Optimising oil spill dispersants as a function of oil type and weathering degree: a multivariate approach using partial least squares (PLS)

Abstract

This is last of three papers concerning multivariate optimisation of oil spill dispersants. Dispersants are used in oil spill response operations to enhance the natural dispersion of an oil slick at sea as small oil droplets in the water column. The first paper in this series proposes a multivariate approach for dispersant optimisation based on simulations with different experimental designs. The second paper verifies the usefulness of this approach using real laboratory data. This multivariate approach is based on designed experiments and response surface methods and represents a new approach within dispersant development. The work described in this third paper shows how the PLS (Partial Least Squares) algorithm can be used to predict optimised dispersant composition as a function of oil type and degree of weathering. This is done by characterisation of the oil type and weathering degree by principal component analysis (PCA). Score values from the first and second principal component are used to select oil type and weathering degree for the calibration samples. Together with selected surfactants the score values are used as parameters for a new 2 5−1 fractional factorial design. The data from this factorial design are used as a calibration set for predicting optimal dispersant composition as a function of oil type and weathering degree. The experimental design used in this study (simplex–centroid for response surface modelling and fractional factorial design) combined with PLS modelling has made it possible to gain new basic knowledge concerning optimal dispersant composition for different oil types and degrees of weathering. The final optimised dispersant was verified to have a high effectiveness on a broad selection of oil types and a low toxicity. It also had the highest effectiveness and the lowest toxicity when compared to a selection of commercially available products.