Stochastic Modeling Approach to Identify Uncertainties of Karst Conduit Networks in Carbonate Aquifers

Abstract

Karst aquifers exist in all climate zones and constitute crucial water resources. Analyses by Goldscheider et al. (2020) showed that approximately 15% of the global ice-free land surface consists of carbonate rock, and Stevanović (2019) estimated that around 9% of the world population consumes water from karst resources. Karst aquifers are characterized by highly permeable conduits embedded in a less permeable porous rock matrix. Most of the groundwater flow in karst aquifers is therefore controlled by the conduits, which form complex, hierarchically organized networks (Ford & Williams, 2007). Thus, the conduit network dominates the hydraulic properties of karst aquifers. To adequately represent these flow dynamics, it is vital to account for the spatial distribution of the conduits. However, this task is often associated with challenges and uncertainties due to the large scale and complexity of karst aquifer systems (Bakalowicz, 2005).Numerical groundwater models are essential tools for addressing groundwater-related problems, for example, water resources assessment, and have been widely applied to study karst aquifers (Anderson et al., 2015;Parise et al., 2018;Zanini et al., 2021). Global model approaches, that link recharge estimations to spring discharge time series, provide information on the hydraulic behavior of the entire karst system, but do not provide detailed information on the hydraulic parameter field and therefore lack predictive power for climatic and geo-hydraulic interactions (Eisenlohr et al., 1997;Kovács & Sauter, 2007). Advanced model approaches for conduit networks are based on the speleogenetic evolution of karst aquifers (Kaufmann et al., 2019;Liedl et al., 2003). These models solve complex coupled process equations of chemical carbonate dissolution and groundwater flow with major computational effort (Bauer et al., 2003). Incorporating stochastic approaches into the modeling process provides the opportunity to evaluate uncertainties related to the resulting spatial distribution and geometry of a conduit network. A stochastic model approach developed by Jaquet et al. (2004) focuses on integrating conservative and reactive transport processes that control the conduit evolution such as advection, dispersion, and dissolution, but does not account for actual field observations of the karst system investigated. Further approaches by Frantz et al. (2021) and Pardo-Igúzquiza et al. (2012) simulate the conduit's geometry stochastically, based on field data of the dimensions of known conduit segments. However, these data are normally laborious to obtain,

Probabilistic Estimation of Tunnel Inflow from a Karstic Conduit Network

“…On the one hand, for a specific site, the geometry and mesh could follow the known topography and 3D geological structure of the study area. The construction of stochastic karst network models with SKS is capable to account for very realistic and complex geometries in 3D (Borghi et al, 2012) as well as a very wide range of situations, from alpine settings (Sivelle et al, 2020;Fandel et al, 2020) to coastal karstic environments (Vuilleumier et al, 2013), while possibly accounting for multiple phases of karstification (Banusch et al, 2022).…”

Karst water resources in a changing world: Review of solute transport modelling approaches

“…For the calibration of karst simulations models, a single variable (i.e., spring discharge, groundwater level) is often used to quantify the misfit between the simulated and observed variables. But when it is possible, introducing multiple sources of information (i.e., isotope concentrations, water age information) and/or soft data (i.e., knowledge and experience on the modelled area) into the model calibration phase is rather preferable, particularly for the improvement of model robustness and parameter reliability (e.g., Banusch et al, 2022;Hartmann et al, 2013aHartmann et al, , 2013bÇallı et al, 2022;Çallı et 2023b).…”

Stochastic simulation and uncertainty analysis of karst conduit network using pyKasso

“…Vuilleumier et.al [10] applied a similar stochastic approach to simulate karst conduit networks, using field mapping data, and acknowledged substantial uncertainties in their model. The studies by Banusch et.al [11] and Borghi et.al [12] further illustrate the complexities involved in modeling karst systems, especially in the context of identifying conduit networks and their properties from various data sources. Fandel et al [13] and Borghi et al [12] contribute to our understanding of the uncertainties in karst modeling, emphasizing the challenges in incorporating unmapped conduits and the effectiveness of inverse methods.…”