Parameter estimation for reactive transport

The chemical parameters used in reactive transport models are not known accurately due to the complexity and the heterogeneous conditions of a real domain. A reactive transport model needs five kinds of inputs: the hydraulic properties of the domain, the chemical reactions occurring and the chemical parameters, the boundary and initial conditions. We will focus here on the chemical part of the problem. To describe a reactive transport model, a set of chemical reactions is first chosen and after that the corresponding parameters are obtained, whatever the way. Nevertheless, the required chemical parameters are not known exactly. Because the described phenomena are non linear, low precision on the determination of the parameters can lead to rejection of an accurate set of reactions. Moreover it is not possible to determine if the divergence between the experimental results and the calculated one is due to a wrong set of reactions or to a insufficiently precise determination of the parameters. Today, parameter estimation is mainly done using batch experiments under unique experimental conditions with tools like FITEQL. Estimated parameters are then accurate for batch reactor and for a given experimental state such as imposed pH or fixed ionic force. Extrapolating these parameters to a natural uncontrolled systems is then very hazardous. We will present the development of an efficient algorithm in order to estimate the chemical parameters using Monte-Carlo method. The objective is to estimate these parameters using multi-conditional experiments. It will then be possible to make predictions under variable conditions (pH, ionic strength) which are more representative of natural systems. (authors)