Predicting long‐term carbon sequestration in response to CO2 enrichment: How and why do current ecosystem models differ?

Large uncertainty exists in model projections of the land carbon (C) sink response to increasing atmospheric CO2. Free‐Air CO2 Enrichment (FACE) experiments lasting a decade or more have investigated ecosystem responses to a step change in atmospheric CO2 concentration. To interpret FACE results in the context of gradual increases in atmospheric CO2 over decades to centuries, we used a suite of seven models to simulate the Duke and Oak Ridge FACE experiments extended for 300 years of CO2 enrichment. We also determine key modeling assumptions that drive divergent projections of terrestrial C uptake and evaluate whether these assumptions can be constrained by experimental evidence. All models simulated increased terrestrial C pools resulting from CO2 enrichment, though there was substantial variability in quasi‐equilibrium C sequestration and rates of change. In two of two models that assume that plant nitrogen (N) uptake is solely a function of soil N supply, the net primary production response to elevated CO2 became progressively N limited. In four of five models that assume that N uptake is a function of both soil N supply and plant N demand, elevated CO2 led to reduced ecosystem N losses and thus progressively relaxed nitrogen limitation. Many allocation assumptions resulted in increased wood allocation relative to leaves and roots which reduced the vegetation turnover rate and increased C sequestration. In addition, self‐thinning assumptions had a substantial impact on C sequestration in two models. Accurate representation of N process dynamics (in particular N uptake), allocation, and forest self‐thinning is key to minimizing uncertainty in projections of future C sequestration in response to elevated atmospheric CO2.