Analysis of the Critical Aspects Associated to Automatic Parameter Identification of Complex Groundwater Flow Systems

The successful use of automatic parameter estimators for calibrating groundwater flow models is strongly conditioned by computational aspects. Since the total number of computations involved in the calibration is influenced by several facts, we consider that minimizing the computer and modeler time involved in the task requires identify the critical aspects associated to the calibrations process. In this sense, the experience got performing models such as El Cabril 6 or El Berrocal 7 allows us to identify these aspects which define the final cost of the calibration in terms of CPU and modeler time. They are: 1intrinsic complexity of the system; 2amount of information available about the system; 3numerical and hydrogeological background of the modeler; 4quality of pre and post processors; 5characteristics of the automatic parameter estimator; 6selection of the space-time discretization and 7computer characteristics. It is then clear that the key codescomputer-modeler defines scope and quality of the expected model. In this paper, we analyze and discuss every one of the aspects mentioned before, pointing out the critical factors associated to the modeling process. One of our conclusions is that relatively big numerical errors can be admitted during a part of the calibration process. This allows save many computations which are not strictly necessary. On the other hand, we show that although the computer velocity is important, it is only one of the factors (in fact, under certain conditions it may be despicable) governing the global cost of the calibration and it is rarely the most important. German Galarza, Pedro Tume and Jesús Carrera