On prior parameter structure investigation to parameter uncertainty
Document Type
Conference Proceeding
Publication Date
12-1-2005
Abstract
This study investigates the significance of the prior parameter structure, which is used as a regularization term in the inverse problem, to estimate the parameter heterogeneity over a field. It has been demonstrated that estimating parameter heterogeneity at all computation nodes are achievable even though the number of observation data is less than that of the unknown nodal parameter values if a regularization term is used in the inverse procedure. However, the quality of heterogeneity estimation at all nodes heavily depends upon the quantity of quality observations and how close the prior parameter structure is to the true heterogeneity. In this study, we use the generalized parameterization (GP) method to find the optimal prior parameter structure with a set of measured data. The estimation error of the prior parameter structure is characterized using a geostatistical method. The parameter value at each node is estimated through the Bayesian approach and the regularization of the prior parameter structure. We investigate the quality of nodal value estimation with different prior parameter structures made by a zonation structure, the ordinary kriging (OK) method, and the GP method. The GP method, in this study, is the combination of the zonation structure and the OK method through a set of weighting coefficients. The best prior parameter structure is obtained by searching for the optimal weighing coefficient values in the GP method via a set of observation data. We use the beta probability function to approximate all possible prior probability density functions of weighing coefficients in the Bayes rule. The proposed methodology is demonstrated at a synthetic confined aquifer, where the true transmissivity heterogeneity is unknown. The GP method is able to provide a good prior transmissivity structure which is better than the zonation structure and the kriged heterogeneity. As a consequence, the GP method is able to uncover the true transmissivity heterogeneity and shows better results than the other two methods. Copyright ASCE 2005.
Publication Source (Journal or Book title)
World Water Congress 2005: Impacts of Global Climate Change - Proceedings of the 2005 World Water and Environmental Resources Congress
First Page
378
Recommended Citation
Tsai, F. (2005). On prior parameter structure investigation to parameter uncertainty. World Water Congress 2005: Impacts of Global Climate Change - Proceedings of the 2005 World Water and Environmental Resources Congress, 378. https://doi.org/10.1061/40792(173)378