Application of multivariate modelling and analysis on an integrated hydrocarbon sensor array system for marine surveys

Abstract

Commercial hydrocarbon sensors play key roles in marine surveys for the detection of dissolved hydrocarbons by providing real time hydrocarbon distribution information. However their full potentials have not been realized with the current commonly used sensor deployment and analysis method where each sensor functions in isolation and only provide very limited information such as relative concentration of certain compound. We recently designed and constructed an array of commercially available hydrocarbon sensors. The system was tested by exposure to 110 diluted water solutions of oil-water extracts in random order. The sensor array data matrix collected from the testing and the analyte solutions' composition matrix was processed with the PLS (Partial Least Squares) regression. The result indicates that from the analyte solution's triggered sensor response pattern, the concentration of individual dissolved oil components, light mono-aromatics in particular, can be predicted with acceptable accuracy.