Tropical forest responses to global change have consequences for Earth’s carbon cycle and climate system. Recent advances in large-scale dynamic vegetation models of forest responses include demographic processes and greater biodiversity, capturing important time lags and forest resilience. Changes in tree recruitment and forest regeneration resulting from land use and climate change are critical to long-term tropical forest change, however even advanced dynamic vegetation models use fairly simple formulations for recruitment. We review current approaches to representing tree recruitment in advanced land surface models and use observations from Barro Colorado Island (BCI) to develop and test a new, computationally efficient formulation for tree regeneration (reproductive allocation, seedling emergence, seedling mortality, and seedling recruitment to the 1 cm dbh adult cohort) to capture stand age-, light- and moisture-dependent recruitment variation within and across years.
Results/Conclusions
Advanced dynamic vegetation models vary in how they capture stages of regeneration, with, for example, LPJ-GUESS, ED2, and CLM(FATES) all assuming constant fractions of net primary production allocated to reproduction and no environmental sensitivity of seedling survival or transition into the smallest adult cohort. Using 10 years of annual data from 12,798 observational plots at BCI, as well as previously published functional relationships, we developed a 2-pool regeneration submodel for implementation in CLM(FATES). Initial tests confirm that size-dependent allocation to reproduction results in a seedbank that grows as adult tree cohorts mature. Implementation of moisture-sensitive seedling emergence results in seasonal variation in seedling emergence, consistent with observed seasonality at BCI. Similarly, by representing moisture- and plant functional type (PFT)-dependent mortality, seedling mortality is highest during the dry season, with greater amplitude in seasonal variation for drought-intolerant PFTs. As a consequence of compensating seasonality in seedling emergence and mortality, the simulated seedling pool exhibits damped seasonality, and also grows over time as the stand matures. Observations show that fractional seedling recruitment into the adult (1cm) cohort is a logarithmic function of light, with early successional species recruiting more strongly at all light intensities above 15 µmol m-2 s-1. Thus, simulated PFT differences in seedling pool sizes reflect PFT variation in light- and moisture-dependent mortality, as well as in transition rates into adult cohorts. Initial tests suggest that our scheme can reasonably reproduce seasonal and interannual variation in tree recruitment rates at BCI; further work is needed to test this representation at other tropical forest sites and confirm the submodel’s generality.