applications, and plays an important role in risk assessment in relation to detection and prediction of extreme values or threshold exceedances. Given the intermittent character of extremal events, a model-based dynamic selection incorporating the historical sample information constitutes a natural way of constructing a time-adaptive monitoring network.
In this context, we formulate a multi-objective optimization criterion, de ned in terms of a convex linear combination involving an entropy measure of the information contained in the data and a penalty function related to the network structure. The criterion proposed is applied to some numerical examples where di erent models for the spatio-temporal dependence as well as for the local variability are assumed.
This work has been supported by grants P05-FQM-00990, Andalusian CICYE, and MTM2005-08597, DGI, Spain