An Automatic Recession Extraction and Parameter Estimation Method for Large-Sample Karst Spring Analysis

Abstract:

Groundwater from karst aquifers serve as major drinking water sources for significant fraction of the earth’s population. Karst aquifers are characterized by complex groundwater flow systems due to the interplay of fast and slow flow and storage processes. Hence, its management requires extensive knowledge of this complex and dynamic hydrogeological behavior. As an integral part of the aquifer, karst spring hydrographs provide good insights about hydrological functioning of the entire aquifer system. Quantitative recession analysis, which is usually manually applied to karst spring hydrographs, is a vital tool for understanding flow dynamics, aquifer characterization and water resource management. In our study, we adapted a range of automated recession extraction methods that were originally developed for streamflow recession and apply them to karst spring hydrographs. Hypothesizing that conduit and matrix drainage produce distinct recessions, we estimate the karst conduit and matrix recession parameters by fitting karst-specific storage-outflow models to automatically extracted recession segments. We evaluate the robustness of recession parameters by using the different extraction techniques within a Monte Carlo uncertainty framework. To test its general applicability, we apply our approach on three karst hydrographs obtained from karst springs with different degrees of karstification and located in different climate regions. Our comparison shows that different starting values of the extracted recessions, which vary among the different extraction techniques, have a strong control on recession parameters estimation, most particular for the conduit recession parameters. However, the Monte Carlo framework is able to account for these uncertainties to some extent providing more robust estimates of conduit and matrix recession coefficients compared to previous recession analysis approaches that did not consider uncertainty. That way, our approach provides new insights for automatic recession analysis of karst springs. While providing the first automatic tool to analyses karst spring recessions, it also emphasizes the uncertainties associated with estimated recession parameters, which will be particularly useful information for groundwater modeling and water resources management.