Séminaire du LHyGeS
Intervenant : Alberto Guadagnini, Professeur - Dipartimento di Ingegneria Civile e Ambientale, Politecnico di Milano
Titre : Ensemble Kalman filter assimilation of transient groundwater flow data via stochastic moment equations.
Lieu : EOST, 1 rue Blessig, amphi. 2
Résumé : The ensemble Kalman filter (EnKF) enables one to assimilate newly available data in transient groundwater and other temporal earth system models through real-time Bayesian updating of system states (e.g., hydraulic heads) and parameters (e.g., hydraulic conductivities). It has become common to treat spatially varying hydraulic conductivities as autocorrelated random fields conditioned on measured conductivities and/or heads. Doing so renders the corresponding groundwater flow equations stochastic. Assimilating data in such equations via traditional EnKF entails computationally intensive Monte Carlo (MC) simulation. We illustrate a methodology to circumvent the need for MC. Our methodology is grounded on (1) an approximate direct solution of nonlocal (integrodifferential) equations that govern the space-time evolution of conditional ensemble means (statistical expectations) and covariances of hydraulic heads and fluxes and (2) the embedding of these moments in EnKF. This provides sequential updates of conductivity and head estimates throughout the space-time domain of interest and, as an additional benefit, obviates the need for computationally intensive batch inverse solution of the moment equations as we have been doing previously. We compare the accuracies and computational efficiencies of our new EnKF approach based on stochastic moment equation and of the traditional Monte Carlo approach. We do so for synthetic scenarios involving a pumping well operating in a two-dimensional randomly heterogeneous porous medium, paying special attention to the effects of number and quality of available data.