Benchmark analysis of parameterization under climate warming for terrestrial carbon cycle model

Background/Question/Methods: Parameterization of terrestrial ecosystem models plays an important role in accurately predicting climate – carbon cycle feedback.  Some studies have shown that climate change shifted parameters compared to those under ambient condition.  However, almost all climate models used a fixed set of parameters to simulate effects of climate change on ecosystem processes.  In this study, we conducted benchmark analysis of a terrestrial ecosystem (TECO) model against a highly accurate data set from mesocosm study under ambient temperature and warmed treatments in Ecologically Controlled Enclosed Lysimeter Laboratories (EcoCELLs) at Desert Research Institute, Reno, Nevada.  We used a Markov chain Monte Carlo (MCMC) technique to estimate parameters of the TECO model and measure the model performance with estimated parameters.

Results/Conclusions: Our analysis showed that the model performance with one set of estimated parameters was poor over a three-year experimental duration for pooled data under ambient temperature and warmed treatments.  The model performance was considerably improved with two sets of estimated parameters from ambient temperature and warmed treatments.  When we considered interannual variability of estimated parameters under both ambient temperature and warmed treatments, the performance was further improved.  Among the three sets of parameter values under both ambient temperature and warmed treatments, some are significantly different, indicating that climate warming and interannual variability caused discontinuous (or discrete) changes in ecosystem processes.  The changes in ecosystem processes pose significant challenges for carbon cycle model parameterizations and generate large uncertainties for model prediction.