“…Estimation of spatially variable parameter fields, such as hydraulic conductivity or transmissivity, is an essential task in groundwater flow and transport modeling. Due to costly collection and sampling of local‐scale cores, field‐scale characterization is typically implemented by inverse modeling of indirect measurements from large‐scale aquifer tests, such as pumping and tracer tests (Bui‐Thanh & Girolami, 2014; Cardiff & Barrash, 2011; Cardiff et al., 2009; Carrera & Neuman, 1986a; Cui et al., 2011; Fienen et al., 2006; Liao & Cirpka, 2011; Tiedeman & Barrash, 2020; Xu & Gómez‐Hernández, 2018; Yeh & Liu, 2000; Zhao et al., 2018; Zhu & Yeh, 2005). Such inverse problems can be conveniently addressed in a Bayesian framework, which formulates the posterior distribution of unknown parameter fields by combining the likelihood function for data fitting and a regularization term, the prior that encodes spatial correlations or smoothness of the unknown fields (Carrera & Neuman, 1986a; Gavalas et al., 1976; Isaac et al., 2015; Kitanidis and Vomvoris, 1983; Liu & Kitanidis, 2011; Neuman, 1980; Rubin et al., 2010; Woodbury & Rubin, 2000).…”