The application of a quantile regression metamodel for salinity event detection confirmation within New York Harbour oceanographic data

This paper presents a continuation of research regarding the utilisation of robust metamodels for uncertainty quantification and event detection within a geophysical system. Using salinity data supplied by the New York Harbour Observing and Prediction System (NYHOPS) for two test datasets and three actual sensors, event detection results from a static threshold method are compared against those of a dynamic uncertainty quantification-based technique and a composite technique that combines both the static and dynamic methods. The results clearly show an appreciable reduction in the number of false positive detections when using the composite event detection method; in test data void of salinity events, false detection rates for low salinity conditions decreased by as much as 80%.