Please note that this session was withdrawn and is no longer available in the respective programme. This withdrawal might have been the result of a merge with another session.
HS8.2.8 | Bayesian Methods for Parameter Inference, Uncertainty Quantification, Error Modelling, Model Learning and Model Choice in Hydrology
Bayesian Methods for Parameter Inference, Uncertainty Quantification, Error Modelling, Model Learning and Model Choice in Hydrology
Convener: Thomas Wöhling | Co-conveners: Anneli GuthkeECSECS, Tobias Karl David Weber, Wolfgang Nowak
Predictions of physical, chemical, and biological processes in soils, aquifers, rivers and across compartments (e.g., stream-aquifer interactions) are strongly affected by uncertainties and errors in model structure, parameters and forcing data. Thus, estimating optimal parameter sets, quantifying uncertainty, correcting for errors, (machine)learning new models and evaluating competing models are crucial for reliable predictions at any scale (lab/field/catchment). We invite contributions on improved concepts, approaches and computational algorithms in these areas that are based on (but not limited to) Bayesian methods, especially on:

- applications and new methods for parameter inference, inverse modelling and data-model fusion,
- robust quantification of predictive uncertainty for model surrogates and machine learning (ML) models,
- novel theories, concepts or frameworks that diagnose, detect and resolve modelling errors and capture conceptual model uncertainty,
- approaches to bring (and extract) sound scientific reasoning into machine learning, be it for model replacement, model error correction or data analysis,
- constraint simulation techniques and novel likelihood formulations as methods to incorporate expert knowledge and soft information into the parameter inference process,
- approaches to define meaningful priors for ML techniques in hydro(geo)logy,
- data worth and optimal experimental design strategies to identify and maximize information from measurements and minimize prediction uncertainty,
- new measurement technologies (from point-scale measurements to remote sensing data and “soft information”) that aid developing or parameterizing models, reduce uncertainties or can help diagnose and detect structural model deficiencies,
- applications and benchmarking efforts in any of these fields.