Modular Models Based on Clustering: Application to Precipitation Downscaling from General Circulation Models

It has been shown in the previous chapters that the application of modular models in the context of rainfall-runoff forecast modelling leads to increase in accuracy. This chapter explores the use of the modular approach technique in yet another area related to forecasting downscaling of information from General circulation models (GCM) and/or numerical weather prediction models (NWP). The use of a version of modular approach, fuzzy committee model ( Solomatine, 2006 ; Solomatine and Price, 2004 ) is used along with the automatic clustering techniques, and compared with statistical downscaling models and single neural network models. This chapter is mainly based on the publication presented by Corzo et al. (2009b)