Abstract

Observational studies, ecological experiments, and modeling are the three foundational approaches in ecological research. With increasing data availability from ecological observations and experiments, data–model integration becomes a critical tool to reduce model uncertainty and gain more reliable predictions of future ecosystem dynamics. In this chapter, we review the need, history, and current status of data–model integration to improve model simulations. Data from long-term manipulative experiments (e.g., free-air CO₂ enrichment experiments), observation networks of eddy covariance measurements (e.g., FLUXNET), and global databases (e.g., IGBP-DIS soil C database) have been successfully incorporated into ecosystem or global biogeochemical models to improve model projections. We also take two ongoing projects, Spruce and Peatland Responses Under Climatic and Environmental Change Experiment (SPRUCE) and Extreme Drought in Grasslands Experiment (EDGE), as examples to illustrate how experiment–model integration approaches are designed to achieve different research goals. In SPRUCE, the operational forecasting system is developed to assimilate data streams in real time to predict ecosystem responses to two global change factors, warming and elevated CO₂, whereas in EDGE, the application of data assimilation is designed to disentangle the role of environmental context versus ecosystem attributes in response to drought and eventually to scale findings obtained at distributed sites to regional scales. Finally, we review the challenges in data–model integration and propose a few strategies to move forward.