Skip to main content
ARS Home » Northeast Area » Beltsville, Maryland (BARC) » Beltsville Agricultural Research Center » Environmental Microbial & Food Safety Laboratory » Research » Publications at this Location » Publication #254802

Title: Sensitivity Analysis of Unsaturated Flow and Contaminant Transport with Correlated Parameters

item PAN, FENG - University Of Maryland
item ZHU, JIANTING - Desert Research Institute
item YE, MING - Florida State University
item Pachepsky, Yakov
item WU, YOUNG - Colorado School Of Mines

Submitted to: Journal of Hydrology
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 10/12/2010
Publication Date: 12/13/2010
Citation: Pan, F., Zhu, J., Ye, M., Pachepsky, Y.A., Wu, Y. 2010. Sensitivity Analysis of Unsaturated Flow and Contaminant Transport with Correlated Parameters. Journal of Hydrology. 397:238-249.

Interpretive Summary: The results of subsurface flow and transport modeling are inherently uncertain because the parameters controlling flow and transport cannot be known in detail. Since the determination of these parameters is extremely laborious and costly, it is important to know which parameters are the most sensitive, i.e. affect modeling results most and therefore have to be determined with higher detail and accuracy. In applying the sensitivity analysis, the common assumption is that flow and transport parameters are independent. Our hypothesis was that parameters of flow and transport in variably saturated soils are actually mutually dependent; consequently, the assumption of parameter independence may substantially distort the sensitivity of parameters. Our work demonstrates that our hypothesis was correct; that by taking into account the correlation of parameters, much more accurate and physically meaningful parameter sensitivity results were obtained. The results of this study are expected to be useful for researchers and engineers who conduct uncertainty and sensitivity analysis in projects related to flow and transport through complex subsurface systems. It can also provide meaningful information on the sampling and monitoring network needed to reduce parameter uncertainties and associated predictive uncertainties in flow and contaminant transport in the unsaturated zone.

Technical Abstract: Relative contributions from uncertainties in input parameters to the predictive uncertainties in unsaturated flow and contaminant transport are investigated in this study. The objectives are to: (1) examine the effects of input parameter correlations on the sensitivity of unsaturated flow and contaminant transport, and (2) assess the relative contributions of parameter uncertainties to the uncertainties of flow and transport at each hydrogeologic layer. Using the unsaturated zone (UZ) of Yucca Mountain (YM) in Nevada as an example, this study considers two cases with independent and correlated input parameters. Global sensitivity analysis is conducted to investigate the relative contributions of input parameters to flow and transport uncertainties based on: (1) sampling-based regression analysis for independent input parameters, and (2) regression-based decomposition method for correlated input parameters. When the input parameters are independent, it is found that the parameter uncertainty in permeability has the largest contribution to the uncertainties in percolation flux and mass arrival of the reactive tracer at the water table. For the percolation flux, the second largest contribution is from the van Genuchten a; the sorption coefficient of the reactive tracer is the second most important parameter on mass arrival tracer uncertainty. The sensitivity to the sorption coefficient is larger in the layers of devitrified and zeolitic tuffs than in the layers of vitric tuff. On the other hand, when the input parameters are correlated, the uncertainties in van Genuchten n and porosity have more contributions to the percolation flux and tracer transport uncertainties due to their correlations with the van Genuchten a and permeability, respectively. These results indicate the significant effects of parameter correlations on the sensitivity of unsaturated flow and transport. The findings are of significance in facilitating future characterizations to reduce the parameter uncertainties and associated predictive uncertainties of flow and contaminant transport in unsaturated, fractured porous media.