Advanced Computing, Mathematics and Data
Testing a Land Model's Water Cycle Simulation Skills
Uncertainty exposed and characterized in model's hydrologic parameters
Understanding how water moves above, below and through the landscape, and successfully modeling that movement, helps scientists predict future impacts on the water cycle and the climate under the pressures of climate change. Photo courtesy of Scott Butner. Enlarge Image.
Results: Scientists at Pacific Northwest National Laboratory and Oak Ridge National Laboratory, exploring new research territory in a popular Earth system model, applied a computational technique to systematically evaluate the relative importance of key parameters for climate prediction and water resources management. Using this technique, they found that the most uncertain parameters are those associated with deep subsurface processes. Their results will apply across multiple climate and site conditions.
Why It Matters: Droughts, floods, super-storms and changing weather patterns are reminders of our vulnerability to uncertain future water resources. Researchers use computational models to understand and predict how water, energy, Earth and biological cycles are changing over time. Uncertainties in water cycle parameters of land surface models could have significant impacts on our understanding of water and energy changes, and land surface that, in turn, affect the atmosphere and how carbon moves through the biosphere. Scientists are working to nail down these uncertainties.
In these publications, researchers are working to fine-tune the parameters in the Community Land Model (CLM4) that represent water runoff from the land surface, and how water travels through the surface, subsurface and water table. With improved predictions of water cycle fluctuation, land-use managers and policy-makers will have increased understanding for future planning.
Methods: In two studies, the researchers integrated CLM4 with an uncertainty quantification (UQ) framework applied to 20 watersheds from the Model Parameter Estimation Experiment (MOPEX) and 13 flux towers from the AmeriFlux network. These studies spanned a wide range of climate and site conditions to investigate the sensitivity of surface fluxes and runoff simulations to major hydrologic parameters in the model. Their goal was to improve the ability of CLM4 to simulate realistic hydrological responses to climate and provide accurate land surface fluxes in Earth system models. The UQ framework features three approaches for efficient and effective analysis of model sensitivity: an entropy-based approach, an exploratory sampling approach and a generalized linear and additive model analysis.
Taking advantage of measurements from the flux towers and relatively undisturbed watersheds, the research team evaluated the runoff generation parameterizations in CLM4 at scales that matter for hydrologic simulations. These studies constitute the first steps toward understanding the uncertainty of the input parameters and how that propagates through the model to affect the runoff and surface energy flux simulations. The results show the need to constrain the hydrologic parameters using surface flux measurements and streamflow records, and provided guidance on how this can be done effectively.
What's Next? In these studies, the researchers developed the CLM UQ capability. Next, they will apply the UQ framework to all 431 U.S. MOPEX sites to provide calibrated model parameters for improving CLM4 hydrologic simulations. They will identify which types of watersheds favor parameter calibration and which types of watersheds share common parameter significance patterns to improve model parameter calibration.
Sponsors: This research was supported by the Climate Science for a Sustainable Energy Future project funded by the U.S. Department of Energy Office of Science Biological and Environmental Research program's Earth System Modeling Program. The PNNL Platform for Regional Integrated Modeling and Analysis (PRIMA) Initiative, funded by the Laboratory Directed Research and Development program at PNNL provided support for the model configuration and datasets used in the numerical experiments.
Research Team: Maoyi Huang, Zhangshuan Hou, Ruby Leung, Yinghai Ke, Ying Liu, Guang Lin, Zhufeng Fang and Yu Sun, Pacific Northwest National Laboratory; Daniel M. Ricciuto, Oak Ridge National Laboratory.
References: Huang M, Z Hou, LR Leung, Y Ke, Y Liu, Z Fang, and Y Sun. 2013. "Uncertainty Analysis of Runoff Simulations and Parameter Identifiability in the Community Land Model - Evidence from MOPEX Basins." Journal of Hydrometeorology 14(6): 1754-1772. DOI: 10.1175/JHM-D-12-0138.1
Hou Z, M Huang, LR Leung, G Lin, and DM Ricciuto. 2012. "Sensitivity of Surface Flux Simulations to Hydrologic Parameters Based on an Uncertainty Quantification Framework Applied to the Community Land Model." Journal of Geophysical Research 117(D15108). DOI: 10.1029/2012JD017521