Summer 2011: Observing the Diurnal SST cycle
over the Mediterranean Sea
Salvatore Marullo
ENEA: Technical Unit Development of Applications of Radiations, Satellite Oceanography Laboratory, Frascati, Italia
Rosalia Santoleri, Daniele Ciani
Gruppo di Oceanografia da Satellite, CNR-Istituto di Scienze dell'Atmosfera e del Clima, Roma, Italia
Pierre Le Borgne, Sonia Pere
Meteo-France/DP/CMS, Lannion, France
Nadia Pinardi
Department of Environmental Sciences, University of Bologna, Ravenna, Italia
Marina Tonani
Istituto Nazionale di Geofisica e Vulcanologia, Bologna, Italia
THE 44th INTERNATIONAL LIEGE COLLOQUIUM ON OCEAN DYNAMICS, Liège, University Campus, 7 to 11 May 2012
1. To Reconstruct the hourly Sea Surface Temperature diurnal cycle over the Mediterranean Sea using numerical model analysis and geostationary satellite measurements.
2. To Evaluate Errors and investigate their
causes
3. To better Understand the Physics of
diurnal warming events
Objectives:
Reconstructing the hourly SST field:
The Mediterranean Optimal Interpolation (OI) Schema
Input: SEVIRI SSTs (O&SI SAF Product) every 1 hour at 0.05 deg resolution
SEVIRI SST anomalies: Subtract the MyOcean Mediterranean model hourly SST to SEVIRI SST
Time Window: Select SEVIRi SST anomalies in a window of ± 24 hours
Run Optimal Interpolation
Add the Model SST of the Central interpolation Time
The first guess: MyOcean Mediterranean Model
The Mediterranean Forecasting System running at INGV is producing every day a new ten days of physical forecast for the Mediterranean Sea. The analyses for the previous fifteen days are produced once a week. The model resolution is 1/16x1/16 degree and 71 unevenly spaced vertical levels. It is the operational nominal product for the Mediterranean (Tonani et al 2010)
Model: NEMO (Nucleus for European Modelling of the Ocean-Ocean PArallelise) version 3.2 off-line coupled with WAM (Wave Analysis Model) Resolution:1/16 deg. x 1/16 deg. horizontal resolution and 71 unevenly spaced vertical levels (Oddo et al., 2009) Assimilation: sea level anomaly, sea surface temperature, in situ temperature profiles by VOS XBTs, in situ temperature and salinity profiles by ARGO floats, and in situ temperature and salinity profiles from CTD (Dobricic et al. 2008). Satellite OA-SST data are used for the correction of surface heat fluxes
A Diurnal Warming Event on July 7th 2011 M
yO
cean M
odel
Hourly O
ISS
T
SE
VIR
I H
ou
rly
°C
Evaluating Errors
Our first attempt was to validate the optimally interpolated diurnal cycle using moored buoy data
provided by MyOcean in situ TAC.
Most of the buoys are very close to
the coast!
As expected, RMS and Bias are larger near the coast.
The Interpolation does not produces significant variation of the basic statistical
parameters of the difference between in situ (buoys) and remotely sensed data.
Bias RMSE R N MyOcean Model -0.35 0.58 0.965 2052
SEVIRI Valid Pixels 0.14 0.56 0.954 992
SEVIRI Interpolated -0.01 0.45 0.984 1060
Buoy # 61002, Depth = ?, Lon=4.7 °E, Lat=42.1 °N, Distance from the coast = 28 Grid points ≈ 85 Nm
Distances are in nm
MTM spectrum of the detrended sea surface temperature
measured by buoy (a), SEVIRI (b) and the MyOcean
Mediterranean Model (c). The associated 90%, 95% and 99%
significance levels are shown by the three smooth curves In
blue, green and red respectively. The band-width parameter is
p=2, and K=3 tapers were used. Periods that pass the 99%
confidence limit are also indicated.
A second attempt….. Using drifters
Summarizing: The Model tends to overestimate drifters SSTs by 1-2 tents of degrees. This bias is due to a night/afternoon/morning bias. SEVIRI tends to underestimate drifters SSTs by 1-2 tents of degrees. The OI interpolation do not significantly change the basic statistics parameters of the differences.
SEVIRI Model
Bias -0.16 0.15
STD 0.49 0.58
R 0.9707 0.9591
N 3336 3336
SEVIRI Model
Bias -0.06 0.14
STD 0.57 0.60
R 0.9567 0.9531
N 485 485
Valid
Pix
els
In
terp
ola
ted P
ixels
The mean Diurnal Cycle over the matchups points
Comparing Model and Satellite SSTs
The Mean SST diurnal cycle as seen by MyOcean Model and the SEVIRI data
SST
An
om
aly
1. The amplitude of the model cycle is less intense than the corresponding SEVIRI, OI and drifters amplitudes due to the different thickness of the surface ocean layer they represent.
2. The model SST seem delayed respect to the satellite SST by 1-2 hours for the same reason.
Only valid Pixels Only Interpolated Pixels
1. The time shift can be partially be ascribed to the different packaging of the SST data in to the hourly files: MyOcean model hourly SSTs are given as the average of data between 00:00 and 00:01, 01:00 and 02:00 ……. While SEVIRI SSTs are given as the best SST measure within time intervals of one hour centered over each single hour of the day (01:30 and 02:30, 02:30 and 03:30 ……)
Model 00:00-01:00 01:00-02:00 02:00-03:00 03:00-04:00
SEVIRI-SAF 23:30-00:30 00:30-1:30 01:30-02:30 02:30-03:30
Evaluating the capability of model and satellite SSTs to catch Diurnal Warming (DW) events over the Mediterranean Sea
We arbitrary defined DW those event where the difference between the
actual, daytime, SST and the mean SST of the previous night is greater than
0.5 °C.
The mean of the previous night SST was defined as the mean of all the valid
(eventually interpolated) SST measurements between midnight and the time
when the solar zenith angle becomes less that 90°.
DT = Actual hourly SST - Mean SST of the previous night
The operational Model cannot reproduce the extreme
diurnal warming events that occurred in the Mediterranean
Sea during Summer 2011.
MyOcean Model
SEVIRI Data
Understanding the physical mechanism that
modulates amplitude of the diurnal cycle
South Tyrrhenian Sea: GOTM Simulation from
July 1st to 10th
1. MyOcean Mediterranean Model (NEMO) used as Initial conditions for
temperature and salinity profiles
2. Heat and Momentum Fluxes from:
ECMWF (as used by the MyOcean Mediterranean Model)
Turbulent fluxes as GOTM but Bignami (1995) for LW and Reed
(1977) for SW
Turbulent fluxes as GOTM but SAF products for LW and for SW
ECMWF Meteorological data digested by GOTM
1. Several turbulence closure
2. Grid zooming applied at the surface
WIN
D S
TR
ES
S
HE
AT
LO
SS
S
HO
RT
WA
VE
Momentum (1 case) and
Heat fluxes (4 cases) Turb.
Method TKE
equation Model dissip. scale
Stability function
second-order model (TKE_3_2_8_3)
dynamic equation (k-epsilon style)
dynamic dissipation rate equation
Schumann and Gerz [1995]
turbulence Model calculating TKE and length scale TKE_2_2_8_1
dynamic equation (k-epsilon style)
dynamic dissipation rate equation
constant stability functions
turbulence Model calculating TKE and length scale TKE_2_2_3_1
dynamic equation (k-epsilon style)
Xing and Davies [1995]
constant stability functions
turbulence Model calculating TKE and length scale MY_2_3_9_1
dynamic equation (Mellor-Yamada style)
dynamic Mellor-Yamada q^2l-equation
constant stability functions
Turbulence Models Investigated
Simulation using Bignami and Reed for
radiative fluxes
Simulation using SAF products for radiative
fluxes
Simulation using fluxes computed by GOTM
Simulation using fluxes used by MyOcean
Mediterranean Model
SEVIRI OISST
Detrended SSTs From GOTM and SEVIRI OISST
GO
TM
SS
T+
SE
VIR
I Gotmturb.nml configuration (2,2,3,1)
Turb method: turbulence Model calculating TKE and length scale
Type of equation for TKE: dynamic equation (k-epsilon style)
Type of model for dissipative length scale: Xing and Davies [1995]
Type of stability function: constant stability functions
Meteorological data (U10m, V10m, T2m,
D2m, TCC, MSLP) from ERA interim, Daily
Fields every 6 hours http://data-portal.ecmwf.int/data/d/interim_daily
SST From SAF (every 1 h)
Fluxes using Bignami and Red
Fluxes using SAF SWR and LWR
Fluxes computed by GOTM
Fluxes used by MyOcean Mediterranean Model
WIN
D S
TR
ES
S
HE
AT
LO
SS
S
HO
RT
WA
VE
MED-MFC FLUX
ECMWF+SAF ECMWF+Bignami+Red
Meteo data
TKE_3_2_8_3 0.490 0.597 0.732 0.747
TKE_2_2_8_1 0.369 0.430 0.515 0.555
TKE_2_2_3_1 0.363 0.416 0.499 0.528
MY_2_3_9_1 0.424 0.547 0.620 0.798
RMS between GOTM simulated SSTs (detrended) obtained using
several configuration and air-sea fluxes and SEVIRI SST (K)
Zooming
over
the first 2 meters (TKE_2_2_3_1)
Jul 3
rd
Jul 7
th
Big
na
mi+
Re
d
SA
F L
W &
SW
R
GO
TM
Flu
xe
s
MyO
ce
an
Flu
xe
s
July 3rd 2011 (4 am to 3 pm) July 7th 2011(4 am to 3 pm)
Drifters depth
Mooring depth
-0.0
-0.4
-0.8
-0.0
-0.4
-0.8
0.0
0
0.0
5
0.1
0
0.1
5
0.0
0
0.0
5
0.1
0
0.1
5
0.0
0
0.0
5
0.1
0
0.1
5
0.0
0
0.0
5
0.1
0
0.1
5
0.0
0
0.0
5
0.1
0
0.1
5
0.0
0
0.0
5
0.1
0
0.1
5
0.0
0
0.0
5
0.1
0
0.1
5
0.0
0
0.0
5
0.1
0
0.1
5
T(z)-T(z=0) Time (UTC)
GOTM Temperature difference between 0.025 m and
0.25 m of depth
Conclusions ①The optimal interpolation schema is able to correctly
reconstruct the diurnal SST cycle including sub-diurnal
components and diurnal warming events.
②The amplitude of the diurnal cycle produced by the MyOcean
Mediterranean Model is less intense than the corresponding
SEVIRI, OI and drifters amplitudes due to the different thickness
of the surface ocean layer they represent.
③1-D models represent an useful tool to investigate the upper
ocean response to air-sea interactions at daily and sub-daily
frequencies.
④ A difference between satellite a buoys (drifters and moored)
SST must be expected as consequence of air-sea exchanges.
⑤ This imply that the “zero bias” goal is not the correct
approach to follow.
⑥ The contribution of dedicated in situ experiments will be very
important.