Data Synthesis in the Community Land Model for Ecosystem Simulation
An approach for extracting fundamental variables from simulated or observed ecosystem data and synthesizing other variables using the fundamental variables.
Sampling theories, data-mining technologies, and virtual-sensor concepts were used to analyze the correlation between model parameters and bridge gaps between observation data streams and modeling data streams.
It is an effort to use sampling theory, data-mining technologies, and virtual-sensor concepts to analyze the correlation between model parameters [e.g., over 60 parameters for the canopy flux module (temperature, air, ground, vegetation, carbon dioxide concentration, photosynthesis, leaf area index, and vcmax)] and to bridge the gaps between observation data streams and modeling data streams. This study is a key step forward in synthesizing model-required data streams from observation or measurable datasets, so that computational experiments can be constructed for direct model-data comparison.
This paper presents a data synthesis model to generate ecosystem data in climate simulations. This model is capable of (1) extracting key features of different physical properties in time and frequency domain, and (2) discovering and synthesizing the physical relationships between ecosystem variables in different feature spaces.
Oak Ridge National Laboratory
The work was supported in part by NSFC grant 61305114, as well as the Terrestrial Ecosystem Science program and Accelerated Climate Modeling for Energy project funded by the U.S. Department of Energy, Office of Science, Office of Biological and Environmental Research. This work also used computing resources at Oak Ridge National Laboratory.
He, H., et al. “Data synthesis in the Community Land Model for ecosystem simulation.” J. Comput. Sci. 13, 83–95 (2016). DOI:10.1016/j.jocs.2016.01.005.