Understanding and predicting the response of vegetated ecosystems to climate change and quantifying the resulting carbon cycle feedbacks requires a coherent program of field and laboratory experiments, data synthesis and integration, model development and evaluation, characterization of knowledge gaps, and understanding of ecosystem structure and function. The U.S. Department of Energy supports such a program, which produces community data, models, and analysis capabilities aimed at projecting the impacts of environmental change on future atmospheric carbon dioxide levels, predicting changes in extreme events, and assessing impacts on energy production and use. Two computational approaches--one for quantifying representativeness of field sites and one for systematically assessing model performance--will be presented.
Resource and logistical constraints limit the frequency and extent of observations, particularly in the harsh environments of the arctic and the tropics, necessitating the development of a systematic sampling strategy to maximize coverage and objectively represent variability at desired scales. These regions host large areas of potentially vulnerable ecosystems that are poorly represented in Earth system models (ESMs), motivating two new field campaigns, called Next Generation Ecosystem Experiments (NGEE) for the Arctic and Tropics, funded by the U.S. Department of Energy. We developed a Multivariate Spatio-Temporal Clustering (MSTC) technique to provide a quantitative methodology for stratifying sampling domains, informing site selection, and determining the representativeness of measurement sites and networks. We applied MSTC to model results and data for the State of Alaska to characterize projected changes in ecoregions and to identify field sites for sampling important environmental gradients.
As ESMs have become more complex, there is a growing need for comprehensive and multi-faceted evaluation, analysis, and diagnosis of model results. The relevance of model predictions hinges in part on the assessment and reduction of uncertainty in predicted biogeochemical cycles, requiring repeatable, automated analysis methods and new observational and experimental data to constrain model results and inform model development. The goal of the International Land Model Benchmarking (ILAMB) project is to assess and improve the performance of land models by confronting ESMs with best-available observational data sets. An international team of ILAMB participants is developing a suite of agreed-upon model evaluation metrics and associated data at site, regional, and global scales. We are developing Open Source software tools for quantifying the fidelity of model performance, allowing modeling groups to assess confidence in the ability of their models to predict responses and feedbacks to global change.