Abstract:In this article, we perform a parameter study for a recently developed karst hydrological model. The study consists of a high-dimensional Bayesian inverse problem and a global sensitivity analysis. For the first time in karst hydrology, we use the active subspace method to find directions in the parameter space that dominate the Bayesian update from the prior to the posterior distribution in order to effectively reduce the dimension of the problem and for computational efficiency. Additionally, the calculated active subspace can be exploited to construct sensitivity metrics on each of the individual parameters and be used to construct a natural model surrogate. The model consists of 21 parameters to reproduce the hydrological behavior of spring discharge in a karst aquifer located in the Kerschbaum spring recharge area at Waidhofen a.d. Ybbs in Austria. The experimental spatial and time series data for the inference process were collected by the water works in Waidhofen. We show that this case study has implicit low-dimensionality, and we run an adjusted Markov chain Monte Carlo algorithm in a low-dimensional subspace to construct samples of the posterior distribution. The results are visualized and verified by plots of the posterior's push-forward distribution displaying the uncertainty in predicting discharge values due to the experimental noise in the data. Finally, a discussion provides hydrological interpretation of these results for the Kerschbaum area.

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is to conduct a parameter study on the karst aquifer model. Specifically: 1. **High - dimensional Bayesian inverse problem**: The Bayesian calibration method is used in the paper to estimate 21 parameters in the karst hydrological model. Since these parameters form a high - dimensional space, traditional methods for solving the Bayesian inverse problem are very computationally expensive and inefficient. Therefore, the active subspace method is introduced in the paper to reduce the dimension of the parameter space and improve computational efficiency. 2. **Global sensitivity analysis**: In addition to parameter estimation, the paper also conducts a global sensitivity analysis to determine which parameters have the greatest impact on model outputs (such as spring flow). Through this method, the importance of each parameter in the model can be better understood and a basis can be provided for further model improvement. ### Main methods and techniques - **Active subspace method**: This is a technique used to identify the dominant directions in the parameter space, which have the greatest impact on the Bayesian update from the prior distribution to the posterior distribution. By finding these dominant directions, high - dimensional problems can be effectively transformed into low - dimensional problems, thereby significantly reducing the computational cost. - **Bayesian inverse problem**: The posterior distribution of model parameters is estimated using the Bayesian framework. Specifically, by combining prior knowledge and observational data (such as time - series data of spring flow), the posterior probability distribution of the parameters is calculated. This helps to quantify the uncertainty in parameter estimation. - **Global sensitivity analysis**: Based on the active subspace method, the global sensitivity index of each parameter is calculated. These indices can help identify which parameters have the greatest impact on model outputs, thereby guiding model optimization and parameter selection. ### Application cases Taking the Kerschbaum spring in the Waidhofen a.d. Ybbs area of Austria as an example, the paper applies the above - mentioned methods to conduct parameter estimation and sensitivity analysis on the LuKARS model. The LuKARS model is a lumped - parameter model used to simulate the impact of land - use change on spring flow in the karst environment. This model contains four main hydrotopes, and each hydrotope has 7 parameters that need to be calibrated. ### Results and discussion - **Dimension reduction**: Through the active subspace method, the 21 - dimensional parameter space is successfully reduced to a lower dimension, significantly improving the computational efficiency. - **Parameter sensitivity**: The results of the global sensitivity analysis show that some parameters have an important impact on the prediction of spring flow, while the impact of other parameters is relatively small. - **Model verification**: The validity and reliability of the model are verified by comparing the predicted values of the posterior distribution with the actual observational data. In conclusion, by introducing the active subspace method, this paper effectively solves the problems of high - dimensional parameter estimation and global sensitivity analysis in the karst hydrological model, providing important scientific support for karst water resource management.

Bayesian calibration and sensitivity analysis for a karst aquifer model using active subspaces

Temporal scale‐dependent sensitivity analysis for hydrological model parameters using the discrete wavelet transform and active subspaces

On the value of hydrochemical data for the interpretation of flow and transport processes in the Baget karst system, France

A data-driven approach for modelling Karst spring discharge using transfer function noise models

Combining noisy well data and expert knowledge in a Bayesian calibration of a flow model under uncertainties: an application to solute transport in the Ticino basin

Assessing streamflow sensitivity to precipitation variability in karst‐influenced catchments with unclosed water balances

Efficient parameter estimation for a methane hydrate model with active subspaces

Combining implicit geological modeling, field surveys, and hydrogeological modeling to describe groundwater flow in a karst aquifer

Advance in Karst Hydrological Model

Investigating Spatial Heterogeneity of Karst Water Storage Capacity and Nonclosure of Underground Watersheds in Karst Hydrological Simulation

Taking into Account both Explicit Conduits and the Unsaturated Zone in Karst Reservoir Hybrid Models: Impact on the Outlet Hydrograph

Conditional Korhunen-Loéve regression model with Basis Adaptation for high-dimensional problems: uncertainty quantification and inverse modeling

Assessing current and future available resources to supply urban water demands using a high-resolution SWAT model coupled with recurrent neural networks and validated through the SIMPA model in karstic Mediterranean environments

Bayesian Poroelastic Aquifer Characterization from InSAR Surface Deformation Data. Part I: Maximum A Posteriori Estimate

Data-driven estimates for the geostatistical characterization of subsurface hydraulic properties

Karst spring discharge modeling based on deep learning using spatially distributed input data

A comprehensive framework for stochastic calibration and sensitivity analysis of large-scale groundwater models

Quantifying the Role of Calibration Strategies on Surface‐Subsurface Hydrologic Model Performance

Modeling of groundwater flow and transport in coastal karst aquifers

V2Karst V1.1: a parsimonious large-scale integrated vegetation–recharge model to simulate the impact of climate and land cover change in karst regions

Integrating field work and large-scale modeling to improve assessment of karst water resources