A key objective of Medwater is the development of tools for the sustainable management of groundwater. A pre-requisite for meeting this objective is a numerical model of the aquifer and the unsaturated zone. The impact of vegetation and soil must also be taken into account. The development of such deterministic models is typically data intense, but often the availability of data is limited. To investigate the potential of statistical approaches in areas with data scarcity, both stochastic and deterministic approaches are taken for the Western Mountain Aquifer (WMA).
Deterministic Approach: Double-continuum model in HydroGeoSphere
The deterministic approach comprises a fully-coupled model of the subsurface and surface domains, created in HydroGeoSphere (HGS). The subsurface domain utilises a double-continuum to represent the primary slow-flowing (matrix) and secondary fast-flowing (conduits) domains. A modified Richards equation is applied for calculation of flux in the primary and secondary domains of the subsurface. The exchange flow between the continua is calculated via a Darcy-type exchange term.
Stochastic Approach: Single-continuum model in MODFLOW with Stochastic Karst Simulator (SKS)
The single-continuum approach is centred on a model of the WMA, created in MODFLOW. The single continuum represents averaged properties of the karst conduit and fracture domains- i.e. these are not considered separately. The Stochastic Karst Simulator (SKS; Borghi et al., 2012) is used to stochastically determine different possibilities for the geometry of the karst conduit network, and these are used to distribute zones of higher (karst) and lower (non-karst) hydraulic conductivity and storage within the single model domain. The SKS algorithm utilises information about the locations of inlets and outlets to the karst system during its development, the locations of faults, and the solubility of different facies. This information is drawn from existing studies of the aquifer’s composition and geological history. In this way, broader geological information may be used to parameterise the groundwater model, reducing the reliance on scarce groundwater head and discharge time-series. For groundwater recharge, analytical functions are applied.
The HGS model has been initially calibrated for the sub-surface continua. During the pre-development phase the only outlets of the aquifer were the Taninim and Yarkon springs. The steady state simulation was configured to represent the water-level and discharge rate of the pre-development period. The calibrated model is able to simulate the spring discharge of the Yarkon and Taninim springs. The simulated discharge reproduces the drying up of the Yarkon spring in the 1970s. In 1991/92 the Yarkon spring was reactivated due to an extremely wet year. The simulation replicates this event.
In the confined part of the aquifer the simulated groundwater levels are in good agreement with the observed data, however there are discrepancies in the unconfined mountainous region.
An initial version of the single-continuum MODFLOW model is undergoing calibration prior to implementation of the SKS. The reactivation of the Yarkon spring after the wet year 1991/92 is also captured in the simulation. However, the reproduction of groundwater levels is currently not as good as in the HGS model.