Mining for ecological thresholds and associations in cytometric data: a coastal management perspective

Abstract

Decision-making in coastal waters management is a complex and interdisciplinary task. Particularly, to find seasonal patterns and ecological thresholds, which are not always clear in tropical areas. Therefore, the ultimate in this activity is to gain knowledge about biogenic element, the biological response, and the selection of indicators which may reveal the trophic status of the system. Under this scenario, this paper applies Data Mining techniques as an alternative approach in order to access hidden patterns of in situ flow cytometry monitoring data. The case studied is the upwelling influenced bay at Cabo Frio Island (Rio de Janeiro-Brazil). A neural network uses phytoplankton and bacterial data of real time monitoring as input variables to forecast marine viruses temporal variability. We also demonstrate that it is possible to access patterns of planktonic community structure in different water masses within a set of association rules.