Forest carbon sink neutralized by pervasive growth-lifespan trade-offs

Terrestrial ecosystems have removed roughly one-third of anthropogenic CO2 emissions over the past 50 years, attributed to afforestation, secondary forest expansion, and possible changes in forest dynamics under rising CO2, nitrogen deposition, and warming. Many Earth System Models (ESMs) project continued net carbon uptake, partly via CO2- and temperature-stimulated growth, even in mature forests. Yet forest carbon balance depends on both growth and mortality; faster growth should eventually increase mortality, offsetting gains. Although monitoring shows widespread mortality increases, linking them mechanistically to growth is challenging due to trees’ longevity and limited monitoring durations. Tree rings provide a means to assess growth–lifespan relationships. While a growth–lifespan trade-off is well known across species, evidence within species and across biomes is incomplete and occasionally contradictory. The study aims to test the universality, causes, and consequences of growth–lifespan trade-offs across species and environments, and to evaluate implications for forest mortality dynamics and the persistence of the forest carbon sink.

Prior work suggests forests have acted as a significant CO2 sink due to regrowth and possible growth stimulation from CO2, warming, and nitrogen deposition. Classic studies identified a cross-species trade-off: fast growth associates with shorter lifespan, consistent with plant strategy spectra from pioneers to long-lived, stress-tolerant species. Within-species evidence has accumulated, often in conifers at high elevations, but coverage across taxa and biomes remained limited, and some studies report no trade-off or highlight higher mortality of slow-growing, suppressed trees prior to death. Mortality is known to increase with tree size and with climate extremes (e.g., drought), and ESMs struggle with representing mortality processes and carbon residence times. The gap addressed here is to determine how widespread within-species growth–lifespan trade-offs are across climates and taxa, whether they are driven intrinsically by growth rather than covarying environment, and how they impact future carbon sink projections.

Data: Compiled tree-ring records for >210,000 trees across 110 species from tropical to arctic regions and >70,000 sites. Major sources include the International Tree-Ring Data Bank (ITRDB) and the Quebec National Forestry Inventory (NFI-Quebec; 156,711 trees from 79,381 sites). Additional tropical datasets and published/unpublished records were included. Data quality controls: removal of duplicates in ITRDB; site selection to ensure old-growth signals (excluding sites with low age variability); for NFI-Quebec, exclusion of managed/disturbed plots; minimum sample sizes imposed for species-level (≥30 trees) and within-species trade-off analyses (typically ≥150 trees and ≥3 sites; 82 species analyzed within-species). Resampling experiments in Picea mariana evaluated sample-size effects on maximum age and trade-off slope estimation. Trade-off estimation: Within species, performed 95th quantile regressions of log(age) versus mean early growth (mean ring width in first 10 years) to estimate the exponential decay constant b describing how lifespan declines with higher early growth. Relative metrics (A/max(A), RW/max(RW)) were used to compare across species, with weighting by cube root of sample size. Artefact checks: (1) Living vs dead trees: compared slopes for trees known dead before 1900 versus living (12 species); no significant differences. (2) Recent growth increases: simulation-based sampling showed only small overestimation of trade-off under recent growth stimulation. (3) Pith offsets/wood decay: comparison of NFI-Quebec (higher quality) vs ITRDB indicated ITRDB trade-offs likely conservative; missing pith would weaken, not strengthen, trade-offs. (4) Sampling biases (big-tree selection): reconstructed population-level size distributions for Picea mariana to test effects of minimum DBH threshold (91 mm) and preferential sampling of large trees; found slight underestimation of the trade-off when using typical sampling. Environmental controls: For nine species across North America and Europe, related site-level early growth and maximum lifespan to WorldClim temperature and precipitation using major axis regressions and mixed-effects models. Also stratified analyses for Picea mariana by temperature, crown cover (competition), and soil type to disentangle direct growth effects from environmental covariates. Simulation of forest dynamics: Built a data-driven stochastic simulator for Picea mariana. Derived size-dependent mortality μ(D) from Quebec forest census size distributions combined with ring-derived ingrowth/outgrowth to match observed mortality increasing with diameter (fourth-order polynomial with D0=91 mm). Also constructed an alternative age-dependent mortality μ(A) for comparison. Created 600-year sequences of annually seeded 1250-member cohorts using observed ring-width trajectories extended to 500 years by matching early growth classes. Applied mortality annually via Monte Carlo draws based on μ(D) or μ(A). Imposed a 50-year growth stimulation (years 300–350) reflecting observed Quebec warming (0.0221 °C yr−1) and temperature sensitivity estimated via space-for-time across Quebec; age-dependent sensitivity Δ(A) applied, with younger trees more responsive. Also tested a “fast baseline” scenario with doubled growth rates. Outputs included mean radial growth, stem mortality rate, mean age of large trees at death (75th percentile), and changes in basal area stocks relative to a no-stimulation baseline.

- Universality of trade-off: Across 110 species, faster early growth associates with shorter maximum lifespan; within species, 74 of 82 examined showed significant negative relationships between early growth and lifespan. On average, lifespan decreased exponentially with a 23% reduction for a 50% increase in early growth. Trade-off strength was similar across taxa (gymnosperms vs angiosperms) and climate zones (boreal, temperate, tropical). - Robustness to environmental controls and artefacts: Negative correlations between lifespan and temperature exist for multiple species, but growth–lifespan trade-offs persisted when controlling for temperature, crown cover, and soil type. For Picea mariana, stratifying by temperature retained strong growth–lifespan relationships, while temperature–lifespan relationships largely disappeared when stratified by growth rate. Comparisons of dead vs living trees, and analyses addressing pith offsets and sampling biases, indicate observed trade-offs are not artefactual. - Mechanistic insight via size constraints: Fast and slow-growing Picea mariana reached similar maximum diameters (~300 mm) but at different ages, consistent with mortality increasing sharply with size and the existence of species/site-specific maximum tree size. Simulations with size-dependent mortality reproduced the observed trade-off, whereas age-dependent mortality did not. - Simulation outcomes for carbon dynamics: A realistic growth stimulation (estimated 29% mean growth increase over 50 years) produced an initial ~20% increase in standing biomass and a similar magnitude increase in mortality, with mortality increases lagging growth by 1–2 decades. Critically, biomass gains were transient and reversed to net losses after stimulation ceased, with mean lifespans of large trees reduced by up to 23 years. Simulations lacking the trade-off (age-dependent mortality) showed no mortality increase and sustained biomass gains. Results suggest faster growth accelerates turnover and does not yield long-term biomass increases. - Consistency with observations: The modeled lagged mortality increase mirrors observed trends in boreal, temperate, and tropical forests where productivity increases have been followed by mortality increases after ~20 years.

The study demonstrates that within-species growth–lifespan trade-offs are pervasive across tree taxa and biomes, and that faster early growth intrinsically shortens lifespan, largely independent of environmental covariates. This provides a mechanistic link between growth stimulation (from CO2, warming, nitrogen deposition, longer growing seasons) and subsequent increases in mortality, helping to explain observed concurrent increases in growth and mortality in long-term forest monitoring. The evidence supports size-dependent mortality and maximum size constraints as key drivers: fast-growing individuals reach vulnerable large sizes earlier, increasing death risk, whereas tree age per se is a weaker predictor. Simulations translating this trade-off into forest dynamics indicate that growth-driven carbon gains are temporary and are neutralized by lagged mortality, challenging ESM projections that do not incorporate such demographic feedbacks. Integrating growth–mortality trade-offs and realistic mortality functions into models is essential for credible forecasts of forest carbon sink persistence under global change.

This work provides strong, global-scale evidence that faster early growth shortens tree lifespans across most species and environments. The trade-off is intrinsic to growth and persists after controlling for climate, competition, and soils. Data-driven simulations show that growth stimulation leads to transient biomass gains that are offset and can reverse due to lagged increases in mortality, reducing the long-term forest carbon sink. These findings imply that current Earth System Model projections likely overestimate the persistence of forest carbon uptake. Future research should: (1) incorporate growth–mortality trade-offs and size-dependent mortality into process-based and Earth system models; (2) elucidate physiological and structural mechanisms (hydraulic, mechanical, defense traits) underlying within-species variation in longevity; (3) assess interactions with recruitment, competition, species turnover, and shifting size–density relationships under changing climate and CO2; and (4) evaluate how maximum potential tree size may shift with warming and vapor pressure deficit. Given likely limits to biomass accrual, rapid reductions in greenhouse gas emissions remain crucial.

- Simulations did not include explicit competition dynamics or recruitment changes; increased biomass could enhance self-thinning, potentially strengthening mortality responses. - Assumed size-dependent mortality curve invariant to CO2 and temperature; real-world shifts in maximum tree size or hydraulic constraints under warming could alter mortality patterns. - Uncertainty about whether maximum potential size increases under elevated CO2; rising vapor pressure deficit may reduce maximum height. - Species distributions and community composition may shift, affecting system-level biomass capacity. - Tree-ring datasets (e.g., ITRDB) can have pith offsets and sampling biases; although analyses indicate these biases likely conservative, residual uncertainties remain. - Age-dependent mortality models used for comparison do not capture observed trade-offs; other mortality formulations might yield intermediate outcomes.