Abstract:The spatial and temporal variations in terrestrial carbon storage play a pivotal role in regulating future climate change. However, Earth system models (ESMs), which have coupled the terrestrial biosphere and atmosphere, show great uncertainty in simulating the global land carbon storage. Here, based on multiple global datasets and a traceability analysis, we diagnosed the uncertainty source of terrestrial carbon storage in 22 ESMs that participated in phases 5 and 6 of the Coupled Model Intercomparison Project (CMIP5 and CMIP6). The modeled global terrestrial carbon storage has converged among ESMs from CMIP5 (1936.9 ± 739.3 PgC) to CMIP6 (1774.4 ± 439.0 PgC) but is persistently lower than the observation-based estimates (2285 ± 669 PgC). By further decomposing terrestrial carbon storage into net primary production (NPP) and ecosystem carbon residence time ( τ E ), we found that the decreased intermodel spread in land carbon storage primarily resulted from more accurate simulations on NPP among ESMs from CMIP5 to CMIP6. The persistent underestimation of land carbon storage was caused by the biased τ E . In CMIP5 and CMIP6, the modeled τ E was far shorter than the observation-based estimates. The potential reasons for the biased τ E could be the lack of or incomplete representation of nutrient limitation, vertical soil biogeochemistry, and the permafrost carbon cycle. Moreover, the modeled τ E became the key driver for the intermodel spread in global land carbon storage in CMIP6. Overall, our study indicates that CMIP6 models have greatly improved the terrestrial carbon cycle, with a decreased model spread in global terrestrial carbon storage and less uncertain productivity. However, more efforts are needed to understand and reduce the persistent data–model disagreement on carbon storage and residence time in the terrestrial biosphere. The spatial and temporal variations in terrestrial carbon storage play a pivotal role in regulating future climate change. However, Earth system models (ESMs), which have coupled the terrestrial biosphere and atmosphere, show great uncertainty in simulating the global land carbon storage. Here, based on multiple global datasets and a traceability analysis, we diagnosed the uncertainty source of terrestrial carbon storage in 22 ESMs that participated in phases 5 and 6 of the Coupled Model Intercomparison Project (CMIP5 and CMIP6). The modeled global terrestrial carbon storage has converged among ESMs from CMIP5 (1936.9 ± 739.3 PgC) to CMIP6 (1774.4 ± 439.0 PgC) but is persistently lower than the observation-based estimates (2285 ± 669 PgC). By further decomposing terrestrial carbon storage into net primary production (NPP) and ecosystem carbon residence time ( τ E ), we found that the decreased intermodel spread in land carbon storage primarily resulted from more accurate simulations on NPP among ESMs from CMIP5 to CMIP6. The persistent underestimation of land carbon storage was caused by the biased τ E . In CMIP5 and CMIP6, the modeled τ E was far shorter than the observation-based estimates. The potential reasons for the biased τ E could be the lack of or incomplete representation of nutrient limitation, vertical soil biogeochemistry, and the permafrost carbon cycle. Moreover, the modeled τ E became the key driver for the intermodel spread in global land carbon storage in CMIP6. Overall, our study indicates that CMIP6 models have greatly improved the terrestrial carbon cycle, with a decreased model spread in global terrestrial carbon storage and less uncertain productivity. However, more efforts are needed to understand and reduce the persistent data–model disagreement on carbon storage and residence time in the terrestrial biosphere.

Carbon Flux Variability from a Relatively Simple Ecosystem Model with Assimilated Data is Consistent with Terrestrial Biosphere Model Estimates

Benchmarking Carbon Fluxes of the Isimip2a Biome Models

Evaluation of Continental Carbon Cycle Simulations with North American Flux Tower Observations

Interannual and Seasonal Drivers of Carbon Cycle Variability Represented by the Community Earth System Model (CESM2)

Terrestrial Biosphere Model Performance for Inter‐annual Variability of Land‐atmosphere CO2 Exchange

Optimizing Carbon Cycle Parameters Drastically Improves Terrestrial Biosphere Model Underestimates of Dryland Mean Net CO2 Flux and its Inter‐Annual Variability

Evolution of Uncertainty in Terrestrial Carbon Storage in Earth System Models from CMIP5 to CMIP6

Evaluation of Terrestrial Carbon Cycle Models for Their Response to Climate Variability and to Co2 Trends

Benchmarking the Seasonal Cycle of CO2 Fluxes Simulated by Terrestrial Ecosystem Models

Intercomparison of global terrestrial carbon fluxes estimated by MODIS and Earth system models

Historically inconsistent productivity and respiration fluxes in the global terrestrial carbon cycle

Carbon cycle confidence and uncertainty: Exploring variation among soil biogeochemical models

How Uncertainties in Future Climate Change Predictions Translate into Future Terrestrial Carbon Fluxes

Evaluation of Modeled Global Vegetation Carbon Dynamics: Analysis Based on Global Carbon Flux and Above-Ground Biomass Data

Tracking 21st century anthropogenic and natural carbon fluxes through model-data integration

Uncertainty in the response of terrestrial carbon sink to environmental drivers undermines carbon-climate feedback predictions.

Intercomparison of Terrestrial Carbon Fluxes and Carbon Use Efficiency Simulated by CMIP5 Earth System Models

Multiscale assessment of North American terrestrial carbon balance

Site-level Model–data Synthesis of Terrestrial Carbon Fluxes in the CarboEastAsia Eddy-Covariance Observation Network: Toward Future Modeling Efforts

Benchmarking Coupled Climate‐carbon Models Against Long‐term Atmospheric CO2 Measurements

CARDAMOM-FluxVal version 1.0: a FLUXNET-based validation system for CARDAMOM carbon and water flux estimates