Exploring the ecological status of human altered streams through Generative Topographic Mapping A. Vellido a, * , E. Martı ´ b , J. Comas c , I. Rodrı ´guez-Roda c , F. Sabater d a Department of Computing Languages and Systems (LSI), Polytechnic University of Catalonia (UPC), C. Jordi Girona, 1e3, 08034 Barcelona, Spain b Centre d’Estudis Avanc ¸ats de Blanes-CSIC, Acce ´s a la Cala de St. Francesc, 14, 17300 Blanes, Girona, Spain c Chemical and Environmental Engineering Laboratory (LEQUIA), University of Girona, Campus Montilivi s/n, 17071 Girona, Spain d Department of Ecology, University of Barcelona (UB), C. Diagonal, 645, 08028 Barcelona, Spain Received 22 March 2005; received in revised form 12 February 2006; accepted 16 June 2006 Available online 14 August 2006 Abstract The STREAMES (STream REAch Management, an Expert System) European project is an international enterprise for the development of a knowledge-based environmental decision support system to assist water managers with their decision making tasks. It involves the evaluation of the effect of substantial nutrient loads on the overall water quality and ecological status of stream ecosystems. Empirical data for the knowl- edge base come from several streams located throughout Europe and Israel, with emphasis on streams from the Mediterranean region. These data comprise several types of variables, including physical, chemical and biological parameters. The complexity of the data limits the amount and completeness of the available information. This study explores how similar the selected streams are on the basis of the ecological descriptors measured. The analysis of these similarities helps us to ascertain whether the same model of ecological status might hold over the variety of streams under consideration. The available data are explored and analysed through reconstruction, visualization and clustering using the Gen- erative Topographic Mapping (GTM), a neural network-based model. Amongst the many advantages of the probabilistic setting of GTM, one is especially relevant to the problem at hand: its ability to handle and reconstruct missing data in a principled way. Ó 2006 Elsevier Ltd. All rights reserved. Keywords: Generative Topographic Mapping (GTM); Ecological status; Clustering; Missing data; Streams; Nutrient retention Software availability Name of the software: GTM Developer and contact details: corresponding author Year first available: 2004 Program language: MatlabÔ Availability: from the corresponding author, on request 1. Introduction The STREAMES (STream REAch Management, an Expert System) European project is an international enterprise for the development and implementation of a knowledge-based envi- ronmental decision support system (KB-EDSS) to assist the de- cision making of water managers. The KB-EDSS involves: a) the evaluation of water quality and, to a larger extent, the eco- logical status of fluvial ecosystems; b) the examination of pos- sible causes of ecosystem impairment; and c) the proposal of ecologically sound management strategies. These strategies aim beyond water quality improvement, to set the more general concept of optimum ecological status of the stream (Water Framework Directive, 2000; United States Environmental Protection Agency (US EPA), 2000) as their target 1 . In doing * Corresponding author. Fax: þ34 93 413 7833. E-mail address: avellido@lsi.upc.edu (A. Vellido). 1 The Water Framework Directive (WFD: Council of the European Commu- nities, 2000) considers five categories of ecological status: bad/poor/moderate/ good/high. Year 2015 was set as their target to achieve a category of ‘‘good’’ for the ecological status of freshwater and coastal ecosystems in Europe. 1364-8152/$ - see front matter Ó 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.envsoft.2006.06.005 Environmental Modelling & Software 22 (2007) 1053e1065 www.elsevier.com/locate/envsoft