Dataset Consistency


THE PROJECT DATA SERVICE DEPLOYMENT	INSTRUMENTATION ACQUISIT° FROM RESEARCH VESSELSAPPLICATIONS & PRODUCTS

	Home > Applications & products > Delayed Mode QC
	Dataset Consistency

Work in progress

The statistic method proposes a salinity correction comparing in situ data with mainly CTD data. Therefore floats are individually analyzed, without considering other drifting profilers. To validate the delayed-mode correction, we first have to test the consistency of the dataset including calibrated floats.

Among the floats available at CORIOLIS data center, 77 with delayed-mode values were chose to perform this study (click here to see the list and the applied correction)

This test is fulfilled applying the objective analysis method. But instead of using raw data (pressure, temperature and salinity) of these 77 floats in standardized values, we now consider their adjusted fields. The study was focused on the North Atlantic Ocean (20°N - 70°N) during the 2001-2005 period.

Fields analysis

The first step was to compare salinity fields from objective analysis including delayed-mode data to salinity fields including only real-time data and to look at the differences.

This evolution of salinity fields (due to the addition of delayed-mode data) is localized in space and time. These changes occur near profiles that were corrected but not systematically. Indeed, some profiles do not modify salinity fields while some do specially in regions with salinity fronts. Besides, concerning numerical values of these differences, they decrease with depth and tend towards zero in bottom layers.


	In order to detect a trend in salinity fields evolution, means have been computed (year by year) and over the 2001-2005 period). We observe the same results as for weekly comparisons: localized differences (up to 0.06 psu in 2004) that decrease in proportion to depth.

Although we observe a difference in salinity fields, that does not necessarily imply that there is an error with the applied correction. To check it, we just have to analyze the evolution of salinity fields in parallel to floats residuals values. If residuals means are low, it means that delayed-mode values are consistent with close CTD data and therefore must be considered as new information in the region. On the contrary, if these residuals values are high, delayed-mode data are then in contradiction with historical data and the float must be reanalyzed. Studies have been made with two different approaches:

on the one hand, we followed the evolution of salinity fields and residuals values in a specific area (a 4°*4° zone) :

on the other hand, we followed the evolution of salinity fields and residuals values along a float displacement (in a 4°*4° zone around the profile location):

In both cases, we saw salinity differences but law values of residuals means tend to prove the consistency of the dataset.

Residuals

To confirm the validation of the dataset, we also looked at all the floats residuals and anomalies (difference between in situ data and climatology) over the bottom layers. The result is rather good since all the drifts have been obviously corrected and the residuals values are smaller than those from analysis with real-time data.

This is confirmed when we calculate the mean of the residuals for the 77 floats. Their mean is lower than the one we got from real-time residuals. Besides the value of the mean is very often inferior to 0.01 psu proving so the consistency of the dataset.