Virtual experimentation on cyanobacterial bloom dynamics and its application to a temperate reservoir (Torrão, Portugal)
The objective of this study was to use artificial neural networks to develop a regression model capable of simulating the complexity of events contributing to the development of cyanobacterial blooms in an important potable water supply located in northern Portugal (Torrão Reservoir). This model was produced with environmental variables collected from the reservoir over a 40-month sampling period. For development of the model, 10 variables were selected from a total group of 18, all with an equal contribution in the final model. The correlations between the predicted and observed values were high (0.970, 0.912 and 0.908 in the training, verification and test sets, respectively). These variables could be grouped into three categories based on their influence on cyanobacterial abundance. In order of decreasing association, these classes were as follows: (i) oxygen stratification; (ii) conductivity, discharge, and minimum air temperature, water evaporation and solar radiation; and (iii) precipitation, phosphate, ammonia, and pH.
Document Type: Research Article
Affiliations: CIIMAR – Centro Interdisciplinar de Investigação Marinha e Ambiental, Porto, Portugal
Publication date: 2008-06-01