Page Banner

United States Department of Agriculture

Agricultural Research Service

Title: Model Abstraction Techniques Related to Parameter Estimation and Uncertainty

Authors
item Pachepsky, Yakov
item Van Genuchten, Martinus
item Cady, Ralph - US NUC REG, ROCKVILLE,MD
item Nicholson, Tomas - US NUC REG, ROCKVILLE, MD

Submitted to: Proceedings of the International Workshop on Uncertainty, Sensitivity, and Parameter Estimation for Multimedia Environmental Modeling
Publication Type: Proceedings
Publication Acceptance Date: March 10, 2003
Publication Date: August 10, 2003
Repository URL: http://www.ars.usda.gov/SP2UserFiles/Place/53102000/pdf_pubs/P2031.pdf
Citation: Pachepsky, Y.A., Van Genuchten, M.T., Cady, R., Nicholson, T.J. 2003. Model abstraction techniques related to parameter estimation and uncertainty. Proceedings of the International Workshop on Uncertainty, Sensitivity, and Parameter Estimation for Multimedia Environmental Modeling. Publisher: U.S. Nuclear Regulatory Commission, pp. 107-109.

Interpretive Summary: Model abstraction is a methodology for reducing the complexity of a simulation model while maintaining the validity of the simulation results with respect to the question that the simulation is being used to address. The need for model abstraction has long been recognised in simulations of complex engineering and military systems that show that increased levels of detail do not necessarily imply increased accuracy of the simulation results, but usually increase computational complexity and may make simulation results more difficult to interpret. Similar observations have been made for simulations of subsurface flow and transport problems. While increased levels of detail in the data currently do not necessarily imply increased accuracy of the simulations, it usually does imply increased data collection density. This paper summarizes ongoing and planned work related to model abstraction carried out in collaboration with the U.S. Nuclear Regulatory Commission (NRC). Main purpose of the research was to provide NRC with an improved analysis capability that would enhance the credibility and defensibility of site performance reviews. Toward this goal, the specific objectives of this study were (1) to identify model abstraction techniques that may be appropriate for characterizing and simulating water flow and contaminant transport in vadose zone, and (2) to develop a case study that will demonstrate the efficiency of some of model abstraction techniques for a specific site. Various methods of model abstraction are being pursued purpose and will be applied to two large databases of field-scale information involving soil water contents and solute concentrations. Both a humid in Belgium and an arid site in Nevada are being considered. Future work with the field data sets will compare the efficiency of model analysis techniques and provide a basis for developing rule-based strategies for model abstraction in the area of subsurface water and solute transport. The model abstraction algorithms described in this paper should be of considerable interest to both theoretical and applied scientists and engineers concerned with the movement of water and a range of contaminants in soils and groundwater. The research aims at making more cost-effective predictions of subsurface contamination from both point and non-point pollution sources.

Technical Abstract: Model abstraction is a methodology for reducing the complexity of a simulation model while maintaining the validity of the simulation results with respect to the question that the simulation is being used to address. The need for model abstraction is recognized in simulations of complex systems where increased level of detail does not necessarily imply increased accuracy, but increased computational complexity, data collection burden, and difficulty of interpreting the simulations results. This paper summarizes ongoing and planned work related to model abstraction carried out in collaboration with the U.S. Nuclear Regulatory Commission (NRC). Main purpose of the research was to provide NRC with an improved analysis capability that would enhance the credibility and defensibility of site performance reviews. Specific objectives of this study were (1) to identify model abstraction techniques that may be appropriate for characterizing and simulating water flow and contaminant transport in the vadose zone, and (2) to develop a case study that will demonstrate the efficiency of some of model abstraction techniques for a specific site. Various methods of model abstraction are being pursued for this purpose and applied to a large database of information involving soil water contents, pressure heads and solute concentrations monitored for several years along a 100-m long trench in a loamy soil at the Bekkevoort experimental site in Belgium. Data from an arid site in the Amargosa Desert, NV, will be analyzed also. The model abstraction analyses will involve (a) the Richards equation and a simpler water budget models, (b) inverse modeling, laboratory measurements and pedotransfer functions to estimate parameters, and (c) layered vs. homogeneous soil. Future work with field data sets will compare the efficiency of model analysis techniques and provide a basis for developing rule-based strategies for model abstraction in the area of subsurface water and solute transport.

Last Modified: 12/21/2014