200 
P. Poli et al.: SVP-BRST: genesis, design, and initial results 
Ocean Sci., 15,199-214, 2019 
www.ocean-sci.net/15/199/2019/ 
temperature (SST) data. Over short timescales, this essen 
tial ocean state variable provides important information on 
the spatial distribution and intensity of dynamic structures, 
such as eddies, coastal currents and upwelling regions, in 
near-real time (within a few hours after acquisition). Over the 
long term (multi-decade), it describes the distribution of heat 
within the Earth system. Long time series of SST datasets 
(e.g., Merchant et ah, 2014) are crucial to provide informa 
tion on global and regional sea surface temperature trends. 
These can be used directly to monitor the evolution of the 
surface ocean on decadal timescales and help quantify the 
intensity of events such as El Nino/La Nina, as well as be 
ing useful to constrain climate reanalyses (e.g., Dee et al., 
2014). For these reasons, the importance of monitoring SST 
was recognized as a priority by the Copernicus program, and 
a sensor aimed at observing SST was included on Sentinel-3 
satellites, the Sea and Land Surface Temperature Radiometer 
(SLSTR; Coppo et ah, 2013). To deliver the SST data product 
service (Bonekamp et ah, 2016), the dual-view capability and 
onboard calibration of SLSTR give it comparable accuracy to 
similar sensors, such as the Advanced Along-Track Scanning 
Radiometer (AATSR; Llewellyn-Jones et al., 2001). 
Satellite sensors measure top-of-atmosphere radiance, 
which has some relation to but is not identical to the phys 
ical temperature of Earth’s emitting surface. The inverse pro 
cess of inference of the surface state tends to amplify un 
certainty. Achieving the desired quality of Earth observa 
tion measurements from SLSTR places stringent require 
ments on the SLSTR sensor calibration (Donlon, 2011). This 
drives a requirement for higher accuracy and better knowl 
edge of uncertainties of the surface measurements used for 
validating the satellite products. This process requires the 
highest-possible quality in situ measurements, with well- 
characterized uncertainties, so that the error budget of SST 
products can be investigated (e.g., Corlett et ah, 2014). Such 
investigation requires covering the various regimes of satel 
lite SST retrievals, mandating in turn that the high-quality 
in situ data be geographically well distributed. 
As a result, concomitantly to the SLSTR development, the 
Copernicus program aims to develop fiducial reference mea 
surement (FRM) initiatives. Among them is the deployment 
of an array of temperature-measuring surface drifters, cov 
ering several SST regimes. The operational nature and cli 
mate quality of Sentinel-3 datasets are expected to deliver 
long-term data records (Donlon, 2011). For consistency, this 
implies that the surface references used for calibration and 
validation must also be homogeneous over time. This FRM 
initiative complements others started lately, such as under 
the European Space Agency (ESA) project Fiducial Refer 
ence Measurements for validation of Surface Temperature 
from Satellites (FRM4STS), which has conducted in particu 
lar a comparison of infrared radiometers with radiation ther 
mometers in a laboratory setting (Theocharous et ah, 2019). 
Beyond comparisons, the goal is to establish the traceabil 
ity of the various sensing techniques to the Système Interna 
tional (SI) unit, as it then guarantees anchoring to interna 
tional physical standards. In such attempt, the importance of 
metadata to define exactly the sensor and its environment is 
essential. For drifters measuring SST, this means knowing in 
particular the SST sensor depth and type, its calibration pro 
cess, and other aspects influencing the buoy behavior (such 
as drogue loss). 
Based on lessons learnt from previous similar initiatives, a 
new type of drifter has had to be developed and submitted to 
a rigorous calibration procedure to meet this goal. In short, 
this new type of drifter must carry a state-of-the-art digital 
temperature sensor coupled to a hydrostatic water pressure 
sensor, allowing for a measurement frequency of up to 1 Hz. 
The value of this new drifter for calibration and validation 
(cal/val) of SST satellite retrievals is expected to be assessed 
through international collaboration. 
The outline of this paper is the following. Section 2 revisits 
the past high-resolution SST drifting buoy initiatives, includ 
ing error budget analysis. Based on the lessons learnt, Sect. 3 
presents the design adopted for a new generation of drifter, 
called the Surface Velocity Platform drifter with Barom 
eter and Reference Sensor for Temperature (SVP-BRST). 
Section 4 shows preliminary measurement results from two 
SVP-BRST prototypes deployed in the Mediterranean Sea. 
Finally, Sect. 5 gives conclusions and prospects for future 
work. 
2 Genesis: lessons learnt from past HRSST drifting 
buoy initiatives 
2.1 Background: the HRSST-1 and -2 requirements 
O’Carroll et al. (2008) compared SST retrievals from 
AATSR with SST retrievals from a microwave sensor and 
with in situ SST from drifters. The drifters were found to 
have a standard deviation of error smaller than the microwave 
SSTs and larger than those from the AATSR. This high 
lighted the need for improved in situ calibrated reference 
temperature data for satellite SST cal/val, particularly in ref 
erence to the validation of high-quality dual-view satellite 
SSTs, and the satellite and in situ communities started a 
dialogue on collaboration and improvements. In 2009, the 
Group for High-Resolution SST (GHRSST) called on the 
Data Buoy Cooperation Panel (DBCP) HRSST Pilot Project 
(HRSST-PP) to implement a number of key requirements 
for buoys to be eligible to support HRSST work (Donlon, 
2009). The buoys would have to provide hourly measure 
ments, nominal or design depth in calm water of the drifting 
buoy SST to an absolute accuracy of 5 cm, location accuracy 
of 500 m, SST with a nominal resolution of 0.01 K or less 
and a total uncertainty of 0.05 K, and measurement time to 
within 5 min.