Karstified carbonate aquifers are highly heterogeneous systems characterised by multiple component porosities and permeabilities. The different porosities are associated with different permeabilities commonly interpreted as (fissured) matrix, fractures and conduits (Worthington, et al., 2000), generally associated with a turbulent quick-flow vs a laminar slow-flow component (Atkinson, 1977, Kiraly, et al., 1995).
Understanding, quantifying and numerically modelling these different flow dynamics is a challenge (Sauter, et al., 2006), yet, at the same time relevant for integrated water resources management.
This applied research comprises three autogenic limestone aquifers in the Republic of Ireland: a low-lying spring catchment dominated by diffuse recharge and influenced by surface water – groundwater interaction; an upland-lowland catchment dominated by quick recharge and flow; and a coastal aquifer discharging via submarine and intertidal springs into the Atlantic Ocean. Hydroclimatic parameters were monitored on an hourly time step, while groundwater catchments were delineated using artificial tracer tests, univariate and bivariate statistical methods, and signal analysis (continuous wavelet transform and wavelet coherence).
Based on the principle that individual flow components may resemble the drainage of linear reservoirs, a systematic approach was employed to separate continuous time series of a low-flow component (LFC) from spring and stream hydrographs, thereby representing drainage of the low permeability domain.
Using the results of the above mentioned methods, conceptual site models were synthesised and represented in a semi-distributed numerical environment in the form of a pipe-network model using InfoWorks ICM® (Innovyze software, Wallingford, version 7.0) (Gill, et al., 2013, Schuler, et al., 2018). This approach breaks down the aquifer into a combination of multiple quasi-linear reservoirs feeding into a network of pipes. Within the pipes, turbulent and open channel flow is modelled using the Saint-Venant equations of conservation of mass and momentum and the Colebrook-White equation, while laminar flow is solved applying Darcy’s empirical law.
The different models were calibrated against discharge, LFC time series and observed head in connected seasonally flooded lakes (turloughs) over at least one year. Validation ranged between six months and five years.
Overall, the model performances show that the conceptual site models can be very well solved numerically. Further, distinct recharge and flow components can be modelled using linear reservoirs, integrated towards the spring outlet within the pipe-network.
This semi-distributed modelling approach is considered as a reasonable compromise in the framework of the heterogeneity of such karst systems, and hence, a promising tool for water resources management.