Multivariate experimental methodology applied to the calibration of a Clark type oxygen sensor

The response of a Clark type oxygen sensor, designed for the study of benthic fluxes in the marine environment, was evaluated with respect to variations in oxygen concentration, temperature, pH, salinity, total pressure, fugacity ([O2]water/[O2]satwater) and stirring rate by means of an experimental design evaluated by partial least squares (PLS). Fugacity and temperature not only explained a major part of the response, but also stirring-rate and the cross term temperature∗fugacity were significant on the 95% level. With the obtained results, a model for predicting absolute oxygen concentrations from the sensor response was created. To verify the practical applicability of the model it was used to predict oxygen concentrations from sensor data obtained in situ during a cruise in the Skagerrak (north-eastern North Sea). Predicted concentrations were then compared, and found to be in accordance (within 5 %) with concentrations obtained from Winkler titrations. The results suggest that fugacity and temperature are the most important factors to consider while using this, and probably other, brands of Clark type oxygen macro sensors.