the site for its entire life. Over three decades, his team documented the expected pattern: microbes in the heated soil worked faster, chewing through the easy-to-digest organic matter and venting CO2. That pulse eventually tapered. Many researchers assumed the system would settle into a new, warmer equilibrium.
It didn’t.
The fourth decade surprise
During the fourth decade of continuous warming, something new happened. The chemically stable carbon — the material bound tightly to minerals, wrapped in resistant molecular structures, and long assumed to be off-limits to microbial appetite — began to degrade.
The peer-reviewed analysis in Science of the Total Environment documents the destabilization of persistent soil organic matter after roughly 30 years of warming. The persistent pool is not inert. It is patient.
According to the Marine Biological Laboratory, Melillo explained that warming affects microbial communities in soil, which play a key role in breaking down organic matter and can accelerate carbon loss. The research showed that warming reshapes microbial communities in ways that can accelerate the loss of carbon from soils.
Why stable carbon was supposed to stay put
Soil carbon comes in flavors. Fresh leaf litter and root exudates are easy calories — microbes tear through them in weeks or months. Deeper down sits the recalcitrant fraction, sometimes hundreds or thousands of years old, protected by physical shielding inside soil aggregates and by chemical bonds to clay and iron minerals.
Climate models have generally treated this deep pool as a slow-turnover reservoir, adjusting mostly to land use change rather than to a few degrees of warming. The Harvard Forest data challenges that framing. Given enough time, warmer temperatures appear to remodel microbial communities in ways that let them reach carbon that was previously locked away.
The mechanism is not fully mapped, but the implicated pathway involves shifts in enzyme production, changes in the fungal-to-bacterial ratio, and the arrival of microbial specialists capable of cracking mineral-associated organic matter. In effect, the underground economy retools itself.
What this does to the climate math
Every major climate model includes some version of a soil carbon feedback. Warmer soils respire more CO2, which warms the atmosphere further, which warms soils again. The question has always been how large the feedback is and how long it lasts.
If short-term experiments were used to calibrate that loop — and most were, because 37-year datasets barely exist — the models likely captured only the early burst from labile carbon. The slower degradation of stable carbon would have been invisible.
Melillo and colleagues argue that folding this delayed release into Earth system models will sharpen projections of 21st-century warming. It will also probably raise them.
The scale of the potential contribution is not trivial. Global soils hold an estimated 1,500 to 2,400 gigatons of carbon in the top two meters, roughly two to three times the amount in the atmosphere. A single-digit percentage loss over decades would rival cumulative fossil fuel emissions from a major economy.
An older experiment, a familiar dynamic
The Harvard Forest result also lands in a longer conversation about how ecosystems respond to sustained CO2 and temperature pressure. A decade-long free-air CO2 enrichment study in the Mojave Desert, published in Nature Climate Change, found that arid ecosystems can accumulate substantially more carbon under elevated CO2 than previously credited. That study identified deserts as unrecognized carbon sinks.
Put side by side, the two experiments sketch a more complicated ledger. Some ecosystems bank more carbon than expected when CO2 rises. Others, once warm enough for long enough, begin to spend down reserves that were thought untouchable. The net global effect depends on which process dominates, and where.
Related work on climate feedbacks is broadening the picture further. Habitat modeling suggests that a third of animal habitats could face overlapping climate disasters by 2085, and physical geography is already shifting in ways that were not forecast — Himalayan rivers are wandering roughly twice as fast as they did a generation ago. Slow systems are speeding up. Stable systems are not.
The policy leverage
The Harvard team is careful to note that the feedback is not fixed destiny. Its magnitude depends on how much warming the planet actually experiences.
Melillo noted that the magnitude of the carbon feedback effect depends on future warming levels and could be reduced through dramatic cuts in CO2 emissions and reduced deforestation.
That sentence carries the whole argument. The stable carbon pool responds to temperature. Temperature responds to emissions. Every tenth of a degree avoided at the smokestack is a tenth of a degree not applied, decade after decade, to a global underground reservoir now known to be less stable than the textbooks claimed.
What comes next for the science
Long experiments are rare because they are expensive, unglamorous, and require someone willing to stay put. The Harvard Forest plots exist because a generation of researchers kept the cables running through funding cycles, retirements, and shifting scientific fashions. Their payoff arrived in year 30.
Several questions remain open. How widely does the fourth-decade destabilization apply beyond temperate hardwood forests? Do boreal soils, tropical soils, and permafrost show similar late-onset shifts, or different ones? How quickly can Earth system models be reparameterized to include a lagged pulse from stable carbon, and what does that do to projected temperatures in 2100?
Field networks are already reporting comparable patterns in other long-term warming plots, and the Harvard team expects follow-up studies to test whether microbial community turnover is the mechanistic key.
For now, the practical takeaway is narrow and sharp. The carbon in forest soils is not as locked away as scientists thought. The lock, it turns out, has a temperature-sensitive tumbler — and heat has been picking it, quietly, for the better part of forty years.
NOTE – This article was originally published in Science Blog and can be viewed here
