Computergestützter Ansatz zur Kalibrierung und Validierung mathematischer Modelle für Strömungen im Untergrund - COMPU-FLOW

• Antragsteller: Professor Dr. Harrie-Jan Hendricks-Franssen; Professor Dr.-Ing. Wolfgang Nowak
• Fachliche Zuordnung: Hydrogeologie, Hydrologie, Limnologie, Siedlungswasserwirtschaft, Wasserchemie, Integrierte Wasserressourcen-Bewirtschaftung
• Förderung: Förderung von 2017 bis 2022
• Projektkennung: Deutsche Forschungsgemeinschaft (DFG) - Projektnummer 359880532

• Erstellungsjahr: 2024

Zusammenfassung der Projektergebnisse

Forecasting phenomena such as groundwater flow and the spread of contaminants in the subsurface is burdened by substantial uncertainties. These uncertainties primarily stem from the inherent heterogeneity of the subsurface, making it practically impossible to characterize and predict the subsurface in full detail and with high confidence. Improving subsurface characterization and mitigating remaining uncertainties requires calibrating flow and transport models based on observed state variables. Stochastic (geostatistical) inverse modeling techniques are indispensable to handle these uncertainties, surpassing conventional model calibration approaches. A plethora of stochastic inverse methods can be found in the scientific literature. However, a comprehensive assessment of their relative strengths and limitations remains absent. This lack poses a formidable challenge to ongoing and future improvements in inverse modeling techniques. The crux of this issue lies in the absence of well-defined benchmark definitions that allow systematically evaluating and comparing diverse methods under controlled, standardized, and reproducible conditions. Accordingly, this project aimed to establish a set of benchmark scenarios endowed with highly precise reference solutions. These benchmark scenarios, along with the associated reference and compared solutions, have been made accessible to the research community for long-term utility. Furthermore, we worked towards a comprehensive community-wide comparative study, utilizing these benchmarks and reference solutions, and provided a set of benchmarking metrics that standardize how benchmarking results are assessed. The proposed benchmark cases encompass fully saturated, transient groundwater flow, cases with low and high spatial variability, and hydraulic conductivity fields characterized by multi-Gaussian distributions. We computed the corresponding reference solutions using specialized algorithms developed in this project, leveraging the preconditioned Crank-Nicholson variant of Markov Chain Monte Carlo. We equipped this framework with adaptive proposal distributions, multi-tempered parallel chains, and extensions tailored for non-multi-Gaussian distributions. Also, we adapted these algorithms for high-performance computing infrastructure to harness the computational power required. A dedicated workshop with the groundwater inverse modeling community was held to finalize the metrics definition, select benchmarking scenarios, and incentivize the comparative study. At the start of the project, a consortium of 12 internationally recognized research groups had already committed to participate.

Projektbezogene Publikationen (Auswahl)

• Simultaneous identification of a contaminant source and hydraulic conductivity via the restart normal-score ensemble Kalman filter. Advances in Water Resources, 112, 106-123.
  Xu, Teng & Gómez-Hernández, J. Jaime
  
• Bayesian Inversion of Hierarchical Geostatistical Models using a tailored MCMC Algorithm, AGU Fall Meeting 2019. San Francisco, CA, DEC. 9-13
  Sebastian Reuschen, Teng Xu & Wolfgang Nowak
  
• Bayesian inversion and visualization of hierarchical geostatistical models. Copernicus GmbH.
  Reuschen, Sebastian; Xu, Teng & Nowak, Wolfgang
  
• Bayesian inversion of hierarchical geostatistical models using a parallel-tempering sequential Gibbs MCMC. Advances in Water Resources, 141, 103614.
  Reuschen, Sebastian; Xu, Teng & Nowak, Wolfgang
  
• High-end solution techniques and accurate reference solutions: towards a community-wide benchmarking effort for stochastic inverse modeling of groundwater flow. Copernicus GmbH.
  Nowak, Wolfgang; Xu, Teng; Reuschen, Sebastian; Hendricks, Franssen Harrie-Jan & Guadagnini, Alberto
  
• Sampling behavioral model parameters for ensemble-based sensitivity analysis using Gaussian process emulation and active subspaces. Stochastic Environmental Research and Risk Assessment, 34(11), 1813-1830.
  Erdal, Daniel; Xiao, Sinan; Nowak, Wolfgang & Cirpka, Olaf A.
  
• Towards a community-wide effort for clean benchmarking in geostatistical inversion, AGU Fall Meeting 2020, Online, DEC. 1-17
  Wolfgang Nowak, Sinan Xiao, Teng Xu, Harrie-Jan Hendricks Franssen & Alberto Guadagnini
  
• Bayesian Inversion of Multi‐Gaussian Log‐Conductivity Fields With Uncertain Hyperparameters: An Extension of Preconditioned Crank‐Nicolson Markov Chain Monte Carlo With Parallel Tempering. Water Resources Research, 57(9).
  Xiao, Sinan; Xu, Teng; Reuschen, Sebastian; Nowak, Wolfgang & Hendricks, Franssen Harrie‐Jan
  
• Efficient Discretization‐Independent Bayesian Inversion of High‐Dimensional Multi‐Gaussian Priors Using a Hybrid MCMC. Water Resources Research, 57(8).
  Reuschen, Sebastian; Jobst, Fabian & Nowak, Wolfgang
  
• Global sensitivity analysis of a CaO/Ca(OH)2 thermochemical energy storage model for parametric effect analysis. Applied Energy, 285, 116456.
  Xiao, Sinan; Praditia, Timothy; Oladyshkin, Sergey & Nowak, Wolfgang
  
• Novel MCMC methods for Bayesian inference of spatial parameter fields. Copernicus GmbH.
  Reuschen, Sebastian; Xu, Teng; Jobst, Fabian & Nowak, Wolfgang
  
• Experimental evaluation and uncertainty quantification for a fractional viscoelastic model of salt concrete. Mechanics of Time-Dependent Materials, 27(1), 139-162.
  Hinze, Matthias; Xiao, Sinan; Schmidt, André & Nowak, Wolfgang
  
• Optimal design of experiments to improve the characterisation of atrazine degradation pathways in soil. European Journal of Soil Science, 73(1).
  Chavez, Rodriguez Luciana; González‐Nicolás, Ana; Ingalls, Brian; Streck, Thilo; Nowak, Wolfgang; Xiao, Sinan & Pagel, Holger
  
• Reliability sensitivity analysis based on a two-stage Markov chain Monte Carlo simulation. Aerospace Science and Technology, 130, 107938.
  Xiao, Sinan & Nowak, Wolfgang
  
• Towards a community-wide effort for benchmarking in subsurface hydrological inversion: benchmarking cases, high-fidelity reference solutions, procedure and a first comparison. Copernicus GmbH.
  Xu, Teng; Xiao, Sinan; Reuschen, Sebastian; Wildt, Nils; Hendricks, Franssen Harrie-Jan & Nowak, Wolfgang