Nong et al. 
10. ' 
TE 
3. 
‚AP | > 1.0_dbar 
6. + 
|AP | > 2.4 dbar 
‚AP | > 5.0_dbar 
"AP ]> 10.0 dbar 
4. 7 
1 
w 
__ 
DD 
3. 
u 
40 
80 
a a a I I I 
120 160 200 240 280 
Cycle 
FIGURE 8 | Percentage of pressure adjustments vs. the number of cycles, 
rom 2.779 APF.X floats 
float types (the auto-correcting floats, e.g., SOLO, PROVOR), but 
is done as part of the real-time and delayed-mode adjustment 
process for other float types (the non-auto-correcting floats, e.g., 
APEX, NAVIS). 
APEX floats, one of the non-auto-correcting float types in 
Argo, report the raw pressure measurements and the SP values 
separately. Thus, examining the SP values from APEX floats is 
an effective way to gauge the long-term stability of pressure data 
from the SBE CTDs. Analysis of delayed-mode pressure data 
from 2,779 APEX floats showed that over the course of 280 cycles, 
about 5% of the SBE CTDs showed pressure sensor drift > 2.4 
dbar, and only about 3% showed pressure sensor drift > 5 dbar 
{Figure 8). 
After delayed-mode adjustment, float pressure data are given 
the accuracy of 2.4 dbar, which is historically (before 2011) 
the manufacturer’s quoted accuracy for pressure. The method 
of using SP values to adjust pressure can eliminate the depth- 
independent error (the offset error) in long-term sensor drift, but 
cannot account for any depth-dependent error (the slope error). 
However, comparisons against ship-based CTD data show that 
the median of possible depth-dependent pressure bias in the Argo 
profiles is within the manufacturer quoted accuracy of 2.4 dbar, 
as will be discussed in section Assessment of Pressure Bias below. 
Problems Encountered 
Pressure measurements from Argo floats have been affected 
by several major sensor issues over the past 20 years (Barker 
2t al., 2011). In 1999-2000, SBE CTDs were fitted with pressure 
transducers manufactured by the Paine Corporation. These 
were discontinued because they showed significant instrument 
drift over the course of a float’s lifetime (e.g., Gouretski and 
Koltermann, 2007). Pressure transducers from Ametek were 
then employed during 2000-2002, but were then discontinued 
when a manufacturing defect, which also caused significant 
instrument drift over time, was discovered. Beginning in 2002, 
SBE used Druck pressure transducers. While the Druck pressure 
sensors typically produce stable measurements, two episodes 
of manufacturing defects affected one generation of Argo 
floats. These are the Druck “snowflakes” problem and the 
Druck “microleak” problem. The Druck “snowflakes” problem 
rontiers in Marine Science | www.frontiersin.orı 
Argo Data 1999-2019 
10| Pain ; 
A Dre 
OD 
MEET 
» 
C 
£ 
* 
x 
7 
Ametek 
aa 
7 | 
Druck 
Er 
—5 
Druck with 
% micro-leak 
° Raw SP’s | 
Calculated SP’s ) 
200 400 3600 800 1000 1200 
Number of days since first profile 
.1C 
FIGURE 9 | Typical SP offsets (dbar} for pressure sensors: Paine (WMO ID 
56501), Ametek (WMO ID 2900089), Druck (WMO ID 3900263}, and a Druck 
microleaker (WMO ID 5901649}. [Source: from (Barker et al., 2011). Copyright 
2011 American Meteoroloagical Society (AMSY 
was due to internal electrical shorting by titanium oxide 
particles (“snowflakes”) in the oil-filled cavity in the pressure 
sensor, causing erratic pressure measurements and thus erratic 
temperature and salinity measurements. The Druck “microleak” 
problem occurred when oil leaked through fine cracks in the 
glass/metal seal of the inner chamber of the sensor, causing 
an internal volume loss and thus an increasing negative offset 
at all pressures. These problems no longer occur: Druck has 
rectified the “snowflakes” problem and SBE has implemented 
procedures that can screen for “microleaks”. Figure 9 shows the 
typical pressure sensor drift patterns from the Ametek, Paine and 
Druck sensors, and an example of a Druck sensor suffering an oil 
microleak. In 2010, due to a supply constraint of Druck sensors, 
SBE started fitting some CTD units with Kistler pressure sensors. 
Presently, SBE use pressure sensors from two manufacturers: 
Druck and Kistler (<10% Kistler as of April 2020). 
Controller board issues have also affected some float 
pressure measurements. In some APEX floats, the SP values 
were restricted to greater than zero. This was done as 
part of the mission control to turn off the CTD pump as 
the float neared the surface. These are APEX floats with 
controller boards identified as APF8 or earlier series. On 
these APEX floats, negative SP values are truncated to zero 
before telemetry. Thus, as a result of this onboard truncation, 
negative pressure drifts cannot be identified and therefore 
cannot be corrected. These data are labeled as having possible 
Truncated Negative Pressure Drifts (TNPDs), and account for 
about 5% of all Argo CTD profiles. Some APF8 controller 
boards were updated specifically to remove this “truncating” 
feature. All later series of controller boards on APEX floats, 
APF9 and above, return raw SP values with no truncation of 
negative values. 
Another pressure problem that has affected Argo data results 
from processing errors onboard the floats. In 2007, it was 
discovered that some SOLO floats from the Woods Hole 
Oceanographic Institution (designated as SOLO-W) returned 
incorrect pressure values because of a bin-average error in the 
Qanteambear 2020 1 Valııme 7 | Article Z01