Parameter estimation for numerical models can synthesize different types of information into a physically plausible narrative. This is of particular relevance for the discipline of hydrogeology, where informed management demands detailed knowledge of the system, but direct measurements of the relevant subsurface properties are scarce and often of limited spatial representativeness (e.g., Rubin, 2003). The process of inferring subsurface properties from dependent information such as hydraulic head, chemical concentrations, or flow is known as inverse modeling (e.g., Carrera et al., 2005).Unfortunately, as a consequence of the exceptional complexity of many hydrogeological systems (Figure 1), there usually exists more than a single plausible explanation for the observed data (Linde et al, 2015(Linde et al, , 2017Moeck et al., 2020). Variations in aquifer depth, sediment properties, atmospheric and hydrogeological forcing, anthropogenic influences, and complex geological features interact with each other and can create similar hydraulic responses in different arrangements. The consequence of this has been summarized succinctly by Poeter and Townsend (1994): "A true evaluation of the possible subsurface configurations and their impact on the decision at hand is the only honest approach to groundwater analyses." and hence surmised that "The era of drawing conclusions on the basis of deterministic flow and transport models has come to a close."Where deterministic models only seek a single promising model configuration, stochastic approaches based on Bayesian statistics explore multiple alternative configurations at once. This process hopes to identify ambiguities in order to endow model predictions with reliable uncertainty estimates. Unfortunately, 25 years later, Poeter and Townsend (1994)'s prediction has yet to fully come to pass. While the need for probabilistic groundwater models has been widely acknowledged (e.g.