_e Menn et al. 
TABLE 3 | Differences between buoys and HRSST ambient and covered 
:emperatures for the buoy n° Y17-07. 
Tbuoy (°C) HRSST (°C) Tbuoy (°C) HRSST (°C) Deviation 
lambient) (ambient) (covered) (covered) (amb. -cover.) 
9.3 
8.7 
20.1 
A 
20.5 
23.0 
23.2 
24.0 
1.0018 
5.0008 
"1.0002 
15.9990 
20.9979 
25.9966 23.1 
30.9969 26.9 
22.9961 7.2 
0.9993 0,0025 
5.9988 0,0020 
11.0030 0.0029 
16.0025 —0.0036 
21.0013 0.0034 
26.0000 0.0034 
30.9986 —0.0017 
2323 9965 0.0004 
TABLE 4 | Uncertainty budget of buovs HRSST measurements. 
Uncertainty budget of HRSST N®° Y17-07 N° Y18-24 
measurements (mK) (mK) 
Zeference temperature (Ujrgf) 
3ath stability (Uparh) 
3uoy HRSST reproducibility (S) 
3u0y HRSST repeatability (Syep) 
=xpanded uncertainty (Ur) 
0.9 
0.3 
2,5 
0.5 
5.5 
0.9 
0.3 
3.4 
0.5 
72 
inertia of buoys. Expanded uncertainties are expressed with a 
coverage factor of 2 including 95.5% of measurements in the case 
of Gaussian distributions. They are inferior to 0.01°C. 
During the two series of HRSST calibration, the temperatures 
of the SST analog sensors have also been recorded. They have 
been calibrated by the manufacturer in the range 5-35°C. 
Figure 8 shows the results of the verification. The deviations are 
inferior to £0.1°C, even for the point at 2°C, which is outside 
the calibration range. If we exclude this point, it is possible to 
improve the trueness and the uncertainty of SST measurements 
öy calculating the coefficients of a straight line. By considering 
this linear correction, it is possible to assess the measurement 
uncertainties of these two SST analog sensors by using the same 
procedure as for the HRSST sensors. However, it is necessary 
to take into account a residual linearity error. The results are 
given in Table 5. The expanded uncertainty of SST analog sensors 
is found to be twice as large as the expanded uncertainty of 
HRSST sensors. One must keep in mind in addition that the 
SST analog sensor is much slower to respond than the HRSST 
sensor, and that it is also more sensitive to radiation effects. All 
these effects contribute to additional larger systematic errors or 
measurement uncertainties. 
UTILITY AND LABORATORY TEST OF THE 
HYDROSTATIC PRESSURE SENSOR 
[n order to try to reduce the uncertainty in the HRSST 
measurement depth, the MoSens have been equipped with a 
hydrostatic pressure sensor located near the HRSST sensor. The 
immersion depth d of the water pressure sensor is given by the 
rontiers in Marine Science | www.frontiersin.ore 
SVP-BRST Fiducial Reference Network 
buoy geometry, d = R cos(@) where R is the radius of the spherical 
buoy and & is the angle of placement of the sensor in the spherical 
buoy (measured from the vertical), but this distance from the 
waterline can vary with the seawater density pw and the traction 
made by the drogue, but also with the variations of x during 
cough sea conditions. 
During calm sea conditions, the air pressure sensor measuring 
Pa is at the level of the waterline. Therefore, d can be obtained 
from the measurement of the pressure p: 
gu @-—Pe) 
DwZ 
(15) 
where g is the acceleration of gravity at the buoy location 
(this value depends on latitude in first approximation, for a 
body that remains on the ocean surface). In this relation, pw 
needs an assessment of the salinity to be determined with a 
suflicient accuracy. 
When the buoy is in rough sea conditions, or oscillating 
(rotating) around its center of gravity, it is submitted also to 
a vertical acceleration ag added to g. If ag values are close to 
g the measurements of water pressure cannot be used directly 
to retrieve depth without ad hoc processing and filtering, but 
the time-series of pressure at high-frequency can provide an 
indication of the sea state. This information can be of use to 
determine whether the water is well-stratified or well-mixed, 
assuming that the air pressure is stationary (this hypothesis 
does not hold if the buoy is oscillating up and down in waves 
with heights of several meters). For comparison with satellite 
measurements, a well-mixed top layer may suggest that emission 
from the surface is at the measured water temperature, whereas a 
stratified top layer may suggest that the radiated temperature may 
need to be corrected (based on wind and radiation conditions). 
Results shown by Poli et al. (2016) corroborate these 
assertions. When two temperatures measured by previous 
HRSST buoys are compared, the differences can be reduced 
to within the digital sensor trueness by considering only well- 
mixed conditions, selected when the waves in the ERA-Interim 
(Dee et al., 2011) reanalysis are above 3m in significant wave 
height. Another application of trying to infer the sea-state is 
to better parameterize the emissivity to be used for simulating 
the radiances seen by the satellite, especially for microwave 
instruments, with rough seas or swell suggesting white caps and 
foam (Niclös et al., 2007). 
During the temperature verifications of the two buoys, the 
MoSens pressure sensors data have been recorded to observe 
their drift with respect to temperature. A reference atmospheric 
pressure Pam has been measured with a recently calibrated 
WIKA CPC 8000 pressure calibrator. It was therefore possible to 
calculate a reference pressure Prof = d + (Patm — 1013.25)/100, 
to observe the pressure drifts of sensors as a function of 
the temperature. 
The immersion depth d was estimated to be 0.13 m in the bath 
(without the weight of the drogue, which remained outside the 
calibration bath). The results show that the external temperature 
of the buoy has no significant effect on the measured pressures, 
but that the temperature of the water leads to maximum 
Qanteambear 2019 | Valııme A| Article R7£