Modeling and prediction of complex environmental systems

Abstract

The last century witnessed the development of numerousapproaches (e.g. linear, nonlinear, deterministic, stochastic,physics-based, data-based) for modeling and prediction ofour complex environmental systems. There is a plethora ofliterature debating whether any one of these approaches issuperior to any other (if at all such an evaluation can bemade), and why and how. While such debates continue, anexamination of environmental literature also reveals theexistence of some serious problems in our modelingpractice (see, for example, Beven 2002), regardless of theapproach adopted. For example: (1) we have a tendency,driven by our technological and methodological advances,to develop more complex models (having too manyparameters and requiring too much data) than that mayactually be necessary; and (2) the models are often devel-oped for specific situations, and their extensions to othersituations and generalizations are normally difficult. Whatis also being increasingly realized is that none of theexisting approaches is adequate for modeling and predic-tion of our environmental systems by itself. With ourincreasing emphasis on specific concepts/methods (see, forexample, Sivakumar 2005) for their ‘individual brilliance’(that normally reflect our extreme and biased views), ratherthan their ‘collective utility’ to address the real environ-mental challenges, the above situation may not changemuch in the future, unless a significant change in paradigmis adopted. The only way to overcome this situation seemsto be to find some ‘common grounds’ in our modelingapproaches, which is the primary motivation for this Spe-cial Issue entitled ‘‘Modeling and Prediction of ComplexEnvironmental Systems.’’ Towards pursuing this, the spe-cific objectives are:1. to disseminate the latest initiatives and developmentsin the modeling and prediction of our environmentalsystems, especially focusing on the applications ofdifferent concepts/methods; and2. to discuss (both philosophically and scientifically) thepotential for the formulation of a ‘middle-ground’approach for more realistic representations of ourenvironmental systems, and also to highlight theimportant challenges in this formulation.This Special Issue consists of 14 papers, contributed bya total of 40 authors. Each paper falls under either of theabove two objectives, and collectively they bring forth theapplications of various approaches/techniques to differentenvironmental systems/problems around the world. Thetechniques applied in these papers include (in the order oftheir appearance), among others: hidden Markov models(Jayawardena et al.; Kwon et al.), wavelet transforms(Jayawardena et al.; Kwon et al.), fractals and multifractals(Cortis et al.), nonlinear dynamics and chaos (Kim et al.;Singh et al.; Sivakumar), artificial neural networks (Aksoyand Dahamsheh; Bagtzoglou and Hossain), nonlinear sto-chastic models (Cayar and Kavvas), numerical methods(Perera et al.), self-organizing maps (Lischeid), statisticalestimation techniques (Bagtzoglou and Hossain; Deanet al.; Vrugt et al.), Bayesian approaches (Bagtzoglou andHossain; Vrugt et al.), and data-based mechanistic models(Young and Ratto). The environmental problems addressedin these papers are, among others: rainfall (Jayawardena