Characterisation and modelling of multi-compartment karst systems by integrated interpretation of spring signals - iKarst

岩溶含水层系统是具有明显各向异性的非均质水文系统。这类系统的特征描述极具挑战性，因为有几个过程尚未完全了解。只要实地调查及其结果没有得到适当的综合建模概念的补充，经典方法（即考虑具体方面的方法组合）的适用性就有限。描述单个岩溶区室（表层带、渗透带和潜水带）的各种工具已经存在，然而，通过基于物理的方法考虑所有相关过程的综合模型仍然缺失。我们的初步工作强调了基于喀斯特特征（输入信号产生、导管-基质相互作用和导管储存过程）的物理建模方面的类似挑战。此外，高参数化模型（模型结构、参数和结果）的非唯一解限制了复杂数值模型的应用。将整个喀斯特系统分离成单独的隔间用于增强现有的基于物理和空间分布特征的分布参数模型。这将导致对流动和运输过程的更好理解。这些工具将与综合建模方法相结合，从而允许通过复杂的逆技术（例如数据加权、目标函数）识别模型的不确定性。理想模型的正向建模允许基于多个弹簧信号（流量，热量和溶质）识别和量化不同的输入信号源。具体过程和参数的相关性通过敏感性分析进行评价。根据信息和数据的可用性，使用多个信号的联合反演来评估所需的模型复杂性。模型复杂度的逐步自适应降低了高参数化复杂岩溶系统模型的计算量。通过自动逆建模工具估计模型的不确定性。最后，这将有助于确定未来的勘探需求，从而促进喀斯特系统的管理。

Karst aquifer systems are heterogeneous hydrological systems with distinctive anisotropy. The characterisation of such systems is extremely challenging, since several processes are yet not fully understood. Classical approaches (i.e. combinations of methods considering specific aspects) are of limited applicability, as long as field investigations and their results are not complemented by appropriate comprehensive modelling concepts. Various tools for describing individual karst compartments (surface zone, vadose zone, and phreatic zone) already exist, however, integrated models considering all relevant processes by physically based approachesare still missing. Our preliminary work highlights similar challenges regarding a physically based modelling of karst features (input signal generation, conduit-matrix interaction, and conduit storage processes). Additionally, non-unique solutions of highly parameterisedmodels (model structure, parameters, and results) limit the application of complex numerical models. The separation of the entire karst system into individual compartments is used to enhance existing distributed-parameter models for a physically based and spatially distributed characterisation. This will lead to improved understanding of flow and transport processes. The tools are to be combined to an integrated modelling approach, which allows to identify model uncertainties through sophisticated inverse techniques (e.g. data weighting, objective functions). Forward modelling of idealised models allows to identify and quantify different input signal sources based on multiple spring signals (flow, heat, and solutes). The relevance of specific processes and parameters is evaluated by sensitivity analysis. Joint inversion of multiple signals is used to evaluate needed model complexity, depending on information and data availability. The stepwise adaption of model complexity reduces computational demands of highly parameterised complex karst system models. Model uncertainties are estimated via automatic inverse modelling tools. Concluding, this will allow to define future exploration demands and, therefore, foster karst system management.