How black is the box: examining the identifiability of pumping schedules based on corresponding boundary fluxes
K.F. McCulloch1, T.P.A. Ferré1
1Department of Hydrology and Atmospheric Sciences
The University of Arizona, Tucson, AZ
Drawdown caused by mine dewatering is often underrepresented in regional groundwater flow models developed by state regulators, despite the hydrogeological significance of these sinks. This underrepresentation arises because modelled drawdown predictions and the pumping schedules (locations and timings) that generate them are treated as proprietary. Nevertheless, the state’s regional models would benefit from information on the boundary fluxes associated with mine dewatering. To encourage mine operators to disclose these predictions to state regulators, this research evaluates the identifiability and uncertainty of pumping schedules from corresponding boundary fluxes using inverse modelling. Inverse modelling theory implies that, as the relative error in observations (i.e., complexity) increases, the identifiability of model parameters decreases and the uncertainty in their estimates increases. A common modern approach to evaluating prediction uncertainty in groundwater flow models is the first-order second-moment (FOSM) method. Hydrogeologists have widely applied FOSM and related linearized uncertainty methods in groundwater modelling using tools such as PREDUNC. For example, Fienen et al. (2010) used PREDUNC to examine the design of a hypothetical hydrologic monitoring network. More recently, White, Fienen, and Doherty (2016) developed pyEMU to support linearized uncertainty analysis and to interface with software such as PEST (Doherty, 2015). This research uses these tools to evaluate the identifiability of locations and timings of asynchronous multi-well pumping in a heterogeneous, anisotropic finite-difference domain, given only the boundary fluxes for the site-specific mine model. The evolution of this identifiability with increasing complexity, starting with an analytical solution of one well in a Theis domain, is presented.