A 37-year soil experiment just rewrote one assumption about a warming planet

For thirty-seven winters the heated plots at the Harvard Forest in Massachusetts have stayed roughly nine degrees Celsius warmer than the forest floor next to them. On 14 July 2026 a team reported what that persistence has cost: the microbes in those plots have begun dismantling soil carbon that, in undisturbed forest, was supposed to stay locked away for centuries.

The finding, drawn from what is now the world's longest-running soil-warming experiment, complicates the basic assumption baked into most global climate models. Soils hold more carbon than the atmosphere and all terrestrial vegetation combined. If that stock decays faster under warming, the planet's carbon budget shrinks faster than the headline numbers suggest, and the time window for capping warming at 1.5°C narrows with it.

What the experiment actually shows

The Harvard Forest warming study began in 1991 as a small, buried-cable installation designed to simulate a future climate. Researchers have sampled the plots on a fixed schedule for more than three decades, providing one of the few datasets that can distinguish a transient microbial response from a permanent shift in how soil stores carbon.

The new analysis, summarised by Phys.org's Latest Science News feed on 14 July 2026, found that warming can cause microbes to break down stable soil carbon that was previously considered resistant. Crucially, the loss did not flatten out. After nearly four decades, the heated plots were still releasing carbon the surrounding forest retained, indicating an ongoing, not a one-off, change in the soil's recycling economy.

In plain terms: forest soils do not simply reach a new equilibrium when warmed. They keep moving.

Why this sits awkwardly with the standard models

Most Earth-system models treat soil carbon as a large, slow-moving reservoir that responds to temperature on a long lag. The intuition is borrowed from laboratory incubations: heat the soil, microbes respire more, the system settles. The Harvard Forest record now suggests that the incubation analogy understates the field reality by a wide margin.

What the experiment is showing, in the language researchers use in the summary, is a hidden climate threat: warming unlocks old carbon. Two structural readings are plausible. The first is that the models need a recalibrated temperature sensitivity for the deep, mineral-protected carbon pool. The second, more uncomfortable reading, is that the relevant carbon is not as protected as the mineralogy implies, and the global stock is more vulnerable on a decadal timescale than current policy pathways assume.

Both readings point in the same direction for policymakers: the carbon budget assumed by current pledges is built on a soil response that the longest direct experiment in the world now calls into question.

The counterweight

There is a counter-narrative worth stating plainly. A single heated-plot experiment at one temperate forest site is not a global verdict. Soil microbial communities differ sharply between pine, oak, boreal peat and tropical moist forest, and the carbon stored in each responds to a different set of chemical and hydrological levers. Critics of alarmist climate framing will point out, fairly, that running the finding forward requires assumptions about microbial acclimation, substrate availability and moisture that the Harvard Forest data alone cannot resolve.

Defenders of the finding will reply, also fairly, that the Harvard record covers more time than almost any alternative, and that the absence of a plateau across nearly four decades is the kind of signal that should change priors. Both positions can be held simultaneously. The honest summary is that the experiment narrows what is plausible; it does not yet close the question.

What to watch next

Three near-term developments will determine how much weight the new finding carries in the next round of national climate plans, due in many jurisdictions over the coming eighteen months. First, whether parallel long-term warming sites in boreal and tropical forests show the same unflattened carbon loss. Second, whether the major Earth-system modelling centres treat the Harvard result as a calibration point or as an outlier in their next assessment cycle. Third, whether soil-carbon responses are folded into the voluntary carbon market's accounting rules, where the financial stakes run into billions of dollars and the methodology has already been criticised for overstating soil removals.

For now, the headline rests on a small forest plot in Massachusetts and thirty-seven winters of buried cables. That is a thin empirical base to rebuild a planetary assumption on. It is also the only empirical base that long.

This article was assembled using dispatches from the Phys.org wire and tagged reporting; Monexus has not conducted independent interviews with the Harvard Forest team and has relied on the published summary for the present analysis.