Atlantic -Iberian Biscay Irish- IBI Production Centre IBI ... · product, currently delivered through the CMEMS catalogue, has been performed. This section is devoted to provide essential

Atlantic -Iberian Biscay Irish- IBI Production Centre

IBI_MULTIYEAR_PHY_005_002

Issue: 5.0

Contributors: Bruno Levier, Pablo Lorente, Guillaume Reffray, Marcos Sotillo.

Approval Date by Quality Assurance Review Group: 07/01/2021

QUID for IBI reanalysis Product


Ref: CMEMS-IBI-QUID-005-002

Date : 10 September 2020

Issue : 5.0

© EU Copernicus Marine Service – Public Page 2/ 50

CHANGE RECORD

Issue

Date § Description of Change Author Validated By

0.0

06/05/2014

All

Creation of the document

Bruno Levier Marcos G. Sotillo Guillaume Reffray

0.1 07/05/2014

All Update with High-Frequency CLASS2 metrics from the validation with data from PdE moorings

Marcos G. Sotillo Roland Aznar

1.0 12/05/2014 All Minor changes. Bruno Levier Marcos G. Sotillo

Enrique Alvarez

1.1 May 1 2015 all Change format to fit CMEMS graphical rules

L. Crosnier

2.0 14/12/2015 all Inclusion of new metrics considering temporal extension: 2012-2014.

Bruno Levier Marcos G. Sotillo

Enrique Álvarez Fanjul

3.0 15/12/2017 all New IBI reanalysis set-up and run. Associated to the CMEMS V4 release

Bruno Levier Pablo Lorente Marcos G. Sotillo

4.0 15/02/2018 all Complete release of the system.

4.1 15/02/2019 all Extension of the reanalysis. 2017 produced, years 2014 to 2016 re-run.

4.2 15/04/2019 all Extension of the reanalysis. 2018 produced, years 2014 to 2017 re-run.

4.3 03/04/2020 Extension of the reanalysis: 2019 produced. Impact of ERA5 atmospheric forcing. New Interim temporal extension.

Bruno Levier Roland Aznar

5.0 10/09/2020 New production of the entire period 1993-2019.

Bruno Levier Roland Aznar

Marcos G Sotillo Missing





Issue : 5.0


TABLE OF CONTENTS

I Executive summary ....................................................................................................................................... 5

I.1 Products covered by this document ........................................................................................................... 5

I.2 Summary of the results ............................................................................................................................... 5

I.2.1 Temperature ...................................................................................................................................... 5

I.2.2 Salinity .............................................................................................................................................. 6

I.2.3 Ocean currents ................................................................................................................................... 6

I.2.4 Sea level ............................................................................................................................................ 6

I.2.5 Transports .......................................................................................................................................... 6

I.2.6 Other.................................................................................................................................................. 6

I.3 Estimated Accuracy Numbers .................................................................................................................... 7

II Production Subsystem description .............................................................................................................. 13

Reanalysis Configuration: the IBIRYSV3R1 System .......................................................................... 13

II.1 The ocean model: NEMO3.6 ................................................................................................................... 13

II.2 Data assimilation method ........................................................................................................................ 15

Adaptive tuning of observation error and implementation ...................................................................... 18

III Validation framework ............................................................................................................................. 19

IV Validation results ......................................................................................................................................... 23

IV.1 Temperature ........................................................................................................................................... 23

IV.1.1 T-CLASS1-T3D-MEAN............................................................................................................. 23

IV.1.2 T-CLASS1-SST-MEAN ............................................................................................................. 24

IV.1.3 T-CLASS1-SST-TREND ........................................................................................................... 25

IV.1.4 T-CLASS2-MOORINGS ............................................................................................................ 25

IV.1.5 T-CLASS3-SST-QNET-2DMEAN ............................................................................................ 27

IV.1.6 T-CLASS3-HC_LAYER ............................................................................................................ 28

IV.1.7 T-CLASS3-IC_CHANGE .......................................................................................................... 29

IV.1.8 T-CLASS4-LAYER .................................................................................................................... 29

IV.1.9 T-DA-INNO-STAT .................................................................................................................... 31

IV.2 Salinity ..................................................................................................................................................... 32

IV.2.1 S-CLASS1-S3D_MEAN ............................................................................................................ 32

IV.2.2 S-CLASS1-SSS-TREND ............................................................................................................ 33

IV.2.3 S-CLASS2-MOORINGS ............................................................................................................ 33

IV.2.4 S-CLASS2-SSS-SWF_2DMEAN .............................................................................................. 35

IV.2.5 S-CLASS3-SC-LAYER .............................................................................................................. 35





Issue : 5.0


IV.2.6 S-CLASS3-IC_CHANGE........................................................................................................... 37

IV.2.7 S-CLASS4-LAYER .................................................................................................................... 37

IV.2.8 S-DA-INNO_STAT .................................................................................................................... 39

IV.3 Currents .................................................................................................................................................. 40

IV.3.1 UV-CLASS1-15 m_MEAN ........................................................................................................ 40

IV.3.2 UV-CLASS2-MOORINGS ........................................................................................................ 41

IV.3.3 UV-CLASS2-MKE_1000 m ....................................................................................................... 42

IV.4 Sea level ................................................................................................................................................... 43

IV.4.1 SL-CLASS1-TREND_2D........................................................................................................... 43

IV.4.2 SL-CLASS2-TIDE_GAUGE ...................................................................................................... 44

IV.4.3 SL-CLASS3-2DMEAN .............................................................................................................. 44

IV.4.4 SL-DA-INNO_STAT ................................................................................................................. 45

IV.5 Transports ............................................................................................................................................... 45

IV.6 Other monthly metrics ........................................................................................................................... 46

IV.6.1 OTH-CLASS1-MLD_CLIM_SEASON ..................................................................................... 46

V Quality changes since previous version ...................................................................................................... 47

VI References .................................................................................................................................................... 49





Issue : 5.0


I EXECUTIVE SUMMARY

I.1 Products covered by this document

This document describes the quality of the CMEMS Atlantic IBI -Iberian Biscay Irish- physical ocean Multi-Year (MY) product (CMEMS product catalogue identifier: IBI_MULTIYEAR_PHYS_005_002). This product includes monthly, daily and hourly mean fields all of them delivered in a regular grid with a horizontal resolution of 1/12° (0.0833° lat x 0.0833° lon) with 50 levels from surface to bottom, for the following variables:

- sea_surface_height_above_sea_level

- sea_water_salinity

- sea_water_potential_temperature

- eastward_sea_water_velocity

- northward_sea_water_velocity

- eastward_sea_barotropic_velocity

- northward_sea_barotropic_velocity

- Sea_water_potential_temperature_at_sea_floor

- ocean_mixed_layer_thickness_defined_by_sigma_theta

The period covered is 01/1993-12/2019

I.2 Summary of the results

The quality of the IBI reanalysis has been mainly assessed validating reanalysed daily and monthly fields with observational sources and comparing the IBI reanalysis products with the global reanalysis one. The headline results for each of the variables assessed are as follows:

I.2.1 Temperature

The IBIRYS reanalysis has a warm bias between surface and 1000 m depth compared to Levitus 2009 climatology (except at 1000 m depth in the pathway of the Mediterranean water, where the bias is cold). The bias is negative at 2000 m depth, except offshore of the western Iberian coast. At surface, the warm bias can be greater than 1°C over the shelf, and the bias is cold along all the western coasts.

The mean SST is close to the observations, but with a negative misfit in average.

The trend is consistent with AVHRR data, positive everywhere.

Compared to in-situ profiles, the bias of the IBIRYS reanalysis is weak (less than 0.1°C except near the surface where it can reach 0.3°C).





Issue : 5.0


I.2.2 Salinity

Compared to the WOA 2009 climatology, the IBIRYS reanalysis is generally saltier between the surface and 1000 m depth. At the surface, IBIRYS is less salty along the European coasts, in the Irish sea and in the Mediterranean Sea (especially at 100 m depth in the Algerian current).

The bias is weak compared to in-situ profiles (maximum at the surface, decreasing towards the bottom). The RMS in average is 0.2 PSU in the 5-200 m layer, less than 0.1 PSU below 200 m.

I.2.3 Ocean currents

IBIRYSV3R1 reproduces well the main ocean currents and is coherent with the AOML surface drifter climatology. However, when compared to fixed moorings measurements, the near-surface currents are poorly correlated to the observations (generally between 0.2 and 0.5). MKE at 1000 m is higher in IBIRYSV3R1 than in ANDRO Argo drift data base (which seems to underestimate the MKE in the IBI area).

I.2.4 Sea level

IBIRYSV3R1 reanalysis is very close to altimetric observations and has a good ability to describe the sea level variability. The globally averaged RMSD of the analysed sea level is about 5~6 cm. Correlations with observations at tide gauges are generally higher than 0.7 and better than GLORYS2V4 reanalysis (not shown). However, the sea level trend is not in agreement with observations.

I.2.5 Transports

The transport through the Gibraltar strait is over-estimated in the surface layer and under-estimated in the bottom layer compared to published values.

I.2.6 Other

The mixed layer depth of IBIRYSV3R1 reanalysis exhibits similar patterns with those in De Boyer Montegut climatology data. There is an overestimation in the north-western part of the domain.





Issue : 5.0


I.3 Estimated Accuracy Numbers

A consistent estimation of the relevant accuracy levels for the IBI-MFC multi-year ocean reanalysis product, currently delivered through the CMEMS catalogue, has been performed. This section is devoted to provide essential statistics on the delivered physical variables, obtained from long-term comparisons with reference information on the ocean dynamics, based on both satellite and in-situ observational data sources.

The SST accuracy numbers, shown in Table 1 and Table 2, were derived from the comparison of IBI best estimates with CMEMS L3 satellite SST data (CMEMS product identifier: SST-EUR-SST-L3S-NRT-OBSERVATIONS-010-009_a). The provided EANs are derived from statistics, averaged over the whole IBI service domain and subregions, computed using analyses of the reanalysis for the years 2009-2017 (period covered by the satellite SST data).

Sea Surface Temperature (K)

IBI Service Domain

Whole period Winter Spring Summer Autumn

Surface

RMSD

0.70 0.60 0.69 0.79 0.68

Surface

Mean difference

0.08 0.11 0.13 0.05 0.05

Canary Island Zone (ICANA)

Surface

RMSD

0.61 0.51 0.54 0.73 0.61

Surface

Mean difference

0.01 0.08 0.02 -0.08 0.03

Table 1: IBI EAN related to SST. Time period: 01/01/2010-31/12/2018. Daily IBI best estimates (hindcast) compared with the observational source used as reference (Daily SST MyOcean/CMEMS L3 satellite SST product). Statistics averaged over the whole IBI Service Domain and other subregions (continues in next page).





Issue : 5.0


Western Mediterranean (WSMED)

Whole

period

Winter Spring Summer Autumn

Surface

RMSD

0.79 0.57 0.76 0.91 0.82

Surface

Mean difference

0.27 0.18 0.33 0.30 0.24

Bay of Biscay (GOBIS)

Surface

RMSD

0.66 0.58 0.65 0.74 0.61

Surface

Mean difference

0.03 0.05 0.09 -0.02 -0.03

Cadiz (CADIZ)

Surface

RMSD

0.93 0.50 0.79 1.32 0.82

Surface

Mean difference

-0.04 0.03 0.03 -0.07 -0.15

Table 2: IBI EAN related to SST. Time period: 01/01/2010-31/12/2018. Daily IBI best estimates (hindcast) compared with the observational source used as reference (Daily SST MyOcean/CMEMS L3 satellite SST product). Statistics averaged over the whole IBI Service Domain and other subregions.

The following Estimated Accuracy Numbers (EAN) numbers have been calculated for a 26-year period (1993-2018), but it depends on the length of each observations time-series.

The EAN for residual sea surface height (the tidal signal has been removed), shown in Table 3, have been computed using as reference observations from data provided by the tide-gauges network on the IBI service domain for a 26-year period (1993-2018) (CMEMS product identifier: INSITU_IBI_REP_OBSERVATIONS_013_040 and INSITU_NWS_TS_REP_OBSERVATIONS_013_043).

Sea Surface Height (cm)

IBI Service

Domain

IBI Service

Domain (lat <

45°N)

IBI Service

Domain (lat >

45°N)

Residual SSH

RMSD

10 5 12

Table 3: IBI EAN related to SSH residuals. Daily IBI best estimates (hindcast) used in the metric computation. Observational source used as reference: Sea level data from in-situ tide-gauge stations avail in the IBI domain (CMEMS IBI-InSituTAC product). Statistics averaged over the whole IBI Service Domain, and also for the south and the north of the domain.





Issue : 5.0


Table 4, Table 5, Table 6 and Table 7 show the EAN set for temperature and salinity, computed using reprocessed (CORA INSITU_GLO_TS_REP_OBSERVATIONS_013_001_b) in-situ observational data sources for the period 1993-2018. The EAN for temperature and salinity have been computed using as reference observational data from any in-situ platform available on the IBI service domain along the period.

Information for the whole water column (full profiles considered) is provided. Likewise, information for different levels is provided, being the specific levels considered in the IBI EAN estimation: 0-5 m, 5-200 m, 200-600 m, 600-1500 m, 1500-2000 m.

Together with the EANs obtained from the comparisons between the IBI solution and the in-situ observations available in the whole IBI service domain, some regional EANs for specific regions within the IBI service domain are provided. Only 4 of these areas (out of the 10 selected for other different validation purposes shown in Figure 1): the one centred in the Canary Island region, the Gulf of Cadiz, the Bay of Biscay and the western Med have been considered for computing these validation indicators. The other IBI regional validation zone (more focused on coastal and shelf areas, have not been considered to compute regional EAN due to the scarcity or unavailability of ARGO floats in those areas).

Figure 1: IBI Service Domain and the different zones selected to make regional metrics for the IBI validation.





Issue : 5.0


Temperature (K)

IBI Service Domain

RMSD difference Mean difference

FULL 0.44 -0.01

0-5 m 0.46 0.08

5-200 m 0.49 -0.04

200-600 m 0.27 0

600-1500 m 0.45 0.01

1500-2000 m 0.37 0.01


FULL 0.34 -0.06

0-5 m 0.32 -0.05

5-200 m 0.41 -0.06

200-600 m 0.30 0.02

600-1500 m 0.32 -0.02

1500-2000 m 0.21 -0.04


FULL 0.29 0.02

0-5 m 0.66 0.19

5-200 m 0.50 0.06

200-600 m 0.12 -0.03

600-1500 m 0.08 0

1500-2000 m 0.04 0.02


FULL 0.40 -0.05

0-5 m 0.43 0.09

5-200 m 0.44 -0.08

200-600 m 0.16 -0.01

600-1500 m 0.30 0.03

1500-2000 m 0.32 0.02

Table 4: IBI EAN related to Temperature for the period 1993-2018. Daily IBI best estimates (hindcast) used in the metric computation. Observational source used as reference: Temperature profiles and time-series from all platforms available in the IBI service domain. Also specific regional EAN in the western Mediterranean, Canary Island, Bay of Biscay and Cadiz regions are shown (continues in next page).





Issue : 5.0


Cadiz (CADIZ)


FULL 0.77 0.00

0-5 m 0.78 0.21

5-200 m 0.83 -0.31

200-600 m 0.38 0.03

600-1500 m 0.58 0.10

1500-2000 m 0.67 -0.23

Table 5: IBI EAN related to Temperature for the period 1993-2018. Daily IBI best estimates (hindcast) used in the metric computation. Observational source used as reference: Temperature profiles and time-series from all platforms available in the IBI service domain. Also specific regional EAN in the western Mediterranean, Canary Island, Bay of Biscay and Cadiz regions are shown.

Salinity (PSU)

IBI Service Domain


FULL 0.11 0

0-5 m 0.28 0.00

5-200 m 0.11 0.01

200-600 m 0.05 0

600-1500 m 0.10 0

1500-2000 m 0.07 0


FULL 0.10 0

0-5 m 0.18 -0.06

5-200 m 0.08 0.03

200-600 m 0.05 0

600-1500 m 0.07 0.01

1500-2000 m 0.04 -0.01

Table 6: IBI EAN related to Salinity for the period 1993-2018. Daily IBI best estimates (hindcast) used in the metric computation. Observational source used as reference: Temperature profiles and time-series from all platforms available in the IBI service domain. Also specific regional EAN in the western Mediterranean, Canary Island, Bay of Biscay and Cadiz regions are shown (continues in next page).





Issue : 5.0




FULL 0.12 -0.03

0-5 m 0.45 -0.21

5-200 m 0.18 -0.06

200-600 m 0.04 -0.01

600-1500 m 0.02 0

1500-2000 m 0.02 0.01


FULL 0.13 0

0-5 m 0.30 -0.01

5-200 m 0.12 0.01

200-600 m 0.03 0

600-1500 m 0.06 0

1500-2000 m 0.06 0.01

Cadiz (CADIZ)

FULL 0.23 -0.13

0-5 m 0.30 -0.22

5-200 m 0.11 -0.01

200-600 m 0.09 0.01

600-1500 m 0.16 0.00

1500-2000 m 0.13 -0.04

Table 7: IBI EAN related to Salinity for the period 1993-2018. Daily IBI best estimates (hindcast) used in the metric computation. Observational source used as reference: Temperature profiles and time-series from all platforms available in the IBI service domain. Also specific regional EAN in the western Mediterranean, Canary Island, Bay of Biscay and Cadiz regions are shown.





Issue : 5.0


II PRODUCTION SUBSYSTEM DESCRIPTION

Production Center: IBI MFC

Production Unit: Mercator Ocean

Dissemination Unit: Puertos del Estado

Scientific Validation Expertise: Mercator Ocean & Puertos del Estado

Reanalysis Configuration: the IBIRYSV3R1 System

The present IBI MFC multi-year reanalysis product has been derived from the IBIRYSV3R1 reanalysis outputs. This reanalysis system relies on the following two main components:

1. the ocean model (including the surface atmospheric boundary condition);

2. the data assimilation method and the assimilated observations.

II.1 The ocean model: NEMO3.6

The IBI model numerical core is based on the NEMO v3.6 ocean general circulation model (Madec, 2008). This model solves the three-dimensional finite-difference primitive equations in spherical coordinates discretized on an Arakawa-C grid. It assumes hydrostatic equilibrium and Boussinesq approximation.

The primitive equations are discretized on an horizontal curvilinear grid which is a refined subset at 1/12° (≈6–9 km) of the so-called “ORCA” tripolar grid, commonly used in other NEMO-based large-scale and global modelling experiments (Barnier et al., 2006). 75 z levels are used in the vertical, with a resolution decreasing from ∼1 m in the upper 10 m to more than 400 m in the deep ocean. A partial step representation of the very last bottom wet cell is used with some constraints on the resulting minimum bottom cell thickness to guarantee model stability (it must be greater than 15 m or 20 % of the reference grid thickness).

In order to properly simulate fast external gravity waves such as tidal motions, a non-linear split explicit free surface (Shchepetkin and McWilliams, 2004) is used. Because of the explicit simulations of tides, sea level elevation can become large on the shelf compared to the local depth. In practice, all model vertical thicknesses are remapped in the vertical at each time step to account for the varying fluid height.

Vertical mixing is parameterized according to a k-ε model implemented in the generic form proposed by Umlauf and Burchard (2003) including surface wave breaking induced mixing. The background turbulent variables are set to their molecular values and are modified in function of the mixing induced by the internal tides not resolved by the model following the parametzrization of de Lavergne et al. (2016). Tracers advection is computed with the QUICKEST scheme developed by





Issue : 5.0


Leonard (1979) conjugated to the limiter of Zalezak (1979). This third-order scheme is well suited to the high resolution used here and the modelling of the sharp front characteristics of coastal environments. Lateral sub-grid-scale mixing is parameterized according to horizontal biharmonic operators for momentum (with a diffusion coefficient of -1.25e10 m4s-1). A vector invariant form of the momentum equations and the energy–enstrophy discretization of vorticity terms are used (Barnier et al., 2006).

The original bathymetry is derived from the 30 arc-second resolution GEBCO 08 dataset (Becker et al., 2009) merged with several local databases (F. Lyard, personal communication, 2010). At open boundaries, within 30-point-wide relaxation areas, bathymetry is exactly set to the parent grid model bathymetry and progressively merged with the interpolated dataset described above. This approach and the identical vertical grid as in the parent system imply that no vertical extrapolation of open boundary data is necessary.

The IBI run has been forced with hourly atmospheric fields (10-m wind, surface pressure, 2-m temperature, relative humidity, precipitations, short-wave and long-wave radiative fluxes) provided by ERA5. IFS empirical bulk formulae (ECMWF, 2014; Brodeau et al., 2017) are used to compute latent sensible heat fluxes, evaporation, and surface stress. Solar penetration is parameterized according to a five-bands exponential scheme (considering the UV radiations) function of surface chlorophyll concentrations using the CMEMS ESA-CCI product covering the North East Atlantic area including the IBI domain at the monthly frequency. The reprocessed product is referenced as OCEANCOLOUR_ATL_CHL_L4_REP_OBSERVATIONS_009_091 while the NRT product is referenced as OCEANCOLOUR_ATL_CHL_L4_NRT_OBSERVATIONS_009_090.

Lateral open boundary data (temperature, salinity, velocities and sea level) are interpolated from the daily outputs from the Global reanalysis at 1/4° of resolution (CMEMS product GLOBAL_REANALYSIS_PHY_001_025). These are complemented by 11 tidal harmonics (M2, S2, N2, K1, O1, Q1, M4, K2, P1, Mf, Mm) built from FES2004 (Lyard et al., 2006) and TPXO7.1 (Egbert and Erofeeva, 2002) tidal models solutions. Atmospheric pressure component, missing in the large scale parent system sea level outputs, is added hypothesizing pure isostatic response at open boundaries (inverse barometer approximation).

The simulation started at rest, with initial temperature, salinity, velocity components and sea surface height provided by the global reanalysis GLORYS2V4 (CMEMS product GLOBAL_REANALYSIS_PHY_001_025). The date of start is 1st January 1992.

Fresh water river discharge inputs are implemented as lateral open boundary condition for 33 rivers. Flow rate data based on daily observations (gathered in PREVIMER project for 9 rivers), simulated data (from SMHI E-HYPE hydrological model (http://e-hypeweb.smhi.se)) and climatology (using monthly climatological data taken from GRDC (http://www.bafg.de/GRDC) and French “Banque Hydro” dataset (http://www.hydro.eaufrance.fr/)) are used in best available basis. An additional climatology for coastal runoff-rate (on a monthly basis) is used as IBI forcing to complement the high frequency forcing associated to the main 33 rivers. This added fresh water input makes the IBI forcing consistent with the ones imposed in the global system (derived from the Dai and al database (2009).

http://e-hypeweb.smhi.se/

http://www.bafg.de/GRDC





Issue : 5.0


NEMO version 3.6

Horizontal resolution 1/12° (5-6 km)

Vertical coord.

z*=f(ssh)

75 levels

Partial bottom cells

Bathymetry Composite (GEBCO_08 + different local

databases)

Free surface Explicit, non-linear, time-splitting

Vertical mixing k-epsilon

Tracer advection QUICKEST + Zalezak

Rivers

As lateral point sources

Merge of daily SMHI & PREVIMER & Monthly climatology (GRDC), 35 rivers

2D flux in complement from derived from the Dai and Trenberth climatology

Atm. Forcing ERA5 (hourly fields)

Surge capability Yes

Tides Yes (11 tidal components, astro pot)

Ocean colour effects Monthly fields of chlorophyll from ESA CCI

IC & OBCs GLORYS2V4 ¼°

Data Assimilation SAM2 (SEEK Filter) + IAU + bias correction +

adaptive error

Table 8: summary of IBIRYS characteristics

More details of physical parameterizations can be found in Maraldi et al. (2013).

II.2 Data assimilation method

The data assimilation scheme relies on a reduced-order Kalman filter. It is based on the Singular Evolutive Extended Kalman Filter (SEEK) formulation introduced by Pham et al. (1998). This approach has been used for several years at Mercator Ocean. It has been implemented in different ocean model configurations with a 5-day assimilation window (e.g. Tranchant et al., 2008; Lellouche et al.,





Issue : 5.0


2013). In all Mercator Ocean forecasting systems, the forecast error covariance is based on the statistics of a collection of predefined 3-D ocean state anomalies. The anomalies are computed from a long numerical experiment (typically around 8 years) with respect to a running mean so that they can provide and estimate of the 7-day scale error on the ocean state at a given period of the year for Temperature (T), Salinity (S), zonal velocity (U), meridional velocity (V) and Sea-Surface-Height (SSH).

More precisely, each temporal anomaly corresponds to the difference between the model state

and a running mean over a fixed time period window ranging from –τ to τ (Figure 2). It should also be noted that the analysis increment is a linear combination of these anomalies and depends on the innovation (observation minus model forecast equivalent as in Ide & al., 1997) and on the specified observation errors. This approach is similar to the Ensemble Optimal Interpolation developed by Oke et al. (2008). The last feature of the model forecast covariance is a localization technique which sets the covariances to zero beyond a distance defined as twice the local spatial correlation scale. Spatial (zonal and meridional directions) and temporal correlation scales (Figure 3) are then used to define an “influence bubble” around the analysis point in which data are also selected. For the IBI reanalysis, these ocean states come from a reference simulation (free simulation without assimilation from 1993 to 2018) carried out with the same ocean model configuration and same forcing. For each analysis cycle, we use 450 ocean state anomalies.

In practice, the SAM2 system is a sequential scheme. Here, we use a 5-day assimilation cycle. The observations minus model forecasts are calculated during the model integration at the right location and at the right time. After each analysis, the data assimilation produces increments of sea surface height (SSH), temperature, salinity and velocity. All these increments are applied progressively in the hindcast run using the Incremental Analysis Update (IAU) method (e.g. Bloom at al., 1996; Benkiran et al., 2008). It avoids inducing a shock in the model after each analysis, because of imbalance between the analysis increment and the model physics.

Figure 2 : Schematic representation of the anomalies calculation along a model trajectory





Issue : 5.0


Figure 3: Temporal (days), zonal and meridional (km) correlation scales used in this Mercator Ocean system

Assimilated observations

Altimeter data, in situ temperature and salinity vertical profiles, and satellite sea surface temperature are assimilated to estimate the initial conditions for numerical ocean forecasting.

Altimeter data consist of along-track Sea Level Anomalies (SLA)(CMEMS product SEALEVEL_GLO_PHY_L3_NRT_OBSERVATIONS_008_043 for near real time and SEALEVEL_GLO_PHY_L3_REP_OBSERVATIONS_008_061 for reprocessed). Observations along tracks are smoothed by several altimetric corrections (Le Traon et al., 2001). A Mean Dynamic Topography (MDT) is also used as a reference for SLA assimilation. This MDT is based on the “CNES-CLS13” MDT (Rio et al., 2014) with adjustments made using high resolution analyses, the last version of the GOCE geoid and an improved Post Glacial Rebound (also called Glacial Isostatic Adjustment).

Temperature and salinity in situ vertical profiles from the CORA 5.2 database tailored for assimilation (INSITU_GLO_TS_ASSIM_REP_OBSERVATIONS_013_051) and near real time in-situ observational data sources (INSITU_GLO_TS_ASSIM_NRT_OBSERVATIONS_013_047) are assimilated (CORA from 1992 to June 2019, NRT from June 2019).

OSTIA satellite SST is assimilated (SST_GLO_SST_L4_REP_OBSERVATIONS_010_011).

Observation operator

In data assimilation, the computation of innovation requires defining an observation operator to represent the model equivalent of the observation. In the present system, prognostic variables are interpolated on a quadrilateral grid, i.e., on the four canvas grid points surrounding the observation. The four weights are calculated with a bi-linear remapping interpolation. In the real time Mercator Ocean systems, the model sea level is filtered from the high frequency barotropic signal in a close





Issue : 5.0


manner than what is done in the altimetric data processing (e.g. Carrère at al. 2003; Dibarboure et al., 2011).

Adaptive tuning of observation error and implementation

In order to refine the prescription of observation errors, adaptive tuning of observation errors for the SLA and SST has been implemented. The method has not been used for temperature and salinity vertical profiles because of the lack of in situ data. Then, 3D fixed observation errors are used for the assimilation of in situ temperature and salinity vertical profiles.

The adaptive tuning method consists in the computation of a ratio which is a function of observation errors, innovations and residuals. It helps correcting inconsistencies on the specified observation errors. Following Desroziers et al. (2005), this ratio can be expressed as:

Ideally, is equal to 1. When the ratio is less (respectively larger) than 1, it means that the observation error is overestimated (respectively underestimated). The objective of this diagnostic is to improve the error specification by tuning an adaptive weight coefficient acting on the error of each assimilated observation. As a first guess of the method, the initial prescribed observation error matches the one used in the previous system (Lellouche at al., 2013) where the observation error variance was increased near the coast and on the shelves for the assimilation of SLA, and increased only near the coast (within 50 km of the coast) for the assimilation of SST. We also use a map of minimum error for altimetry in order to prevent the method to decrease to much the error.

BIAS Correction

In addition to the assimilation scheme, a method of bias correction has been developed. This method is based on a 3-D-Var approach which takes into account cumulative innovations over the last 3-month period in order to estimate large-scale temperature and salinity biases when enough observations are available. The aim of the bias correction is to correct the large-scale, slowly evolving error of the model, whereas the SAM assimilation scheme is used to correct the smaller scales of the model forecast error. For IBIRYS the correction is applied on a time-window of 90 days.





Issue : 5.0


III VALIDATION FRAMEWORK

The validation period is 1st January 1993 – 23 December 2018 (the period spanned by the reanalysis). The times series are based on monthly fields except for innovation diagnostics (see T-DA_INNO_STAT, S-DA_INNO_STAT, SL-DA-INNO_STAT, SI-CONC-DA-INNO_STAT sections), and the specific High-frequency IBI-Reanalysis validation performed with independent in-situ observational data (in hourly frequency basis) from the Puertos del Estado moorings located along the Spanish coast.

The validation methodology defined and used to validate the ocean reanalysis products is compliant with the WP18 validation plan. It consists in a series of diagnostics defined in common by all ocean reanalysis / reference forced simulation producers. The list of metrics applied following the common guidelines for Reanalysis validation (based on monthly output data) are:

Temperature

Metric name Description

T-CLASS1-T3D_MEAN Maps of long-term annual mean difference between WOA 2009 climatology and REA temperature at 0 m, 200 m, 600 m, 1500 m and 2000 m (closest model vertical levels).

T-CLASS1-SST_MEAN Maps of long-term annual mean difference between SST reference data set (Reynolds AVHRR 0.25°x0.25°) and REA SST.

T-CLASS1-SST_TREND Map of SST trend (computed from annual average) over the REA period.

T-CLASS1_SST_DAILY IF DAILY SST FIELDS ARE PRODUCED. Map of the std dev of difference between daily reference SST and REA SST.

T-CLASS2-MOORINGS Monthly mean values of REA temperature profiles collocated on observed temperature profiles for reference moorings. The following quantities are displayed: - Model and observation time series - Correlation between model and observations - Model – Observation mean bias and RMS difference

T-CLASS3-SST_QNET_2DMEAN

- Domain averaged monthly SST and surface net heat flux as a function of time. - Long term mean value of the surface net heat flux.

T-CLASS3-HC_LAYER Domain average of temperature monthly fields as a function of time in the layers [0 m; 200 m], [200 m; 600 m], [600 m; 1500 m], [1500 m; 2000 m].

T-CLASS4-LAYER Time series of the bias and rms difference of the observation minus REA (difference for in situ temperature averaged) in different layers and for each month using CORA4.1 in situ profile observation data set. The time series are calculated over the whole domain in the following layers: [0 m; 200 m], [200 m; 600 m], [600 m; 1500 m], [1500 m; 2000 m] using reanalysis monthly fields.

T-DA-INNO_STAT - SST innovation bias and rms difference averaged over the whole domain. - In situ temperature innovation bias and rms difference averaged over the whole domain as a function of time.

Table 9 Description of the metrics used for temperature in the IBI-Rea validation.





Issue : 5.0


Salinity


S-CLASS1-S3D_MEAN Maps of long-term annual mean difference between WOA 2009 climatology and REA salinity at 0 m, 200 m, 600 m, 1500 m and 2000 m (closest model vertical levels).

S-CLASS1-SSS_TREND Map of SSS trend (computed from annual average) over the REA period

S-CLASS2-MOORINGS Monthly mean values of REA temperature profiles collocated on observed temperature profiles for reference moorings. The following quantities are displayed: - Model and observation time series - Correlation between model and observations - Model – Observation mean bias and RMS difference

S-CLASS3-SSS_SFW_2DMEAN

- Domain averaged monthly SSS and surface net fresh water flux as a function of time. - Long term mean value of the net surface fresh water flux.

S-CLASS3-SC_LAYER Domain average of salinity monthly fields as a function of time in the layers [0 m; 200 m], [200 m; 600 m], [600 m; 1500 m], [1500 m; 2000 m]

S-CLASS3-IC_CHANGE Domain average salinity change with respect to initial condition as a function of depth and time.

S-CLASS4-LAYER Time series of the bias and rms difference of the observation minus REA (difference for in situ salinity averaged) in different layers and for each month using CORA4.1 in situ profile observation data set. The time series are calculated over the whole domain in the following layers: [0 m; 200 m], [200 m; 600 m], [600 m; 1500 m], [1500 m; 2000 m] using reanalysis monthly fields.

S-DA-INNO_STAT - In situ salinity innovation bias and std dev averaged over the whole domain as a function of time.

Table 10 Description of the metrics used for salinity in the IBI-Rea validation.

Currents


UV-CLASS1-15 m_MEAN NOAA AOML drifter-derived climatology for total current, including Ekman component (see http://www.aoml.noaa.gov/phod/dac/drifter_climatology.html)

UV-CLASS2-MOORINGS Monthly mean values of REA temperature profiles collocated on observed temperature profiles for reference moorings. The following quantities are displayed: - Model and observation time series - Correlation between model and observations - Model – Observation mean bias and RMS difference

UV-CLASS2-MKE_1000 m

Mean value of kinetic energy (KE=0,5x(U2+V

2)) at 1000 m depth estimated from ANDRO

Argo drift data base (http://wwz.ifremer.fr/lpo/Produits/ANDRO ) over the years 2002-2009. To be consistent, the averaging period for the reanalysis will also be 2002-2009.

Table 11: Description of the metrics used for currents in the IBI-Rea validation.

http://www.aoml.noaa.gov/phod/dac/drifter_climatology.html





Issue : 5.0


Sea level


SL-CLASS1-TREND_2D Map of sea level trend as estimated from MyOcean altimetry delayed time L4 product. Same time period as the reanalysis to be fully consistent.

SL-CLASS2-TIDE_GAUGE Monthly mean values of sea level at tide gauges locations. The following quantities are displayed on a map at tide gauge location: - REA sea level minus observation std dev, - correlation, - percentage of explained variance (100x(1-variance(model-obs)/variance(obs) ))

SL-CLASS3-2DMEAN Domain averaged sea level monthly mean time series.

SL-DA-INNO_STAT Sea level innovation bias and rms difference averaged over the whole domain as a function of time.

Table 12 Description of the metrics used for sea level in the IBI-Rea validation.

Transport


UV-CLASS3-VOL_TRANSP

Volume transport through sections: long term mean value and STD from monthly means. Published estimation for Gibraltar, South Portugal and Bay of Biscay.

Table 13 Description of the metrics used for transport in the IBI-Rea validation.

High Frequency IBI reanalysis validation: List of metrics.

The high frequency validation performed with the IBI reanalysis outputs that come to complement the previous described validation framework (focused on monthly scales), is based on the following metric lists.


SST-CLASS2-HF-MOORINGS

IBI-RE Hourly mean SST values collocated at buoy mooring positions are compared with hourly measured SST data from 16 PdE moorings. Following statistics and metrics are computed and displayed:

IBI-RE & OBS time series (both for absolute values and anomalies)

Bias, RMS error, time correlation index and scatter index Quantile-quantile plots and Taylor diagram are generated.

UV-CLASS2-HF-MOORINGS

U- and V- current components from IBI-RE hourly outputs collocated at buoy mooring positions are compared with hourly zonal and meridional current components observed at the 16 PdE moorings. Following statistics and metrics are computed and displayed:

IBI-RE & OBS time series.

Current distribution by direction (current roses plots)

Bias, RMS error, time correlation index and scatter index.

Quantile-quantile plots and Taylor diagram generated.

Progressive vector diagrams displayed

Subinertial currents computed and displayed.

Table 14: Description of the CLASS2 metrics used in the IBI-Rea validation with the Puertos del Estado moorings (continues in the next page).





Issue : 5.0



SSS-CLASS2-HF-MOORINGS

IBI-RE Hourly mean Sea surface salinity values collocated at buoy mooring positions are compared with hourly measured SSS data from 16 PdE moorings. Following statistics and metrics are computed and displayed:

IBI-RE & OBS time series (both for absolute values and anomalies)

Bias, RMS error, time correlation index and scatter index

Quantile-quantile plots and Taylor diagram are generated. SSS values together with the SST ones are depicted in T/S diagrams allowing characterization of main surface water masses occurred at each location.

Table 15: Description of the CLASS2 metrics used in the IBI-Rea validation with the Puertos del Estado moorings.

Other

Comparison to Mixed Layer Depth climatology.





Issue : 5.0


IV VALIDATION RESULTS

IV.1 Temperature

IV.1.1 T-CLASS1-T3D-MEAN

Compared to the climatology, the temperature of the RAN is higher between surface and 400 m depth (Figure 4). At 1000 m depth, the RAN temperature is higher between 40 and 48°N, but lower in the southern part of the domain. At 2000 m depth, the RAN temperature is higher in the Cadiz sea and west of the Iberian Peninsula, and lower elsewhere.

Figure 4: Mean differences (model-climatology) in temperature between IBIRYSV3R1 and the World Ocean Atlas 2009 climatology (WOA09) at 0.5 m, 97 m, 300 m, 773 m and 1945 m depth, respectively. For this comparison, model data have been averaged over the period 1993 – 2018.





Issue : 5.0


IV.1.2 T-CLASS1-SST-MEAN

Compared to the satellite observations, the RAN temperature is colder except along the south coast of North Sea, and in the Mediterranean Sea along the French coast and between Spain and Ibiza Islands. On the contrary, the GLO RAN (right panel) exhibits a warm difference, with higher temperature than the observations in the Mediterranean Sea and west of the Iberian Peninsula.

Figure 5: Mean differences (observations-model) in sea surface temperature between IBIRYSV3R1 and the AVHRR SST data (left), and GLORYS2V4 and the AVHRR SST data (right). For this comparison, model data have been averaged over the period 1993 – 2018.





Issue : 5.0


IV.1.3 T-CLASS1-SST-TREND

The general pattern of the SST trend of IBI RAN is similar to the observations, but the IBI RAN slightly underestimates this positive trend.

Figure 6: SST trend of IBIRYSV3R1 and AVHRR data. Covered time period: 1993 – 2018.

IV.1.4 T-CLASS2-MOORINGS

Figure 7, Figure 8, Figure 9 and Figure 10 display the monthly time series of near surface temperature at selected buoys. The annual cycle is the dominant signal, and the IBI RAN is close to the observations.

Figure 7: monthly time series of near surface temperature (°C)(black line: IBIRYSV3R1; red squares: observations) at buoys K1 and Britany. Mean difference, RMSD and correlation are estimated over the period 1993-2018 (depending on the in-situ time series length).





Issue : 5.0


Figure 8: monthly time series of near surface temperature (°C)(black line: IBIRYSV3R1; red squares: observations) at buoys Gascogne and Estaca de Bares. Mean difference, RMSD and correlation are estimated over the period 1993-2018 (depending on the in-situ time series length).

Figure 9: monthly time series of near surface temperature (°C)(black line: IBIRYSV3R1; red squares: observations) at buoys Tenerife and Cadiz. Mean difference, RMSD and correlation are estimated over the period 1993-2018 (depending on the in-situ time series length).





Issue : 5.0


Figure 10: monthly time series of near surface temperature (°C)(black line: IBIRYSV3R1; red squares: observations) at buoys Cabo de Gata and Tarragona. Mean difference, RMSD and correlation are estimated over the period 1993-2018 (depending on the in-situ time series length).

IV.1.5 T-CLASS3-SST-QNET-2DMEAN

The domain average of SST and surface net heat flux of IBI RAN is in good agreement with the observations. This is also the case of the GLO RAN (shown in Figure 11).

Figure 11: Domain averaged of SST (left, °C) and surface net heat flux (right, W.m-2) over the period 1993 – 2018: observations (black), GLORYS2V4 (green), IBIRYSV3R1 (red).

Long term mean value of the surface net heat flux: GLORYS2V4: -9.7 W.m-2, IBIRYSV3R1: -5.5 W.m-2.





Issue : 5.0


IV.1.6 T-CLASS3-HC_LAYER

The IBI RAN temperature averaged by level is close to the GLO RAN, and higher than the climatology. One can notice a sudden unrealistic decrease of the temperature starting in 1998. The IBI and GLO RAN come back to the mean values of 1998 only in 2003 (shown in Figure 12).

Figure 12: Time evolution of domain average temperature in different layers ([0-200 m], [200 m-600 m], [600-1500 m], [1500-2000 m], [0-2000 m], [2000 m-bottom], [0-bottom]). WOA 09 climatology (black), WOA 98 climatology (magenta), GLORYS2V4 (green), IBIRYSV3R1 (red), IBIRYSV3R1 free (blue).





Issue : 5.0


IV.1.7 T-CLASS3-IC_CHANGE

As noticed previously there is a strong decrease of the temperature compared to the initial conditions during the 1998-2003 period (shown in Figure 13). After that there is an increase of the temperature between the surface and 1000 m depth until 2015.

Figure 13: Time evolution of temperature change with respect to initial conditions over the period 1993 – 2018: IBIRYSV3R1.

IV.1.8 T-CLASS4-LAYER

Looking at the RMS of the difference between the RAN and the observations (Figure 14 and Figure 15), the IBI RAN exhibits a cold bias in the surface layer and in the layer [0-200m]. The annual cycle is clearly visible in the first layers (surface to 200 m) with higher RMSD in summer. One can observe that the RMSD values decrease after 2005 with the increase of the number of in-situ profiles thanks to the ARGO program.





Issue : 5.0


Figure 14: Time evolution of the difference (right) between the reanalysis and CORA in-situ temperature profiles (top) over the period 1993 – 2018 (in °C).

Figure 15: Time evolution of the RMS of the difference between the reanalysis and CORA in-situ temperature profiles (top) over the period 1993 – 2018 (in °C).





Issue : 5.0


IV.1.9 T-DA-INNO-STAT

Here we compare the IBI RAN temperature during the assimilation process with the assimilated observations (Argo profiles only). For the surface temperature, the mean difference and RMS difference values are lower than 0.5°C (Figure 16). Under the surface, the maximum differences are located in the thermocline layer as expected, with a clear annual cycle. There is also a secondary maximum of difference near 1000 m, depth of the Mediterranean waters in the Atlantic Ocean. As noticed previously, the differences with observations slightly decrease after 2005 between 100 and 800 m.

Figure 16: Time evolution of the domain average innovation in term of bias (left) or rmsd (right): SST (top) and CORA in-situ temperature profiles (bottom). Top: OSTIA SST data (in °C). Covered time period: 2001 – 2018 for vertical Argo profiles, 1993-2018 for SST.





Issue : 5.0


IV.2 Salinity

IV.2.1 S-CLASS1-S3D_MEAN

Compared to the climatology (shown Figure 17), the salinity of the RAN is higher between surface and 400 m depth (but not in the Algerian current, especially at 100 m depth). At 1000 m depth, the RAN temperature is higher between 38 and 48°N (and in the Mediterranean Sea), but lower in the southern part of the domain. At 2000 m depth, the RAN salinity is lower in the northern part of the domain, and higher elsewhere.

Figure 17: Mean differences (model-climatology) in salinity between IBIRYSV3R1 and the World Ocean Atlas 2009 climatology (WOA09) at 0.5 m, 97 m, 300 m, 773 m and 1945 m depth, respectively. For this comparison, model data have been averaged over the period 1993 – 2018.





Issue : 5.0


IV.2.2 S-CLASS1-SSS-TREND

The salinity trend is positive in the northern part of the domain (especially in the southern part of the North Sea) and in the Mediterranean Sea (Figure 18).

Figure 18: SSS trend of IBIRYSV3R1. Covered time period: 1993 – 2018.

IV.2.3 S-CLASS2-MOORINGS

Figure 19 and Figure 20 display the monthly time series of near surface salinity at selected buoys. The RMS difference between IBI RAN and observations varies from 0.2 to 0.6. The correlation is generally small.

Figure 19: monthly time series of near surface salinity (psu) (black line: IBIRYSV3R1; red squares: observations) at buoys Estaca de Bares and Cadiz. Mean difference, RMSD and correlation are estimated over the period 2002-2018 (depending on the in-situ time series length).





Issue : 5.0


Figure 20: monthly time series of near surface salinity (psu) (black line: IBIRYSV3R1; red squares: observations) at buoys Tarragona and Cabo de Gata. Mean difference, RMSD and correlation are estimated over the period 2002-2018 (depending on the in-situ time series length).





Issue : 5.0


IV.2.4 S-CLASS2-SSS-SWF_2DMEAN

The domain average of SSS is slightly lower in IBI RAN (and consistently the surface net fresh water flux is slightly higher in IBI RAN).

Figure 21: Domain averaged of SSS (left, psu) and surface net fresh water flux (right, mm.d-1) over the period 1993 – 2018: GLORYS2V4 (green), IBIRYSV3R1 (red).

The long term mean value of the net surface fresh water for GLORYS2V4 is 0.43 mm.d-1, and for IBIRYSV3R1 is 0.55 mm.d-1.

IV.2.5 S-CLASS3-SC-LAYER

The IBI RAN salinity averaged by level is close to the GLO RAN (shown in Figure 22), and higher than the climatology. One can notice a sudden unrealistic decrease of the salinity starting in 1998 (consistent with the decrease in temperature previously noticed). The IBI and GLO RAN come back to the mean values of 1998 only in 2003.





Issue : 5.0


Figure 22: Time evolution of domain’s average salinity in different layers ([0-200 m], [200 m-600 m], [600-1500 m], [1500-2000 m], [0-2000 m], [2000 m-bottom], [0-bottom]). WOA 09 climatology (black), WOA 98 climatology (magenta), GLORYS2V4 (green), IBIRYSV3R1 (red), IBIRYSV3R1 free (blue).





Issue : 5.0


IV.2.6 S-CLASS3-IC_CHANGE

As noticed previously there is a strong decrease of the salinity compared to the initial conditions during the 1998-2003 period between 1000 and 2000 m depth (shown in Figure 23). On the contrary the salinity increases near the bottom during the same period. After that there is an increase of the salinity between the surface and 1000 m depth until 2015.

Figure 23: Time evolution of salinity change with respect to initial conditions over the period 1993 – 2018: IBIRYSV3R1.

IV.2.7 S-CLASS4-LAYER

One can observe that the mean differences and RMSD values decrease after 2005 with the increase of the number of in-situ profiles thanks to the ARGO program.





Issue : 5.0


Figure 24: Time evolution of the difference between the reanalysis and CORA in-situ salinity profiles (top) over the period 1993 – 2018 (in psu).

Figure 25: Time evolution of the RMS of the difference between the reanalysis and CORA in-situ salinity profiles (top) over the period 1993 – 2018 (in psu).





Issue : 5.0


IV.2.8 S-DA-INNO_STAT

Here we compare the IBI RAN salinity during the assimilation process with the assimilated observations (Argo profiles only; Figure 26). As for temperature the maximum differences with observations are located in the thermocline layer. There is also a secondary maximum of difference near 1000 m, depth of the Mediterranean waters in the Atlantic Ocean.

Figure 26: Time evolution of the domain average innovation in term of bias (left) or rmsd(right) (CORA in-situ salinity profiles). Covered time period: 2001 - 2018.





Issue : 5.0


IV.3 Currents

IV.3.1 UV-CLASS1-15 m_MEAN

At large scale the IBI RAN and observations are in good agreement (show in Figure 27). The major current can be seen: one branch of the North Atlantic Drift, the Azores current and the Algerian current.

IBIRYSV3R1 mean zonal velocity (m/s)

Climatology zonal velocity (m/s)

IBIRYSV3R1 mean meridional velocity (m/s)

Climatology meridional velocity (m/s)

Figure 27: Zonal (top) and meridional (bottom) 15 m depth current climatology. IBIRYSV3R1 (left) and NOAA AOML drifter-derived climatology (right). For this comparison, model data have been averaged over the period 1993 – 2018.





Issue : 5.0


IV.3.2 UV-CLASS2-MOORINGS

Figure 28 to Figure 31 display the monthly time series of zonal and meridional component of the near surface current for IBI RAN and selected observations. The RMS differences for both components vary between 4 and 10 cm/s.

Figure 28: monthly time series of near surface zonal (left) and meridional (right) velocity (m.s-1) (black line: IBIRYSV3R1; red squares: observations) at buoy Estaca de Bares. Mean difference, RMSE and correlation are estimated over the period 1993-2018 (depending on the in-situ time series length).

Figure 29: monthly time series of near surface zonal (left) and meridional (right) velocity (m.s-1) (black line: IBIRYSV3R1; red squares: observations) at buoy Cadiz. Mean difference, RMSD and correlation are estimated over the period 1993-2018 (depending on the in-situ time series length).





Issue : 5.0


Figure 30: monthly time series of near surface zonal (left) and meridional (right) velocity (m.s-1) (black line: IBIRYSV3R1; red squares: observations) at buoy Cabo de Gata. Mean difference, RMSD and correlation are estimated over the period 1993-2018 (depending on the in-situ time series length).

Figure 31: monthly time series of near surface zonal (left) and meridional (right) velocity (m.s-1) (black line: IBIRYSV3R1; red squares: observations) at buoy Tarragona. Mean difference, RMSD and correlation are estimated over the period 1993-2018 (depending on the in-situ time series length).

IV.3.3 UV-CLASS2-MKE_1000 m

The MKE is higher in IBI RAN than in the Andro product (Figure 32). This can perhaps be explained by the few number of observations for this area, which does not allow to catch all the variability of the currents.





Issue : 5.0


Figure 32: Mean value of kinetic energy at 1000 m. Left: MKE estimated from ANDRO Argo drift data base; right: IBIRYS. Covered time period: 1993-2018.

IV.4 Sea level

IV.4.1 SL-CLASS1-TREND_2D

IBI RAN sea level trend is in good agreement with observations, and performs slightly better than the GLO RAN (Figure 33). The trend is positive almost everywhere. IBI RAN slightly underestimates the trend in the north sea.

Figure 33: SLA trend of MyOcean altimetry delayed time L4 product (left), IBIRYSV3R1 (middle) and GLORYS2V4 (right), respectively. Covered time period: 1993-2018.





Issue : 5.0


IV.4.2 SL-CLASS2-TIDE_GAUGE

Comparisons with tide gauges measurements show a good agreement between observations and IBI RAN are shown in Figure 34. Most of the correlations are higher than 0.7, and the mean differences are lower than 5 cm in the southern part of the domain and lower than 10 cm in the northern pas of the domain. These results are for residual sea level (that means that the sea level is de-tided).

Figure 34: From left to right, rms difference(cm), correlation and explained variance (%) between reanalysis sea level and tide gauges observations (from CMEMS in-situ TAC).

IV.4.3 SL-CLASS3-2DMEAN

IBI RAN sea level is in good agreement with the observations (and also with GLO RAN sea level). As previously seen, there is a positive trend (Figure 35).

Figure 35: Domain averaged sea level monthly mean time series: Aviso SLA CCI observations (black), GLORYS2V4 (green), IBIRYSV3R1 (red). Covered time period: 1993 - 2018.





Issue : 5.0


IV.4.4 SL-DA-INNO_STAT

Next figures show that the reanalysis performs well compared to along track sea level anomalies. The mean difference is centered and the RMS difference is smaller than 7 cm in average (Figure 36).

Figure 36: Seal level innovation bias (left) and rmsd (right) over the whole domain (cm). Colours represent different altimetry satellites (ERS, T/P, Jason, Envisat, GFO, …). Covered time period: 1993 - 2018.

IV.5 Transports

The inflow transport (from Atlantic Ocean to Mediterranean Sea) in the Gibraltar strait is over-estimated, while the outflow is under-estimated. As a consequence the net transport is over-estimated (Table 16).

Density classes 0-37.25 37.25-60 0-60

Transport (Sv) 1.12 0.49 0.63

Published estimations

(Sv)

0.66 – 0.97 -0.84 – -0.57

Table 16: Mean volume transport across the Gibraltar section in IBIRYSV3R1 (1993-2018 mean) and published estimations (Tsimplis and Bryden, 2000; Baschek et al, 2001, Lafuente et al, 2002).





Issue : 5.0


IV.6 Other monthly metrics

IV.6.1 OTH-CLASS1-MLD_CLIM_SEASON

The comparison of IBI RAN with De Boyer Montegut mixed layer depth climatology shows similar pattern with a meridian gradient in MLD values (the MLD is deeper in the north) (Figure 37). In March the convection occurring in the Gulf of Lion is visible in IBI RAN, but not in the climatology due to the lack of observations (that is why the climatology is very smooth compared to the RAN).

Figure 37: Climatological mixed layer depth (m) in March (top) and September (bottom). IBIRYSV3R1 (left) and De Boyer Montegut climatology data (right).





Issue : 5.0


V QUALITY CHANGES SINCE PREVIOUS VERSION

IBIRYSV3R1 is the new version of the reanalysis.

The previous version (IBIRYSV2R1) covered the period 1992-2016, and has then been extended several times (it covers now 1992-2019; see Table 17 for a detailed view of the evolution of the product).

System Version (Project/Service)

Operational launch

Novelties

IBI-V2 (MyOcean2)

FIRST RELEASE

19/04/2016 First release of the IBI-MFC PHY-V2 MY hindcast. The IBI physical Product, delivered to users from CMEMS V3 release (temporal coverage 2002-2014).

IBI-V4 (CMEMS-I)

FIRST RELEASE

26/04/2018 First release of the IBI-MFC MY PHY/BGC-V4 product.

Delivered to users from CMEMS V4 release, with a temporal coverage from 01/01/1992 to 31/12/2016.

Both physical and biogeochemical models updated (NEMO/PISCES V3.6). Data Assimilation system released and new observational data sources included.

CMEMS Apr 2019 release (CMEMS-II) TEMPORAL EXTENSION

16/04/2019 Extension of the IBI-MFC MY PHY/BGC-V4 product to cover the period 01/01/1992 to 24/12/2017. In addition to the year 2017, the years 2014 to 2016 were run again to take into account new reprocessed observational and forcing data.

CMEMS Jul 2019 release (CMEMS-II) TEMPORAL EXTENSION

09/07/2019 Extension of the IBI-MFC MY PHY/BGC-V4 product to cover the period 01/01/1992 to 24/12/2018 In addition to the year 2018, the years 2014 to 2017 were run again to take into account new reprocessed observational and forcing data.

CMEMS Jul 2020 release (CMEMS-II) TEMPORAL EXTENSION

07/07/2020 Extension of the IBI-MFC MY PHY/BGC-V4 product to cover the period 01/01/1992 to 24/12/2019. In addition to the year 2019, the year 2018 was re-run.

CMEMS Dec 2020 release (CMEMS-II)

FIRST RELEASE

15/12/2020 First release of the IBI-MFC MY PHY/BGC-V5 product Update of the whole historic catalogue to cover the period 01/01/1993 to 24/12/2019 Both physical and biogeochemical models updated (NEMO/PISCES V3.6). The Data assimilation system was released and new observational data sources were included. New Interim temporal extensions.

Table 17: Historical evolution of the IBI MFC Multi-Year Physical Reanalysis System along the Service. Time coverage of each model set-up used to produce the IBI_MULTIYEAR_PHY_005_002 product, as well as main novelties introduced, are provided.





Issue : 5.0


The new version shows comparable results than the previous version. An important feature is the release of the atmospheric forcing from ERA interim to ERA5. In term of comparison to altimetry, the new version has a negative difference with satellite measurements. Consequently, the temperature in the surface and sub-surface layers has also a negative difference with satellite and in-situ measurements. Despite these features, the new version corrects some spurious biases in temperature and salinity that sometimes happen in the previous version.

New Interim temporal extension

Since July 2020, the IBI PHY reanalysis is planned to be updated following a new catalogue policy destined to ensure the dissemination of a quasi-near real time reanalysis timeseries.

For this purpose, every month (M), interim MY products covering the previous month (M-1) are generated with the same model and configuration used for the current reanalysis. For producing M-1 outputs, the model is forced with ERA5T atmospheric data from the ECMWF and it assimilates near real time (NRT) upstream data, since reprocessed (REP) data for such a short term are non-available.

Then, at a biannual frequency, the model is rerun from M-18 to M-13 months in order to use REP upstream data and ERA5 ECMWF atmospheric forcing, thus ensuring that the best available MY products are generated.





Issue : 5.0


VI REFERENCES

Barnier, B., Madec, G., Penduff, T., Molines, J.-M., Treguier, A.-M., Le Sommer, J., Beckmann, A., Biastoch, A., Boning, C., Dengg, J., Derval, C., Durand, E., Gulev, S., Remy, E., Talandier, C., Theetten, S., Maltrud, M., McClean, J., and De Cuevas, B.: Impact of partial steps and momentum advection schemes in a global ocean circulation model at eddy-permitting resolution, Ocean Dynam., 56, 543–567, doi:10.1007/s10236-006-0082-1, 2006.

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., and Weatherall, P.: Global bathymetry and elevation data at 30 arc seconds resolution: SRTM30 PLUS, Mar Geod., 32, 355–371, 2009.

Benkiran M.,Greiner E. (2008), Impact of the Incremental Analysis Updates on a Real-Time System of the North Atlantic Ocean. Journal of Atmospheric and Oceanic Technology 25(11): 2055.

Bloom, S.C., Takas, L.L., Da Silva, A.M., Ledvina, D. (1996), Data assimilation using incremental analysis updates. Monthly Weather Review, 124, 1256–1271.

Carrère, L. and Lyard, F.: Modelling the barotropic response of the global ocean to atmospheric wind and pressure forcing - comparisons with observations, Geophys. Res. Let., 30(6), pp 1275, 2003.

Dai A., Trenberth K.E. (2002), Estimates of Freshwater Discharge from Continents: Latitudinal and Seasonal Variations. J. of Hydrometeorology, 3 pp 660-687.

Dai, A., Qian, T., Trenberth, K., and Milliman, J. D.: Changes in Continental Freshwater Discharge from 1948 to 2004, J. Climate, 22, 2773–2792, 2009.

Desroziers, G., Berre, L., Chapnik, B., and Polli, P., 2005: Diagnosis of observation, background and analysis-error statistics in observation space, Q. J. R. Meteorol. Soc., 131, pp. 3385–3396, doi: 10.1256/qj.05.108.

Dibarboure G. M.-I. Pujol, F. Briol , P. Y. Le Traon , G. Larnicol , N. Picot , F. Mertz , M. Lellouche, J.-M., Le Galloudec, O., Drévillon, M., Régnier, C., Greiner, E., Garric, G., Ferry, N., Desportes, C., Testut, C.-E., Bricaud, C., Bourdallé-Badie, R., Tranchant, B., Benkiran, M., Daudin, A. and De Nicola, C. : Evalution of global monotoring and forecasting systems at Mercator Océan. Ocean Sci., 9, 57-81, 2013.

Egbert, G., Bennett, A., and Foreman, M.: TOPEX/Poseidon tides estimated using a global inverse model, J. Geophys. Res., 99, 24821–24852, 1994.

Garric G., Verbrugge N, and C Bricaud (2011), Large scale ERAinterim radiative and precipitation surface fluxes assessment, correction and application on 1/4° global ocean 1989-2009 hindcasts. EGU2011-10892. XY136, EGU 2011.

Large, W. G. and Yeager S. G.: Diurnal to decadal global forcing for ocean and sea-ice models: the data sets and flux climatologies, NCAR Technical Note: NCAR/TN-460+STR, CGD Division of the National Center for Atmospheric Research, 2004.

Lellouche, J.-M., Le Galloudec, O., Drévillon, M., Régnier, C., Greiner, E., Garric, G., Ferry, N., Desportes, C., Testut, C.-E., Bricaud, C., Bourdallé-Badie, R., Tranchant, B., Benkiran, M., Daudin, A.





Issue : 5.0


and De Nicola, C. : Evalution of global monotoring and forecasting systems at Mercator Océan. Ocean Sci., 9, 57-81, 2013.

Leonard, B. P.: A stable and accurate convective modelling procedure based on quadratic upstream interpolation, Comp. Method. Appl. M. 19, 59–98, 1979.

Le Traon, P-Y., G. Dibarboure, and N. Ducet, 2001: Use of a high-resolution model to analyze the mapping capabilities of multiple-altimeter missions. J. Atmos. Oceanic Technol., 18, 1277–1288.

Lyard, F., Lefevre, F., Letellier, T., and Francis, O.: Modelling the global ocean tides: modern insights from FES2004, Ocean Dynam., 56, 394–415, 2006.

Madec, G., (2008), NEMO ocean engine. Note du Pôle de modélisation, Institut Pierre-Simon Laplace(IPSL), France, No 27, ISSN No 1288-1619.

Maraldi, C., Chanut, J., Levier, B., Reffray, G., Ayoub, N., De Mey, P., Lyard, F., Cailleau,S., Drévillon, M., Fanjul, E.A., Sotillo, M.G., Marsaleix, P., and the Mercator team: NEMO on the shelf: assessment of the Iberia-Biscay-Ireland configuration. Ocean Sci., 9, 745-771, 2013.

Oke, P. R., G. B. Brassington, D. A. Griffin, and A. Schiller (2008), The Bluelink Ocean Data Assimilation System (BODAS), Ocean Modell., 21, 46– 70, doi:10.1016/j.ocemod.2007.11.002.

Pham, D., Verron, J., and Roubaud, M., (1998).A Singular Evolutive Extended Kalman filter for data assimilation in oceanography, J. Mar. Syst., 16(3-4), 323-340.

Rio, M. H., S. Guinehut, and G. Larnicol (2011), New CNES-CLS09 global mean dynamic topography computed from the combination of GRACE data, altimetry, and in situ measurements, J. Geophys. Res., 116, C07018, doi:10.1029/2010JC006505.

Shchepetkin, A. F. and McWilliams, J. C.: The regional ocean modelling system (ROMS): a split-explicit, free-surface, topography-following-coordinate oceanic model, Ocean Model., 9, 347–404, 2004.

Tranchant B., C.-E., Testut, R. Bourdallé-Badie, C. Derval, O. Le Galloudec, Y. Drillet, C. Bricaud and G. Garric, The Global 1/12° Mercator Ocean forecasting system: new insights, Proceeding of the EUROGOOS Exeter Conference, May 2008, Exeter, UK.

Umlauf, L. and Burchard, H.: A generic length-scale equation for geophysical turbulence models, J. Mar. Res., 61, 235–265, 2002.

Zalezak, S. T., Fully multidimensional flux-corrected transport algorithms for fluids, J. Computat. Phys., 31, 335–362, 1979.

Documents

Atlantic -Iberian Biscay Irish- IBI Production Centre IBI ... · product, currently delivered through the CMEMS catalogue, has been performed. This section is devoted to provide essential