The long-term free-air CO2 enrichment experiments in a loblolly pine forest in North Carolina (Duke FACE) and a sweetgum forest in Tennessee (ORNL FACE) provide a rich source of data describing forest responses to future concentrations of atmospheric CO2 and the feedback they provide to the global carbon cycle. Model predictions of feedbacks between the terrestrial biosphere and the atmosphere are critical components of coupled carbon-climate models. To have confidence in their predictions it is important that the models be benchmarked against relevant experimental data. Using 12 ecosystem process and land surface models and 10-year data sets from the two FACE experiments, we are asking a series of questions: (1) Are the models good enough to replicate the measured processes of the carbon, water, and nitrogen cycles at the scale for which the models are intended? (2) Can the models explain the observed differences in response between the two experiments? (3) Will the comparison between model and experimental results lead to improvements in the models and give increased confidence in their predictive capacity? The first steps in this process were to assemble site information, meteorological and CO2 concentration records, and ecosystem response data, and develop a protocol for model implementation.
The modeling community found the data sets from the two experiments to be well suited for a model intercomparison study. The experiments are similar in many important respects – similar climatic zone, experimental protocol, forest stand stature. Both experiments documented increased net primary productivity in response to elevated CO2, but the allocation of the additional assimilated carbon has differed – fine-root production is enhanced in the ORNL FACE, whereas aboveground wood production is enhanced in the Duke FACE. Representing this difference correctly is a key challenge to the models because the distribution of carbon to tissues with very different turnover rates has important implications for carbon cycling and the feedback to the atmosphere. Another key challenge identified during protocol development is the representation of the tree-soil interface, including changes in soil properties with depth, differential root deployment, and nitrogen uptake. Improving ecosystem models to appropriately capture lessons learned from FACE studies, and continued progress in data-model fusion, represent significant steps in the longer-term goal of improving climate and earth system modeling.