Bayesian assimilation of resistivity and lithologic logs for updating hydraulic conductivity

Abstract

Characterizing spatially heterogeneous hydraulic conductivity (K) plays a crucial role in groundwater resources management and subsurface contaminant remediation Since the direct measurements of K are sparse the uncertainty is inherent in the estimated parameters We propose to decrease the uncertainty by assimilating secondary data (related with K to some degree) with the primary data (K) using Bayesian statistical method Different from classical geostatistical methods both linear and nonlinear relations between the primary and secondary data can be considered in Bayesian statistical method A synthetic example is designed to evaluate the method Results show that increasing numbers of secondary data improves K estimates efficiently if higher correlation exists for primary and secondary data However the relation type has little influence on the accuracy and uncertainty Also we explore the use of resistivity logs and lithologic description for K estimation at Choushui River alluvial fan using the method Results indicate jointly assimilating both types of secondary data can obtain the most accurate and least uncertain estimates By combining both types of data with K data a less uncertain K distribution of Choushui River alluvial fan is obtained The K distribution of aquifer 1 has similar spatial variation with the result of facies distribution from previous study The reduction efficiency of uncertainty is 37 3 % for aquifer 1 and 9 6 % for aquifer 2 compared to the uncertainty by ordinary kriging

Date of Award	2019
Original language	English
Supervisor	Kuo-Chin Hsu (Supervisor)