Parameter estimation for an island aquifer considering tidal overheight (Norderney, Germany)

Parameter estimation for coastal aquifers generally is a time-consuming and computationally expensive task. It requires compromises on the number of parameters to estimate as well as how and if to incorporate variable-density driven flow and transport. While locations in coastal aquifers where certain data types (e.g., hydraulic head or salinity data) are most informative for inverse modeling have been established, data at such locations may not be routinely collected. This adds difficulty to the process of estimating a unique parameter set for a coastal aquifer.

A further challenge at coastal sites influenced by ocean tides and episodic sea-level variations (e.g., caused by storm events) is the consideration of tidal overheight, which can elevate the groundwater table well above mean sea level. Due to the computational expenses of simulating tidal influences in regional-scale groundwater flow models, tidal overheight is often neglected in such models. During a parameter estimation procedure, the neglected tidal overheight would be erroneously compensated for by lower hydraulic conductivities to match observed groundwater levels.

Our objective was to include the effects of tidal overheight in a parameter estimation procedure to characterize the hydraulic properties of the island aquifer below Norderney (Germany). For this purpose, a phase-averaged tidal boundary condition and routinely collected groundwater observation data were used. The model was implemented in MODFLOW-2005 and depicts the freshwater lens of the island as a steady-state groundwater flow model. The freshwater/saltwater interface, estimated using the Ghyben-Herzberg relation, is assumed a no-flow boundary due to a lack of salinity data describing the transition zone between fresh- and saltwater. Observed data were hydraulic heads averaged over a time frame of 10 years (2006-2015) and respective vertical differences in hydraulic heads at multi-level observation wells. Observation weights were defined based on measurement uncertainty and standard error of the mean.

Parameter estimation was performed using PESTPP-GLM with Tikhonov regularization and first-order second-moment (FOSM) uncertainty analysis.  Estimated parameters were: horizontal hydraulic conductivities and anisotropy factors for different zones based on a hydrogeological structural model for the island; conductances for river and drainage boundary conditions, which describe surface waterbodies and drainage channels present on the island; water levels for the river boundary condition; a scaling factor for production well skin sediment.

Results suggest that simulated heads match observed heads reasonably well, while prior parameter uncertainties were only reduced for horizontal hydraulic conductivities and vertical anisotropy factors of certain zones. The observed head data show pronounced variability on a smaller scale likely originating from locally present confining clay lenses and areas of lower permeability, which are known to exist from borehole data. For validation, transient simulations were performed with MODFLOW-2005 and the saltwater intrusion package (SWI2) to simulate the salt-/freshwater interface.