Importance of Reference Dataset Improvements for Argo Delayed-Mode Quality Control

Abstract

For the Argo Project, monitoring the global upper ocean by a large number of profiling floats, maintaining the quality of salinity data is critical; the goal for measurement accuracy is ±0.01. Experiments using the method of Wong et al. (2003), the standard delayed-mode quality control (dQC) for the Project, show that its performance depends critically on the reference datasets used. This study concludes that the method is useful for Argo and has sufficient potential to achieve the goal for salinity measurement in the North Pacific, when suitable reference datasets are prepared. Considering the Wong et al. (2003) algorithms, we suggest that reference datasets with the following characteristics will be most suitable for Argo dQC: They should be basically derived from the most extensive datasets, such as the latest World Ocean Database; in regions with denser observations, datasets with carefully quality controls should be used; in the regions with subsurface temperature inversions, such as the subarctic North Pacific, the profiles used for the reference must extend below the deepest temperature maximum to prepare proper salinities for the deep layer reference.