P.  Poli  et  al.:  SVP-BRST:  genesis,  design,  and  initial  results

209

www.ocean-sci.net/15/199/2019/

Ocean  Sci.,  15,199-214,  2019

Figure  7.  Trajectories  of  the  two  SVP-BRST  prototypes  after  deployment ­
  on  26  April  2018.  The  two  buoys  separated  on  22  May
2018.  Map  data:  SIO,  NOAA,  US  Navy,  NGA,  GEBCO;  map  image: ­
  Landsat/Copernicus.

mean  temperature  over  5  min  reported  by  the  HRSST  sensor,
5  percentiles  of  the  SST  distribution  within  that  time  interval
(10  %,  30  %,  50  %  or  median,  70  %,  and  90  %),  and  the  mean
and  the  standard  deviation  of  the  hydrostatic  water  pressure
during  5  min.
These  parameters  are  shown  in  Fig.  8,  where  atmospheric
pressure,  SST,  and  significant  wave  height  from  the  ECMWF
operational  analyses  have  been  added.  This  information  was
co-located  to  the  buoy  dates,  times,  and  locations  using  the
same  procedure  as  described  in  Sect.  2.2  (albeit  at  different
horizontal  and  temporal  resolutions).  For  the  sake  of  comparing ­
  results,  the  time  series  are  only  for  as  long  as  both  buoys
were  freely  drifting  (until  11  June).
The  information  from  ECMWF  analyses,  although  at  a
horizontal  resolution  of  around  10  km,  is  independent  from
the  buoys.  It  hence  provides  interesting  information  to  consider ­
  when  assessing  the  buoy  data.  For  air  pressure  (Fig.  8a),
both  buoys  agree  with  the  ECMWF  analyses  to  within
0.8  hPa  rms.  This  is  comparable  to  the  state-of-the-art  SVP-B
deployed  in  this  region.
For  SST  (Fig.  8b),  the  comparison  to  ECMWF  analyses
only  suggests  that  the  latter  are  typically  lagging  behind  the
buoy  evolution  by  24  h,  until  5  June  2018.  It  must  be  remembered ­
  that  the  SST  is  not  currently  analyzed  in  the  ECMWF
prediction  system,  but  this  system  was  upgraded  on  6  June,
including  a  component  to  include  atmosphere-ocean  coupling ­
  (Buizza  et  ah,  2018).
The  depth  inferred  from  the  HRSST  hydrostatic  pressure
sensor  (Fig.  8c)  shows  values  around  15  to  18  cm  (which  is
the  design  location  of  the  HRSST  sensor).  The  spread  between ­
  the  two  estimates  is  stable  in  time,  around  4  cm.  The
calibration  procedure  of  the  pressure  sensors  may  explain

this  difference.  This  remains  however  close  to  the  design
depth  of  18  cm  below  the  flotation  line  of  the  buoy.
The  spread  in  the  SST  percentiles,  shown  in  Fig.  8d,  is  usually ­
  within  0.1  K  but  sometimes  exceed  0.3  K.  In  such  situations, ­
  the  calibration  accuracy  of  the  sensor  is  not  of  much
help  to  help  exploit  the  data  for  precise  comparison  with
other  sources.  However,  the  availability  of  five  estimates  of
SST,  instead  of  just  the  mean,  should  help  users  move  their
applications  to  a  small  (five-member)  ensemble  and  better
understand  how  the  spread  in  input  in  situ  SST  impacts  their
products.
Figure  8e  shows  the  standard  deviation  of  depth  (inferred
assuming  hydrostatic  equilibrium).  This  estimate  varies  between ­
  1.5  and  3.5  cm.  It  is  largest  when  the  significant  wave
height  (estimated  by  the  ECMWF  analyses)  is  largest,  in  line
with  stronger  winds  at  the  same  times  (Fig.  8f).  This  is  expected ­
  from  the  buoy  dynamics  (as  the  pressure  measured
will  be  affected  by  positive  and  negative  accelerations),  and
confirms  that  the  ECMWF  wind  and  wave  height  analysis
appears  to  be  correct.  Given  this  result,  the  larger  spread  in
SST  percentiles  appears  to  be  well  correlated  with  situations
where  the  wave  heights  and  wind  speeds  are  smaller.  This
would  seem  to  validate  the  conjectures  formed  earlier  by  revisiting ­
  the  HRSST-2  SVP-BS  data  record,  namely  that  the
sea  state  is  an  important  parameter  to  consider  when  exploiting ­
  the  in  situ  SST  data.
Regarding  the  SST  data,  we  see  that  both  buoys  capture ­
  fairly  well  the  diurnal  warming/cooling  cycle,  a  feature
that  is  generally  clearly  missing  from  the  ECMWF  analyses.
What  is  more,  the  amplitude  of  the  daily  cycle  is  variable,
suggesting  that  the  local  ocean  and  atmospheric  dynamics
impacts  the  SST  measured  by  the  buoys.  This  is  indeed  the
case  for  the  period  from  29  April  to  5  May  (time  period  A
in  Fig.  9):  the  observed  SST  is  slightly  cooler  and,  crucially,
is  missing  the  diurnal  cycle  found  in  the  rest  of  the  time  series. ­
  Looking  at  co-located  wind  data  (not  shown),  we  do  not
find  any  clear  modification,  suggesting  that  the  reason  for
this  behavior  in  the  SST  data  is  principally  oceanic  and  not
atmospheric.  Indeed,  if  we  look  at  the  buoys’  location  during ­
  that  time  period,  we  see  that  they  are  trapped  within  an
eddy  core  (Fig.  10),  and,  significantly,  it  is  a  cold  eddy.  It
is  known  that  these  eddies  generate  an  upwelling  within  their
core,  leading  to  colder  and  vertically  more  homogeneous  surface ­
  and  near  surface  waters.  The  buoy  data  suggest  that  this
upwelling  more  than  compensates  the  diurnal  warming  and
eliminates  the  near  surface  stratification.  During  time  period
A,  the  average  diurnal  cycle  measured  by  the  two  buoys  is
rather  weak  (Fig.  11a  and  b).
Once  the  buoys  move  out  of  the  eddy  core  (Fig.  12),  the
diurnal  cycle  is  once  again  found  in  the  data.  This  is  visible
in  Fig.  9  during  time  period  B,  and  in  Fig.  1  lc  and  d,  where
the  daily  amplitude  in  SST  exceeds  0.5  K  (when  it  was  less
than  0.2  K  in  time  period  A).
Looking  at  the  evolution  of  SST  5  min  percentiles  enables
to  gauge  the  small-scale  variations  in  temperature  near  the