Our capability to predict future climate is strongly limited by the poor representations of carbon-cycle responses to climate changes in terrestrial ecosystem models. In the past three decades, although >1600 field manipulative experiments have been conducted globally to explore the response of terrestrial ecosystem to global changes, their findings have not been widely used for facilitating model evaluation or development. Here, we present two cases using experimental studies to benchmark terrestrial ecosystem models. First, we evaluated nine models (CLM4.5, GDAY, ORCHIDEE, CABLE, SDGVM, JULES, TECO, LPG-GUESS and OCN) on their simulated responses of ecosystem carbon storage capacity to CO2 elevation based on results from six Free-Air CO2 enrichment (FACE) experiments in North America. We further applied a traceability analysis to trace the model-to-model difference in CO2 effect on ecosystem carbon storage capacity back to the key processes and parameters in the models. Second, we evaluated two models (DAYCENT and TECO) on their simulated responses of ecosystem CO2 fluxes to experimental warming and increasing precipitation in a temperate steppe in northern China.
Results/Conclusions
In the first case, we found that C input and residence time contributed equally to the huge model-to-model difference in simulated effect of elevated CO2 on ecosystem C storage capacity. In the second case, we show that although the models can be tuned to mimic the CO2 dynamics observed by the local eddy flux tower and in the control plots, they cannot well capture the observed experimental effects of warming and increasing precipitation. These results indicate that the widely distributed field experiments can provide useful benchmarks for facilitating evaluation and improvements of terrestrial ecosystem models.