QUALITY INFORMATION DOCUMENT For Global Ocean …€¦ · QUALITY INFORMATION DOCUMENT For Global Ocean Reanalysis Multi-model Ensemble Products GREP-V1 GLOBAL-REANALYSIS-PHY-001-026

QUALITY INFORMATION DOCUMENT

For Global Ocean Reanalysis Multi-model Ensemble Products GREP-V1

GLOBAL-REANALYSIS-PHY-001-026

Issue: 1.1

Contributors: Charles Desportes, Gilles Garric, Charly Régnier, Marie Drévillon, Laurent Parent, Gilles Garric, Yann Drillet, Simona Masina, Andrea Storto, Isabelle Mirouze, Andrea Cipollone, Hao Zuo, Magdalena Balmaseda, Drew Peterson, Richard Wood, Laura Jackson, Sandrine Mulet, Eric Greiner

Approval Date by Quality Assurance Review Group :

Quality Information Document

For product


Ref :CMEMS-GLO-QUID-001-026

Date : Novembre 2017

Issue : 1.1

© EU Copernicus Marine Service – Public Page 2/ 40

CHANGE RECORD

Issue Date § Description of Change Author Checked

By

1.0 3 January 2017

all First version of document Charles Desportes, Marie Drévillon

Yann Drillet

1.1 7 novembre 2017

add 2016 T&S and velocities validation

Charles Desportes Yann Drillet


For product




Issue : 1.1


TABLE OF CONTENTS

I EXECUTIVE SUMMARY ............................................................................................................................ 4

Products covered by this document ........................................................................................................... 4 I.1

Summary of the results ............................................................................................................................... 4 I.2

Temperature ....................................................................................................................................... 4 I.2.1

Salinity ................................................................................................................................................ 5 I.2.2

Velocities ............................................................................................................................................. 5 I.2.3

Estimated Accuracy Number ..................................................................................................................... 6 I.3

II Production Subsystems description ............................................................................................................... 7

III Validation framework ............................................................................................................................... 8

IV Validation results of GREP-V1 ................................................................................................................... 10

Temperature ........................................................................................................................................... 10 IV.1

T-CLASS1-T3D-MEAN .................................................................................................................. 10 IV.1.1

T-CLASS1-SST-MEAN ................................................................................................................... 12 IV.1.2

T-CLASS1-SST-TREND ................................................................................................................. 13 IV.1.3

T-CLASS2-MOORINGS ................................................................................................................. 14 IV.1.4

T-CLASS3-SST_2DMEAN ............................................................................................................. 15 IV.1.5

T-CLASS3-HC_LAYER ................................................................................................................. 15 IV.1.6

T-CLASS3-IC_CHANGE ............................................................................................................... 17 IV.1.7

T-CLASS4-LAYER ......................................................................................................................... 18 IV.1.8

T-STDEV .......................................................................................................................................... 22 IV.1.9

Salinity ..................................................................................................................................................... 24 IV.2

S-CLASS1-S3D_MEAN .................................................................................................................. 24 IV.2.1

S-CLASS1-SSS-TREND .................................................................................................................. 27 IV.2.2

S-CLASS2-MOORINGS ................................................................................................................. 28 IV.2.3

S-CLASS3-SC_LAYER ................................................................................................................... 29 IV.2.4

S-CLASS3-IC_CHANGE ................................................................................................................ 30 IV.2.5

S-CLASS4-LAYER .......................................................................................................................... 31 IV.2.6

S-STDEV .......................................................................................................................................... 35 IV.2.7

Currents .................................................................................................................................................. 37 IV.3

UV-CLASS1-15m_MEAN ............................................................................................................... 37 IV.3.1

UV-CLASS2-MKE_1000m ............................................................................................................. 39 IV.3.2

V references ..................................................................................................................................................... 40


For product




Issue : 1.1


I EXECUTIVE SUMMARY

Products covered by this document I.1

The products assessed in this document are referenced as: GLOBAL-REANALYSIS-PHY-001-026 for Global Reanalysis Multi-Model Ensemble Product GREP.

Global ocean reanalyses are homogeneous 3D gridded descriptions of the physical state of the ocean spanning several decades, produced with a numerical ocean model constrained with data assimilation of satellite and in situ observations. The multi-model ensemble approach allows uncertainties or error bars in the ocean state to be estimated. The ensemble mean may even provide, for certain regions and/or periods, a more reliable estimate than any individual reanalysis product.

Four reanalyses covering the 1993-2016 period during which altimeter altimetry data observations are available; GLORYS2V4 from Mercator Ocean (Fr), ORAS5 from ECMWF, GloSea5 from Met Office (UK), and C-GLORS05 from CMCC (It) provided four different time series of global ocean simulations 3D monthly estimates, which were post-processed to create the new product called GREP (Global Reanalysis Ensemble Product), and for which the following variables are available:

* sea_water_potential_temperature * sea_water_salinity * eastward_sea_water_velocity * northward_sea_water_velocity

Other variables like ssh and ice variables will be available in next releases.These reanalyses are built to be as close as possible to the observations (i.e. realistic) and in agreement with the model physics. It covers the “altimetric era” (namely 1

st of January 1993 until 31 December 2016). The numerical

products available for users are monthly mean averages describing the ocean from surface to bottom (5900 m).

The product comprises 12 datasets (6 datasets for temperature, and 6 for salinity), which consist for T and S of 4 global ocean reanalyses datasets, their ensemble mean and standard deviation, on a regular grid at 1° horizontal resolution.

Summary of the results I.2

Although all the individual members and the ensemble mean have been assessed independently, this summary only describes the quality of the GREP ensemble mean product. For more details on individual members results, please see the references listed at the end of the document, and section IV of this document.

Temperature I.2.1

The GREP-V1 ensemble mean has very small biases (less than 0.1°C on global average) and RMS errors (less than 1°C on global average) in temperature with respect to CMEMS CORA in situ data. The largest biases, RMS errors and disagreement between the four members occur during the 1993-2002 period, in the deep layers (under 800 m) and in the northern Atlantic and Mediterranean Sea, in the polar oceans, in the Indonesian region, and in the western side of the Pacific warm pool.


For product




Issue : 1.1


The heat content of the 0-700 m increases with time during the whole period. Only the deepest layers (below 700m depth) show a disagreement between individual members, indicating uncertainty on the heat content variability at these (far less observed) depths.

The thermal structure largely improves with the deployment of Argo buoys (after 2002), mainly in the upper layers (depth < 300m). The disagreement between members (GREP-V1 standard deviation) also decreases with time in accordance with the observations network density.

The GREP-V1 temperature estimates are not deemed reliable either in the Indonesian region, and the western side of the Pacific warm pool.

Salinity I.2.2

The sea surface salinity (SSS hereafter) is generally fresher (less than 0.02 psu on global average) than in situ observations. Large fresh anomalies are found in the western Tropical Atlantic and the Indonesian Throughflow. Saltier surface waters are also found in the Arctic Ocean compared to the Levitus climatology. A general salty bias (less than 0.1 psu) is found in the [50m-800m] layer. While no global average trend is detected, positive trends of SSS appear in the Tropical Pacific and Indian Oceans. Negative trends are found in the Arctic Ocean and in the Indonesian Throughflow.

Discrepancies in between members, biases and RMSE against all the in-situ observations drastically decrease with time in accordance with the increasing observations network density.

As for temperature the largest biases, RMS errors and disagreement between the four members occur during the 1993-2002 period, in the deep layers (under 800 m) and in the northern Atlantic, Mediterranean Sea, and in the polar oceans.

The GREP-V1 salinity estimates are not deemed reliable in the Mediterranean Sea, in the Mediterranean Sea outflow, the Indonesian region, and the western side of the Pacific warm pool.

Velocities I.2.3

The GREP-V1 mean zonal current is in very good agreement with the drifters climatology from CMEMS (including corrections of biases due to drogue loss). Deep currents energy is overestimated with respect to the drifters, especially in the western Tropical Pacific.


For product




Issue : 1.1


Estimated Accuracy Number I.3

Estimated Accuracy Numbers are calculated using mean and RMS departures of the ensemble mean from in situ observations.

Temperature (Celsius degrees)

Level Period 1993-2015 1993-2004 2005-2015

Mean RMS Mean RMS Mean RMS

0-5 0.05 0.76 0.07 0.86 0.04 0.70

5-100 0.04 0.98 0.06 1.09 0.03 0.92

100-300 0.04 0.87 0.05 0.95 0.03 0.82

300-800 0.05 0.52 0.06 0.59 0.05 0.48

800-2000 0.04 0.21 0.04 0.26 0.04 0.19

2000-5000 0.01 0.14 0.006 0.14 0.01 0.12

0-5000 0.04 0.49 0.05 0.61 0.04 0.45

Salinity (psu)

Level Period 1993-2015 1993-2004 2005-2015

Mean RMS Mean RMS Mean RMS

0-5 0.013 0.22 0.021 0.28 0.011 0.20

5-100 -0.004 0.17 -0.011 0.20 -0.002 0.16

100-300 -0.006 0.11 -0.013 0.13 -0.004 0.11

300-800 -0.001 0.07 -0.003 0.07 -0.001 0.07

800-2000 -0.002 0.03 -0.004 0.03 -0.002 0.03

2000-5000 -0.003 0.02 -0.003 0.02 -0.003 0.01

0-5000 -0.003 0.07 -0.005 0.08 -0.002 0.06


For product




Issue : 1.1


II PRODUCTION SUBSYSTEMS DESCRIPTION

Global ocean reanalyses are homogeneous 3D gridded descriptions of the physical state of the ocean spanning several decades, produced with a numerical ocean model constrained with data assimilation of satellite and in situ observations. The four reanalyses used here are based on the NEMO model, on the ORCA025 grid (1/4° horizontal resolution), with 75 vertical levels, all are forced at the surface by ERA interim, and all assimilate sea surface temperature SST, sea level anomalies SLA, sea ice concentrations SIC and in situ temperature and salinity profiles T/S(z) The main differences in between the four individual products are summarized in Table 1.

The reader will find more detailed descriptions in the following documents:

GLORYS2V4: http://marine.copernicus.eu/documents/QUID/CMEMS-GLO-QUID-001-025-011-017.pdf, and description of Mercator Ocean’s ocean analysis system: Lellouche et al (2013)

GLOSEA5v13: MacLachlan et al (2014), and description of FOAM: Blockley et al (2014)

C-GLORS: Storto et al (2016)

ORAS5: Zuo et al (2017), is an upgrade of ORAP5 (Zuo et al 2015)

Reanalysis Production

centre

COMMON Model version

Surface Forcing ASSIMILATION

GLORYS2V4 Mercator

Océan

NEMO, ORCA1/4°

75 vertical

levels

TKE

Altimetry ERA : 1993-2015

ERAinterim and

bulk formulae

Observations : SST, SLA, T/S profiles, SIC

Multivariate assimilation, monovariate for the SIC

NEMO3.1

LIM2

No surface nudging

precipitation, flux correction

Climatological runoff +

ice shelf and iceberg melting

SAM2 (SEEK)

Large scale bias correction

7-day assimilation window

Merge MDT (obs+model)

Reynolds SST, CORA

FOAM-GLOSEA5v13

(hereafter FOAM)

UK Met

Office

NEMO3.4

CICE4.1

Non linear free surface

SST, SSS surface nudging

NEMOVAR (3Dvar)

Bias correction


EN4

C-GLORS CMCC NEMO3.4

LIM2

SST, SSS, SIC surface nudging

OceanVar (3Dvar)

Large scale bias correction


Model MDT

Reynolds SST, EN4

ORAS5 ECMWF NEMO3.4.1

LIM2

Surface waves

SST, SSS surface nudging

NEMOVAR (3Dvar)


HadlSSTv2 SST, EN4

Table 1: synthetic view of the four reanalyses subsystems components

http://marine.copernicus.eu/documents/QUID/CMEMS-GLO-QUID-001-025-011-017.pdf

http://marine.copernicus.eu/documents/QUID/CMEMS-GLO-QUID-001-025-011-017.pdf


For product




Issue : 1.1


III VALIDATION FRAMEWORK

The validation methodology defined and used to validate the global ocean reanalysis products is described by Ferry et al (2013) and consists in a series of diagnostics defined in common by ocean reanalysis producers that were involved in the MyOcean projects.

The reference validation period is 1st January 1993 – 31 December 2014 (year 2015 is missing for one member at the time of writing). The times series are based on monthly fields.

Name.. Description Reference data set used for metric

T/S-CLASS1-T/S3D_MEAN

Maps of long-term annual mean difference between reference climatological data set and REA temperature at different depths (see reference data sets table for the chosen depths).

WOA 2009 annual mean interpolated on the model grid at 0m, 100m, 300m, 800m and 2000m (closest model vertical levels).

T-CLASS1-SST-MEAN

Maps of long-term annual mean difference between SST reference data set and REA SST.

REYNOLDS1/4° OR OSTIA reanalysis SST mean field.

T/S-CLASS1-SST/SSS_TREND

Map of SST/SSS trend (computed from monthly average) over the REA period

REYNOLDS1/4° OR OSTIA reanalysis monthly mean SST field

T/S-CLASS2-MOORINGS

Monthly mean values of REA temperature profiles collocated on observed temperature profiles for reference moorings. Correlation, stdev, visual inspection.

Moorings from (see annex for more details): TAO/TRITON: 0N-140W, 0N-110W, 0N-165E, 0N-147E PIRATA: 0N-23W, 0N-0E RAMA: 0N-90E

T-CLASS3-HC_LAYER

/SC

Domain average of temperature monthly fields as a function of time in different layers (see reference data sets table for the layers definition)

HC computation in the layers [0-700m], [0-2000m], [2000m-bottom], [0-bottom] HC estimates derived from CORA3.4 OI product (INSITU_GLO_TS_REP_OBSERVATIONS_013_001_b). INSITU TAC will provide the HC estimates in [0-700m], [0-2000m] layers.

T/S-CLASS3-IC_CHANGE

Domain average temperature change with respect to initial condition as a function of depth and time.

None


For product




Issue : 1.1


T/S-CLASS4-LAYER

Time series of the bias and std dev of the observation minus REA (difference for in situ temperature averaged) in different layers

Temperature from CORA3.4 data base (CMEMS delayed time in situ T and S profiles data base). CLASS4 are computed in the following layers: [0m; 100m], [100m; 300m], [300m; 800m], [800m; 2000m] using reanalysis monthly fields.

T/S-STDEV Global or regional average profile of standard deviation between the four members of the GREP-V1 ensemble

None

UV-CLASS1-15m_MEAN

Zonal and meridional 15m depth current climatology. CMEMS drifter-derived climatology for total current, including Ekman component

UV-CLASS2-MKE_1000m Mean value of kinetic energy (KE=0,5x(U

2

+V2

)) at 1000m depth estimated from monthly means.

Mean value of kinetic energy at 1000m 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.


For product




Issue : 1.1


IV VALIDATION RESULTS OF GREP-V1

Temperature IV.1

T-CLASS1-T3D-MEAN IV.1.1

As shown in Figure 1 the mean anomaly of temperature with respect to climatology is consistent with the observed global warming of the ocean on the whole water column for all individual member as for the ensemble mean. Most anomaly features are similar in between individual members. Strong differences occur mostly at the Mediterranean waters outflow, in the Indonesian throughflow. FOAM/GloSea tends to be warmer than the average, and C-GLORS colder than the average


For product




Issue : 1.1



For product




Issue : 1.1


Figure 1: Differences of temperature (in °C) between the ensemble mean, then the four members, and the World Ocean Atlas 2009 climatology (WOA09) at 0.5m, 97m, 300m, 773m and 1945m depth, respectively. For this comparison, model data have been averaged over the period 1993 – 2014.

T-CLASS1-SST-MEAN IV.1.2

The bias with respect to AVHRR (assimilated) time series is very small as shown in Figure 2. FOAM/GloSea is warmer than the ensemble mean, especially in the Indian basin, while C-GLORS is colder than the ensemble mean.


For product




Issue : 1.1


Figure 2: Mean 1993-2014 difference in sea surface temperature (°C) between the ensemble mean then the 4 members, and the AVHRR SST data.

T-CLASS1-SST-TREND IV.1.3

As seen in Figure 3, the SST trend features of the ensemble mean closely follow that of the AVHRR observations (assimilated) except in the Eastern Pacific, where the ensemble mean diagnoses a cold trend which is not observed by AVHRR, where a warm trend is diagnosed in the North Western Tropical Pacific. This latter signal is week in all individual members, especially in FOAM. Figure 5 shows that the global average SST of all members and ensemble mean is very close to the observations.

AVHRR ENSEMBLE ORAS5


For product




Issue : 1.1


FOAM/GLoSea GLORYS2V4 CGLORS

Figure 3: Linear 1993-2015 SST trend of AVHRR data (left panel), then the ensemble mean and the four members. NB: trend 1993-2014 for FOAM/GloSea, 2015 not available at the moment.

T-CLASS2-MOORINGS IV.1.4

The ensemble mean and individual members are close to (assimilated) TAO, PIRATA and RAMA profiles on average on the 1993-2014 period, as seen on Figure 4.

TAO/TRITON 0N-147E TAO/TRITON 0N-165E TAO/TRITON 0N-140W TAO/TRITON 0N-110W

PIRATA 0N-23W PIRATA 0N-0E RAMA: 0N-90E

Figure 4: Mean equatorial temperature estimated over the period 1993 – 2014: observations (black), GLORYS2V4 (green), free-GLORYS2V4 (red), ORAS5 (pink), CGLORS (cyan), FOAM (blue). Moorings observations are coming from


For product




Issue : 1.1


TAO/TRITON: 0N-140W, 0N-110W,0N-165E,0N-147E, PIRATA:0N-23W,0N-0E and RAMA: 0N-90E.

T-CLASS3-SST_2DMEAN IV.1.5

Figure 5: Global spatial average of SST (°C) over the period 1993 – 2015: observations (black), ensemble mean (red), ORAS5 (pink), CGLORS (cyan), FOAM (blue), GLORYS2V4 (green).

T-CLASS3-HC_LAYER IV.1.6

The 0-700 heat content increase (Figure 6) is well captured by the ensemble mean and individual members. C-GLORS is colder than the other members, which solutions converge after 2005 and the setting of the ARGO observing network. In the 700-2000 m layer, individual members are somewhat correlated at low frequency and all show an increasing trend, except C-GLORS. FOAM/GloSea variability is noticeably stronger than that of the other reanalyses. Under 2000 m the reanalyses estimates disagree strongly, highlighting large uncertainties.


For product




Issue : 1.1



For product




Issue : 1.1


Figure 6: Time evolution of global average of heat content (in Joules) in different layers: [0-700m], [700-2000m], [2000m-bottom] and [0-bottom], for the ensemble mean (black), GLORYS2V4 (red), ORAS5 (blue), CGLORS (cyan), FOAM (green).

T-CLASS3-IC_CHANGE IV.1.7

The evolution of global average temperature after January 1993, beginning of the GREP-V1 time series, is diagnosed on Figure 7. Note that the individual members’ initial conditions are different (date and values). Consistently with previous diagnostics, the warming diagnosed between the surface and 2000m is stronger in GLORYS and weaker in C-GLORS compared to the average. Strong periodic cold events are diagnosed in FOAM during the first decade.


For product




Issue : 1.1


Figure 7: Time evolution of spatially averaged temperature changes with respect to January 1993 over the period 1993 – 2014.

T-CLASS4-LAYER IV.1.8

The cold bias of C-GLORS is confirmed by the comparison to temperature profiles of Figure 8. The RMS differences show that ORAS5 is less accurate than the other products under 300m.Some RMS error peaks can be noticed for C-GLORS in the 800-2000m layer.


For product




Issue : 1.1



For product




Issue : 1.1


Figure 8: time evolution of mean (left) and rms (right) error with respect to observations from the Coriolis database. RMS and mean differences are computed in the observations’ space, between monthly ocean reanalyses estimates and daily observations for temperature in °C

Complementary regional time series are shown in Figure 9, with the green envelope showing the spread of the RMS errors of individual members. In most cases, and all along the time series, the ensemble mean GREP-V1 reaches the accuracy of the most accurate individual member. The regions which display lower performances are the Mediterranean Sea, the North Atlantic and the Arctic.


For product




Issue : 1.1



For product




Issue : 1.1


Figure 9: RMS differences as in Figure 8 but on the 0-2000m layer, and for major ocean basins. The black line here shows the RMS of the ensemble mean product GREP-V1, the green shading indicates the spread of the RMS of the four members of GREP-V1, and the red line shows the RMS obtained using the WOA 13 monthly climatology for the 2005-2012 decade. The grey shading indicates the number of individual observations used to compute the statistics.

T-STDEV IV.1.9

The profiles of basin average standard deviation of salinity (Figure 10) confirm that the strongest disagreement between the members is located in the Mediterranean Sea, and during the first decade for all regions. Propagating deep signals may bring disagreement in the whole Atlantic basin. Note that the 2000-2005 is also a transition phase for the observation network, in between WOCE and ARGO.

The standard deviation displays a seasonal signal, with more discrepancies between systems during summer in the surface layer (stratification problems)


For product




Issue : 1.1


GLO

NAT

NPA

IND

MED

SAT

TPA

SPA


For product




Issue : 1.1


TAT

ACC

ARC

Figure 10: Time evolution of the profile of standard deviation between the four members, global average (left), North Atlantic (right), in °C.

Salinity IV.2

S-CLASS1-S3D_MEAN IV.2.1

At the surface, C-GLORS is fresher than the climatology and than the average of the other members (Figure 11), while ORAS5 is saltier on average. GLORYS is saltier in the Beaufort Sea, and displays a fresh pattern in the Eastern Tropical Atlantic. In subsurface, the individual members’ anomaly patterns differ in many places. ORAS5 is saltier than other products on average and a strong salty anomaly is diagnosed at depth in the Mediterranean Sea and outflow. Differences in between members are strong for instance in the Indonesian throughflow, with a strong signal appearing in FOAM/GloSea.


For product




Issue : 1.1



For product




Issue : 1.1



For product




Issue : 1.1


Figure 11: Differences of salinity (PSU) between the ensemble mean, then the four members, and the World Ocean Atlas 2009 climatology (WOA09) at 0.5m, 97m, 300m, 773m and 1945m depth, respectively. For this comparison, model data have been averaged over the period 1993 – 2014.

S-CLASS1-SSS-TREND IV.2.2

The regional surface salinity trends reach a meaningful agreement in the tropical Pacific and Indian Ocean, as seen in Figure 12. GLORYS displays a strong positive trend in the Tropical Pacific which is not seen by the other members. This signal is thus filtered out by the ensemble mean.

ENSEMBLE ORAS5 FOAM


For product




Issue : 1.1


GLORYS2V4 CGLORS

Figure 12: Linear 1993-2015 SSS trend of the ensemble mean and the four members. NB: trend 1993-2014 for FOAM, 2015 not available at the moment.

S-CLASS2-MOORINGS IV.2.3

The ensemble mean and individual members are close to (assimilated) TAO, PIRATA and RAMA profiles on average on the 1993-2014 period, as seen on Figure 13.

TAO/TRITON 0N-147E TAO/TRITON 0N-165E TAO/TRITON 0N-140W TAO/TRITON 0N-110W

PIRATA 0N-23W PIRATA 0N-0E RAMA: 0N-90E

Figure 13: Mean equatorial salinity estimated over the period 1993 – 2014: observations (black), GLORYS2V4 (green), free-GLORYS2V4 (red), ORAS5 (pink), CGLORS (cyan), FOAM (blue). Moorings observations are coming from


For product




Issue : 1.1


TAO/TRITON: 0N-140W, 0N-110W,0N-165E,0N-147E, PIRATA:0N-23W,0N-0E and RAMA: 0N-90E.

S-CLASS3-SC_LAYER IV.2.4

As can be seen in Figure 14, the individual members salt content estimates only start to converge after the setting of the global ARGO observation network in 2005. C-GLORS stays fresher than the other members through the whole period. No significant trend can be diagnosed from these results.


For product




Issue : 1.1


Figure 14: Time evolution of global average of salt content (PSU) in different layers: [0-700m], [700-2000m], [2000m-bottom] and [0-bottom], for the ensemble mean (black), GLORYS2V4 (red), ORAS5 (blue), CGLORS (cyan), FOAM (green).

S-CLASS3-IC_CHANGE IV.2.5

With respect to January 1993 (start of the GREP-V1 time series) , the ensemble mean gets saltier at the surface in the 0-500m layer, while it gets slightly fresher from 500 to 2000m (Figure 14). A strong freshening is diagnosed in ORAS5 and in C-GLORS between 500 and 1000 m. During the first decade, FOAM and GLORYS display opposite evolutions with respect to January 1993.


For product




Issue : 1.1


Figure 15: Time evolution of spatially averaged salinity changes with respect to January 1993 over the period 1993 – 2014.

S-CLASS4-LAYER IV.2.6

The salty bias of ORAS5 is confirmed by the comparison to temperature profiles of Figure 8. The RMS differences show that ORAS5 is less accurate than the other products under 300m.Some RMS error peaks can be noticed for C-GLORS in the 800-2000m layer.


For product




Issue : 1.1



For product




Issue : 1.1


Figure 16: time evolution of mean (left) and rms (right) error with respect to observations from the Coriolis database. . RMS and mean differences are computed in the observations’ space, between monthly ocean reanalyses estimates and daily observations for salinity in psu

Complementary regional time series are shown in Figure 17, with the green envelope showing the spread of the RMS errors of individual members. In most cases, and all along the time series, the ensemble mean GREP-V1 reaches the accuracy of the most accurate individual member. As for temperature, the regions which display lower performances are the Mediterranean Sea, the North Atlantic and the Arctic. The setting of the ARGO network considerably reduces the spread and improves the skill of salinity estimates after 2002.


For product




Issue : 1.1



For product




Issue : 1.1


Figure 17: RMS differences as in Figure 16 but on the 0-2000m layer, and for major ocean basins. The black line here shows the RMS of the ensemble mean product GREP-V1, the green shading indicates the spread of the RMS of the four members of GREP-V1, and the red line shows the RMS obtained using the WOA 13 monthly climatology for the 2005-2012 decade. The grey shading indicates the number of individual observations used to compute the statistics.

S-STDEV IV.2.7

The profiles of basin average standard deviation of salinity (Figure 18) confirm that the disagreement between the members is strongest in the Mediterranean Sea, and during the first decade for all regions. Propagating deep signals may bring disagreement in the whole Atlantic basin. Note that the 2000-2005 is also a transition phase for the observation network, in between WOCE and ARGO.


For product




Issue : 1.1


GLO

NAT

NPA

IND

MED

SAT

TPA

SPA


For product




Issue : 1.1


TAT

ACC

ARC

Figure 18: Time evolution of the profile of standard deviation between the four members, global average (left), North Atlantic (right), in PSU.

Currents IV.3

UV-CLASS1-15m_MEAN IV.3.1

As can be seen in Figure 19, the mean zonal current of all members and ensemble mean is in very good agreement with the drifters climatology from CMEMS (including corrections of biases due to drogue loss). GLORYS and FOAM both display a spurious strong westward zonal current in the Papua New Guinea area, which affects the ensemble mean.


For product




Issue : 1.1


Figure 19: Zonal 15m depth current climatology. CMEMS TAC IN SITU drifter-derived climatology (top left), then the four members and the ensemble mean. Data have been averaged over the period 1993-2014.

Figure 20: Meridional 15m depth current climatology. NOAA AOML drifter-derived climatology (top left), then the ensemble mean and the four members. Data have been averaged over the period 1993-2014.


For product




Issue : 1.1


UV-CLASS2-MKE_1000m IV.3.2

Figure 21 shows that deep currents energy is overestimated in FOAM, which affects the ensemble mean, especially in the western Tropical Pacific.

Argo drifters

ENSEMBLE MEAN

CGLORS

FOAM

ORAS5

GLORYS2V4

Figure 21: Mean value of the kinetic energy at 1000m from the ANDRO Argo drift data base (top left) then the ensemble mean and the four members. Data averaged over 1993-2014.


For product




Issue : 1.1


V REFERENCES

Blockley, E. W., et al.(2014) Recent development of the Met Office operational ocean forecasting system: an overview and assessment of the new Global FOAM forecasts. Geoscientific Model Development, 7 (6), 2613–2638, doi:10.5194/gmd-7-2613-2014

N. Ferry, L. Bertino, K. Haines, G. Larnicol, A. Lars, S. Masina, S. Salon, S. Simoncelli, C. Solidoro, A. Storto, M. Valdivieso, H. Zuo + B. Levier + all WP18 partners. Calibration / Validation definition for ocean reanalyses CMEMS internal report ref. MYO2-WP18-CALVALdef-v1.7 available on request by writing to [email protected]

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., Drillet, Y., Daudin, A., and De Nicola, C. (2013) Evaluation of global monitoring and forecasting systems at Mercator Océan, Ocean Sci., 9, 57-81, doi:10.5194/os-9-57-2013

MacLachlan, C., A. Arribas, K.A. Peterson et al. (2014) Global Sesonal forecast system version 5 (GloSea5): a high-resolution seasonal forecast system. Quart. J. Royal Meteorol. Soc, 14, 1072-1084.

Storto, A., Masina, S. and Navarra, A. (2016), Evaluation of the CMCC eddy-permitting global ocean physical reanalysis system (C-GLORS, 1982–2012) and its assimilation components. Q.J.R. Meteorol. Soc., 142: 738–758. doi:10.1002/qj.2673

Zuo, H., M.A. Balmaseda, K. Mogensen, (2015) The new eddy-permitting ORAP5 ocean reanalysis: description,evaluation and uncertainties in climate signals. Clim. Dyn. 10.1007/s00382-015-2675-1

Zuo, H., M.A. Balmaseda, Boisseson, E., Hirahara, S., (2017). A new ensemble generation scheme for ocean reanalysis. ECMWF Tech Memo (795), to be submitted.

Documents

QUALITY INFORMATION DOCUMENT For Global Ocean …€¦ · QUALITY INFORMATION DOCUMENT For Global Ocean Reanalysis Multi-model Ensemble Products GREP-V1 GLOBAL-REANALYSIS-PHY-001-026