Larval presence prediction through logistic regression: an early warning system against Mytilopsis leucophaeata biofouling

Abstract

Mytilopsis leucophaeata is a biofouling bivalve causing major problems in the cooling water system of BASF, Antwerp NV, Belgium, a large water-using industrial facility. This study aimed to develop a statistical model to predict the response of M. leucophaeata larvae to environmental conditions in estuarine ecosystems. Multiple logistic regression, taking into account temporal autocorrelation, was applied on a large dataset allowing the prediction of the probability of occurrence of M. leucophaeata larvae at BASF NV as a response to the environmental variables. The final model made it possible to predict larval presence in the water column solely by monitoring water temperature. The results from subsampling indicated that the model was stable. The model was tested with 2005 data, demonstrating a 98% precise prediction of the occurrence of M. leucophaeata larvae in the water column, with a sensitivity of 100% and a specificity of 97%, even though autumn 2005 was exceptionally warm, which led to an extended presence of the larvae.

Keywords:

• Mytilopsis leucophaeata
• biofouling control
• population dynamics
• larvae
• logistic regression
• predictive statistics

Acknowledgements

The first author was financially supported by a BOF-project (contract 011D13503) of Ghent University and especially appreciates the logistic and financial support of BASF, Antwerp NV and ONDEO Nalco (Contract d.d. 21/10/2001). This research is part of the UGent project GOA 01G00705. We would also like to thank the chemical measuring network MWTL (The Netherlands) for the use of their chlorophyll a data.