Newsletter - Mercator Ocean · 2015-05-04 · Mercator Ocean Quarterly Newsletter #39 – October...

#39

QU

AR

TE

RLY

New

slet

ter

The regional seas within the MyOcean project are the Arctic Ocean, the Baltic Sea, the Atlantic- European North West Shelf- Ocean, the Atlantic- Iberian Biscay Irish- Ocean, the Mediterranean Sea and the Black Sea. A focus is here put on the Black Sea area, as well as on the Atlantic- Iberian Biscay Irish- Ocean and the Atlantic- European North West Shelf-

Ocean. Credits: [email protected]

Editorial – October 2010 – MyOcean Physical Systems in Regional Seas

Greetings all,

This month’s newsletter is devoted to some of the regional seas within the MyOcean project http://www.myocean.eu/ and to the numerical

systems that allows describing the ocean physics in those areas. A focus is here put on the Black Sea area, as well as on the Atlantic- Iberian

Biscay Irish- Ocean and the Atlantic- European North West Shelf- Ocean, with the description of new products that will be part of the MyOcean

catalogue from June 2011 on. Next January 2011 issue of the newsletter will also display papers about the Myocean project, focusing this time

on the ecosystem products in the Mediterranean Sea, the Arctic Ocean, the Black Sea as well as the Global Ocean.

In the present issue, Durand et al. are introducing this newsletter telling us about the Ferrybox data within MyOcean which are handled by the In-

Situ Thematic Assembly Centre (TAC). Environment sensor package onboard of ship-of-opportunity are referred to as Ferrybox system.

Ferrybox data products as temperature, salinity, oxygen and chlorophyll will be available from the MyOcean portal in near real time from June

2011. Delayed time products will be made available later on, around December 2011.

Scientific articles are then displayed as follows: First, Cailleau et al. are telling us about the numerical system that allows describing the ocean

physics in the MyOcean Atlantic- Iberian Biscay Irish- Ocean (referred to as the IBI area) from June 2011 on. Then, Demyshev et al. are talking

about the Black Sea MyOcean physics products used to investigate the Black Sea climatic changes. Results of the regional model improvement

and NEMO implementation in the Black Sea are also discussed. Finally, O’Dea et al. are presenting the next system for the Atlantic- European

North West Shelf- Ocean that will be part of the MyOcean catalogue from June 2011 on. A

The next January 2011 newsletter will also display papers about the Myocean project, focusing this time on the ecosystem products.

We wish you a pleasant reading!

Quarterly Newsletter #39 – October 2010 –

Mercator Ocean

Contents

The FERRYBOX component in MYOCEAN ......................................................................................................... 3

By Dominique Durand, Are Folkestad, Kai Sørensen

The new regional generation of Mercator Ocean system in the Iberian Biscay Irish (IBI) area.......................... 5

By Sylvain Cailleau, Jérôme Chanut, Bruno Levier, Claire Maraldi, Guillaume Reffray

The MyOcean Black Sea from a scientific point of view.................................................................................. 16

By Sergey Demyshev, Vasily Knysh, Gennady Korotaev, Alexander Kubryakov, Artem Mizyuk

NEMO-Shelf, towards operational oceanography with SST data assimilation on the North West European Shelf

....................................................................................................................................................................... 25

By Enda J. O’Dea, James While, Rachel Furner, Patrick Hyder, Alex Arnold, David Storkey, John R. Siddorn, Matthew Martin,

Hedong. Liu, James T. Holt

Notebook....................................................................................................................................................... 29

Mercator Ocean Quarterly Newsletter #39 – October 2010 –

The FERRYBOX component in MYOCEAN

The FERRYBOX component in MYOCEAN By Dominique Durand 1, Are Folkestad 2, Kai Sørensen 2

1Norwegian Institute for Water Research (NIVA), Bergen, Norway 2 Norwegian Institute for Water Research (NIVA), Oslo, Norway

Ships of opportunity have been used for many years to observe the ocean and coastal seas and to complement monitoring

capabilities of the marine environment. In the last 10 years a rapid improvement of ocean observing systems has taken place, in

which observations from existing commercial ships such as ferries and cargo ships have raised increased interest. Nowadays

environment sensor package onboard of ship-of-opportunity are often referred to as Ferrybox system. The most advanced

Ferrybox integrate measurements of physical, chemical and biological parameters of the marine environment, and observations of

optical properties of the ocean and atmosphere. These systems enable to observe at high frequency, and along repetitive tracks,

the most important ocean parameters in surface waters, and to deliver them in quasi-real time to users.

MyOcean is the implementation project of the GMES Marine Core Service, aiming at deploying the first concerted and integrated

pan-European capacity for Ocean Monitoring and Forecasting. In this context, Ferrybox data present an important source of in-situ

data. Within MyOcean, a great effort is made to facilitate access to Ferrybox data as one of the key source of operational in-situ

data. Special attention is given to harmonization and quality control of data both in real time mode and in consolidated delayed

mode. Qualified Ferrybox data will contribute to validation of models and satellite observations.

At the present stage, several European

institutions deliver Ferrybox data

operationally to the MyOcean system,

encompassing physical (temperature

and salinity) and biochemical (e.g.

chlorophyll and oxygen) measurements.

However, MyOcean’s ambition is to

gather and deliver all Ferrybox data

available from all European waters

(Figure 1). Any institution performing

Ferrybox measurements are therefore

invited to deliver their data to MyOcean

and thereby contribute to this common

effort of providing the best possible data

network.

Ferrybox data within MyOcean are

handled by the In-Situ Thematic

Assembly Centre (TAC). The role of the

In-Situ TAC is to collect oceanographic

measurements from all possible

sources, and to calibrate, validate, edit,

archive and distribute them.

Figure 1 - Map of Ferrybox systems in

Europe operated by NIVA (NO), GKSS

(GER), IMR (NO), BCCR/UoB (NO),

NOCS (UK), POC (UK), Marlab/FRS

(UK), NIOZ (NL) SYKE (FIN), SMHI

(SWE), TTU (EST), LOMI (EST).The

ambition of MyOcean is to gather and

deliver all Ferrybox data available from

all European waters.


The FERRYBOX component in MYOCEAN

Ferrybox data originating from the various lines and operators are transferred to a common server where data are harmonized

with regards to data and metadata formats and standards. The approach followed in MyOcean is that data delivery should require

as little effort as possible for the data provider. Therefore the Ferrybox data provider is not required to perform any pre-processing

of the data, and the data and metadata can be provided in the format in which it is delivered by the Ferrybox system. Automatic

transfer of data can be set up either directly from the ship, from a local ftp server or from another centralized site (e.g. Oceansite).

All Ferrybox data delivered to MyOcean are imported to a common global database hosted by the Norwegian Institute for Water

Research (NIVA). Quality controlled data are thereafter exported in OceanSites NETCDF format and delivered to the central

MyOcean In Situ TAC database at Ifremer, within 24 hours after data acquisition. Ferrybox observations will thereby be made

available to the MyOcean Monitoring and Forecasting Centres for assimilation into models and for model validation. Data are also

made available to users of the marine core service under special agreements.

MyOcean is developing quality control procedure for real time data and consolidated assessment procedure for time series of

data. The procedures apply also to Ferrybox data, and necessary adjustments have been made to match the specificity of

Ferrybox systems. Real time quality control procedures are built on the heritage from previous efforts performed by e.g. the

ARGO, GOSUD, MERSEA, SeaDataNet, and PABIM projects. Data provided are then flagged following a simple 4-class quality

scale. Data products are also to be available in delayed mode, after having been screened by a more thorough examination with

regards to calibration, manual quality control and a check of the performance of the automatic flagging procedures.

In summary, the MyOcean project facilitates operational delivery of Ferrybox data from any vessel on a standard format and with

harmonised quality control system and flags. Furthermore, a common place for collection and distribution of Ferrybox data is of

high benefit for the whole marine community, including the data providers and the data users. Ferrybox data products

(temperature, salinity, oxygen and chlorophyll) will be available from the MyOcean portal in near real time from June 2011.

Delayed time products will be made available later on, around December 2011.


The new regional generation of Mercator Ocean syste m in the Iberian Biscay Irish (IBI) area

The new regional generation of Mercator Ocean syste m in the Iberian Biscay Irish (IBI) area By Sylvain Cailleau 1, Jérôme Chanut 1, Bruno Levier 1, Claire Maraldi 2, Guillaume Reffray 1

1 Mercator Ocean, Toulouse, France 2 LEGOS, Toulouse, France

Introduction

During the past years, in the framework of French project (Reffray et al. 2008) and European projects such as ECOOP (

www.ecoop.eu ), EASY ( www.project-easy.info ) and mainly now MyOcean ( http://www.myocean.eu.org ), Mercator Ocean has

developed its own regional system on the Iberian Biscay Irish area (henceforth called the IBI system). Outputs from the IBI system

will be distributed through the MyOcean data access portal from June 2011 on and provide an unprecedented high resolution

picture of ocean forecasts fields for the whole IBI area to end users for several applications such as fishery, search and rescue, oil

spill drifting, routing, research, education … among other. Secondly, this system will answer the increasing needs of partners who

deal with shelf and coastal ocean modelling. Their local systems will use as boundary and initial conditions the IBI system. The

quality of forecasts of these local systems also depends on the quality of their ocean forcings.

Robustness has been an essential driver in the design of the IBI operational system. To ensure service completeness and

timeliness, two identical systems will be operated redundantly by MyOcean IBI core partners (Puertos del Estado and Mercator

Ocean). The nominal system will be operated in Puertos del Estado (Spain) and the back-up one in Mercator Ocean. Thus if the

nominal system can not provide data in real time, the back-up one can take over. A strong collaboration between Mercator Ocean

and Puertos del Estado has been carried out in order Puertos del Estado to install and run the IBI system on their machine with

the same configuration as the Mercator Ocean one. For the MyOcean V1 stream2 version (June 2011), Puertos del Estado will

also replace the ESEOAT system used for Myocean V0 version on the IBI area by the new IBI system developed at Mercator

Ocean. Other partners involved in the IBI system in the framework of the MyOcean project are the Laboratoire d’Etudes en

Géophysique et Océanographie Spatiale (France) and the Proudman Oceanographic Laboratory (UK).

In conclusion, the new Mercator Ocean regional IBI system will complete all the other Mercator systems and will be the MyOcean

V1 stream2 system in the IBI area from June 2011 on. It will hence replace the Puertos del Estado ESEOAT system used during

the MyOcean V0 and V1 stream1 versions (from April 2009 untill June 2011). The upgrade of the IBI system for MyOcean V2 is

expected at the end of the MyOcean project early 2012.

This paper presents the IBI system developed at Mercator Ocean. A first part is dedicated to the description of the model

configuration including the latest coastal physics and new forcings, as well as the operational protocol based on an initialization

from the 1/12° resolution North Atlantic PSY2 Merca tor Ocean system. A second part concerns the validation which compares

system and observations and shows that the IBI system displays better scores than the ESEOAT system and better scores than

the North Atlantic Mercator Ocean systems (PSY2v3 and its release PSY2v4).

System description

Numerical setup

The numerical set up of the IBI forecasting system is based on the v2.3 NEMO/OPA9 Ocean General Circulation Model (Madec et

al. 1998; Madec 2008). The numerical core is very similar to other large scale forecasting systems in operation at Mercator

Océan, so that only important changes are described hereafter (ie: the main differences Table 1 - Model comparison between the

North Atlantic PSY2v3 and the Iberian Biscay Irish (IBI) Mercator Ocean systems are given in Table 1). These changes arised

from the need to improve the model physics on the shelf, since NEMO has been originally designed for the deep ocean. Proper

simulation of tidal waves implied the replacement of the default “filtered” free surface formulation (Roullet and Madec, 2000) by a

time-splitting procedure (implemented in the form proposed by Shchepetkin and McWilliams (2005)): the barotropic part of the

dynamical equations is integrated explicitly with a short time step while depth varying prognostic variables (baroclinic velocities

and tracers) that evolve more slowly are solved with a larger time step. This choice is indeed required to properly simulate tidal

waves without unrealistic damping (Levier et al. 2007). In addition, the default linear free surface approximation has been relaxed.

In practice the vertical coordinate is rescaled from the sea level height, becoming the so-called ‘z*’ coordinate as described by

Adcroft and Campin (2004). On a theoretical point of view, this is indeed required whenever the sea level variations are of the



order of the local water depth, which is often the case on the shelf due to the occurrence of large tidal amplitudes. This also adds

non linear feedbacks in dynamical equations that greatly enhance compound tidal waves and bottom stresses, thus improving the

representation of tidal waves in coastal regions.

Vertical mixing processes on the shelf are of primary importance, in particular because they dominate a large part of the water

column. This is mainly explained by tidal motions that maintain a significant level of turbulence by enhancing bottom stresses. This

induces bottom boundary layers over a large part of the water column, that eventually intersect surface boundary layers. River

plumes generation and their subsequent spreading, another fundamental physical process on the shelf, are largely controlled by

subtle mixing processes. Although simple and less costly one equation turbulence models (such as the default “TKE” closure used

in NEMO) have often been used with success on the shelf, two-equations models appear more adapted to the various processes

mentioned above. After extensive testing, the turbulence closure retained here is a k- ε version of the generic length scale (GLS)

formulation (Umlauf and Burchard, 2003) with the Canuto A stability functions (Canuto et al. 2001).

PSY2v3 IBI

Domain North Atlantic North East Atlantic

NEMO code version 1.09 2.3

Resolution 1/12° 1/36°

Time step 720 s 150 s

Vertical coordinate Z (50 levels) Z* (50 levels)

Free surface Explicit, filtered Split explicit

Vertical mixing 1.5 TKE closure

(Gaspar, 1990)

k-ε

(Umlauf and Burchard, 2003 )

Tracer horizontal advection TVD Quickest (Leonard, 1979)

Tides No Yes (including potential)

Short wave radiation penetration 2 bands scheme. Constant water type I 2 bands scheme. Variable

climatological PAR absorption depth.

Atmospheric forcing Daily ECMWF outputs 3h ECMWF outputs including

atmospheric pressure forcing

Open boundaries No (Buffer zones) Daily outputs from PSY2v3 system

Initial conditions Levitus climatology Output from PSY2v3

Table 1 - Model comparison between the North Atlant ic PSY2v3 and the Iberian Biscay Irish (IBI) Mercat or Ocean systems.



Model configuration

The IBI model configuration covers the North East Atlantic ocean from the Canary Islands to Iceland. It encompasses the Western

Mediterranean Sea and the Skagerrak Strait that connects the Baltic Sea to the North Sea. The primitive equations are discretized

on a 1/36° (~2 km) curvilinear grid extracted from the global ORCA tripolar grid used by other Mercator systems. This makes the

chosen grid an exact refinement by a factor 3 of the North Atlantic system that provides initial and boundary conditions. The

reference vertical grid has the same 50 geopotential levels as other Mercator systems, with a resolution decreasing from ~1 m

near the surface to more than 400 m in the abyssal plain. A partial step representation of the bathymetry is used which improves

the model solution in the North Atlantic Ocean (Barnier et al. 2006). The bathymetry is derived from the 30 arc-second resolution

GEBCO 08 data set (Becker et al. 2009) merged with regional bathymetries provided by IFREMER, the French Navy (SHOM,

Service Hydrographique et Océanographique de la Marine) and the NOOS community (North West Shelf Operational

Oceanographic System, http://www.noos.cc) (work in collaboration with F. Lyard, LEGOS). Local grids cover the Mediterranean

Sea, the Gibraltar Strait, regions off Portugal and off North Spain, regions along the French coasts, the Southern Celtic Sea, part

of the North Sea and the Baltic Sea. The new bathymetry as well as GEBCO 08 were confronted to independent ICES

bathymetric data (http://www.ices.dk) to illustrate significant improvements in shelf sea regions.

Forcing

Meteorological fields from the European Centre for Medium-Range Weather Forecasts (ECMWF) with a 3 hour and 0.25°x 0.25°

resolution are used to force the model. According to Bernie et al. (2005), this temporal resolution is enough to simulate diurnal

variations of Sea Surface Temperature. Wind stresses and heat surface fluxes are computed from CORE bulk formulae (Large

and Yeager, 2004) using a set of atmospheric variables that consists in air temperature, relative humidity, atmospheric pressure

and the wind components at 10 meters. Radiative heat fluxes, freshwater flux and atmospheric pressure are also used to force the

momentum and scalar equations. Short wave radiation penetrates the surface layer according to an extension law whose

extinction coefficient is dependent of the ocean color. This is necessary to correctly simulate the diurnal cycle in the surface

oceans (Dai and Trenberth, 2004). In the model we use a monthly ocean colour climatology based on SeaWiFS data. In addition

to atmospheric forcing, the model includes astronomical tidal forcing. It also includes river fresh inputs from a monthly runoff

climatology built by averaging data from the Global Runoff Data Center (http://grdc.bafg.de) and the French hydrographic

database “Banque Hydro” (http://hydro.eaufrance.fr). Finally, 35 major rivers are taken into account and applied at lateral open

boundaries over the model area.

Initial and open boundary conditions (temperature, salinity, velocity components and sea surface height) are originally derived

from the Mercator Ocean operational system over the North Atlantic at 1/12° (PSY2v3) (Hurlburt et al. 2009) solution and have

been recently switched to the new PSY2v4 system. The new PSY2v4 system is operated in real time since December 2010

(MyOcean V1stream1 version). The PSY2v4 system benefits from many enhancements: multivariate data assimilation with

incremental analysis updates method of bias correction, ECMWF 3-hourly forcings. In particular the incremental analysis update

of the assimilation scheme allows us to restart from an equilibrate state, which was not the case with the PSY2v3 system.

As both PSY2 systems do not include tidal forcing and atmospheric pressure forcing, these signals are added at the open

boundaries. Tidal open boundary data for 11 constituents (M2, S2, K2, N2, K1, O1, P1, Q1, M4, Mf, Mm) are provided according

the protocol described by Chanut et al. (2008). Elevations due to atmospheric pressure static effects, also known as inverse

barometer effects (Wunsch and Stammer, 1997) are computed from the ECMWF pressure fields.

Operational Protocol

The downscalling methodology used here and depicted in figure 1 is inherited from the strategy developed for the ESEOAT

system (Sotillo et al. 2007). Every week, the regional system is initialized in the past from analysed outputs (3d temperature,

salinity velocities and sea-level) taken from the PSY2 system and bilinearly interpolated on the refined grid. The model is then

integrated until D0 to allow the spin up of small scales and the convergence of physical processes that are not resolved by the

parent system. Impact on the forecasts results of the duration of this phase (typically a couple of weeks) is discussed in the next

sections. The analysed output at D0 is then used to provide one week forecasts until the next analysis stage. Note that in the

future operational context, the scenario will be somehow different: to account for the best available atmospheric forcing, a 5 days

forecast will be performed every day from D0 to D0+7.



Figure 1 - MyOcean V1stream2 (from June 2011 on) I BI system Operational protocol scheme

Results

In order to investigate the results of this operational protocol on the physics of the model, a three months simulation was

performed covering the spring 2009 period. Every week from the 1st April 2009 to the 17th June 2009 (this period of time is 13

weeks long), the model was restarted from initial conditions provided by PSY2v3 for a five weeks period (PSY2v4 was not yet

available at the time of this study). Thus we obtained 13 runs of 5 weeks. Each run could be decomposed as a 1 week spin-up

and 4 weeks forecast, or 2 weeks spin-up and 3 weeks forecast, or 4 weeks spin-up and 1 week forecast. In that way, the period

May-June 2009 was simulated four times, from one week spin-up to four weeks spin-up. This allowed us to investigate the impact

of the spin-up period on the model results. The second part of the letter deals with the results of this simulation.

Most of the results of the IBI simulation presented here come from the simulation with 2 weeks spin-up (2WSP hereafter),

corresponding to the operational protocol. The IBI simulation is compared to the Puertos del Estado ESEOAT system and the

MERCATOR PSY2 systems (PSY2v3 the parent system and its release PSY2v4). The outputs of PSY2v4 were not available

when the spring 2009 simulation was performed, and they are just used in this paper as a comparative criterion. However, it is

important to keep in mind that IBI will be nested inside PSY2v4 for the V1 stream 2 version of the MyOcean project (from June

2011on).

Validation Protocol

We present here some diagnostics to evaluate the behavior of the model depending on the number of weeks of spin-up. We first

present an illustration of the consequence of the spin-up duration on the Sea Surface Temperature representation. Then we

assess the frontal positions from Iroise to Irish Sea, and the tidal currents in the Iroise Sea. The stratification over the Bay of

Biscay is investigated. Finally we compare the SST representation by the different systems, and also the residual currents at

buoys.

Figure 2 presents the RMS of the Sea Surface Temperature difference between high resolution satellite-based observations (L3

multi-sensor product from Météo France CMS center in Lannion, France) and the model, for (a) the IBI 1WSP (one week spin-up)

simulation and (b) the IBI 4WSP (four weeks spin-up ) simulation, for the period May-June 2009, for the Bay of Biscay and

Channel region. For the IBI 1WSP simulation, the largest errors occur in the Channel and along the shelf break, linked with the

tides dynamics, and north of the Galicia coast probably due to the initialization temperature field. The IBI 4WSP simulation exhibits

substantial error decrease in the Channel. The RMS error pattern north of Galicia has disappeared. The RMS error is also

reduced over the French shelf. On the other hand, the RMS error is slightly increased along the shelf break probably due to a lack

of internal tides mixing; it also increases close to the Landes coast. In some specific areas (in particular areas where the tides are

important), a long spin-up time allows us to reduce the RMS error. Figure 2c presents the RMS error evolution by regions and for

each IBI simulations (from 0 to 4 weeks spin-up). As expected, the RMS decreases strongly with the 1WSP simulation (except in

the Mediterranean Sea) thanks to the improved physics of the model compared to the parent model. Then the model drifts and the

RMS error tends to slightly increase (except in the Bay of Biscay region).



a) b) c)

Figure 2 - RMS (observation minus model) Sea Surfac e Temperature (SST) difference (°C) for the period May-June 2009

for (a) the 1 week spin-up IBI simulation and (b) t he 4 weeks spin-up IBI simulation, for the Bay of B iscay and Channel region. (c) Spatial average RMS (observation minus model) SST difference (°C) for the period May-June 2009 for the 0 to

4 weeks spin-up IBI simulations, for the IBI domain region (GLOB), the Iberic Peninsula region (IBER: 13°W-5°W, 34°N-

45°N), the North Sea region (NSEA: 12°W-14°E, 48°N -62°N) and the Mediterranean Sea region (MEDI: from Gibraltar strait to 9°E).

Frontal positions

From the Sea Surface Temperature (SST) gradient we can simply estimate the positions of thermal fronts. In order to make a

direct comparison of model results with observations, we interpolate in space and time the model SST onto the observations grid.

We calculate then the temperature gradient along the x and y axis and sum their norms. Finally we map the maximum gradients

for the model and the observations, as seen in Figure 3 in the Channel and Celtic Sea region. These fronts are located at the

boundaries between stratified areas and mixed areas, where the tides interact with bathymetry. On June 24th 2009, the IBI 2WSP

simulation and satellite-based observations (L3 multi-sensor from Météo France CMS) show a very good agreement (Figure 3).

Almost all the main fronts are reproduced at the right position by the model, along the Brittany coast, across the Channel, or along

the Cornwall coast. The main discrepancy comes from the front crossing from Ireland to Whales: this front goes deep into the Irish

Sea in the IBI 2WSP simulation. This is not the case with the IBI 4WSP simulation (not shown), which indicates (as seen in the

previous paragraph) that for this area a longer spin-up period is necessary to correctly reproduce the tidal dynamics. The PSY2v3

simulation (Figure 3c) and PSY2v4 (not shown) which do not include the tidal forcing can not reproduce the fronts.

a) b) c)

Figure 3 - Sea Surface Temperature (°C) on 24 th June with maximum horizontal gradients of SST over laid for (a)

observation, (b) IBI and (c) PSY2v3.



Tidal currents validation: comparisons with HF rada r

In order to validate the representation of tidal currents in the model, we use surface current data from the High Frequency WERA

(Wellen Radar) radars (Gurgel et al. 1999) at the Brezellec headland, 48°04N/4°40W, and at the Garchin e headland,

48°30N/4°46W (Figure 4). This system is operated by SHOM and provides radial surface currents every 20 minutes with a

resolution of 1.5 km and an accuracy of few centimeters per second (Le Boyer et al. 2009). The resulting radial currents have

been interpolated on a regular grid with a 2 km spatial resolution extending from 6°78W to 4°65W and f rom 47°30N to 49°26N

(Muller et al. 2009). Given intense tidal component dominating currents over the Iroise Sea, the 15 cm.s-1 uncertainty still allows

us to compute the tidal current components (Le Boyer et al. 2009). We used a 6 month dataset to perform the harmonic analysis

of 9 tidal components (M2, S2, K2, N2, K1, O1, P1, Q1 and M4). Comparisons with radar HF in the Iroise Sea allow us to

investigate the ability of the model to reproduce tidal surface ellipses in a region characterized by intense tidal currents. We use

the three months simulation of the IBI 2WSP to perform the harmonic analysis.

a) b)

Figure 4 - Modeled (black ellipses) and observed (r ed ellipses) tidal ellipses for (a) the M2 and (b) M4 components. The background colour fields (in correspondence with th e colour palette) represent the difference between the observed and

modeled semi major axis (cm/s). Observations come f rom coastal HF radar.

Figure 4a represents the observed and modelled tidal ellipses and the difference between the semi major axis (background fields)

for the M2 component. The tidal currents are globally well reproduced. The model semi-major axes are slightly overestimated

almost everywhere but with discrepancies depending on the location. The phase and inclination differences are higher on the

northern and eastern edges of the measurements areas. Discrepancies may results from both data quality and modeling.

Projection of radar HF velocities onto a Cartesian grid may increase errors due to the radar orientation. In addition, intersection of

beams at small angles may affect the precision of Cartesian components of the current vectors (Chapman et al. 1997; Sentchev et

al. 2009). The orthogonality of radar beams to dominant current direction may also decrease the data quality at far ranges

(Sentchev et al. 2009). Radar measurements accuracy is also limited in the vicinity of islands. Thus radar data may not be reliable

on edges and near islands which may explain errors over these areas. The M4 component is more difficult to be modeled but the

tidal currents are globally well reproduced (Figure 4b). They are generally underestimated and the major discrepancies are

located, as for M2, in the vicinity of islands.

Stratification over the Bay of Biscay shelf

We use the in-situ profiles measurements from the oceanographic cruise PELGAS 2009 (IFREMER; the data have been extracted

from data holdings at the SISMER data centre) accomplished in the Bay of Biscay in May 2009 for comparisons with IBI and

PSY2 models. The data are irregularly distributed in space and time and we used a collocalisation tool (adapted from Juza 2008)

to build modeled profiles at the same location and same day than observed profiles. Observed and modeled profiles are then

interpolated on a 3D grid for visualization. Figures 5 and 6 present the interpolated temperature and salinity fields for the

observations, the IBI model and the PSY2 models. Each profile’s date is different so the fields are not snapshots of a particular



day but for whole period of May 2009. The ship track goes from the South to the North. At the end of May, measurements were

made again in the South of the shelf (the results of this part of the cruise are presented in little window).

The observed thermal stratification shows two minima (Figure 5). One along the Aquitaine coast: the measurements were made in

this area at the Beginning of May when the thermal stratification is not yes established; surface waters in this area can also be

impacted by the Gironde plume. The other minimum is observed west of the Brittany coast, in the Ushant front; in this area, the

thermal stratification can not set up due to the strong tidal mixing. A maximum of thermal stratification is observed near 47°N,

2.5°W. At the end of May (small window on Figure 5) the thermal stratification is set up and difference between surface and

bottom can reach 5°C over the shelf. The IBI simula tion is closed to the observations: the minimum and maximum are well

reproduced. The thermal stratification is slightly over-estimated offshore the 200 m isobath. The PSY2v3 and PSY2v4 simulations,

which do not use the tidal forcing, do not reproduce the minimum of stratification in the Ushant front. The thermal stratification is

too weak for these two simulations.

The observed haline stratification (Figure 6) shows an along-shore gradient over the shelf with a maximum in the Gironde estuary,

a secondary maximum near the Loire estuary, and a minimum near Brittany. The difference between surface and bottom salinity is

negative all over the shelf until Brittany. This is due to the rivers plumes which spread over the shelf. In the southern part of the

shelf, the IBI simulation reproduces the haline stratification pattern, but the Gironde estuary maximum is under-estimated. The

positive difference near Brittany and offshore the 200 m isobath is not reproduced. However, the IBI simulation slightly improves

the results compared to the PSY2v3 one by which it was initialized. The PSY2v4 simulation reproduces well the north-south

gradient of haline stratification, but the maximum area is too close to the coast and not wide enough.

As a conclusion, the IBI system ability to reproduce haline stratification is reduced by its initial conditions. The thermal stratification

is well reproduced, despite problem in the Gironde plume near field due to a too strong mixing.

a) b) c) d)

Figure 5 - Surface minus near bed temperature diffe rence (°C) for (a) observations, (b) IBI, (c) PSY2v 3 and (d) PSY2v4.

a) b) c) d)

Figure 6 - Same as Figure 5 but for salinity (psu).

Sea Surface Temperature

Figure 7 represents the difference of the Sea Surface Temperature between satellite-based observations (Météo France CMS L3

multi-sensor SST) and model outputs for the period May-June 2009. Compared to PSY2v3, the IBI 2WSP simulation reduces the

errors almost everywhere, except along the shelf slope and around the Cape St Vincent. Compare to the ESEOAT system, the

errors are reduced over the whole area. Table 2 gathers statistics calculated for the three models for different areas of the IBI

region for the same period; the IBI 2WSP simulation presents most of the best scores.



Difference (°C) Correlation RMS error (°C)

IBI PSY2v3/v4 ESEOAT IBI PSY2v3/v4 ESEOAT IBI PSY2v3/v4 ESEOAT

BOBI -0.03 0.30/0.30 0.52 0.43/0.44 0.63 0.73/0.73

CANA -0.08 0.05/0.09 0.64 0.52/0.66 0.59 0.55/0.55

ESEO -0.07 0.25/0.13 -0.35 0.55 0.49/0.51 0.47 0.61 0.65/0.60 0.73

IBER -0.08 0.19/0.08 -0.47 0.53 0.47/0.48 0.45 0.65 0.67/0.61 0.78

ISHF -0.19 .03/-0.02 -0.50 0.48 0.44/0.48 0.42 0.89 0.84/0.77 1.10

MEDI -0.11 -.04/-.37 0.65 0.65/0.54 0.69 0.66/0.85

NSEA -0.11 0.23/0.43 0.70 0.64/0.62 0.60 0.73/0.84

GLOB -0.07 0.16/0.09 0.59 0.53/0.53 0.58 0.62/0.64

Table 2 – Spatial average (observation minus model) Sea Surface Temperature difference (°C) (left), co rrelation (middle)

and RMS (observation minus model) SST error (°C) ( right) for the period May-June 2009 in the Bay of B iscay region (BOBI: 11°W-3°E, 43°N-51°N), the Canary Islands reg ion (CANA: 21°W-8°W, 25°N-33°N), the ESEOAT region (ESEO: 14°W-

0°E, 33°N-47°N), the Iberic Peninsula region (IBER: 13°W-5°W, 34°N-45°N), the Iberic Peninsula shelf r egion (ISHF: IBER

region over the shelf), the Mediterranean Sea regio n (MEDI: from Gibraltar strait to 9°E), the North S ea region (NSEA: 12°W-14°E, 48°N-62°N) and the IBI domain region (GL OB). Best scores are in bold. We subtracted the lin ear trend of the

time series before calculating the correlations.

a) b) c) d)

Figure 7 - Sea Surface Temperature difference (°C) for the period May-June 2009 for (a) PSY2V3, (b) PS Y2V4, (c) IBI and (d) ESEOAT. Observations come from CMS L3S SST. The Mediterranean Sea is masked. ESEOAT model does not cover

the region north of 48°N.

Residual currents

We compare observed currents measurement time series from the Puertos del Estado deep sea network (http://www.puertos.es)

with currents from the models outputs. Buoys considered hereafter are located around the north-west Spanish coasts: Cabo

Silleiro (9.39°W, 42.13°N), Villano-Sisargas (9.21° W, 43.5°N), Estaca de Bares (7.62°W, 44.06°N), Cabo de Penas (6.17°W,

43.74°N). In order to remove the tidal signal from the current time series, we perform an harmonic analysis on ESEOAT and IBI,

but not on the PSY2 systems which do not include the tidal forcing and which outputs are daily averaged. Figure 8 represents the

zonal and meridional components of the near surface residual current (hourly measurements, one-day filtered and smoothed) at

(a) the Cabo Silleiro and (b) Estaca de Bares buoys. The Cabo Silleiro buoy is located on the west Spanish coast (north of

Portugal boundary) above the continental slope (above the 600 m isobath) which is oriented north-south, so the meridional

component is equivalent to the along-shore component, and the zonal component is equivalent to the cross-shore component.

This is the contrary for the Estaca de Bares buoy which is located north of Cap Ortegal above the continental slope (above the

400 m isobath). During the May 2009 period, both IBI and ESEOAT systems are well correlated with observed measurements,

especially for the cross-shore component. The along-shore component is over-estimated by ESEOAT while IBI under-estimates it.

Table 3 presents the RMS error for the zonal and meridional components of the current at different locations. The RMS error is

calculated with ESEOAT and IBI hourly currents and PSY2V3 and PSY2V4 daily currents. Detrended correlations are presented

in table 4. ESEOAT and IBI show similar results, the RMS of ESEOAT is smaller and correlation values are close.



a) b)

Figure 8 - Time series of zonal (top) and meridiona l (bottom) components of the current (m/s) at 3 met er depth for the

period May-June 2009 at (a) Cabo Silleiro buoy and (b) Estaca de Bares buoy. Observation is in black, ESEOAT in red, IBI

in blue and PSY2V3 in green.

ESEOAT IBI PSY2V3 PSY2V4

Cabo Penas 7/6 7/7 18/9 12/8

Cabo Silleiro 5/7 5/9 8/10 5/8

Estaca bares 9/5 9/5 22/10 12/5

Villano Sisargas 6/9 11/9 14/9 11/7

Table 3 - RMS difference observation-model (cm/s) f or zonal/meridional component of currents at 3 mete r depth for the

period May-June 2009 at different buoys. Statistics for ESEOAT and IBI are calculated with 1 day filte red residual

currents. Statistics for PSY2v2 and PSY2V4 are calc ulated with daily averaged currents.

ESEOAT IBI PSY2V3 PSY2V4

Cabo Penas 66/0 73/8 54/-31 59/-4

Cabo Silleiro 64/40 67/39 56/67 68/32

Estaca Bares 44/63 52/72 69/24 68/57

Villano Sisargas 62/26 40/40 49/25 54/67

Table 4 - Same as table 3 but with correlations.

Conclusion

In a first part, this paper has presented the description of the new Iberian Biscay Irish (IBI) Mercator system which has been

chosen as next version of IBI MyOcean system. A strong effort of development has been carried out to add further coastal physics

in the ocean code NEMO usually used for larger model configuration. Now 11 tidal components can be taken into account thanks

to a new non-linear free surface, the time splitting which permits to separate high frequency barotropic signals and baroclinic

signals, and new formulations of open boundary conditions. The K-epsilon turbulent scheme has improved the vertical mixing too.

High frequency 3-hourly atmospheric fluxes have been used to force surface model. And a special hand made merged bathymetry

has been generated from different regional data basis. The 1/36° horizontal resolution has allowed the model to resolve

mesoscale and sub-mesoscale structures and specially improve the representation of local front areas. For the operational

scenario, the system is initialized by the Mercator North Atlantic system PSY2. All these developments have allowed the model to

adapt itself to coastal dynamics.

In a second part, this paper has shown the validation of the IBI system against data and other systems: ESEOAT from Puertos del

Estado (i.e the MyOcean V0 and V1stream1 version of IBI system from April 2009 untill June 2011), and PSY2, the system used



to initialize and force to the open boundaries the new IBI system. A comparison between observed and system (initialized with 0 to

4 weeks spinup) SST has shown a strong decrease in RMS during the first week due to better regional physics in the IBI system

against PSY2. In the end, a 2 weeks spinup has been finally chosen for the operational protocol. Besides, a better representation

of frontal zones in the IBI system is showing the impact of tide in the model. The Ushant thermal front is particularly well

represented in the IBI system against radar data. The same conclusion can be drawn for corresponding tidal ellipses.

A use of in-situ data (PELGAS 2009) has allowed validating the stratification on the Biscay shelf. The thermal stratification of the

IBI system corresponds to data, whereas the haline stratification is further away from observations probably because the IBI

system is too influenced by initialization of PSY2. The use of climatological monthly runoffs could also play a role. About SST

statistics (RMS and correlations) by subregions, The IBI system has shown globally the best results against ESEOAT and PSY2.

Regarding dynamics, the IBI system currents are better than in PSY2 but close to ESEOAT.

This new system is now almost ready to become the V1 stream2 version for MyOcean and should better answer the need of

coastal modellers (in term of initial and boundary conditions). Following the first conclusions from the validation and in order to

improve the system for the V2 version for MyOcean next year, several upgrades have been planned:

- MetNO (Norway) can provide to Mercator new wind stress which takes into account more wave components than the stress from

ECMWF presently used. As a result a more realistic vertical mixing should be reached.

- New daily runoffs from a data base used by PREVIMER and real time runoffs modelled by HYPE (from SMHI, Sweden) and/or

ISBA (from MétéoFrance) hydrological models should improve the haline stratification on shelves.

- Another operational scenario with a decreasing influence of PSY2 initial condition on shelves should improve the solution in

these areas since IBI system physics is more adapted.

- A part of the large scale SST bias seems to be triggered by large scale radiation fluxes bias. So a correction of ECMWF fluxes

with CMS satellite fluxes should reduce this bias.

References

Adcroft A., Campin J.M., 2004. Rescaled height coordinates for accurate representation of free-surface flows in ocean circulation

models. Ocean Modelling 2004, Vol 7, Issues 3-4, Pages 269-284.

BarnierB., Madec G., Penduff T., Molines J.-M., Treguier A.-M., Le Sommer J., Beckmann A., Biastoch A., Böning C., Dengg J.,

Derval C., Durand E., Gulev S., Remy E., Talandier C., Theetten S., Maltrud M., McClean J., De Cuevas B., 2006. Impact of

partial steps and momentum advection schemes in a global ocean circulation model at eddy-permitting resolution. Journal Ocean

Dynamics, 56(5-6), 543-567, doi .1007/s10236-006-0082-1.

Becker J. J., Sandwell D. T., Smith W. H. F., Braud J., B. Binder B., Depner J., Fabre D., Factor J., Ingalls S., Kim S.-H., Ladner

R., Marks K., Nelson S., Pharaoh A., Trimmer R., Von Rosenberg J., Wallace G., Weatherall P., 2009. Global Bathymetry and

Elevation Data at 30 Arc Seconds Resolution: SRTM30_PLUS. Marine Geodesy, 32(4), 355 - 371.

Bernie D. J., Woolnough S. J., Slingo J. M., Guilyardi E., 2005. Modelling diurnal and intraseasonal variability of the ocean mixed

layer. Journal of Climate, 18(8), 1190-1202.

Canuto V. M., A. Howard, Y. Cheng, M. S. Dubovikov, 2001. Ocean Turbulence. Part I: One-Point Closure Model—Momentum

and Heat Vertical Diffusivities, Journal of Physical Oceanography 2001 31:6, 1413-1426.

Chapman R. D., Shay L. K., Graber H. C., Edson J. B., Karachintsev A., Trump C. L., Ross D. B., 1997. On the accuracy of HF

radar measurements: intercomparison with ship-based sensors. Journal of Geophysical Research, 102, 18737-18748.

Dai A., Trenberth K. E., 2004. The Diurnal cycle and Its Depiction in the Community Climate System Model. Journal of Climate,

17(5), 930-951.

Egbert G., Bennett A., Foreman M., 1994. TOPEX/Poseidon tides estimated using a global inverse model. Journal of Geophysical

Research, 99(C12), 24821-24852.

Gaspar, P., Y. Grégoris, and J.-M. Lefevre, 1990 : A simple eddy kinetic energy model for simulations of the oceanic vertical

mixing Tests at station papa and long-term upper ocean study site. J. Geophys. Res, 95(C9).

Gurgel K.-W., Antonischki G., Essen H.-H., Schlick T., 1999. Wellen Radar (WERA): a new ground-wave HF radar for ocean

remote sensing. Coastal Engineering, 37, 219-234.

Hurlburt H. E., Brassington G B., Drillet Y., Kamachi M., Benkiran M., Bourdallé-Badie R., Chassignet E. P., Jacobs G. A., Le

Galloudec O., Lellouche J.-M., Metzger E. J., Oke P. R., Pugh T. F., Schiller A., Smedstad O. M., Tranchant B., Tsujino H., Usui

N., Wallcrat A. J., 2009. High-Resolution Global and Basin-Scale Ocean Analyses and Forecasts. Oceanography, 22(3).



Juza M., 2008. Collocation and validation tools for the DRAKKAR ensemble of simulations. Technical report, MEOM-LEGI,

Grenoble.

Large W. G. and Yeager S. G., 2004. Diurnal to decadal global forcing for ocean and sea-ice models: the data sets and flux

climatologies. NCAR technical notes.

Leonard, B. P., 1979. A stable and accurate convective modelling procedure based on quadratic upstream interpolation. Comp.

Meth. in Appl. Mech. and Eng., 19, 59-98.

Le Boyer A., Cambon G., Daniault N., Le Cann B., Marié L., Morin P., 2009. Observations of the Ushant tidal front in September

2007. Continental Shelf Research, 29(B), 1026-1037.

Levier, B., Tréguier A. M., Madec G., Garnier V., 2007. Free surface and variable volume in the NEMO code. MERSEA IP report

WP09-CNRS-STR03-1A, 47pp.

Madec G., Delecluse P., Imbard M., Lévy C., 1998. OPA 8.1 Ocean general circulation model reference manual. Institut Pierre-

Simon Laplace, 20, p. 91.

Madec G., 2008. NEMO Ocean General Circulation Model Reference Manuel. Internal Report. LODYC/IPSL, Paris.

Muller H., Blanke B., Dumas F., Lekien F., Mariette V., 2009. Estimating the Lagrangian residual circulation in the Iroise Sea.

Journal of Marine Systems, 78(1), S17-S36.

Reffray G., Levier B., Marsaleix P., Lazure P., Garnier V., 2008. Intercomparaison de modèles sur le Golfe de Gascogne pour

l’année 2004. Rapport SHOM/Mercator Océan.

Roullet G., Madec G., 2000. Salt conservation, free surface and varying levels: a new formulation for ocean general circulation

models. Journal of Geophysical Research, 105 (C10), 23927-23942.

Sentchev A., Forget P., Barbin Y., 2009. Residual and tidal circulation revealed by HVF radar surface current measurements in

the southern Channel Isles region (English Channel). Estuarine, Coastal and Shelf Science, 82, 180-192.

Shchepetkin A., McWilliams J., 2005. The regional oceanic modeling system (ROMS): a split-explicit, free-surface, topography-

following-coordinate oceanic model. Ocean Modelling, 2005, Volume 9, Issue 4, Pages 347-404.

Sotillo, M.G., Jordi, A., Ferrer, M.I., Conde, J., Tintoré, J., Álvarez-Fanjul, E., 2007. The ESEOO Regional Ocean Forecasting

System. Proceedings of the seventeenth (2007) International Offshore And Polar Engineering Conference, Vol 1 – 4, 1716 –

1722.

Umlauf L., Burchard H., 2003. A generic length-scale equation for geophysical turbulence models, Journal of Marine Research,

Volume 61, Number 2, 1 March 2003 , pp. 235-265(31)

Wunsch C., Stammer D., 1997. Atmospheric loading and the oceanic “inverse barometer” effect. Reviews of Geophysics, 35(1),

79-107.


The MyOcean Black Sea from a scientific point of vi ew

The MyOcean Black Sea from a scientific point of vi ew By Sergey Demyshev 1, Vasily Knysh 1, Gennady Korotaev 1, Alexander Kubryakov 1, Artem

Mizyuk 1

1Marine Hydrophysical Institute, Sevastopol, Ukraine

Abstract

The Black Sea MyOcean products are studied in this paper in order to investigate climatic changes. Results of the regional model

improvement and NEMO implementation in the Black Sea are also discussed.

The role of the MyOcean project for the Black Sea o ceanography

The Black Sea is a semi-enclosed basin of the oceanic type with simple configuration and a depth up to 2 km. Observations of the

Black Sea started at the end of 19th century. Hydrographic observations covered the basin area more or less uniformly in all

seasons between the end of fifties and mid of 1990-ies. Also were collected lots of interdisciplinary data concerning the Black Sea

biogeochemistry. Collapse of the Soviet Union has resulted in collapse of the Black Sea observing system. A new cost efficient

observing system is built by consolidated efforts of riparian countries in the framework of the Black Sea GOOS project. It is mainly

based on remote sensing, tide gauges and free drifting buoys data but includes also observation from platforms and vessels

surveys near the shore.

The development of operational forecasting system in the basin started in 2003 in the framework of the FP5 ARENA project. The

pilot system of the Black Sea forecasting was built and tested under the aegis of the project in summer 2005. Then its improved

version served as a support for coastal forecasting in the framework of FP6 ECOOP project. The MyOcean project which started

in 2009 plays a significant role for the development of the Black Sea operational oceanography. Regular high quality products

which are based on the processing of the Earth remote sensing data permit to consider evolution of SSH, SST and sea colour

fields with high resolution. New insight to the Black Sea dynamics and ecosystem is provided by the MyOcean project. The Black

Sea nowcasting and forecasting products are already available from the MyOcean catalogue (http://catalogue.myocean.eu.org).

The Black Sea reanalysis products which will be available from the MyOcean catalogue in June 2011 are partially presented in

this paper.

Thus MyOcean provides a set of new regional products which make possible to consider the Black Sea basin as a whole.

Particularly, Reanalysis of the Black Sea dynamics provides a powerful tool for the investigation of the climatic changes in the

basin and their influence on the marine ecosystem. Semi-enclosed nature of the basin provides good opportunity for the different

component budget estimations. New operational products provide another important contribution to the process studies in the

basin including multidisciplinary investigations. The study of small-scale phenomena or biogeochemical processes is possible now

under controlled background dynamics.

MyOcean is a user driven project. It means that the quality of the MyOcean products should satisfy user requirements. End –user

demands in the framework of MyOcean are translated to the research community requests for the improvement of operational

oceanography tools.

The Black Sea has a simple coast line and its hydrography is typical for the oceanic basin. Many observations have been

collected during more than century which makes possible providing assessment of efficiency of different models and assimilation

schemes. Therefore MyOcean has an opportunity to consider the Black Sea basin as a natural laboratory for the model

development and improvement.

Thus, the MyOcean project opens new field for the scientific problem formulation and for new knowledge in the region. This paper

presents scientific development of the Black Sea Monitoring and Forecasting Center (MFC). Its first part presents the Black Sea

dynamic reanalysis for 1971 – 1993 obtained by means of assimilation of archive hydrography in the Princeton Ocean model.

Then, in the second part of this paper, we present assessment of the improved version of circulation model which will be used for

operational nowcast and forecast starting June 2011. Comparison results of two numerical simulations with observations

demonstrate the model skill for reproduction of the upper layer thermodynamics and structure of the permanent pycnocline.

Finally, the third part of the paper presents the transition occurring within MyOcean towards the use of the NEMO code to simulate

the Black Sea circulation.



Part 1 : MyOcean Black Sea contribution to investig ation of the long-term basin evolution - Reanalysis of the Black Sea dynamics fo r the study of climatic changes

A Reanalysis simulation of the Black Sea dynamics covering the time period 1971-1993 on a grid with resolution 8.1 km in zonal

and 6.95 km in meridional directions and 26 σ -surfaces has been carried out using POM (Princeton Ocean Model). Simulations

have been fulfilled with assimilation of monthly temperature and salinity arrays optimally interpolated on the model grid. The

regular seasonal climate of the Black Sea circulation and hydrography constructed by Knysh et al. (2005) was used as the base

field to interpolate relatively regular in space and time hydrographic surveys of the Black Sea which were carried out between

1971 and 1993. Spatial and temporal autocorrelation functions for the climatic fields were estimated and approximated by ellipses

(Knysh et al. 2010) bearing in mind high correlation of temperature and salinity on climatic scales. The atmospheric forcing has

been compiled from the ERA-40 global reanalysis (Uppala et al. 2006).

Figure 1 - Total heat flux (ERA-40) (° С·m/s) (top panel) and distribution of basin-average d temperature within upper layer

0-300m (°С) (bottom panel).

Space-time diagram of the temperature averaged over the basin area presents variability of the temperature field in the upper 0-

300 m layer during 23 years (Figure 1, bottom panel). The analysis of the diagram permits to establish such physical processes of

water masses formation in the Black Sea as the autumn-winter cooling of the sea, the Upper Mixed Layer (UML) formation,

renewal of the Cold Intermediate Layer (CIL), spring-summer heating of waters, the seasonal thermocline and new CIL formation,

its heating and breaking.

The depth of UML boundary varies from 20 to 64 m (isotherm 7°С). The largest thickness of the UML is observed in 1976, 1985,

1987, 1989 and 1992. The seasonal thermocline which forms with the spring-summer heating is observed between the depths

from 10 to 40 m. The upper and lower boundaries of the CIL are usually identified according to the position of the 8°С isotherm.

Temperature of this layer increases till the autumn.

Years 1987 and 1989 are characterized by normal and years 1976, 1985, 1992 and 1993 by cold thermal conditions (Titov 2003).

During the latter ones, the thickness and intensity of the CIL is larger in cold years due to the low total downward heat flux during

winter (Figure 1). Abnormally high values of the downward heat flux also have an influence on the CIL behaviour. There was no

trace of cold waters in the basin-averaged temperature in the autumn of 1971, 1972, 1975, 1977, 1980–1982 and 1984.

Globally, the CIL thickness tends to increase from the end of 70-ies (Figure 1). The CIL thickness and intensity is lower in 1981,

than in 1985, as Figure 2 shows. The summer 1981 temperature on the 50 m depth (abnormally warm winter conditions, Titov

2003) is larger than in summer 1985 (cold thermal conditions). Figure 3 shows that an averaged temperature on the 50 m depth is



larger than climatic values within almost the whole year (except October - December). During cold year of 1985 temperature is

lower than the climatic ones.

The analysis of annual mean temperature averaged in different layers is presented in Figure 4. The linear trends in the layers 0-50

and 50-100 m are negative (Figure 4b) in consistency with the negative trend of the total downward heat flux (Figure 4a). Trend in

the layer 100-300 m is slightly positive (Figure 4c). Deeper (500-1000 m layer) trends are positive (not shown).

The structure of the temperature fields in the Black Sea is influenced by several factors: heat fluxes through the interface sea-

atmosphere and through lateral boundaries, lateral and vertical mixing.

Figure 2 - Temperature (° С) cross sections across 43° N in summer 1981 (a) an d 1985

Figure 3 - Annual cycle of temperature (° С) at 50 m depth. Figure 4 - Interannual variability and its linear trend

(red) of: annual mean total downward heat flux

(°С·m/s) (a), annual mean temperature (° С) at 50m (b) and 200m depth (c).

Part 2. Towards high quality modelling of the Black Sea dynamics

The current MyOcean V0 version of the Black Sea MFC is based on the Marine Hydrophysical Institute (MHI) basin-scale primitive

equation circulation model. It is an eddy-resolving model with horizontal resolution 5x5 km and 38 non-uniformly spaced z-levels in

vertical. Finite-difference approximation of the model equations is done on the C-grid. The second order numerical scheme

conserves energy, carefully describes kinetic and potential energy exchange within every box and admits nonlinear dependence



of density from temperature and salinity. Pacanowski and Philander (1981) parameterization is used to describe turbulence in the

upper layer of the basin. Model testing has shown that the Pacanowski and Philander (1981) parameterization works well on the

seasonal scales but is insufficient during intense storm events. The improved version of the Black Sea circulation model is

supplemented by Mellor-Yamada (1982) turbulence closure which should have broader application and which is now on the stage

of testing and validation.

Two prognostic numerical simulations are conducted using the improved version of MHI circulation model. Tangential wind stress,

heat fluxes and precipitations and evaporation on the sea surface are the same as in operational model of the Black Sea MFC

(Ratner et al. 2005, 2006). The goal of simulations was to assess the model skill in the upper mixed layer and in the permanent

pycnocline.

The RUN1 begins from 1 January 2006 using the Black Sea MFC output as the initial conditions. Integration has continued until

the end of the year and included both detrainment and entrainment processes. The simulations were carried out with relaxation to

the measured SST on the sea surface to evaluate the quality of the upper mixed layer simulation

The RUN2 starts from 1 January 2007 also with initial conditions provided by the Black Sea MFC. Integration was carried out

during two years. Only heat flux was specified on the sea surface for the RUN2.

Model & 56093-drifter temperature profiles

Temperature (°C) Temperature (°C) Temperature (°C) Temperature (°C)

Figure 5 - Temperature (° С) profiles in the upper layer of the Black Sea. Bl ue line is observations and red line is model simulations.

The improved model validation is carried out by means of comparison of simulations with the free floating buoys data. The RUN1

simulation is compared with data of drifters with thermistor chain (Figure 5) allowing to consider the upper layer thermodynamics.

The model provides reasonable results both for detrainment and entrainment phases (Figure 5).

Figure 6- Temperature section (° С) along 43,7° N in the upper active layer of the Bl ack Sea simulated by new version circulation model during 2008: a – February 26, b – May 16, c – August 14, d – November 2.



Deep stratification is evaluated by the comparison of the RUN2 results with profiling floats data. The second year integration of

RUN2 results are presented on Figure 6. Intense convection occurs during the winter season, which penetrates down to the depth

of 100 m (Figure 6a). Cold intermediate layer (CIL) is formed on the depth of 80 m (Figure 6b) in spring. Core of the CIL with the

temperature of less than 7°С is clearly defined there. The near-surface warm layer is formed then in summer. The CIL is located

below it (Figure 6c). The temperature of CIL slowly increases and becomes higher than 7.5°С in the autumn keeping the

qualitative structure. The near-surface warm layer depth increases during the fall and that leads to the deepening of CIL core

(Figure 6d). Thus, the simulated upper layer stratification is consistent with observations.

Quantitative model validation was conducted on the base of comparison of the simulated temperature and salinity fields and the

floating buoy data (Korotaev et al. 2006). Two buoys (4900489 and 4900541) were in the deep sea zone, two other (4900540 and

4900542) moved along the Caucasus and Anatolian coast respectively. Let us mention that the dynamics in the coastal zone and

in the central part of the basin is strongly different. Hence the comparison of the simulated fields and buoys data allows presenting

the model skill sufficiently well. Salinity profiles for two of four buoys in January 2008 are shown in Figure7. The highest errors

take place on the sea surface and in the halocline area. The error is small and do not exceed 0.03 ‰ below 100 m. The largest

error in summer and spring takes place in the upper 50-meter sea layer. The error profile in the autumn is similar to that in winter

season. Let us mention that the model describes the vertical structure of the halocline qualitatively well and keeps it during two

years of integrations.

Salinity (psu) Salinity deviations (psu) Salinity (psu) Salinity deviations (psu)

Station: Buoy 4900489. Latitude=41.5(N). Longitude=37.8(E).

Date=2007-02-08. Time=07:06(UTC) 19.0,06.0 =−= ss σµ

Station: Buoy 4900542. Latitude=41.4(N). Longitude=39.6(E).

Date=2008-01-27. Time=12:56(UTC) 24.0,24.0 == ss σµ

Figure 7 - Left panels present salinity (psu) profi les in the upper 1500m of the Black Sea simulated b y model (red line)

and measured by profiling floats (blue line). Right panels present the difference between the simulate d and observed

salinity (psu).

Similarly were considered the temperature profiles (not shown). Temperature may be different of several degrees at the surface in

winter. Data of all four buoys confirms good evolution of temperature profiles during the warming stage. The highest errors which

are less than 2 °C are observed on the sea surface at that time. The upper mixed layer, seasonal thermocline, CIL and the

permanent thermocline are well pronounced in the temperature profiles after two years of integration.

Assimilation of restricted number of temperature an d salinity profiles

Favourable situation with available hydrographic observations from 1971 to 1993 has turned to the worth in the following years.

Limited number of cruises was arranged after 1995. The launches of the PALACE profiling floats after 2002 has changed situation

to the best. However even for those years a small number of available temperature and salinity profiles does not allow applying

traditional data assimilation algorithms for the correction of simulations inside the permanent pycnocline. Therefore a new method

is elaborated to assimilate a limited number of temperature and salinity profiles. The method takes into account the Black Sea

specificities. Due to a sharp and narrow pycnocline, a broad range variability of temperature and salinity fields can be presented

by only one empirical orthogonal function (EOF) which describes about 80 percents of energy of oscillations. Therefore the



deviation of density (temperature or salinity) field from the basin-averaged stratification along the vertical is approximately

proportional to the sea surface elevation at the same point as the sea level topography can be evaluated with high accuracy by

dynamical method. Then if we have even one measured density (temperature or salinity) profile, it is possible to extrapolate it

along the sea level contour. The obtained pseudo-observations can be then assimilated by means of usual approach. Twin

experiments are carried out to evaluate efficiency of the above mentioned approach.

Figure 8 - Potential density anomaly fields (kg/m³) : a, c, d– “truth”; e, f –pseudo-observations. (b) - sea level elevation (cm).

The model simulations are used to produce pseudo-observation fields of density (salinity, temperature). The procedure of pseudo-

fields formation is the following. Average density profiles along the selected set of the sea level contours are specified at some

moment of time. Then the same density profiles are attached to proper sea level contours at any following or backward time

moment. This procedure permits to extrapolate 3D density forward and backward in time. Accuracy of such extrapolation is shown

below.

The average density profiles for the set of the sea level contours are evaluated on April 22nd 2007 from the model run (Figure 8a).

The pseudo-observations of density are produced on 21st July 2007 according to the described above procedure using the sea

level field simulated by the model for that day (Figure 8c and d). Consistency of the “truth” and pseudo-observations is rather

good. Some statistics was evaluated at 10, 20, 30…110,120 days in order to obtain quantitative characteristic of extrapolation

accuracy and estimate optimal interval of extrapolation.

The dispersions of the difference between “truth” field and pseudo-observations are much smaller than the natural dispersion of

the “truth” field below 60 m whereas it has the same order of magnitude for shallow layers. Dispersion of pseudo-density formed

within 30 to 40-days differs from “truth” density on 3.5% and 3.6 % in winter, 2.6% and 2.6% in spring, 12% and 12.2% in summer,

4.6% and 4.5% in autumn respectively. Thus a 30-40 days period can be chosen as an optimal solution for constructing of

pseudo-observations (density, salinity, temperature). Analysis of the depth dependences of the correlation coefficient between

“truth” and “pseudo” density fields for various seasons reveals the following. Correlation coefficient depends weakly on interval of



extrapolation in layer from 50 to 700 m. Its values for 10-day interval change from 0.85 to 0.99 since January to July. For the rest

of the year values are in the range 0.8-0.95. Value of the coefficient for 30-day period at 350 m depth is equal 0.94 in winter, 0.95

in spring, 0.84 in summer and 0.9 in autumn. Maximum of the correlation coefficient is observed in spring at 94 m depth and

equals 0.99.

Thus, the twin experiments show rather high consistency of pseudo-observations to the “truth” data. The simulation of pseudo-

observations should help to build reanalysis fields after 1996. The similar procedure can be applied for the assimilation of profiling

float data by the forecasting system to ensure consistency of model results with observations in the permanent pycnocline area.

Part 3. Pan-European integration

So far two ocean code models have been used in the Black Sea MFC: the Marine Hydrophysical Institute circulation model

(Demyshev and Korotaev 1992) and the Princeton Ocean Model (Blumberg and Mellor 1987; Kubryakov 2004). The MyOcean

project provides opportunity to use NEMO (Madec et al. 1998), a more flexible modelling tool consisting of the ocean general

circulation model, ice thermodynamics, biogeochemical tracers transport model and nesting tool. Implementation of the NEMO in

the Black Sea MFC will provide possibility to use all innovations available to NEMO customers. Some results of the NEMO

simulations at the Black Sea MFC are presented below.

The regular-grid spacing with the horizontal resolution 0.1°×0.0625° is used in simulations of circula tion. The mesh dimension is

141×88 points with 35 z-levels. Vertical grid-spacing is determined according to the hard-wired hyperbolic tangent stretching

function proposed by Madec et al. (1998). Sea level free surface is calculated according to the procedure proposed by Roullet and

Madec (2000) for the case of linearized sea level equation. Second order central scheme is used for tracer advection. The lateral

diffusion Laplace operator is used with a diffusion coefficient equal to 60 m²/s. The so-called invariant vector form is used for

momentum advection. Diffusion of momentum is described by laplacian with horizontal eddy viscosity equal to 300 m²/s. Vertical

turbulent motions are produced with embedded 1.5 turbulent closure model of Blanke and Delecluse (1993). The Black Sea is

considered as closed basin, i.e. neither river runoffs nor Bosphorus and Kerch straits are taken into account.

The model is forced by atmospheric climatology (Belokopytov 2004). Monthly fields of zonal and meridional wind stress

components, evaporation minus precipitation, total downward heat flux and penetrative solar radiation are used as the boundary

conditions. The model is initialized by climatic temperature and salinity fields reconstructed by Knysh et al. (2005). A retroaction

term with climatological SST was added to the surface heat flux to overcome overheating in the upper layer of the model.

Numerical simulations are carried for 10 years. Analysis is based on the results of last year of integration.

Figure 9 - Cross section of temperature (° С) field across 43°N for winter (a), spring (b), sum mer (c) and autumn (d).



One of the important features of the Black Sea thermal structure is the cold intermediate layer (CIL). Its boundaries are usually

determined by location of 8°C isotherm. The section s across 43°N are presented on the Figure 9 for dif ferent seasons. It is

evident the upper mixed layer formation, winter renewal of the CIL, spring-summer isolation of the CIL by heating, and autumn

convective cooling. The thickness and location of the CIL are close to observations. The model produces well-pronounced upper

mixed layer in winter with thickness of 40 meters. The formation of seasonal thermocline is observed in summer (Figure 10). The

model reproduces in adequate way the eastern and western gyres of the Black Sea, Rim current and anticyclonic mesoscale

eddies to the right of the jet. The salinity field corresponds to the Black Sea gyre structure.

Thus NEMO-OPA simulates reasonably well the Black Sea climatic circulation. However, several important problems still remain

unsolved. The turbulence closure model should be better tuned for the regional modelling to achieve quantitative consistency of

simulations and observations. Also river runoffs and water flow through the Strait of Kerch and Bosphorus Strait should be

implemented into the model.

Figure 10 - Temperature (° С) profiles in winter and summer

Conclusion

Many scientific questions concerning the Black Sea dynamics are still unresolved in spite of long-term history of investigations.

We do not know definitely how the Black Sea stratification is formed, why there are the eastern and western gyres, how waters of

the deep Bosphorus current propagate to the basin, why mesoscale lenses like “meddies” are not observed in the Black Sea near

the Bosphorus Strait mouth, what is the reason of coastal anticyclones formation, what induces decadal changes of the Black Sea

stratification. MyOcean proposes new tools permitting to consider the above mentioned and many other problems. Final efficiency

of the MyOcean tools depends on the quality of observations and simulations. Thus, the MyOcean project stimulates improvement

of the Black Sea observing system and models which is crucial for the development of the regional oceanography.

Acknowledgements

The research leading to these results has received funding from the European Community's Seventh Framework Programme

FP7/2007-2013 under grant agreement n°218812 (MyOce an).

References

Belokopytov V.N. 2004: Thermohaline and hydrological-acoustical structure of the Black Sea, PhD thesis, Sevastopol, MHI NAN

Ukraine, 160 pp. (in Russian)

Blanke B. and P. Delecluse 1993: Low frequency variability of the tropical Atlantic Ocean simulated by a general circulation model

with mixed layer physics. J. Phys. Oceanogr., 23, 1363-1388.

Blumberg A.F., Mellor G.L. 1987: A description of a three-dimensional coastal ocean model. In Three Dimensional Shelf Models,

Coastal Estuarine Sci., vol. 5, edited by N. Heaps, AGU, Washington D.C.,1–16.



Demyshev S.G.., Korotaev G.K. 1992: Numerical energy-balanced model of the baroclynic currents with non-flat bottom on

Arakawa’s “C” grid. Numerical models and results of calibration calculations of the Atlantic Ocean circulation. Moscow: 163–231

(in Russian).

Knysh V.V., Kubryakov A.I., Inyushina N.V., Korotaev G. K. 2005: Reconstruction of the climatic seasonal Black Sea circulation by

means of sigma-coordinate model and assimilation of the temperature and salinity data. Ecological safety of coastal and shelf

zone and complex use of their resources, Sevastopol, Vol. 12: 243–265 (in Russian).

Knysh V.V., Korotaev G.K., Moiseenko V.A., Kubryakov A.I., Belokopytov V.N., Inyushina N.V. 2010: Seasonal to interannual

variability of the Black Sea hydrophysical fields, reconstructed in the reanalysis for 1971– 1993. Izvestiya RAN. Fizika atmosfery i

okeana. (delivered for printing) (in Russian)

Korotaev G., T. Oguz, S. Riser 2006: Intermediate and deep currents of the Black Sea obtained from autonomous profiling floats.

Deep-Sea Research, 2, 53: 1901–1910

Kubryakov A.I. 2004: Implementation of the nesting technology within building of the monitoring system of the hydrophysical fields

in the Black Sea coastal regions. Ecological safety of coastal and shelf zone and complex use of their resources, Sevastopol, Vol.

11: 31–50 (in Russian).

Madec G., P. Delecluse, M. Imbard and C. Levy 1998: OPA 8 ocean general circulation model – reference manual. Tech. rep.,

LODYC/IPSL Note 11.

Mellor G.L. and Yamada T. 1982: Development of a turbulence closure model for geophysical fluid problem. Rev. Geophys. and

Space Physics, vol. 20, No. 4. 851– 875.

Pacanowski, R.C., Philander, S.G.H. 1981. “Parametrization of vertical mixing in numerical models of tropical oceans”. J. Phys.

Oceanogr., 11, 1443-1451.

Ratner Y.B., Martynov M.V., Bayankina T.M. et al. 2005: Information flows in the Black Sea monitoring system and automation of

its processing. Systems of the environment control, Sevastopol: 140–149 (in Russian).

Ratner Y.B., Martynov M.V., Bayankina T.M. et al. 2006: Structure and control of the calculation process in the system of the

Black Sea dynamic simulation. Systems of the environment control, Sevastopol: 150–158 (in Russian).

Roullet G. and G. Madec 2000: Salt conservation, free surface, and varying levels: a new formulation for ocean general circulation

models. J. Geophys. Res., 105, 23,927– 23,942.

Titov V.B., 2003: Impact of the long-term variability of climatic conditions to the hydrological structure and interannual renewal of

the intermediate cold layer in the Black Sea. Oceanology, 43, No. 2, 176-184 (in Russian).

Uppala, S.M., P. W. KÅllberg, A. J. Simmons, U. Andrae, V. Da Costa Bechtold, M. Fiorino, J. K. Gibson, J. Haseler, A.

Hernandez, G. A. Kelly, X. Li, K. Onogi, S. Saarinen, N. Sokka, R. P. Allan, E. Andersson, K. Arpe, M. A. Balmaseda, A. C. M.

Beljaars, L. Van De Berg, J. Bidlot, N. Bormann, S. Caires, F. Chevallier, A. Dethof, M. Dragosavac, M. Fisher, M. Fuentes, S.

Hagemann, E. Hólm, B. J. Hoskins, L. Isaksen, P. A. E. M. Janssen, R. Jenne, A. P. Mcnally, J.-F. Mahfouf, J.-J. Morcrette, N. A.

Rayner, R. W. Saunders, P. Simon, A. Sterl, K. E. Trenberth, A. Untch, D. Vasiljevic, P. Viterbo, J. Woollen, 2006, The ERA-40 re-

analysis, Quarterly Journal of the Royal Meteorological Society, Volume 131, Issue 612, pages 2961–3012, October 2005 Part B.


NEMO-Shelf, towards operational oceanography with S ST data assimilation on the North West European She lf

NEMO-Shelf, towards operational oceanography with S ST data assimilation on the North West European Shelf By Enda J. O’Dea 1, James While 1, Rachel Furner 1, Patrick Hyder 1, Alex Arnold 1, David Storkey 1,

John R. Siddorn 1, Matthew Martin 1, Hedong. Liu 2, James T. Holt 2

1Met Office, Exeter, U.K.

2 NOC, Liverpool, U.K

Abstract

In the deep ocean data assimilation has proven itself for several years as a valuable constituent in operational ocean forecast

systems. However, data assimilation for the tidally driven shelf presents significant additional challenges. The Met Office is

developing a new operational forecast system for the North West European Shelf that incorporates data assimilation of SST. This

system will replace the existing non-assimilative operational forecast system based on POLCOMS. The physical model utilized in

the new system is a modified version of NEMO suitable for modelling the highly dynamic shelf seas. The system incorporates data

assimilation of SST using a modified version of the existing FOAM system. Preliminary hindcast runs have shown that the new

system provides good skill compared to the existing extensively validated POLCOMS system for the same region. Additionally, the

data assimilation has not had an adverse effect on the simulated water column structure both in well mixed and seasonally

stratified waters. This system is running pre-operationally at the Met Office and will constitute the V1 MyOcean Forecast system

for the North West European Shelf.

Introduction

The North West European Shelf is one of the most studied shelf seas systems in the world owing to the keen economic and

environmental interests of the surrounding nations. Such interests include marine transport, petrochemical exploitation, fishing,

aquaculture and more recently the renewable energy industry in the form of wind, wave and tidal generation of power. Such

interests have motivated the development of a variety of forecast systems for the region. Such systems have evolved from 2D tide

(Flather 1976) and surge models to fully 3D (Holt and James 2001) systems at ever increasing spatial resolution, including ever

more complex processes.

However, until recently the problem of including data assimilation into an operational forecast system for the highly dynamic shelf

seas region has not been tackled. Data assimilation has been used extensively for a number of years in the deep ocean with great

success leading to systems with much greater forecast skill (Martin et al. 2007). Motivated by the potential additional skill of data

assimilation, a new forecast system is under development at the Met Office that in the first instance includes data assimilation of

SST. The new forecast system is based upon the existing FOAM system (Storkey et al. 2010) for the deep ocean which utilizes

NEMO for the core physical ocean model, and an optimal interpolation method of data assimilation.

Currently the Met Office provides an operational forecast system based upon POLCOMS for the North West European Shelf. The

system has been validated over a number of years with continual developments to improve the system. This existing system

provides a firm reference point against which any replacement system can be compared. The old system consists of an outer

domain, the Atlantic Margins Model (AMM) at 12km resolution, and a nested inner domain, the Medium-Resolution Continental

Shelf domain at 7km resolution. The new system aims to replace both domains with a single AMM domain covering the original

AMM region, but at 7km resolution. Both systems are nested into the FOAM 1/12th degree North Atlantic model that provides

temperature, salinity, sea surface height and depth integrated current information at the open boundaries.

The two systems are compared for a 2 year hindcast period against observations. The new NEMO system is run twice, once with

and once without data assimilation. This is to ensure the underlying physical model provides similar or improved skill compared

the existing POLCOMS system without data assimilation and also provides a reference to compare with our assimilative run. Such

a comparison allows us to understand how the data assimilation of SST changes the solution and if it produces any unrealistic

modifications to the internal dynamics away from the surface.

Physical System

NEMO (Madec 2008) was originally developed to model the deep ocean rather than the shelf seas. Thus, a number of important

modifications were required to ensure that the NEMO physical model is suitable for application in shelf seas. The first modification

is the inclusion of tidal forcing both on the open boundary conditions via a Flather radiation condition (Flather 1976), and the



inclusion of the equilibrium tide. Tidal modelling also requires a non linear free surface and this is facilitated in NEMO by using a

variable volume layer approach. The short time scales associated with tidal propagation and the free surface require a time

splitting approach, splitting modes into barotropic and baroclinic components. Additionally, the bottom boundary condition now

includes a log layer representation, and a k-epsilon turbulence scheme is implemented with the generic length scale option

developed at Mercator Ocean (France).

The coordinate system is a modified version of S-coordinates (Song and Haidvogel 1994). The coordinates are pure sigma on the

shelf and stretch off the shelf to maintain vertical resolution near the surface and bottom. Additionally, where the bathymetry is

particularly steep the coordinates can intersect the bottom. The loss of vertical resolution at these points is more than

compensated by reducing errors related to the horizontal pressure gradient term and steep coordinate surfaces. The horizontal

pressure gradient scheme itself is also updated to a pressure Jacobian scheme and it performs very well in classical tests

including sea mounts in specified uniform stratification.

Other modifications include

• Allowing the bilaplacian and laplacian diffusion operators to work on geopotential or isopycnal surfaces separately. • River inputs to be mixed to prescribed water depths. • A POLCOMS style light attenuation coefficient that varies dependant on total water depth. • The addition of an inverse barometer effect from atmospheric pressure forcing.

With these modifications the physical modelling system has been validated in a constant density case for the evaluation of tides.

Additionally, a fully baroclinic but non-assimilative simulation is used to assess the forecast system without assimilation.

Assimilation System

Data assimilation within NEMO-shelf uses the Analysis Correction method of (Martin et al. 2007) to assimilate SST data. In

essence this is an iterative Optimal Interpolation scheme, which treats both model and observation errors, with their associated

covariances, as constant. Operationally, assimilation proceeds in three steps. Firstly a 1 day model forecast is performed, within

which observations are compared to model output at the nearest time-step; this is a First Guess at Appropriate Time FGAT

system. In the second stage observation minus model differences are converted to SST increments by solving the Best Linear

Unbiased Estimator (BLUE) equations. In solving these equations, we use precalculated values for observation error (assumed

uncorrelated), model error, and model error covariances. Finally, to produce the analysis the model is reran for the same day with

the increments added onto the SST field using the Incremental Analysis Update (IAU, see Bloom et al. 1996) method. Increments

are added into the model down to the base of the instantaneous mixed layer, where the mixed layer depth is determined by a

0.20C temperature difference from the surface.

In the present set-up assimilated satellite observations are taken from the infra-red SEVIRI, AATSR, METOP and AVHRR

instruments and from the AMSRE microwave sensor. Because of biases in satellite data, a bias correction scheme is used to

correct satellite measurements, with the AATSR sensor, which is considered to be less biased than the other satellite instruments,

and available in-situ data used as reference ‘unbiased’ data. In addition to satellite measurements, we also assimilate available in-

situ data from drifting buoys, moorings and ships. All data are quality controlled using a Bayesian system (Lorenc and Hammon,

1988) before being assimilated.

Results

Barotropic Results

A constant density simulation with only tidal forcing at the boundaries with online harmonic analysis of the main tidal constituents

confirms that the overall skill of the simulated tides is similar to POLCOMS. The SSH RMS errors for M2 are 0.188m for NEMO

and 0.179m for POLCOMS and the mean error is -0.014m and 0.055m for NEMO and POLCOMS respectively. Figure 1 displays

the SSH amplitude and phase errors between NEMO and observations for the dominant M2 constituent. It should be noted that

the underlying bathymetry of the POLCOMS and NEMO systems are the same. Further refinement of the tides should be possible

with more accurate bathymetry at 7km resolution.

Preliminary baroclinic and assimilative results

The model system both with and without data assimilation is compared against POLCOMS for the hindcast period 2007-2008 for a

variety of observation types. For 2008 the non-assimilative NEMO system has an SST RMS error of 0.660C and for POLCOMS it

is 0.690C. However, there is a warm bias in NEMO of 0.30C compared to 0.20C in POLCOMS. With assimilation of SST the errors

are much reduced, with an RMS SST error of 0.380C and mean of 0.10C respectively. Both systems are also compared through



the water column using data at profile points and the ICES data set for the North Sea. Figure 2 gives the mean surface minus bed

temperature difference for 2008 for the non assimilative NEMO system, POLCOMS and the ICES data set. The NEMO results do

appear to be closer to ICES than POLCOMS.

Figure 3 displays the surface-bed temperatures for 2007 from the assimilative and non assimilative NEMO systems the ICES data

set. Through most of the North Sea the effect of data assimilation is small on the stratification with the exception being in the

Norwegian trench where stratification is intensified.

Figure 1 - SSH amplitude in metres (top left panel) and phase errors in degrees (top right panel) betw een NEMO and

observations for the dominant M2 constituent for NE MO constant density run and absolute SSH amplitude in metres

(bottom left panel) and phase in degrees (bottom ri ght panel) for model plotted against observations a lso for the M2 constituent.

Figure 2 - Surface minus bed temperature ( 0C) for 2008 in NEMO (left panel), POLCOMS (middle p anel) and ICES (right

panel)



Figure 3 - Surface minus bed temperature ( 0C) for 2007 in NEMO with (left panel) and without a ssimilation (middle panel) and ICES (right panel)

Conclusions

A new shelf seas operational forecasting system for the North West European Shelf has been developed based on NEMO with OI

data assimilation of SST. The SST data assimilation has improved the forecast skill of the model without marked disruption of the

3D structure of the water column. The new system is undergoing extensive validation tests and is running pre-operationally at the

Met Office to ensure the system’s robustness before full operational implementation.

In addition to the physical system described here, the ecosystem component ERSEM is also being coupled to the system with a

view to replacing the existing POLCOMS-ERSEM operational system at the Met Office. Future upgrades to the system include

bathymetry, improved light attenuation, profile data assimilation, river inputs from E-HYPE and Baltic inflow from a Baltic model in

place of climatology.

References

Bloom, S. C., Takacs, L. L., Da Silver A. M. and Ledvina, D., 1996: Data assimilation using incremental analysis updates. Monthly

Weather Review, 124, 1256-1271

Flather, R. A., 1976: A tidal model of the North West European continental shelf, Mem. Soc. R. Sci. Liege, 10, 141-164

Holt, J.T., James, I.D., 2001: An s-coordinate density evolving model of the northwest European continental shelf Part 1 model

description and density structure. Journal of Geophysical Research 106, 14015–14034.

Lorenc, A. C. and Hammon, 1988: Objective quality control of observations using Bayesian methods. Theory, and a practical

implementation. Q. J. Roy Met Soc, 114, 515-543

Madec G. 2008. NEMO ocean engine. Note du Pole de modélisation, Institut Pierre-Simon Laplace (IPSL), France, No 27 ISSN

No 1288–1619

Martin, M.J., Hines, A. and Bell, M.J., 2007: Data assimilation in the FOAM operational short-range ocean forecasting system: a

description of the scheme and its impact. Q. J. Roy Met Soc, 133, 981-995

Song, Y., and D. Haidvogel, 1994: A semi-implicit ocean circulation model using a generalized topography-following coordinates

system, J. Comput. Phys., 115, 228-244

Storkey, D., E.W. Blockley, R. Furner, C. Giuavarc'h, D. Lea, M.J. Martin, R.M. Barciela, A. Hines, P. Hyder, J.R. Siddorn, 2010.

Forecasting the ocean state using NEMO: The new FOAM system. J. Operational Oceanography, 3, 3-15.


Notebook

Notebook

Editorial Board:

Laurence Crosnier

Secretary:

Fabrice Messal

Articles:

The FERRYBOX component in MyOcean

By Dominique Durand, Are Folkestad, Kai Sørensen

The new regional generation of Mercator Ocean syste m in the

Iberian Biscay Irish (IBI) area

By Sylvain Cailleau, Jérôme Chanut, Bruno Levier, Claire Maraldi, Guillaume

Reffray


By Sergey Demyshev, Vasily Knysh, Gennady Korotaev, Alexander

Kubryakov, Artem Mizyuk

NEMO-Shelf, towards operational oceanography with SST data

assimilation on the North West European Shelf

By Enda J. O’Dea, James While, Rachel Furner, Patrick Hyder, Alex Arnold,

David Storkey, John R. Siddorn, Matthew Martin, Hedong. Liu, James T.

Holt

Contact :

Please send us your comments to the following e-mail address: [email protected]

Next issue: January 2011

Newsletter - Mercator Ocean · 2015-05-04 · Mercator Ocean Quarterly Newsletter #39 – October...

Documents

Transcript of Newsletter - Mercator Ocean · 2015-05-04 · Mercator Ocean Quarterly Newsletter #39 – October...