A Massachusetts forest plot has been artificially warmed by 5°C for decades. A new peer-reviewed analysis finds that even its most chemically stable soil carbon, once assumed safe from decomposition, began breaking down during the trial's fourth decade.
For 35 years, a team led by Jerry Melillo of the Marine Biological Laboratory has kept a set of plots at Harvard Forest, in central Massachusetts, running 5°C hotter than the surrounding ground — every day, every season, without a pause.3 Melillo describes his own work on Harvard Forest warming trials as spanning 37 years; the continuously heated plots documented in the experiment's own data archive have run since 1991, 35 years to date. Noodle News could not verify the extra two years from public materials.3
The Prospect Hill Soil Warming Experiment is the longest-running trial of its kind. By 2017, after 26 years, the heated plots had lost 17% of the carbon stored in the top 60cm of soil, in an uneven pattern of loss alternating with years of no detectable change.2 That result already reshaped expectations of how forest soils behave as the planet warms.
That 17% loss did not arrive as a steady decline. It came in bursts: the 2017 paper documents four distinct phases over the 26-year study, with soil respiration in the heated plots alternating between measurably higher than the unwarmed control plots and statistically indistinguishable from them. About three-quarters of the total carbon loss happened in the first decade, as microbes chewed through easily-accessed organic matter. Respiration then fell back in line with the controls for years — until years 18–23, when a second, smaller wave of loss appeared, this time linked to accelerated breakdown of lignin, a tougher plant compound. The remaining quarter of the 26-year loss happened in that second wave; the two flat periods in between show no measurable loss at all.2
A new analysis, published in April 2026, adds a twist the 2017 data could not yet see. Using molecular biomarkers — plant-derived compounds called n-alkanes, chemically bound into soil in ways long assumed to resist microbial attack — researchers found that this "persistent" carbon pool also began breaking down, roughly three decades into the trial.1
The stakes are large because of scale, not rate. Globally, soil organic matter holds more than three times as much carbon as the atmosphere or all land vegetation combined — so a small proportional loss from that pool, sustained over decades, is not a small number.4
The new paper does not convert its finding into a tonnage or percentage figure for the fourth-decade loss — it is a molecular-level chemistry analysis (using nuclear magnetic resonance and mass spectrometry), not a mass-balance update, so no comparable headline number yet exists to sit alongside the 26-year 17% figure.1 It is also unclear whether the pattern holds in other forest types, in colder or wetter soils, or under smaller warming than this trial's 5°C, which sits at the upper end of century-scale projections rather than the middle of them.3