Integrated and process-based modeling of flow and transport in multi-compartment karst systems with thick vadose zones

The hydraulic characterization of karst systems remains a high challenge given their heterogeneous nature and large range of hydrogeological properties. In this study, a methodological approach is presented that demonstrates to what extent the temporal variation of spring signals, such as discharge rate, dissolved constituents and water temperature can be employed to characterize the karst system and to differentiate the individual contributions of the different physical compartments, as well as to derive hydraulic properties of the individual compartments by integrated inverse modelling of the spring signals.

Each compartment – (i) surface zone, (ii) vadose zone, and (iii) phreatic zone – imposes a complex transformation of the input signals (e.g., flow rate, temperature, concentration) that are routed through the whole system. However, numerical approaches to reproduce flow and transport dynamics in karst systems often lack the physical representation of controlling processes (e.g., preferential flow dynamics in the vadose zone) and therefore struggle to provide unique solutions. Therefore, this study aims at the identification of parameter sensitivities and hence reduction of model uncertainty employing an integrated approach for the modeling of karst systems. In test scenarios artificial rain events deal as model input for the Precipitation Runoff Modeling System (PRMS) coupled to a dual-domain type vadose zone and discrete karst conduit network system embedded in a porous matrix within the phreatic zone in order to account for fast and slow flow components in each compartment. In the vadose zone diffuse flow through the porous matrix is modeled by standard bulk effective approaches (MODFLOW UZF or simple transfer functions) and rapid fluxes via preferential flow paths are represented by a source-responsive infiltration model governed by film flow dynamics. In the phreatic zone diffuse and conduit flow are represented by a discrete-continuum model (MODFLOW CFPv2). The model geometry is kept simple (i.e., one model layer and a single conduit connecting a single sinkhole with the spring) while vadose zone properties (e.g., overall thickness) and input signals are altered to focus on their impact on the flow signal and on the sensitivity of parameters.