KALMAN CALIBRATION APPROACH TO ASSESS HYDRAULIC IMPACT INDUCED BY THE CONSTRUCTION AND OPERATION OF THE PLANNED IHLRW REPOSITORY - CIGEO (FRANCE)

摘要

Prediction of hydraulic perturbations induced by theconstruction and the operation of deep geologicalradioactive waste repository is need to support: (i) theengineering and monitoring operations, and (ii) theassessment of the consequences on groundwaterresources. Andra (French National Radioactive WasteManagement Agency) has developed a three-dimensionalhigh-resolution numerical model of variably saturatedflow in order:1. to understand the local flow behavior and to improvethe characterization of lithological structures andheterogeneities of the Oxfordian aquifer system atMeuse/Haute-Marne site.2. to assess 3D extension of the perturbation and estimategroundwater volumes drained by underground ramps andshafts.Novel calibration method was applied to reproduce themonitored transient flow and map the heterogeneity of theOxfordian aquifer formation affected by undergroundstructures (shafts). Simulations of the predictive transientflow behavior induced by the construction and theoperation of the planned repository allow the assessmentof the extension of perturbation and the estimation ofdrained water discharge. Robustness of the numericalmodel as well as reliability of the prediction resultsdepend on the quality of the transient flow calibrationbased on inverse approach, called “Iterative EnsembleSmother (IES)” for parameter estimation. By assimilatingall monitoring data at once, the ensemble smootherallows the estimation of poorly known parameters bysampling of an approximate Gaussian posteriordistribution which unknown moments estimated fromupdated realizations. Computation of Kalman gain matrixdepends on covariance matrices approximated from theensemble describing a linear relationship between theupdated realizations and the observed data.