Abstract. The length of time that carbon remains in forest biomass
is one of the largest uncertainties in the global carbon cycle, with both
recent historical baselines and future responses to environmental change
poorly constrained by available observations. In the absence of large-scale
observations, models used for global assessments tend to fall back on
simplified assumptions of the turnover rates of biomass and soil carbon
pools. In this study, the biomass carbon turnover times calculated by an
ensemble of contemporary terrestrial biosphere models (TBMs) are analysed to
assess their current capability to accurately estimate biomass carbon
turnover times in forests and how these times are anticipated to change in
the future. Modelled baseline 1985–2014 global average forest biomass
turnover times vary from 12.2 to 23.5 years between TBMs. TBM differences in
phenological processes, which control allocation to, and turnover rate of,
leaves and fine roots, are as important as tree mortality with regard to
explaining the variation in total turnover among TBMs. The different
governing mechanisms exhibited by each TBM result in a wide range of
plausible turnover time projections for the end of the century. Based on
these simulations, it is not possible to draw robust conclusions regarding
likely future changes in turnover time, and thus biomass change, for
different regions. Both spatial and temporal uncertainty in turnover time
are strongly linked to model assumptions concerning plant functional type
distributions and their controls. Thirteen model-based hypotheses of
controls on turnover time are identified, along with recommendations for
pragmatic steps to test them using existing and novel observations. Efforts
to resolve uncertainty in turnover time, and thus its impacts on the future
evolution of biomass carbon stocks across the world's forests, will need to
address both mortality and establishment components of forest demography, as
well as allocation of carbon to woody versus non-woody biomass growth.