Ecosystem composition changes over the past millennium: model simulations and comparison with paleoecological observations

Abstract:

Over multi-decadal to multi-centennial timescales, ecosystem function and carbon storage is largely influenced by vegetation composition. The predictability of ecosystem responses to climate change thus depends on the understanding of long-term community dynamics. Our study aims to quantify the influence of the most relevant ecological factors that control plant distribution and abundance, in contemporary terrestrial biosphere models and in paleo-records, and constrain the model processes and parameters with paleoecological data. We simulated vegetation changes at 6 sites in the northeastern United States over the past 1160 years using 7 terrestrial biosphere models and variations (CLM4.5-CN, ED2, ED2-LU, JULES-TRIFFID, LINKAGES, LPJ-GUESS, LPJ-wsl) driven by common paleoclimatic drivers. We examined plant growth, recruitment, and mortality (including other carbon turnover) of the plant functional types (PFTs) in the models, attributed the responses to three major factors (climate, competition, and disturbance), and estimated the relative effect of each factor. We assessed the model responses against plant-community theories (bioclimatic limits, niche difference, temporal variation and storage effect, and disturbance). We found that vegetation composition were sensitive to realized niche differences (e.g. differential growth response) among PFTs. Because many models assume unlimited dispersal and sometimes recruitment, the “storage effect” constantly affects community composition. Fire was important in determining the ecosystem composition, yet the vegetation to fire feedback was weak in the models. We also found that vegetation-composition changes in the simulations were driven to a much greater degree by growth as opposed to by turnover/mortality, when compared with those in paleoecological records. Our work suggest that 1) for forecasting slow changes in vegetation composition, we can use paleo-data to better quantify the realized niches of PFTs and associated uncertainties, and 2) for predicting abrupt changes in vegetation composition, we need to better implement processes of dynamic turnover and fire in current ecosystem models.