Correcting method of Argo data based on HydroBase I
― Introduction of Potential Conductivity ―

Taiyo Kobayashi1, Yasuko Ichikawa2, Yasushi Takatsuki2, Toshio Suga1, Naoto Iwasaka1, Kentaro Ando2, Keisuke Mizuno2, Nobuyuki Shikama1, and Kensuke Takeuchi1


New method to correct conductivity data measured by Argo float is developed. In this method, many climatological profile data of HydroBase can be used as the reference to correct Argo data by introducing a new variable named "potential conductivity". The corrected Argo data by this method become much better in comparison with the CTD data independently observed by R/V Mirai at almost the same time and location. Also the corrected data suggest that this method does not strictly depend on the quality and quantity of the climatological data.

Key words: Argo project, Quality Control, HydroBase, Conductivity sensor correction

1: Frontier Observational Research System for Global Change, Climate Variations Observational Research Program
2: Japan Marine Science and Technology Center, Ocean Research Development

In this study, we introduce a correcting-method of Argo (salinity) data, which is planed to be used in Japan Argo delayed-mode database in JAMSTEC/FORSGC.

Firstly, we show the horizontal movements of two Argo floats, WMO 29032 and 29033, used in this study. They were deployed in March 2000 and drifted in the sea for more than one year.

Fig. 1: Horizontal distribution of locations observed by the Argo floats (WMO 29032 and 29033) with climatological bottle sampling data. Red circles represent stations of WOCE-CTD data. Note that the location of the profile #11 of WMO29033 was not fixed.

Examples of the salinity/conductivity sensor drift equipped with Argo floats.

Example 1: Sudden drift (WMO 29033)
Salinity drift occurred between profiles #10 and #12 when it was drifting near sea surface for about 3 months in summer due to the insufficiency of its buoyancy (here after "summer vacation").

Fig. 2: T-S diagram of WMO 29033 observations. The red and blue circles represent the profiles #1-10 and #12-19, respectively.

Example 2: Slowly drift (WMO 29032)
Salinity drift became large between profiles #8 and #11. This float also took 3 moths "summer vacation", and its influence to the sensor should appear the data between profiles #6 and #7. Its sensor, however, was hardly damaged at that time, but about 2 months later. The sensor drifting was very slowly, and continued for about 4 months.

Fig. 3: T-S diagram of WMO 29032 observations. The red, blue, and green circles represent the profiles #1-8, #9-10, and #11-13, respectively

Development of correcting method

Detect of conductivity sensor drift

Argo floats observe temperature, conductivity, and pressure of seawater, and then calculate salinity and transmit its value to us via ARGOS system. So it is considered that we can obtain better-corrected Argo salinity data by correcting its conductivity values.

Conductivity observed by Argo floats is calculated with the following equation:
C = C( S, T, P)
S: salinity
T: in situ temperature
P: in situ pressure

In this study, we use conductivity ratio R, which is written with C as follows:
R = C( S, T, P) / C( 35, 15, 0 ).

Here after, it is written as
R = sal78( S, T, P). .......(1)

Firstly, we examine that sensor drift can be detected in the conductivity data.

However, clear differences between the data before and after of the "summer vacation" cannot be detected in theta-R diagram of WMO 29033.

Fig. 4: Relation of potential temperature and conductivity ratio for the data observed by WMO29033. Red and blue circles represent the data before and after "summer vacation", respectively.

The conductivity ratio is also affected by pressure, so we plotted the Argo data at the isobar surfaces. At each isobar, the data are almost arrayed on two lines of before and after "summer vacation", and the two data groups are clearly discriminated. However, the lines of the before/after "summer vacation" in different isobars are not arrayed on one line, therefore the difference of data drift can not be detected in the whole view (see Fig. 5b).

Following to the relation of in situ temperature and potential temperature, we introduced a new value "potential conductivity ratio", R_theta:
theta = theta( S, T, P, Pref), ........(2)
R_theta = sal78( S, theta, Pref). .....(3)
theta: potential temperature
Pref: Reference pressure
In this study, we use simply Pref = 0 dbar.

After that, the conductivity sensor drift can be clearly discriminated in the whole view (see Figs. 5c and 6).

Fig. 5: Relations of (a) potential temperature and salinity, (b) potential temperature and conductivity ratio, and (c) potential temperature and potential conductivity ratio for WMO 29033 data on the isobar surfaces. Circles and stars represent data before and after "summer vacation", respectively. Light-blue, blue, red, and green represent data observed at 1200, 1300, 1400, and 1500 dbar, respectively.

Fig. 6: Same as Fig. 4 but for potential conductivity ratio.

Correcting method

  1. A linear local climatological relation of theta and R_theta in the middle/deep layers is estimated by the least squares method from the climatological profile data set, HydroBase (Fig. 7a).

  2. "True" values of R_theta (R_theta_"true") are calculated from its observed theta based on the local climatological relation (Fig. 7a).

  3. A linear relation of the observed and "true" values of R_theta, namely a correcting-equation, is estimated by the least squares method (Fig. 7c).
    R_theta_"true" = a * R_theta_obs + b........(4)
    Note that value of a in the above equation is restricted within 1 plus/minus 0.001 to obtain a reasonable and stable correction.

  4. Finally, corrected R_theta is calculated from the observed values by the equation (4). Then, corrected salinity data are obtained.

Thus, Argo data in deep layers are directly corrected by the local climatological data. On the other hand, those in middle and surface layers are extrapolated by the relation in the deep layers.

Fig. 7: An example to correct Argo data (yellow) based on the climatology (black). (a) Estimation of a local climatological relation of potential temperature and potential conductivity ratio (the light-blue line), (b) drift of potential conductivity ratio from the local climatological relation, and (c) estimation of a correcting-equation of potential conductivity ratio of Argo data (the light-blue line).

Corrected results and its comparison with CTD observation by R/V Mirai

In order to confirm the sensor drift of WMO 29033, a CTD cast was conducted at 27.502°N 148.086°E by R/V Mirai on February 17, 2001.
This CTD station is apart six days in time and 75 km in space from 18th station of WMO 29033 at 27.498°N 148.859°E on February 11, 2001 (see Fig. 1).
By using this CTD data, we examine a performance of the above correcting method, in which the local climatology was historical profile data in area 1214_ABh.

Correcting result 1

Corrected Argo data almost overlap with R/V Mirai CTD data except for those in the sea surface layer, where time and spatial variations are very large. Therefore, this correcting method seems to work well.

Fig. 8: Corrected Argo data (red) based on the climatology with WOCE-CTD data (black). Orange and blue circles represent the raw Argo data and CTD data observed by R/V Mirai, respectively.

Correcting result 2

As shown in Fig. 1, historical profile data in the area 1214_ABh contains WOCE-CTD data set, which has a huge number of highly quality data. Thus, to examine dependence of the correcting method performance on quality/quantity of the local climatology, we conducted another Argo data correction with the same way except for without WOCE-CTD data for the local climatology. Corrected results are relatively good although the corrected data is slightly saline in pycnocline layer.

Fig. 9: Same as Fig. 8 but for climatology without WOCE-CTD data.

Salinity differences from the CTD data by R/V Mirai are shown in Fig. 10. The raw Argo data shows about 0.029 plus/minus 0.011 psu higher salinity than the Mirai CTD data. On the other hand, the corrected Argo data with WOCE-CTD data set reduces its salinity difference to 0.007 plus/minus 0.010 psu, that without WOCE-CTD to 0.014 plus/minus 0.016 psu.

Fig. 10: Salinity difference from the CTD data observed by R/V Mirai. Red and blue circles represent the corrected Argo data based on climatology with and without WOCE-CTD data, respectively. Light-blue circles represent the raw Argo data.

Considering to the differences in the time and space between the observations of the Argo float and the R/V Mirai CTD cast, the correcting-method developed in this study is sufficient to correct the Argo salinity data, in which its conductivity sensor seems to be drifted.
Also it is suggested that this method is almost independent on the quality and quantity of the climatological data set.

The method described in this study is a prototype; therefore, we will continue to improve the method to achieve the goal of the Argo program for data accuracy in salinity, within 0.01 psu.


We wish to express thanks to Prof. G.C. Johnson for his discussions.