-e Menn et al. 
& 
& 
= 
0,16 
2,08 
0,06 
0,04 
0,02 
0,00 
-0,02 ( 
0,04 
0,06 
0,08 
0.10 
T 
1* 
TeTyT707- Tre 
3Ty1707 - Tref (covered) 
ATy1824 - Tref 
XTy1824 - Tref (covered) 
„Co 
25 - 
X 
= 
St 
SVP-BRST Fiducial Reference Network 
Tref (°C 
FIGURE 8 | Deviations obtained during the verification of SST sensors of two buoys during the two series, with the expanded uncertainty of the verification 
TABLE 5 | Uncertainty budget of buoys SST measurements in the range 6-35°C. 
Uncertainty budget of SST measurements N° Y17-07 N? Y18-24 
(analog sensor) (mK) (mK) 
Zeference temperature (Ujrgf) 
3ath stability (Upaza) 
3uoy SST reproducibility (S) 
3uoy SST repeatability (Srep) 
3uoy SST linearity error (u) 
Expanded uncertainty (Ur) 
systematic errors of 40.15 dbar between 0 and 35°C. They also 
show that this effect is linear and that this technology of sensor 
reacts with a good reproducibility to temperature variations. This 
behavior can hence be corrected for each buoy: a straight line 
correction curve yields residual errors inferior to 0.004 dbar 
(or 4mm of water). Since the slopes and offsets for the two 
sensors are very similar, it is possible to consider correcting the 
two instruments with average coefficients. The residual errors 
obtained with the average coefficients slope = —0.0077 dbar/°C 
and offset = 1.499 dbar are inferior to 0.02 dbar, which is close 
to the accuracy claimed by the manufacturer, and in relation 
with the technology of the sensors. The same coefficients may be 
applied to future buoys. 
TRIALS AT SEA AND COMPARISON WITH 
A CTD PROFILER 
While the initial concept of buoy development using a pre- 
calibrated sensor was demonstrated to work in the laboratory, 
two further elements are needed to build a FRM drifter network. 
The first element is to demonstrate that the trueness estimate still 
holds, once the buoys are deployed in the target environment 
(at sea). The second element is to demonstrate that the trueness 
estimate remains valid for the lifetime of the buoy (ie., no 
significant temporal drift). While it is too early to study the 
second element, this section explores the first one. 
rontiers in Marine Science | www.frontiersin.or 
The two prototype buoys were deployed at sea during an 
oceanographic cruise in Mediterranean Sea. After unpacking on 
the ship deck to test the transmission and the good transmission 
of the Iridium SBD messages, a comparison with a CTD profiler 
(SBE 9117) was set up. The CTD profiler was fixed under a Multi. 
Bottle Sampling Array (MBSA), with a reference thermometer 
SBE 35 calibrated in the fixed point cells of the ITS-90. The 
drifters were held in place near the ship by means of a line. 
When the MBSA was completely immersed, the CTD and 
the SBE 35 temperature measurements were recorded at about 
Im under the surface. After the surface measurements were 
collected, the MBSA was lowered to 15m depth in order to 
estimate the temperature profile of the first layers. This profile 
showed that in the four first meters, the temperature of the 
water was homogeneous and close to 16.4°C, allowing fair 
comparison with the HRSST buoy. Between 4 and 5.5m (the 
depth where Argo floats surface temperature measurements are 
sometimes used as reference) there was a strong temperature 
gradient of —1.25°C m”7!. Until 15m depth, the temperature 
was still very stable but shifted by about 2°C as compared to 
the surface. 
During the subsurface waiting time, 5 measurement series 
of 29, 57, 113, 53, and 149 values were made with the SBE 35 
sampling at 1 Hz, giving an average temperature of 16.3968°C 
with a standard deviation of 0.0057°C. The CTD temperature at 
1.08 m depth is very close to this value: 16.398°C. Table 6 gives 
the results of the comparison with the values transmitted by the 
buoys. The first two lines show the results for the SST analog 
sensors and the following for the HRSST sensors. In the second 
column of this array, SST temperatures from the analog sensors 
have been corrected with the slopes and offsets coeflicients of 
the straight lines of Figure 8, in order to fairly compare with 
what may be expected, at best, from low-cost SST sensors with 
ideal calibration. 
The results of this comparison show that without any 
correction, HRSST values are in the standard dispersion range 
of the SBE 35 and the deviations compared to CTD and SBE 35 
are inferior to 0.01°C. Without corrections, SST deviations from 
analog sensors are close to —0.05°C and with the corrections 
Qantembear 2019 1 Valııme A 1 Article R7£