Formal representation and diagnostics of environmental systems

Abstract

Statistical methodologies appropriate to certain interesting study areas of environmental systems are presented. Three approaches are described which are useful for: i) the pooling of the multi-source information coming from a number of stations, i-e., the simultaneous formalization of real and/or simulated data together with the quantification of expert knowledge using a globally informative probability density function.ii) the prediction of the class to which an environmental variable belongs among several pre-established classes. Two non-parametric discriminant approaches are presented; both were derived from projection pursuit techniques.iii) the modelling of the time evolution of dichotomic environmental variables; the modelling is focused on statistical inference methods, such as estimation of the probabilities of state transitions and of the stay in a particular state.