Automatic calibration of CODESA-3D using PEST

Immagine di anteprima
Lecca, Giuditta
Titolo del periodico
Titolo del volume
We describe here our experience in using the Model Independent Pameter ESTimation (PEST) free software tool [Doherty, 2002] to perform the automatic calibration of the COupled DEnsity-dependent variably SAturated flow and miscible transport (CODESA-3D) groundwater model [Gambolati et al., 1999]. Generally speaking, calibration of a model requires that a suitable method of spatial parameter characterization be defined in order to adjust model parameters until model outputs correspond well to specific laboratory and/or field measurements of the system which is simulated. In particular, for groundwater models the adjustable parameters are usually given by main hydrogeological properties (e.g. hydraulic permeability) and/or system excitations (e.g. abstraction volumes) while control data are represented by piezometric heads and/or salt concentrations measured in the field. Model calibration is a complex task. To perform it for a 3D fully-distributed physically-based hydrological model we need to build up a chain of interdependent software tools and data through the interdisciplinary expertise of GIS experts, modelers and hydrogeologists (Figure 1). The newly generated optimization model is comprised by the two pieces of software CODESA-3D and PEST with the latter wrapping the former up. The optimization model is not restricted in its use solely to the calibration of the groundwater model, through this tool modeler can gain valuable insight into the strengths and weakness of the input dataset allowing future data gathering to be undertaken in an optimal manner. In addition, lessons learned will be applicable also to the estimation of the degree of uncertainty associated with a given calibrated model prediction and to make decisions regarding appropriate levels of model complexity. In the following we discuss in detail the optimization model development and test using synthetic observations generated by the groundwater model itself.
model parameter estimation , optimization tool , groundwater model