WaterTalk | Hydrogeology: What's the Use?

Hydrogeologists are faced with the challenge of building models of water and solute movement through complex, sparsely characterized geologic systems. Every step of this process introduces opportunities for error and uncertainty, from the choice of where and what to measure, to the conceptualization of the flow system, to the construction, parameterization, and calibration of the numerical model. There is increasing acceptance among hydrogeologists that 'all models are imperfect'. But there is little to no agreement about how (or if) model uncertainty should be communicated to clients and how that uncertainty might be included in water resources planning. In this talk, I will present a very simple hydrogeologic problem and use it to explore the following topics. How does measurement uncertainty affect model prediction uncertainty and how can we determine if those uncertainties matter?  How can we determine how much hydrologic observations are worth in a decision context? How can we account for stakeholder biases and, importantly, are they always something to be avoided? The talk will draw on Bayes' Law, Pareto optimization, and multi-model analysis.  But the take home messages will be clear and simple to explain to any audience.