P. Poli et al.: SVP-BRST: genesis, design, and initial results
207
www.ocean-sci.net/15/199/2019/
Ocean Sci., 15,199-214, 2019
0.00 0.05 0.10 0.15 0.20 0.00 0.05 0.10 0.15 0.20
SST uncertainty (K) SST uncertainty (K)
(SLSTR 3-channel nighttime) (SLSTR 3-channel nighttime)
Figure 5. SLSTR SST uncertainty validation plot for (a) all drifters
and (b) a subset of HRSST-1 and HRSST-2 drifters, with uncertainty
bins of 0.001 K. An uncertainty of 0.05 K is assumed for the
drifter SST.
Also, when in that situation, the float is more likely to reach
wave crests. There, the sky visibility is improved, reducing
the GPS time to first fix (TTFF), which can serve as an additional
indicator of drogue loss (Petolas, 2013).
To investigate the influence of the drogue, the SVP-BS
data record is revisited. These buoys used submergence sensors,
whereas drifters nowadays use strain gauges, e.g., as
indicated by Rio (2012), who developed a advanced method
to identify drogue loss using drifter currents, satellite altimetry,
and wind reanalysis data. The submergence (or tether
strain gauge) readings are neither straightforward to interpret
nor fully reliable on their own (Rio, 2012). Flowever,
the SVP-BS drifter data considered here (available from the
Coriolis In Situ Thematic Assembly Center) are not found
in the drifter dataset of Rio and Etienne (2018), which includes
drogue presence flags. Consequently, for this analysis,
we use the submergence and GPS TTFF data. A visual
inspection indicates that 10 of the 20 buoys in Table 2 have
lost their drogues during their mission. For these buoys, two
series of data records are extracted: (1) before drogue loss
and (2) after drogue loss.
During daytime, the median of the differences between
the twin SST measurements is —0.04K in (1), whereas it is
—0.03 K in (2). The reduction in differences may appear insignificant,
but it is consistent with the CT sensor being more
often exposed to depths similar to the sensor integral to the
hull when the drogue is lost than when the drogue is present.
Similarly, the robust standard deviation of the differences between
the twin SST measurements is 0.03 K in (1), whereas
it is 0.01 K in (2). Again, this reduction is consistent with
drogue loss for the same reasons.
During nighttime, no influence of the drogue loss is expected
if the temperatures are homogeneous just below the
surface. This is indeed what is observed. The median of the
differences is —0.04K in both (1) and (2), and the robust
standard deviation of the differences is 0.03 K in both (1)
and (2).
In other terms, the SVP-BS data record confirms the expectation
that once the drogue is lost, the SST probes on a
drifter are more likely to be exposed to water immediately
below the surface than when the drogue is present, and this
effect is more visible in the presence of stratification (e.g.,
during daytime). To keep track of the drogue effect on SST
measurements, it is important to monitor drogue loss as well
the immersion depth and its variations.
2.6 Limited traceability
Adopting a more general point of view for SST observations,
several works have already attempted to document the uncertainties
in the various in situ SST measurement methods.
The present paper does not attempt to review all these efforts
but cites relevant results from the comprehensive review of
Kennedy (2014). While the focus of this earlier work was on
the creation on long time series, with the largest issues identified
at the time of World War II (transition on ships from
bucket to engine-room intake), the quality of SST buoys was
found to be the subject of several concerns. The first concern
is the spread in quality between buoys, depending on
the source of the uncertainty estimate, with no reliable link
to the actual metrological reference. The second concern is a
suggested improvement in quality over time, though without
quantified evidence or clear a priori reason for it that would
be explained by metrological documentation. Both points
stem from an insufficient knowledge of the sensor technology,
and of the calibration procedure that was actually used,
for each drifting buoy deployed. The results shown earlier,
showing differences in SST quality between general drifters
versus HRSST drifters, reinforce the importance of enhancing
the knowledge of drifter metrology and metadata.
3 Design of the SVP-BRST
The HRSST-2 efforts were initiated by the cal/val needs of
AATSR SST retrievals. With the demise of this instrument
after 10 years of service in 2012 (ESA Communications Department,
2012), the HRSST-2 developments were put to a
halt, until the replacement sensor (SLSTR on Sentinel-3) was
launched. However, this gap gave time to finish all HRSST-2
deployments and review the lessons learnt from them. Coupled
with the need to assert long consistent time series of SST
at an accuracy level compatible with SLSTR requirements,
sound bases were used to imagine a novel sensor package
for reference SST. The result is the SVP-BRST, based on the
SVP-B design (Sybrandy et ah, 2009), with a strain gauge
to detect drogue loss. In addition, the HRSST-2 requirements
presented earlier are included, as well as others, described
hereafter.
The first additional requirement is to employ an additional
HRSST sensor, in addition to the regular SST sensor.
The HRSST sensor collects data within the 5 min before the
round hour, when the position is updated by means of GNSS.
The mean SST is to be computed from 1 Hz SST measurements.
In addition, the data can be relayed at 1 Hz frequency
