Building evidence-based models: Bridging the gap between experimental data and vegetation models
Ecosystem-scale observations and experiments provide rich datasets but it is not always clear how these data can best be used to inform models of vegetation function. We present a new approach to model intercomparison and evaluation that we have successfully applied to ecosystem-scale Free-Air CO2Enrichment (FACE) experiments. In this approach, we recognise that ecosystem models contain many different assumptions and that we need to evaluate the individual assumptions rather than the models as a whole. By identifying and evaluating the main assumptions causing differences among models, it is possible to produce a clear roadmap for reducing model uncertainty.
We used this approach to evaluate eleven ecosystem models against two completed FACE experiments in plantation forests (Duke FACTS-I experiment on Pinus taeda and Oak Ridge National Laboratories experiment on Liquidambar styraciflua). We evaluated assumptions related to productivity, water use, and carbon allocation. For some processes, including stomatal conductance, nitrogen limitation, and allocation, we could identify which assumptions were supported by data and which were not. For other processes, we identified the reasons for the differences among the models and the research required to resolve these differences.
The approach can also be used in advance, in order to help guide experimental design. We describe how the approach has been applied to a new FACE experiment in a mature Eucalypt woodland (EucFACE experiment on Eucalyptus tereticornis). This approach identified several key research questions that would allow the experiment to best inform models.
We strongly encourage the application of this approach in other model intercomparison projects in order to fundamentally improve predictive understanding of the Earth system.