P. Poli et al.: SVP-BRST: genesis, design, and initial results 
201 
www.ocean-sci.net/15/199/2019/ 
Ocean Sci., 15,199-214, 2019 
These requirements were adopted on a number buoys 
deployed by the Economic Interest Group (EIG) EUMET- 
NET Operational Service for surface marine observations 
(E-SURFMAR) and European partners. This brought about 
four major technical improvements, as compared to standard 
practices at the time. 
First, the location accuracy was increased, thanks to GPS 
instead of Argos for estimating position, and several buoys 
adopted Iridium instead of Argos for the transmission, to en 
sure regular hourly data reports. Second, the temperature was 
reported and transmitted to shore at a resolution of 0.01 K. 
These technical improvements are collectively known as 
“HRSST-1”. While only few buoys adhered to the HRSST- 
1 requirement in 2009, it has now become the standard, at 
the time of writing, for almost all drifters deployed globally. 
From there, a third requirement appeared, namely the adop 
tion of a new Binary Universal Form for the Representation 
of meteorological data (BUFR) template in 2015, to encode 
the SST data at the resolution of 0.01 K, and transmit to oper 
ational data users via the World Meteorological Organization 
(WMO) Global Telecommunications System (GTS), with 
out loss of information. That template became operational 
at most data-originating centers by the end of 2016; before 
that, many data transmitted on the GTS were sent at reduced 
SST resolution of 0.1 K. At the time of writing, all these three 
improvements are standard for most operational drifters. 
The fourth technical improvement was for each buoy to 
use an individually calibrated temperature probe, instead of 
one picked from a batch calibration, in order to guarantee 
the more stringent total uncertainty requirement of 0.05 K, as 
well as traceability to national standards. This requirement 
(on top of previous ones) was called “HRSST-2”. In total, 
46 such HRSST-2 buoys fitted with all three technical ad 
vances, as well as including each a barometer, were deployed 
between 2012 and 2017. These buoys are listed in Table 1 be 
low. They were manufactured by Metocean (Petólas, 2016), 
using Yellow Springs Instrument Company (YSI Inc.) sen 
sors described in the table. One buoy was redeployed after 
running ashore. 
In addition, several other HRSST-2 buoys were manufac 
tured for experimental purposes, also by Metocean. Each 
buoy carried a conductivity-temperature (CT) probe man 
ufactured by Sea-Bird Electronics (SBE) in order to mea 
sure salinity. Each HRSST-2 SVP buoy with barometer and 
salinity (SVP-BS) hence included two individual-calibrated 
SST probes: one integrated with the buoy hull (around 17 cm 
depth), and one in the CT probe (around 45 cm depth). 
This twin-sensor configuration offered near-optimal horizon 
tal and temporal co-location by virtue of the buoy design. The 
only major differences between the two sensors were the ver 
tical positioning and the housing of the sensors (one digital 
SST sensor integral with the hull, the other CT sensor im 
mersed entirely in water). In total, there were 19 such buoys 
deployed between 2012 and 2015 (one buoy was redeployed 
after beaching). Table 2 shows the list of such buoys, the de- 
Number of points 
275 280 285 290 295 300 305 
Hull SST (K) 
Figure 1. Density plot of the scatter between hull SST measure 
ments (horizontal axis) and CT SST measurements (vertical axis) 
from HRSST-2 SVP-BS buoys. 
ployment areas, and the mission dates. Most buoys were de 
ployed in the North Atlantic. 
2.2 HRSST-2 SVP-BS data record revisited 
In order to exploit the co-located information from two indi 
vidually calibrated SST probes, the data record from the sec 
ond set of HRSST-2 buoys, SVP-BS fitted with CT probes, is 
addressed here. The record consists of about 87 000 data re 
ports between 2012 and 2016. Figure 1 shows a scatter den 
sity plot of the two temperatures. The twin measurements 
are highly correlated, and the robust standard deviation of 
the difference is 0.03 K. This result is compatible with un 
certainty in a difference of two sensors with total uncertain 
ties better than 0.05 K (or possibly 0.02 K). However, Fig. 1 
shows a small fraction of outliers in both directions, espe 
cially for warmer temperatures. In fact, the root mean square 
(rms) of the differences is quite large, at 0.36 K. 
The differences between the two measurements are not 
only due to sensor accuracy but also to the placement of the 
sensors: vertical location and housing (one integral with the 
buoy hull, the other underneath the buoy). To better under 
stand the sources of differences. Fig. 2a shows the differences 
between the two sensor temperatures as a function of solar el 
evation angle. Differences that are out of range (below —IK 
or above 1 K) are also shown for completeness (at — 1 and 
+ 1 K, respectively); they represent about 0.5 % of the entire 
data record. We find, as expected, that most large-magnitude 
differences (absolute value greater than 0.2 K) are positive