A novel statistical approach for Near-Real Time Quality Control of hydrographic observations

This study concerns a quality check (QC) test in which temperature and salinity observations are compared to the range of variability as known from a reference dataset.

For delayed-time QC, Gourrion et al. (2019) have shown that a validity range built from local minimum and maximum values performs better than the standard one estimated from the local mean +/- N standard deviations. It allows to account for the assymetry and peakedness of the local parameter distribution. The performance of the test is significantly improved, reducing strongly the number of hydrographic stations to be checked by a delayed-time operator.

Nevertheless, for near-real-time applications, the available operator time is severely reduced ; the method needs to be automatized and the number of erroneous detections further reduced.

Here, we first present the impact on the performance brought by the increase of the temporal extent of the reference dataset. Then, an adhoc widening of the validity interval is proposed and shown to further improve the performance, reaching a point of applicability for NRT processing. Finally, performance differences between temperature and salinity are highlighted.