Simulation of Flow and Transport Processes in Karst Systems: LuKARS 3.0 Unveiled

Karst water resources play a vital role in global water supply, providing drinking water to 10–25% of the world´s population. Karst systems exhibit complex hydrological behavior, with fast-flow pathways and highly variable storage capacities. Hydrological models are essential for effective water resource management. However, modelling of karst systems is still a difficult task due to the heterogeneity of these systems and the uncertainties in karst structure.

LuKARS, a semi-distributed hydrological model for karst systems, addresses some of these challenges by allowing the consideration of multiple hydrotopes (i.e. distinct landscape units characterized by similar land use and soil types and thus by homogeneous hydrological properties) within a catchment. Despite its low computational cost, LuKARS faces challenges in the context of sensitivity analysis and uncertainty quantification due to its large number of parameters. Compared to the original LuKARS version developed by Bittner et al., (2018), the newly developed version of LuKARS 3.0 presented in this study allows much faster computational times with reduction in runtime of approximately 99.39% (from 1.14 seconds per test run down to 7 milliseconds), for the same model structure. In addition, LuKARS 3.0 allows an easy implementation of the model on clusters and a flexible model structure characterized by an arbitrary number of hydrotopes as well as by the possibility of activating/deactivating different model compartments, i.e., epikarst, matrix and conduit.  

In this study, we leverage the low computation time of LuKARS 3.0 to apply Morris’ sensitivity analysis method, demonstrating its comparability to dimensional reduction techniques like the active subspace method. The efficient runtime also facilitates the investigation of combined parameter and structural uncertainties. We calibrate different model structures for the Kershbaum spring in Austria, with parameter estimation and uncertainty quantified via the GLUE method. The best-performing model structure is then coupled with PHREEQC to create an initial solute transport model based on the complete mixing assumption accounting for the posterior distributions of the parameter of the selected model structure of LuKARS 3.0.

 

Reference

Bittner, D., Narany, T.S., Kohl, B., Disse, M., and Chiogna, G. (2018). Modeling the hydrological impact of land use change in a dolomite-dominated karst system. Journal of Hydrology 567:267–279.