Burial driven recycling is an important process in the natural gas hydrate (GH) systems worldwide, characterized by complex multiphysics interactions like gas migration through an evolving gas hydrate stability zone (GHSZ), competing gas-water-hydrate (i.e. fluid-fluid-solid) phase transitions, locally appearing and disappearing phases, and evolving sediment properties (e.g., permeability, reaction surface area, and capillary entry pressure). Such a recycling process is typically studied in homogeneous or layered sediments. However, there is mounting evidence that structural heterogeneity and anisotropy linked to normal and inclined fault systems or anomalous sediment layers have a strong impact on the GH dynamics. Here, we consider the impacts of such a structurally complex media on the recycling process. To capture the properties of the anomalous layers accurately, we introduce a fully mass conservative, high-order, discontinuous Galerkin (DG) finite element based numerical scheme. Moreover, to handle the rapidly switching thermodynamic phase states robustly, we cast the problem of phase transitions as a set of variational inequalities, and combine our DG discretization scheme with a semismooth Newton solver. Here, we present our new simulator, and demonstrate using synthetic geological scenarios, a) how the presence of an anomalous high-permeability layer, like a fracture or brecciated sediment, can alter the recycling process through flow-localization, and more importantly, b) how an incorrect or incomplete approximation of the properties of such a layer can lead to large errors in the overall prediction of the recycling process.

Global sensitivity analysis of model parameters is an important step in the development of a hydrological model. If available, time series of different variables are used to increase the number of sensitive model parameters and better constrain the model output. However, this is often not possible. To overcome this problem, we coupled the active subspace method with the discrete wavelet transform. The Haar mother wavelet is the most appropriate for this purpose in case of homoschedastic measurement error, since it avoids any loss of information through the discrete wavelet transform of the signal. With this methodology, we study how the temporal scale dependency of hydrological processes affects the structure and dimension of the active subspaces. We apply the methodology to the LuKARS model of the Kerschbaum spring discharge in Waidhofen a.d. Ybbs (Austria). Our results reveal that the dimensionality of an active subspace increases with increasing hydrologic processes which are affecting a temporal scale. As a consequence, different parameters are sensitive on different temporal scales. Finally, we show that the total number of sensitive parameters identified at different temporal scales is larger than the number of sensitive parameters obtained using the complete spring discharge signal. Hence, instead of using multiple data time series to identify more sensitive parameters, we can also obtain more information about parameter sensitivities from one single, decomposed time series.