Do Earth System Models Overestimate Forest Carbon Sequestration? The Evidence Says Yes.

The claim is that current Earth system models (ESMs) — the same models used to project climate futures and set carbon budgets — systematically overestimate how much carbon forests will pull out of the atmosphere. Based on converging evidence from field experiments, satellite observations, and the IPCC's own assessment, this claim is true, though the precise magnitude of overestimation varies by region and model.

The strongest evidence comes from nutrient cycle research. Standard ESMs largely assume that forests can keep absorbing more CO2 as atmospheric concentrations rise — a process called CO2 fertilization — without being constrained by soil nutrients. That assumption is wrong. Wieder et al. in Nature Climate Change (2015) showed that when nitrogen and phosphorus limitations are properly built into models, projected terrestrial carbon storage by 2100 falls by 50–75% compared to nutrient-unlimited model runs. Terrer et al. in Science (2021) reinforced this with a meta-analysis of 138 CO2-enrichment experiments, finding that phosphorus limitation alone cuts forest carbon uptake by roughly 50% relative to what standard ESMs predict. A real-world test in a phosphorus-limited Australian eucalyptus forest — the EucFACE experiment, published in Nature Geoscience (2019) by Fleischer et al. — found near-zero additional carbon sequestration under elevated CO2 over six years, directly contradicting model predictions of substantial uptake.

The strongest version of the counter-argument is that forests globally are measurably greening and absorbing more carbon as CO2 rises, and that satellite data confirm vegetation expansion. This is partially true — the land sink is real. But Jiang et al. in Nature Communications (2020) used that same satellite data to show that CMIP6 models overestimate the vegetation greening response to CO2 fertilization by approximately 30–40% compared to observed trends. The models are moving in the right direction; they are simply moving too fast and too far.

A second structural flaw compounds the nutrient problem: ESMs largely omit disturbance feedbacks. Bugmann and Bigler in Global Change Biology (2011) demonstrated that drought-induced tree mortality, bark beetle outbreaks, and wildfire — all of which are intensifying under climate change — can flip forests from carbon sinks to carbon sources. These dynamics are systematically underrepresented, creating a positive bias in sequestration projections on top of the nutrient bias. The Global Carbon Budget 2023, coordinated by Friedlingstein et al., confirms that the land sink uncertainty remains very large at plus or minus 0.9 gigatons of carbon per year, and that dynamic vegetation models show high spread in future projections precisely because these structural gaps are unresolved. The IPCC AR6 Working Group I report (2021), Chapter 5, states explicitly that nutrient limitation, permafrost feedbacks, and disturbance regimes are insufficiently represented in current ESMs, and that the land carbon sink could be substantially weaker than central estimates.

The manipulation pattern to watch for here is not deliberate deception — this is a genuine scientific finding, not a fringe claim. The risk is selective citation in policy debates: using the high-end model projections as reliable baselines for carbon offset schemes or national climate targets without disclosing that those projections carry a known, directional bias. If forests sequester 50–75% less carbon than standard models project, every strategy that leans on forest offsets to meet a net-zero target becomes significantly less reliable. When you see forest carbon sequestration cited as a confident number rather than a range with acknowledged downside risk, that is the signal to ask which model version was used and whether nutrient and disturbance constraints were included.