QUality Information Document for Med biogeochemistry...

COPERNICUS MARINE ENVIRONMENT MONITORING SERVICE

MED MFC

QUality Information Document for Med biogeochemistry analysis and forecast

product: MEDSEA_ANALYSIS_FORECAST_BIO_006_006

Reference: CMEMS-Med-QUID-006-006-V2

Validated by: Mercator Ocean Document release number: 1.4 Date: 11/09/2017

Contributors : G. Cossarini, S. Salon, G. Bolzon, P. Lazzari, E. Clementi

MED MFC QUality Information Document for Med biogeochemistry analysis and forecast product

Ref : CMEMS-Med-QUID-006-006-V2

Date : 16-05-2017

Issue : 1.4

COPERNICUS MARINE ENVIRONMENT MONITORING SERVICE PAGE 2/ 36

CHANGE RECORD

Issue Date § Description of Change Author

1.0 26/01/2016 all First version of document at CMEMS V2 G. Cossarini

1.1 04/04/2016 all Upgraded after V2 acceptance review G. Cossarini

1.2 17/10/2016 IV Update after event with impact on data availability

G. Cossarini

1.3 22/11/2016 IV.2 Update after change in Med-Currents of GLO-LOBC

S. Salon

1.4 16/05/2017 IV.3 Update after change in NRT/DT surface chlorophyll data from OC TAC

G. Cossarini



Date : 16-05-2017

Issue : 1.4


TABLE OF CONTENTS

I EXECUTIVE SUMMARY............................................................................................................................... 9

I.1 Products covered by this document ........................................................................................................... 9

I.2 Summary of the results ............................................................................................................................... 9

I.3 Estimated Accuracy Numbers .................................................................................................................. 11

II PRODUCTION SUB-SYSTEM DESCRIPTION ....................................................................................... 13

II.1 Production centre details ........................................................................................................................ 13

II.2 Description of the 3DVAR-OGSTM-BFM model system .................................................................... 14

II.3 Description of Data Assimilation scheme .............................................................................................. 15

II.4 Upstream data and boundary condition of the 3DVAR-OGSTM-BFM model ................................. 16

III VALIDATION FRAMEWORK .................................................................................................................. 18

IV VALIDATION RESULTS........................................................................................................................... 23

V Recent events having an impact on data availability and/or quality ........................................................... 35

VI Quality changes since previous version ...................................................................................................... 36



Date : 16-05-2017

Issue : 1.4


LIST OF TABLES

TABLE I.1. MEAN AND RMS OF THE DIFFERENCE BETWEEN MODEL AND SEAWIFS SATELLITE REFERENCE DATA (COMPUTED ON NATURAL AND LOG10 VALUES OF SURFACE CHLOROPHYLL FIELD ON GRID POINTS WITH DEPTH LARGER THAN 200M). WINTER (LEFT) REFERS TO NOVEMBER TO APRIL, SUMMER (RIGHT) REFERS TO MAY TO OCTOBER. 11

TABLE I.2. MEAN RMSD AND CORRELATION OF PHOSPHATE, NITRATE AND OXYGEN ESTIMATED BY COMPARING V2 REANALYSIS AND IN-SITU 1999-2014 OBSERVATIONS. 12

TABLE I.3. UNCERTAINTY OF PH (TOTAL SCALE) AND PCO2 FOR THE SURFACE LAYER, ESTIMATED INDIRECTLY BY ERROR PROPAGATION APPROACH USING UNCERTAINTY OF DIC AND ALKALINITY COMPUTED ON THE PRE-OPERATIONAL QUALIFICATION RUN 2014-2015. 12

TABLE III.1. LIST OF METRICS. 19 TABLE III.2. LIST OF DATASETS USED TO BUILD CLIMATOLOGY OF THE CARBONATE SYSTEM

VARIABLES. TRANSMED: SCIENTIFIC CRUISE THAT COVERED THE MEDITERRANEAN SEA FROM THE WESTERN SUB-BASIN TO THE EASTERN SUB-BASIN; ANC.VARS: ANCILLARY VARIABLES (T, S, OXYGEN AND NUTRIENTS). 21

TABLE IV.1. ANNUAL AVERAGED VERTICALLY INTEGRATED PRIMARY PRODUCTION (GC M−2 YR−1) FOR SOME SELECTED SUB-REGIONS. FROM MULTI-ANNUAL SIMULATION (LAZZARI ET AL., 2012), FROM SATELLITE MODEL (COLELLA, 2006), FROM CMEMS V2 SYSTEM (PRE-OPERATIONAL QUALIFICATION RUN) 24

TABLE IV.2. MEAN CORRELATION AND BIAS BETWEEN SURFACE CHLOROPHYLL MODEL MAPS AND SATELLITE MAPS. ON THE RIGHT, THE SKILL INDEXES ARE COMPUTED ON THE MODEL AND SATELLITE CHLOROPHYLL LOG-TRANSFORMED. “WIN” REFERS TO THE JANUARY-JUNE PERIOD, WHILE “SUM” REFERS TO THE JULY-DECEMBER PERIOD. 25

TABLE IV.3. AVERAGED CORRELATION BETWEEN MODEL AND BIO-ARGO CHLOROPHYLL PROFILES. AVERAGES ARE COMPUTED FOR MATCHED PROFILES CONSIDERING EACH SUB-BASIN AND A MONTHLY TEMPORAL RESOLUTION. 27

TABLE IV.4. AVERAGED RMS BETWEEN MODEL AND BIO-ARGO CHLOROPHYLL PROFILES. AVERAGES ARE COMPUTED FOR MATCHED PROFILES CONSIDERING EACH SUB-BASIN AND A MONTHLY TEMPORAL RESOLUTION. 28

TABLE IV.5. SKILL METRICS FOR THE COMPARISON OF ALKALINITY AND DIC ON THE 1X1 GRID DOMAIN AND SELECTED LAYERS. 31

TABLE IV.6. SKILL METRICS OF THE COMPARISON OF ALKALINITY AND DIC ON THE 18 SELECTED AREAS 4°X4° WIDE. 32

LIST OF FIGURES

FIGURE II.1. SCHEME OF THE FUNCTIONING OF THE MED-MFC-BIOGEOCHEMISTRY SYSTEM FOR ANALYSIS AND FORECAST: GREY BOXES REPRESENT THE DAYS OF ANALYSIS, ORANGE BOXES REPRESENT THE 6 (7) DAYS OF HINDCAST FOR WEDNESDAY (SATURDAY) PRODUCTION DAY, YELLOW BOXES REPRESENT THE 10 DAYS OF FORECAST. THE PRODUCTION WEEK DAYS WHEN THE SYSTEM IS RUN AT THE CINECA (WEDNESDAYS OR SATURDAYS) FACILITY ARE REPORTED ON THE LEFT COLUMN. .................................................................................................................................................................... 13

FIGURE II.2: THE MED-BIOGEOCHEMISTRY MODEL SYSTEM AND INTERFACES WITH OTHER COMPONENTS OF COPERNICUS SYSTEM. ........................................................................................................................ 14

FIGURE III.1. LOCATION OF MOORINGS WITH AVAILABLE BIOGEOCHEMICAL DATA FROM CMEMS PRODUCT INSITU_MED_NRT_OBSERVATIONS_013_035 (ONLY PYLOS AND E1M3A HAVE DATA FOR A RELIABLE MODEL-OBSERVATION COMPARISON). ................................................................................................. 20

FIGURE III.2. LOCATION OF THE CARBONATE SYSTEM VARIABLES DATA, AND IDENTIFICATION OF 18 AREAS 4°X4° WIDE USED FOR THE VALIDATION ACTIVITIES. ............................................................................. 21

FIGURE III.3. SUBDVISION OF THE MODEL DOMAIN IN SUB-BASIN USED FOR MODEL VALIDATION. ....... 22



Date : 16-05-2017

Issue : 1.4


FIGURE IV.1. PHOSPHATE SPATIAL DISTRIBUTIONS ANNUAL AVERAGE AND VERTICALLY AVERAGED OVER THE 0-50 M LAYER (MMOL P.M-3), FROM CMEMS V2 SYSTEM (PRE-OPERATIONAL QUALIFICATION RUN, LEFT) AND FROM A VALIDATED MULTI-ANNUAL SIMULATION (FROM FIG. 4 OF LAZZARI ET AL., 2016, RIGHT). ...................................................................................................................................... 23

FIGURE IV.2. NITRATE SPATIAL DISTRIBUTIONS ANNUAL AVERAGE AND VERTICALLY AVERAGED OVER THE 0-50 M LAYER (MMOL N.M-3), FROM CMEMS V2 SYSTEM (PRE-OPERATIONAL QUALIFICATION RUN, LEFT) AND FROM FROM A VALIDATED MULTI-ANNUAL SIMULATION (FROM FIG. 4 OF LAZZARI ET AL., 2016, RIGHT). ............................................................................................................................................. 23

FIGURE IV.3. ANNUAL AVERAGED VERTICALLY INTEGRATED PRIMARY PRODUCTION (GC M−2 YR−1) CMEMS V2 SYSTEM (PRE-OPERATIONAL QUALIFICATION RUN, LEFT) AND FROM MULTI-ANNUAL SIMULATION (LAZZARI ET AL., 2012, RIGHT) ............................................................................................................... 24

FIGURE IV.4. TREND OF BIAS AND RMS BETWEEN MODEL AND SATELLITE DATA FOR DIFFERENT SUB-BASINS. ............................................................................................................................................................................................. 25

FIGURE IV.5. CHLOROPHYLL PROFILES FROM MODEL (BLACK) AND BIO-ARGO FLOAT (BLUE) FOR THE NORTH WESTERN MEDITERRANEAN SUB-BASIN. MATCHED MODEL AND BIO-ARGO PROFILES ARE GROUPED BY MONTHS ............................................................................................................................................................. 26

FIGURE IV.6. CHLOROPHYLL PROFILES FROM MODEL (BLACK) AND BIO-ARGO FLOAT (BLUE) FOR THE EASTERN PART OF THE SOUTH-WESTERN MEDITERRANEAN SUB-BASIN. MATCHED MODEL AND BIO-ARGO PROFILES ARE GROUPED BY MONTHS ...................................................................................................... 26

FIGURE IV.7. CHLOROPHYLL PROFILES FROM MODEL (BLACK) AND BIO-ARGO FLOAT (BLUE) FOR THE IONIAN SUB-BASIN. MATCHED MODEL AND BIO-ARGO PROFILES ARE GROUPED BY MONTHS. ........ 27

FIGURE IV.8. CHLOROPHYLL PROFILES FROM MODEL (BLACK) AND BIO-ARGO FLOAT (BLUE) FOR THE LEVANTINE SUB-BASIN. MATCHED MODEL AND BIO-ARGO PROFILES ARE GROUPED BY MONTHS. .............................................................................................................................................................................................................. 27

FIGURE IV.9. CHLOROPHYLL TIMESERIES AT DIFFERENT DEPTHS AT THE E1M3A (LEFT) AND PYLOS (RIGTH) IN-SITU MOORINGS. MODEL (RED) AND IN-SITU OBSERVATION (BLUE) ARE REPORTED. 28

FIGURE IV.10. OXYGEN TIMESERIES AT SURFACE LAYER OF THE E1M3A IN-SITU MOORINGS. MODEL (RED) AND IN-SITU OBSERVATION (BLUE) ARE REPORTED. ................................................................................ 29

FIGURE IV.11. MEAN ANNUAL MAP OF ALKALINITY (LEFT) AND DIC (RIGHT) SIMULATED BY THE OGSTM-BFM MODEL (UPPER PANELS) AND RECONSTRUCTED BY THE 1°X1° CLIMATOLOGY (LOWER PANELS) FOR THE LAYER 0-50M. .................................................................................................................... 30

FIGURE IV.12. MEAN ANNUAL MAP OF ALKALINITY (LEFT) AND DIC (RIGHT) SIMULATED BY THE OGSTM-BFM MODEL (UPPER PANELS) AND RECONSTRUCTED BY THE 1°X1° CLIMATOLOGY (LOWER PANELS) FOR THE LAYER 200-500M. ............................................................................................................ 30

FIGURE IV.13. ALKALINITY PROFILES: MEAN MONTHLY MODEL PROFILES (COLORED LINES) AND CLIMATOLOGICAL (±STANDARD DEVIATION) PROFILES (BLACK LINES) FOR THE AREAS 2, 6, 15. . 31

FIGURE IV.14. DIC PROFILES: MEAN MONTHLY MODEL PROFILES (COLORED LINES) AND CLIMATOLOGICAL (±STANDARD DEVIATION) PROFILES (BLACK LINES) FOR THE AREAS 1, 5, 18. . 32

FIGURE IV.15. MEAN MONTHLY VALUES OF PH AT IN SITU CONDITION AND REPORTED IN SEA WATER SCALE (SWS) AT THE 0-50M LAYER. .................................................................................................................................. 33

FIGURE IV.16. MEAN MONTHLY VALUES OF PCO2 [PPM] FOR THE 0-50M LAYER. .............................................. 33



Date : 16-05-2017

Issue : 1.4


GLOSSARY AND ABBREVIATIONS

Additional terms:

ALK Alkalinity

ASSIM Data Assimilation

CAL/VAL Pre operational qualification/VALidation

CHL Chloroophyll

CMCC Centro EuroMediterraneo per i Cambiamenti Climatici

CORR Correlation

DA Data Assimilation

DIC Dissolved Inorganic Carbon

ECMWF European Centre for Medium range Weather Forecasting

GLO Global

GODAE Global Ocean Data Assimilation Experiment

HCMR Hellenic Centre for Marine Research

INGV Istituto Nazionale di Geofisica e Vulcanologia

INS IN-Situ

Med Mediterranean

MERSEA Marine Environment and Security for the European Area

MFC Monitoring and Forecasting System

MonGOOS Mediterranean Operational Network for the Global Ocean Observing System

NRT Near Real Time

OBS Observation

OGS Istituto Nazionale di Oceanografie e Geofisica Sperimentale

PQ Product Quality

PQWG Product Quality Working Group

RMS Root Mean Square

TAC Thematic Assembly Center



Date : 16-05-2017

Issue : 1.4


Applicable and Reference Documents

Ref Title Date / Version

DA 1 CMEMS-PQ-MGT CMEMS Product Quality Management Plan Not detailed yet

RA 1 Lazzari et al., 2010 Lazzari P., Teruzzi A., Salon S., Campagna S., Calonaci C., Colella S., Tonani M., Crise A. 2010. Pre-operational short-term forecasts for the Mediterranean Sea biogeochemistry. Ocean Science, 6, 25-39.

RA 2 Lazzari et al., 2012; Lazzari, P., Solidoro, C., Ibello, V., Salon, S., Teruzzi, A., Béranger, K., Colella, S., and Crise, A., 2012. Seasonal and inter-annual variability of plankton chlorophyll and primary production in the Mediterranean Sea: a modelling approach. Biogeosciences, 9, 217-233.

RA 3 Teruzzi et al., 2013; Teruzzi A., Dobricic S., Solidoro C., Cossarini G. 2013. A 3D variational assimilation scheme in coupled transport biogeochemical models: Forecast of Mediterranean biogeochemical properties, Journal of Geophysical Research, doi:10.1002/2013JC009277.

RA 4 Cossarini et al., 2015 Cossarini G., Lazzari P., Solidoro, C., 2015. Spatiotemporal variability of alkalinity in the Mediterranean Sea. Biogeosciences, 12(6), 1647-1658.

RA 5 Foujols et al., 2000 Foujols, M.-A., Lévy, M., Aumont, O., Madec, G., 2000. OPA

8.1 Tracer Model Reference Manual. Institut Pierre Simon Laplace, pp. 39.

RA 6 Thingstad and Rassoulzadegan, 1995

Thingstad T.F., Rassoulzadegan F., 1995. Nutrient limitations, microbial food webs, and 'biological C-pumps': suggested interactions in a P-limited Mediterranean. Marine Ecology Progress Series, 117: 299-306.

RA 7 Krom et al., 1991; Krom M.D., Kress N., Brenner S., Gordon L.I., 1991. Phosphorus limitation of primary productivity in the eastern Mediterranean Sea. Limnology and Oceanography, 36(3) 424-432.

RA 8 Bethoux et al., 1998 Bethoux, J. P., Morin, P., Chaumery, C., Connan, O., Gentili, B., and Ruiz-Pino, D., 1998. Nutrients in the Mediterranean Sea, mass balance and statistical analysis of concentrations with respect to environmental change, Mar. Chem., 63, 155–169.

RA 9 Orr et al., 1999 Orr J.C., Najjar R., Sabine C.L., Joos F., 1999. Abiotic-HOWTO. Internal OCMIP Report, LSCE/CEA Saclay, Gifsur-Yvette, France, 25 pp.

RA 10 Dobricic and Pinardi, 2008

Dobricic S., Pinardi N., 2008. An oceanographic three-dimensional variational data assimilation scheme. Ocean Modelling, 22, 3-4, 89-105.

RA 11 Zeebe and Wolf-Gladrow, 2001

Zeebe, R. E. andWolf-Gladrow, D.: CO2 in Seawater: Equilibrium, Kinetics, Isotopes, Elsevier Oceanography Series, Elsevier, Amsterdam, the Netherlands, 2001

RA 12 Somot et al., 2008 Somot S., Sevault F., Déqué M., Crépon M., 2008. 21st century



Date : 16-05-2017

Issue : 1.4


climate change scenario for the Mediterranean using a coupled atmosphere–ocean regional climate model, Global and Planetary

Change, 63, 2–3: 112–126.

RA 13 Chopin-Montegut, 1993

Copin-Montegut C., 1993. Alkalinity and carbon budgets in the Mediterranean Sea. Global Biogeochemical Cycles, 7(4), pp. 915-925.

RA 14 Ludwig et al., 2009 Ludwig W., Dumont E., Meybeck M., Heussner S., 2009. River discharges of water and nutrients to the Mediterranean and Black Sea: Major drivers for ecosystem changes during past and future decades?. Prog. Oceanogr., 80 (3-4): 199-

217.

RA 15 Loye-Pilot et al., 1990; Loÿe-Pilot, M. D., J. M. Martin, and J. Morelli, 1990. Atmospheric input of inorganic nitrogen to the western

Mediterranean. Biogeochem., 9: 117-134.

RA 16 Guerzoni et al., 1999; Guerzoni, S., Chester, R., Dulac, F., Herut, B., Loÿe-Pilot, M.-D., Measures, C., Migon, C., Molinaroli, E., Moulin, C., Rossini, P., Saydam, C., Soudine, A., Ziveri, P., 1999. The role of atmospheric deposition in the biogeochemistry of the Mediterranean Sea. Prog. Oceanogr., 44 (1-3): 147-190.

RA 17 Herut and Krom, 1996; Herut, B. and Krom, M.: Atmospheric input of nutrients and dust to the SE Mediterranean, in: The Impact of Desert Dust

Across the Mediterranean, edited by: Guerzoni, S. and Chester, R., Kluwer Acad., Norwell, Mass., 349–358, 1996.

RA18 Cornell et al., 1995; Cornell S., Rendell A., Jickells T., 1995. Atmospheric inputs of dissolved organic Nitrogen to the oceans, Nature, 376, 243–246.

RA 19 Bergametti et al., 1992 Bergametti G., Remoudaki E., Losno R., Steiner E., Chatenet B., 1992. Source, transport and deposition of atmospheric Phosphorus over the northwestern Mediterranean, J. Atmos. Chem., 14, 501–513.

RA 20 Ribera d’Alcalà et al., 2003

Ribera d'Alcalà M., Civitarese G., Conversano F., Lavezza R., 2003. Nutrient ratios and fluxes hint at overlooked processes in the Mediterranean Sea. Journal of Geophysical Research, 108(C9), 8106, doi:10.1029/2002JC001650.

RA 21 Lazzari et al., 2016 Lazzari et al., 2015. Seasonal variability of phosphate and nitrate in the Mediterranean Sea: a modelling approach. Deep Sea Research, Part I: Oceanographic Research Papers, Volume 108, February 2016, Pages 39-52, ISSN 0967-0637



Date : 16-05-2017

Issue : 1.4


I EXECUTIVE SUMMARY

I.1 Products covered by this document

This document describes the quality of the product MEDSEA_ANALYSIS_FORECAST_BIO_006_006, the nominal product for the analysis and forecast of the biogeochemical state of the Mediterranean Sea. The MED Biogeochemistry product includes 3D daily fields at 1/16° horizontal resolution of 8 variables grouped in 4 datasets:

NUTR: phosphate and nitrate;

PFTC: chlorophyll and phytoplankton biomass;

BIOL: dissolved oxygen concentrations and net primary production;

CARB: ocean pH and ocean pCO2.

I.2 Summary of the results

The present document describes the quality of the MED biogeochemical analysis and forecast system at the horizontal resolution of 1/16°, which roughly corresponds from 5 km (at 45°N) to 6 km (at 30°N) for the Mediterranean basin.

V2 version of NUTR, PFTC, BIOL datasets has not changed from the previous version V1, therefore the quality of V2 products remains unchanged, however, new validation metrics have been implemented. CARB dataset is a new feature of the CMEMS-MED-MFC-Biogeochemistry system.

The main results of the MEDSEA_REANALYSIS_BIO_006_008 quality product assessment are summarized in the following points:

a) PFTC dataset

Chlorophyll: the surface chlorophyll is the only variable that can be subjected to a quantitative and in-depth skill assessment analysis. Results give evidence of the model capability of reproducing spatial patterns, seasonal cycle and interannual variability. Model chlorophyll error is positive both in winter (mean BIAS equals to 0.004 mg/m3) and in summer (BIAS equals to 0.006 mg/m3). Western sub-basins are characterized by higher uncertainty and variability than eastern ones. EANs of surface chlorophyll computed from the pre-qualification run are similar to those of the previous estimate. New data sources for validation and qualification have been used: chlorophyll data from Bio-Argo floats and near real time (NRT) in situ moorings. It is worth to report that the model reproduces shape and evolution of the vertical profiles observed in data. However, at the present stage, the skill assessment analysis based on in-situ and Bio-Argo data must be considered qualitatively and results preliminary, since availability of data is low and reliability of the metrics and statistics has to be tested and verified.

Phytoplankton carbon biomass: no validation metrics are feasible due to the lack of a reliable reference dataset.

b) NUTR dataset

At the present stage, phosphate and nitrate cannot be validated with NRT model-observation comparisons. The use of new datastream (Bio-Argo and in situ platforms) has been tested during the V2 implementation, however the lack of enough available data and the ongoing development of proper and published quality standard procedures of the new data prevent at present any NRT validation



Date : 16-05-2017

Issue : 1.4


approach. Therefore, the estimate of model product quality for these variables has been inferred by the consistency of the model results with published literature. The reader is, then, referred to the QUID of the reanalysis product (CMEMS-Med-QUID-006-008) that provides a consistent validation of the OGSTM-BFM model by presenting an in-depth skill performance analysis of the aforementioned biogeochemical variables for the simulated period 1999-2014. Skill performance analyses are based on a point-to-point comparison approach; therefore the results are a conservative estimate of the model predictive performance, and calculated metrics are used as EANs (Table I.2) for the analysis and forecast products.

c) BIOL dataset

Net primary production displays a good accordance with previous basin-wide estimates from literature, with pre-operational results within the range of variability of the in-situ estimates. Given the available dataset, a qualitative estimate of the uncertainty of this important ecosystem indicator is the only feasible and reliable assessment.

The NRT quantitative uncertainty estimates of Oxygen, as already reported for nutrients, is not available at the present stage. The reader is, then, referred to the QUID of the reanalysis product (CMEMS-Med-QUID-006-008) that provides a skill performance analyses based on a point-to-point comparison approach. Therefore the results are a conservative estimate of the model predictive performance for oxygen, and calculated metrics are used as EANs (Table I.2) for the analysis and forecast products.

d) CARB dataset

The carbonate system and the variables pH (total scale) and pCO2 are new features of the MED biogeochemistry system, and version V2 is the first attempt of product quality evaluation. The model system reproduces the basin scale variability of carbonate system and the mean vertical profiles of the key variables for the different sub-basins. Mean RMS values of DIC and Alkalinity are about 20

mol/kg at surface. Uncertainty of pH and pCO2 is indirectly estimated by an error propagation approach using the uncertainty of DIC and Alkalinity.



Date : 16-05-2017

Issue : 1.4


I.3 Estimated Accuracy Numbers

The Estimated Accuracy Numbers (EANs) of surface chlorophyll were computed using a pre-operational run (period April 2014 - June 2015) produced by the V2 version of 3DVAR-OPATM-BFM model system.

The EANs are the mean and RMS of the difference between model values and MODIS satellite reference data. Statistics are computed on both natural and on log transformed data for grid points with depth higher than 200m. EANs are computed for several sub-regions (Figure III.3) and two seasons: winter (from November to April) and summer (from May to October).

Surface (0-10m) chlorophyll Mod-Sat

Surface (0-10m) chlorophyll log10(Mod)-log10(Sat)

RMS BIAS

RMS BIAS

win Sum win sum win sum win sum

alb 0.157 0.171 0.031 0.029 0.511 0.650 0.109 0.111

sww 0.066 0.021 0.014 0.004 0.147 0.099 0.038 0.032

swe 0.067 0.018 0.004 0.004 0.160 0.095 0.006 0.027

nwm 0.093 0.023 -0.012 0.002 0.190 0.102 -0.029 0.017

tyr 0.052 0.013 -0.006 0.001 0.141 0.081 -0.011 0.015

adn 0.039 0.024 0.004 0.010 0.142 0.135 0.024 0.067

ads 0.039 0.025 -0.007 0.010 0.141 0.100 -0.027 0.044

aeg 0.107 0.020 0.021 0.002 0.140 0.091 0.024 0.007

ion 0.031 0.015 -0.003 -0.001 0.107 0.089 -0.022 -0.014

lev 0.023 0.013 -0.003 -0.003 0.104 0.095 -0.028 -0.033

Table I.1. Mean and RMS of the difference between model and SeaWiFS satellite reference data (computed on natural and log10 values of surface chlorophyll field on grid points with depth larger than 200m). Winter (left) refers to November to April, summer (right) refers to May to October.

The RMS values are lower during summer for all the sub-regions, indicating that model forecasts reproduce fairly well the low chlorophyll values and the variability observed in satellite data. Higher values of RMS during winter indicate lower accuracy of the winter model forecasts. During this period, the surface chlorophyll values are characterized by patchy bloom events and high spatial variability, which are not completely resolved by the model forecast. Moreover, since the blooms are strongly related to the local mixing conditions of the water column, a possible further source of error for the model forecast arises from the uncertainty of the physical forcing which can erroneously drive the vertical mixing conditions that fertilize the surface layer.

It is important to remind that the EANs of chlorophyll data reported in Table I.1 are computed and given on both natural and on log transformed data. Log-transformed metrics have some issues to consider: the statistics are not intuitively readable and emphasize the differences at small concentrations between model results and observations.

The Estimated Accuracy Numbers (EANs) of phosphate, nitrate and oxygen were computed using the reanalysis run produced by the V2 version of 3DVAR-OPATM-BFM model system. The EANs are the RMS and the correlation computed on a point-to-point model-observation approach over selected layers and sub-basins. The overall average of the sub-basins is reported.



Date : 16-05-2017

Issue : 1.4


RMS Corr

Variables 0-10 10-50 50-100 100-150

150-300

300-600

600-1000

Vertical

Phosphate [mmol/m3]

0.08 0.08 0.10 0.11 0.14 0.13 0.12 0.63

Oxygen [mmol/m3] 37.7 50.1 40.5 27.8 18.4 17.1 14.1 0.53

Nitrate [mmol/m3] 1.03 1.47 1.99 2.34 2.57 2.38 1.87 0.69

Table I.2. Mean RMSD and correlation of phosphate, nitrate and oxygen estimated by comparing V2 reanalysis and in-situ 1999-2014 observations

1.

The Estimated Accuracy Numbers (EANs) of pH and pCO2 were computed using a pre-operational run produced by the V2 version of 3DVAR-OPATM-BFM model system. The EANs of pH and pCO2 is indirectly estimated by an error propagation approach using the uncertainty of DIC and Alkalinity.

Variables Uncertainty

pH 0.050

pCO2 50.0 atm

Table I.3. Uncertainty of pH (total scale) and pCO2 for the surface layer, estimated indirectly by error propagation approach using uncertainty of DIC and Alkalinity computed on the pre-operational qualification run 2014-2015.

1 This table refers to Table I.2 of CMEMS-Med-QUID-006-008-V2.V1.1 dcoument.



Date : 16-05-2017

Issue : 1.4


II PRODUCTION SUB-SYSTEM DESCRIPTION

II.1 Production centre details

a) Production centre name: Med-MFC

b) Production subsystem name: Med-MFC-biogeochemistry

c) Production Unit: OGS

d) Production centre description for the version covered by this document

The biogeochemical analysis and forecasts for the Mediterranean Sea are produced by the OGS Production Unit by means of the 3DVAR-OGSTM-BFM model system (described below), using as physical forcing the output of the Med-MFC-currents products managed by the INGV Production Unit. The forecasting system is automatically run at the CINECA supercomputing centre (Bologna, Italy). Ten days of forecast are produced twice per week on Wednesday and on Saturday, and data assimilation of surface chlorophyll satellite observations (provided by the CMEMS-OCTAC managed by GOS-ISAC-CNR) is performed once a week in the Wednesday run (Fig. II.1).

Figure II.1. Scheme of the functioning of the Med-MFC-biogeochemistry system for analysis and forecast: grey boxes represent the days of analysis, orange boxes represent the 6 (7) days of hindcast for Wednesday (Saturday) production day, yellow boxes represent the 10 days of forecast. The production week days when the system is run at the CINECA (Wednesdays or Saturdays) facility are reported on the left column.

The Med-MFC-biogeochemistry system run is composed by several steps:

1. Preprocessing: download of the physical forcings, download of the satellite chlorophyll observations and 7 days centred average for the assimilation in the Wednesday run.

2. Data assimilation (only for the Wednesday run): a 3DVAR scheme for the assimilation of surface chlorophyll observations is performed (details on the assimilation scheme are provided in Teruzzi et al., 2013).

3. Model run: the biogeochemical model is the OGSTM-BFM (details are described in Lazzari et al. 2012). The model run is composed by: 1 day of analysis (with data assimilation), 6 days of hindcast and 10 days of forecast for the Wednesday run; 7 days of hindcast and 10 days of forecast for the Saturday run.

4. Post processing: the model output is processed in order to obtain the products for the CMEMS catalogue.



Date : 16-05-2017

Issue : 1.4


II.2 Description of the 3DVAR-OGSTM-BFM model system

The Med-biogeochemistry system consists of the 3DVAR-OGSTM-BFM model (Lazzari et al., 2010, 2012; Teruzzi et al., 2013; Cossarini et al., 2015, and references thereby).

OGSTM-BFM is designed with a transport model based on the OPA system and a biogeochemical reactor featuring the Biogeochemical Flux Model (BFM), while 3DVAR is the data assimilation scheme for the correction of phytoplankton functional type variables (Figure II.2).

Figure II.2. The Med-biogeochemistry model system and interfaces with other components of Copernicus system.

The transport OGSTM model is a modified version of the OPA 8.1 transport model (Foujols et al., 2000), which resolves the advection, the vertical diffusion and the sinking terms of the tracers (biogeochemical variables). The meshgrid is based on 1/16° longitudinal scale factor and on 1/16°cos(φ) latitudinal scale factor. The vertical meshgrid accounts for 72 vertical z-levels: 25 in the first 200m depth, 31 between 200 and 2000 m, 16 below 2000 m. The temporal scheme of OGSTM-BFM is an explicit forward time scheme for the advection and horizontal diffusion terms, whereas an implicit time step is adopted for the vertical diffusion. The sinking term is a vertical flux which acts on a sub-set of the biogeochemical variables (particulate matter and phytoplankton groups). Sinking velocity is fixed for particulate matter and dependent on nutrients for two phytoplankton groups (diatoms and dinoflagellates).

The physical dynamics that are off-line coupled with the biogeochemical processes are pre-computed by the Med-currents which supplies the temporal evolution of the fields of horizontal and vertical current velocities, vertical eddy diffusivity, potential temperature, salinity, in addition to surface data for solar shortwave irradiance and wind stress (see section on upstream data and boundary conditions for further details).

The features of the biogeochemical reactor BFM (Biogeochemical Flux Model) have been chosen to target the energy and material fluxes through both “classical food chain” and “microbial food web” pathways (Thingstad and Rassoulzadegan, 1995), and to take into account co-occurring effects of multi-nutrient interactions. Both of these factors are very important in the Mediterranean Sea, wherein microbial activity fuels the trophodynamics of a large part of the system for much of the year and both phosphorus and nitrogen can play limiting roles (Krom et al., 1991; Bethoux et al., 1998). The model presently includes nine plankton functional types (PFTs). Phytoplankton PFTs are diatoms, flagellates, picophytoplankton and dinoflagellates. Heterotrophic PFTs consists of carnivorous and omnivorous mesozooplankton, bacteria, heterotrophic nanoflagellates and microzooplankton.



Date : 16-05-2017

Issue : 1.4


BFM model describes the biogeochemical cycles of 4 chemical compounds: carbon, nitrogen, phosphorus and silicon through the dissolved inorganic, living organic and non-living organic compartments. Nitrate and ammonia are considered for the dissolved inorganic nitrogen. The non living compartments consists of 3 groups: labile, semilabile and refractory organic matter. The last two are described in terms of carbon, nitrogen, phosphorus and silicon contents. The model is fully described in Lazzari et al. (2012, 2015), where it was corroborated for chlorophyll, primary production and nutrients in the Mediterranean Sea for a 1998-2004 simulation.

The BFM model is also coupled to a carbonate system model (Cossarini et al., 2015), which consists of two prognostic state variables: alkalinity (ALK) and dissolved inorganic carbon (DIC). DIC evolution is driven by biological processes (photosynthesis and respiration, and precipitation and dissolution of CaCO3) and physical processes (exchanges at air-sea interface and dilution-concentration due to evaporation minus precipitation process). Alkalinity evolution is affected by biochemical processes that alter the ions concentration in sea water (nitrification, denitrification, uptake and release of nitrate, ammonia and phosphate by plankton cells, and precipitation and dissolution of CaCO3). DIC exchange at the air-sea is resolved by computing the seawater pH, pCO2 and gas transfer formula (OCMIP II model, Orr et al., 1999).

II.3 Description of Data Assimilation scheme

The data assimilation of surface chlorophyll concentration is performed through a variational scheme (3D-VAR) once a week in the Wednesday run (see details on Teruzzi et al., 2013). The surface chlorophyll concentration is provided by satellite observations produced by the OCTAC. The data assimilation corrects the four phytoplankton functional groups included in the OGSTM-BFM. The 3D-VAR scheme decomposes the error covariance matrix using a sequence of different operators that account separately for the vertical covariance (VV), the horizontal covariance (VH) and the covariance among biogeochemical variables (Vb). Full details of formulation can be found in the previous issues.

VV is defined by a set of synthetic profiles that are evaluated by means of an Empirical Orthogonal Function (EOF) decomposition. EOF has been applied to the validated multi-annual (1998-2004) OGSTM-BFM run (Lazzari et al., 2012) considering 12 months and 9 sub-regions in order to account for the variability of 3D chlorophyll fields.

VH is built using a Gaussian parameterization whose correlation radius modulates the smoothing intensity and the horizontal spatial areas influenced by the operator (Dobricic and Pinardi, 2008). A radius of 15 km has been chosen as the optimal criteria that increases the coverage of the innovation in points where there is no observation and avoids excessive smoothing of the solution.

Vb operator maintains the ratio among the phytoplankton groups and preserves the physiological status of the phytoplankton cells: in particular the internal ratios of chlorophyll-carbon and chlorophyll-nutrient are preserved, and the innovations are proportionally applied to all of the components of the phytoplankton functional types. In case of positive innovation Vb operator is relaxed in order to keep the optimal growth condition, preserving the internal contents of nutrients or increasing it as necessary. Moreover, positive innovation at depths with no light and nutrient internal ratio very far from optimal ratios (starvation stage of phytoplankton groups) is not applied.

On the Wednesday run, the assimilation scheme operatively consists of a sequence of five steps:

1. Available chlorophyll maps in a range of ±3 days are downloaded, temporally averaged and spatially interpolated on the model grid.

2. The misfit is evaluated as the difference between the satellite chlorophyll and the daily mean value of the sum of the four phytoplankton type chlorophylls.

3. The 3D-VAR provides the innovation for the four phytoplankton group variables.

4. The new initial conditions are produced, and the set-up for the following simulation is prepared.

5. The analysis/hindcast (7 days) and the new forecast (10 days) are produced using the new initial conditions derived from the data assimilation (Fig. 1), together with the physical forcings provided by Med-MFC-Currents.



Date : 16-05-2017

Issue : 1.4


II.4 Upstream data and boundary condition of the 3DVAR-OGSTM-BFM model

The CMEMS – MED-MFC-Biogeochemistry analysis and forecast system uses the following upstream data:

1. physical forcing and atmospheric forcing are those of the official product MEDSEA_ANALYSIS_FORECAST_PHYS_006_001, version V2 which has been run for the period from 1

st January 2013 till present.

2. data assimilation uses satellite surface chlorophyll from the dedicated OC TAC product OCEANCOLOUR_MED_CHL_L4_NRT_OBSERVATIONS_009_060

2.

3. initial conditions are those of the start of the V2 archive analysis and forecast run at 1st

January 2013. The initialization uses the outputs of the reanalysis product (MEDSEA_REANALYSIS_BIO_006_008) at the starting date.

4. boundary conditions (BCs) of the pre-operational qualification task are set equal to those of the CMEMS analysis and forecast V1 version for all variables but carbonate system variables. BCs of dissolved inorganic carbon and alkalinity are calculated from historical data and published scientific estimates.

In particular the boundary conditions of Gibraltar Strait, Dardanelles, rivers and atmosphere deposition are provided as following:

A Newtonian dumping term regulates the Atlantic buffer zone western of the Strait of Gibraltar, where the tracer concentrations are relaxed to the seasonally varying profiles. Seasonal profiles of phosphate, nitrate, silicate, dissolved oxygen are derived from climatological MEDAR-MEDATLAS data measured outside Gibraltar. For ALK and DIC, the Atlantic buffer profiles are obtained from the recent dataset (Huertas et al., 2009; de la Paz et al., 2011; Alvarez et al., 2014).

Atmospheric deposition rates of inorganic nitrogen and phosphorus were set according to the synthesis proposed by Ribera d’Alcalà et al. (2003) and based on measurements of field data (Loye-Pilot et al., 1990; Guerzoni et al., 1999; Herut and Krom, 1996; Cornell et al., 1995; Bergametti et al., 1992). Atmospheric deposition rates of nitrate and phosphate were assumed to be constant in time during the year, but with different values for the western (580 Kt Nyr

−1 and 16

Kt Pyr−1

) and eastern (558 K tNyr−1

and 21 Kt Pyr−1

) sub-basins. The rates were calculated by averaging the “low” and “high” estimates reported by Ribera d’Alcalà et al. (2003).

Nutrient loads from rivers and other coastal nutrient sources were based on the reconstruction of the spatial and temporal water discharge variability estimated following the method described by Ludwig et al. (2009). These values are based on available field data for nutrient concentrations, climate parameters that have been made available since the early 1960s. The nutrient discharge rates for the major rivers (Po, Rhone and Ebro) take into account seasonal variability on a monthly scale and are calculated on the basis of direct observations. All other inputs are treated as constants throughout the year due to a lack of data.

Terrestrial inputs of ALK and DIC are derived with a procedure based on two phases. In the first phase the typical concentration of ALK and DIC per water mass are computed for 10 macro coastal areas covering the entire Mediterranean Sea coastline (as defined by Ludwig et al., 2009). In the second phase the total terrestrial input of mass of ALK and DIC is derived by multiplying the estimates of water discharge (Ludwig et al., 2009) by the concentration derived in the phase 1. The Dardanelles inputs where considered as river inputs (Somot et al., 2008): also in this case the total inflow was derived considering typical water mass concentration of ALK and DIC for Marmara Sea (Chopin-Montegut, 1993) multiplied by the net water mass fluxes.

Surface evaporation-precipitation flux, which affects ALK and DIC by producing a dilution or concentration, was implemented as a virtual flux at the surface layer. The evaporation-

2 This dataset is a re-gridding of the official product OCEANCOLOUR_MED_CHL_L3_NRT_OBSERVATIONS_009_040 at the

resolution of 1/16°.



Date : 16-05-2017

Issue : 1.4


precipitation flux, provided as monthly variable climatology, was corrected in order to be balanced with Dardanelles, Gibraltar Straits, and fresh water budget.



Date : 16-05-2017

Issue : 1.4


III VALIDATION FRAMEWORK

The performance of the CMEMS-Med-MFC-biogeochemistry subsystem version V2 has been evaluated by means of a simulation covering the period April 2014 - June 2015. The simulation uses the MED-MFC-BIO subsystem described in the previous section, that will be operational from V2 version.

V2 version of NUTR, PFTC and BIOL model datasets has not changed from the previous version. The quality of these datasets of version V2 remains unchanged. Nevertheless, the present document reports the qualification of the model products computed using the semi-independent data (satellite chlorophyll) and new independent data (data sources of the Bio-Argo float network and of near real time (NRT) in situ data).

V2 version of CARB model dataset is a new feature of the product MEDSEA_ANALYSIS_FORECAST_BIO_006_006. A specific qualification task has been dedicated to both the new model variables, dissolved inorganic carbon (DIC) and alkalinity, and to the new CARB dataset variables, ocean pH and ocean pCO2.

The product quality activities used both quantitative metrics and qualitative assessment approaches. The metrics used for the quantitative validation are reported in the following table (Table III.1). The qualitative approaches are thought for those variables that cannot be quantitatively assessed such as primary production, oxygen and nutrients. Mean vertically-integrated map of net primary production is compared with previous estimation and literature values. Nutrient variables cannot be validated at NRT level due to the lack of available appropriate data sources. Therefore, validation consists of the assessment of model consistency with literature.



Date : 16-05-2017

Issue : 1.4


Name Description Ocean parameter

Supporting reference dataset Quantity

CHL-SURF-W-CLASS4-SIMG-RMSD- <BASIN>

Surface chlorophyll comparison between model output and satellite estimates

Chlorophyll OCEANCOLOUR_MED_CHL_L4_NRT_OBSERVATIONS_009_060

Weekly chlorophyll root mean square of the differences between surface satellite estimates and model output for the Mediterranean Sea and selected sub-basins. Metric is reported as time series over the pre operational qualification testing period.

CHL-SURF-W-CLASS4-SIMG-BIAS- <BASIN>

Surface chlorophyll comparison between model output and satellite estimates

Chlorophyll OCEANCOLOUR_MED_CHL_L4_NRT_OBSERVATIONS_009_060

Weekly chlorophyll average of the differences between surface satellite estimates and model output for the Mediterranean Sea and selected sub-basins. Metric is reported as time series over the pre operational qualification testing period.

CHL-PROF-W-CLASS4-

PROF-RMSD- <BASIN> Chlorophyll comparison with float profiles

Chlorophyll Biofloat platform (ftp.ifremer.fr/ifremer/argo/dac/coriolis)

Chlorophyll correlation (and RMS of differences) between Biofloat profiles and model output at the float locations. Metrics are reported for selected sub-basins and averaged over the pre operational qualification testing period.

CHL-<X-Ym>-W-CLASS4-INS-RMSD- <BASIN>

Chlorophyll comparison between model forecast and mooring and vessel datasets

Chlorophyll INSITU_MED_NRT_OBSERVATIONS_013_035/monthly/

Chlorophyll root mean square of the differences between available datasets and model output at the in situ observation locations. The metric is calculated for selected layers (0-10, 10-50, 50-100, 100-150, 150-300, 300-600, 600-1000) and sub-basins and averaged over the pre operational qualification testing period.

O2-<X-Ym>-M-CLASS4-INS-RMSD- <BASIN>

Oxygen comparison between model forecast and mooring and vessel datasets

Oxygen INSITU_MED_NRT_OBSERVATIONS_013_035/monthly/

Oxygen root mean square of the differences between available datasets and model output at the in situ observation locations. The metric is calculated for selected layers (0-10, 10-50, 50-100, 100-150, 150-300, 300-600, 600-1000) and sub-basins and averaged over the pre operational qualification testing period.

ALK-<X-Ym>-W-CLASS4-INS-RMSD- <BASIN>

Alkalinity comparison between model output and a climatology from historical datasets

Alkalinity Climatology from historical datasets Alkalinity root mean square of the differences between a climatological dataset interpolated on a 1°x1° grid and the model output averaged over the pre operational qualification test period. The metric is computed for selected layers (0-50, 50-100, 50-150, 150-200, 200-500, 500-1000, 1000-1500, 1500-4000).

DIC-<X-Ym>-W-CLASS1-CLIM-RMSD- <BASIN>

Dissolved inorganic carbon (DIC) comparison between model output and a climatology from historical datasets

DIC Climatology from historical datasets Dissolved inorganic carbon root mean square of the differences between a climatological dataset interpolated on a 1°x1° grid and the model output averaged over the pre operational qualification test period. The metric is computed for selected layers (0-50, 50-100, 50-150, 150-200, 200-500, 500-1000, 1000-1500, 1500-4000).

Table III.1. List of Metrics.

ftp://ftp.ifremer.fr/ifremer/argo/dac/coriolis



Date : 16-05-2017

Issue : 1.4


The datasets of observations used for the qualification task are listed below.

1) Surface chlorophyll from satellite from CMEMS OCTAC

Weekly maps of surface chlorophyll from MODIS satellite have been produced by averaging daily 1/16̂ chlorophyll fields from the dedicated CMEMS-OCTAC product: OCEANCOLOUR_MED_CHL_L4_NRT_OBSERVATIONS_009_060.

2) Chlorophyll profiles of Bio-Argo Float from Ifremer/Coriolis service

This dataset consists of Bio-Argo float chlorophyll observations retrieved from Coriolis/IFREMER webportal (ftp://ftp.ifremer.fr/ifremer/argo/dac/coriolis). Only “adjusted” profiles are used, being covered by a quality check procedure and appropriate documentation (http://archimer.ifremer.fr/doc/00243/35385/). Oxygen and nitrate data from Bio-Argo floats have not “adjusted” data and are not used at the present stage of qualification activity for V2 version.

27 Bio-Argo floats have collected a total number of 1312 “adjusted-data” chlorophyll profiles for the period covered by the pre-operational run (April 2014 to June 2015).

3) Near Real Time biogeochemical data from CMEMS INS TAC

Near Real Time in-situ data used in the pre-operational qualification task consist of chlorophyll observations available thought the CMEMS-INS TAC (ftp://medinsitu.hcmr.gr/Core/INSITU_MED_NRT_OBSERVATIONS_013_035/monthly).

For the period April 2014 – June 2015 the data retrieved from the CMEMS product INSITU_MED_NRT_OBSERVATIONS_013_035 are the following:

- Chlorophyll at 4 depths from 2 moorings: Pylos (Ionian Sea), E1M3A (Aegean Sea); - Oxygen at 4 depths from one mooring: E1M3A (Aegean Sea).

Figure III.1. Location of moorings with available biogeochemical data from CMEMS product INSITU_MED_NRT_OBSERVATIONS_013_035 (only Pylos and E1M3A have data for a reliable model-observation comparison).

4) Carbonate system dataset for the Mediterranean Sea.

The collected dataset covers the period from 1999 to 2013 and includes 4 scientific cruises covering the whole Mediterranean Sea, several scientific cruises in marginal seas and local areas, and a single fixed station (Dyfamed) continuously monitored for several years. Usually, the scientific cruises were conducted during one month mostly in spring or autumn. Data have sparse and scarce spatial and temporal coverage (Table III.2, and Fig. III.2), they do not resolve the annual cycle and barely work out

http://archimer.ifremer.fr/doc/00243/35385/



Date : 16-05-2017

Issue : 1.4


the basin wide gradients. The most observed variables are DIC and alkalinity (up to 90% of the samplings), while pH was collected only in less than 30% of the samplings.

Therefore, the data have been used to compute two kinds of reference climatologies for DIC and alkalinity:

mean maps computed on an 1°x1° grid for selected layers (0-50, 50-100, 50-150, 150-200, 200-500, 500-1000, 1000-1500, 1500-4000).

mean profiles computed over 18 selected areas 4°x4° wide. Areas (Figure II.2) have been selected according to the location and abundance of data.

Name Variables Period Location # data Reference

METEOR51 DIC, ALK, anc. vars Oct-Nov 2001 TransMed 253 Schneider et al., 2007

BUOM2008 DIC, ALK, anc. vars June-July 2008 TransMed 567 Touratier et al., 2011

PROSOPE DIC, pH@25, anc. vars Sep-Oct 1999 West Med 188 Begovic and Copin, 2013

METEOR 84/3 DIC, ALK, pH@25, anc. Vars Apr 2011 TransMed 845 Tanhua, et al., 2012.

SESAME-EGEO DIC, ALK, T,S Apr and Sep 2008 Aegean Sea 265 http://isramar.ocean.org.il/PERSEUS_Data/

SESAME regina_maris

ALK, pH@25, anc. vars Apr 2008 Alboran Sea 254 http://isramar.ocean.org.il/PERSEUS_Data/

SESAME Garcia del Cid

ALK, pH@25, anc. vars Sep 2008 Alboran Sea 331 http://isramar.ocean.org.il/PERSEUS_Data/

SESAME Adriatic ALK, pH@25, anc. vars Apr and Sep 2008 Adriatic Sea 333 http://isramar.ocean.org.il/PERSEUS_Data/

CARBOGIB ALK, DIC, pH@25, anc. Vars May, Sept, Dec 2005; Mar, May, Dec 2006

Alboran Sea 229 Huertas, 2007a

GIFT ALK, DIC, pH@25, anc. Vars Jun, Nov 2005 Alboran Sea 30 Huertas, 2007b

DYFAMED Station ALK, DIC Almost monthly from 1999 to 2004

North West Med 707 Copin-Montegut and Begovic, 2002

MEDSEA 2013 DIC, ALK, T,S May 2013 TransMed 462 Goyet et al., 2015

Table III.2. List of datasets used to build climatology of the carbonate system variables. TransMed: scientific cruise that covered the Mediterranean Sea from the Western sub-basin to the Eastern sub-basin; anc.vars: ancillary variables (T, S, oxygen and nutrients).

Figure III.2. Location of the carbonate system variables data and identification of 18 areas 4°x4° wide used for the validation activities.

The 4°x4° gridded profile climatology has been used to qualify the vertical dynamics of the carbonate system, while the 1°x1° grid climatology has been used to computed the metrics of DIC and alkalinity reported in the next section. Using the uncertainty of DIC and alkalinity in the surface layer (which is the most interesting for climate and acidification processes) and the error propagation approach (see next section), the uncertainty of the new product variables pCO2 and pH is estimated.



Date : 16-05-2017

Issue : 1.4


The Mediterranean Sea is nominally subdivided into a selected number (10) of sub-basins (Figure III.3). Biogeochemistry validation task is carried out considering the previous list. However, according to data availability and to ensure consistency and robustness of the metrics, a subset of the sub-basin list or some combinations among the given sub-basins can be used for the different metrics.

Figure III.3. Model bathymetry and subdvision of the model domain in sub-basins used for model validation.



Date : 16-05-2017

Issue : 1.4


IV VALIDATION RESULTS

Following the discussion of section III, the results of the validation can be resumed as follows.

I. Horizontal maps of fundamental biogeochemical variables and sub-basin estimates

At near real time, several products variables (nutrients, oxygen, primary production) cannot be validated due to the lack of available and reliable data for the model-observation comparison. Therefore, the evaluation of the model consistency with published literature performs the estimate of product quality for these variables. The qualitative comparison of model results of nutrients and primary production with other estimates is shown in the following figures. Relevant gradients and average values in the different sub-basins are well reproduced by the V2 version of the CMEMS MED-MFC-Bio model system. Additional evidences of the model consistency with the known dynamics are given in the QUID of the reanalysis product (CMEMS-MED-QUID-006-008).

Figure IV.1. Phosphate spatial distributions annual average and vertically averaged over the 0-50 m layer (mmol P.m

-3), from CMEMS V2 system (pre-operational qualification run, left) and from a

validated multi-annual simulation (from Fig. 4 of Lazzari et al., 2016, right).

Figure IV.2. Nitrate spatial distributions annual average and vertically averaged over the 0-50 m layer (mmol N.m

-3), from CMEMS V2 system (pre-operational qualification run, left) and from from a

validated multi-annual simulation (from Fig. 4 of Lazzari et al., 2016, right).



Date : 16-05-2017

Issue : 1.4


Figure IV.3. Annual averaged vertically integrated primary production (gC m−2

yr−1

) CMEMS V2 system (pre-operational qualification run, left) and from multi-annual simulation (Lazzari et al., 2012, right)

Lazzari et al. (2012)

Colella (2006)

CMEMS V2 – system range of variability

Mediterranean (MED) 9882 9048 70-200

Alboran Sea (ALB) 274155 179116 160-230

South West Med (SWW) 16089 11343 130-180

North West Med (NWM) 11679 11567 115-170

Levantine (LEV) 7661 7221 80-125

Table IV.1. Annual averaged vertically integrated primary production (gC m−2

yr−1

) for some selected sub-regions. From multi-annual simulation (Lazzari et al., 2012), from satellite model (Colella, 2006), from CMEMS V2 system (pre-operational qualification run).

II. Surface chlorophyll validation using satellite data

Validation of surface chlorophyll is performed using semi-independent data. In fact, within a cycle of assimilation, re-initialization and simulation, the misfit computed before the assimilation is here used as a measure of the model accuracy of surface chlorophyll. Indeed, the satellite data are used for assimilation and validation but at different moments.

Metrics (see definition in Table III.1) are computed considering both the natural and the log-transformation of the chlorophyll. Time series of bias and RMS of the differences are plotted in Figure IV.4 for selected sub-basins, and their averages considering two seasons are reported in table IV.2. These averages serve as Estimated Accuracy Numbers reported in section I.3. The errors are quite low in all sub-basins and seasons. RMS exceeds 0.1 mg/m

3 only in ALB, and in NWM (actually RMS

is equal to 0.093) and AEG in winter.. During spring bloom in the NWM sub-basin, model depicts a delay in bloom dynamics of about 3-4 weeks. In the ALB sub-basin, model error is due to a not well-resolved spatial local dynamics, although values of this very productive area are well reproduced. In this area, blooms are strongly related to the presence of local, sub-mesoscale patches, fronts, horizontal circulation structures and local mixing conditions of the water column. A possible source of error for the model forecast arises from the uncertainty of the physical forcing which can erroneously drive the position of the high-chlorophyll patchiness.



Date : 16-05-2017

Issue : 1.4


Figure IV.4. Trend of bias and RMS between model and satellite data for different sub-basins.

Surface (0-10m) chlorophyll Mod-Sat

Surface (0-10m) chlorophyll Log10(Mod)-log10(Sat)

RMS BIAS

RMS BIAS

win sum win sum win sum win sum

alb 0.157 0.171 0.031 0.029 0.511 0.650 0.109 0.111

sww 0.066 0.021 0.014 0.004 0.147 0.099 0.038 0.032

swe 0.067 0.018 0.004 0.004 0.160 0.095 0.006 0.027

nwm 0.093 0.023 -0.012 0.002 0.190 0.102 -0.029 0.017

tyr 0.052 0.013 -0.006 0.001 0.141 0.081 -0.011 0.015

adn 0.039 0.024 0.004 0.010 0.142 0.135 0.024 0.067

ads 0.039 0.025 -0.007 0.010 0.141 0.100 -0.027 0.044

aeg 0.107 0.020 0.021 0.002 0.140 0.091 0.024 0.007

ion 0.031 0.015 -0.003 -0.001 0.107 0.089 -0.022 -0.014

lev 0.023 0.013 -0.003 -0.003 0.104 0.095 -0.028 -0.033

Table IV.2. Mean correlation and bias between surface chlorophyll model maps and satellite maps. On the right, the skill indexes are computed on the model and satellite chlorophyll log-transformed. “win” refers to the January-June period, while “sum” refers to the July-December period.



Date : 16-05-2017

Issue : 1.4


III. Chlorophyll vertical profiles validation using Bio-Argo float data

The present section reports the comparison of chlorophyll model output with Bio-Argo profiles. The figures IV.5-8 show the matching between model and Bio-Argo profiles grouped by months and sub-basins, while the averaged skill indexes are reported in Tables IV.3 and 4. An overall RMS of 0.24 mg/m

3 and an overall correlation of 0.77 give evidence of the good quality of the model in reproducing

the shape of the chlorophyll profiles throughout the months. Dynamics of surface blooms, development and destruction of the deep chlorophyll maximum are well simulated by the model. The eastern sub-basins (LEV, ION, AEG, ADS) have low values of RMS (almost lower than 0.20), while the model presents values of RMS that can be quite high in the western sub-basins, due to an under-estimation of surface and sub-surface values. The comparison of Bio-Argo float data represents the first attempt of a quantitative and systematic validation of the 3D field of chlorophyll. The use of this new datastream needs further refinement in terms of quality check of the observations (i.e. no internal quality check procedure is implemented except for the a-priori selection of the “adjusted” data) and in terms of selection of the appropriate metrics.

Figure IV.5. Chlorophyll profiles from model (black) and Bio-Argo float (blue) for the North Western Mediterranean sub-basin. Matched model and Bio-Argo profiles are grouped by months from July 2014 to June 2015.

Figure IV.6. Chlorophyll profiles from model (black) and Bio-Argo float (blue) for the eastern part of the South-Western Mediterranean sub-basin. Matched model and Bio-Argo profiles are grouped by months from August 2014 to April 2015.



Date : 16-05-2017

Issue : 1.4


Figure IV.7. Chlorophyll profiles from model (black) and Bio-Argo float (blue) for the Ionian sub-basin. Matched model and Bio-Argo profiles are grouped by months from July 2014 to June 2015.

Figure IV.8. Chlorophyll profiles from model (black) and Bio-Argo float (blue) for the Levantine sub-basin. Matched model and Bio-Argo profiles are grouped by months from July 2014 to June 2015.

CORR J14 A S O N D J15 F M A M J

alb 0.92 0.55 0.58 0.76 0.76 0.91 0.68 0.89 0.75 0.75 0.76 0.84

nwm 0.90 0.93 0.94 0.81 0.84 0.81 0.86 0.66 0.75 0.76 0.76 0.75

sww 0.63 0.59 0.62 0.71 0.88 0.90 0.87 0.86

0.68 0.78 0.83

swe

0.82 0.68 0.70 0.73 0.93 0.87 0.87 0.87 0.86

tyr 0.33 0.53 0.72 0.90 0.90 0.88 0.93 0.98 0.84 0.84 0.85 0.84

ads 0.69 0.85 0.84 0.75 0.90 0.85

ion 0.76 0.78 0.80 0.84 0.91 0.84 0.86 0.71 0.88 0.83 0.79 0.69

lev 0.77 0.60 0.60 0.65 0.86 0.74 0.90 0.68 0.65 0.58 0.69 0.80

aeg 0.68 0.57

Table IV.3. Averaged correlation between model and Bio-Argo chlorophyll profiles. Averages are computed for matched profiles considering each sub-basin and a monthly temporal resolution.



Date : 16-05-2017

Issue : 1.4


RMS J14 A S O N D J15 F M A M J

alb 0.29 0.32 0.30 0.27 0.19 0.06 0.10 0.13 0.15 0.09 0.09 0.08

nwm 0.31 0.22 0.27 0.26 0.21 0.24 0.26 0.23 0.46 0.46 0.32 0.27

sww 0.53 0.37 0.28 0.40 0.31 0.33 0.45 0.39 0.00 0.35 0.26 0.32

swe 0.21 0.25 0.24 0.30 0.22 0.30 0.37 0.28 0.24

tyr 0.44 0.35 0.30 0.24 0.25 0.23 0.23 0.31 0.28 0.27 0.17 0.18

ads 0.28 0.25 0.18 0.21 0.21 0.13

ion 0.16 0.16 0.16 0.14 0.14 0.16 0.13 0.18 0.19 0.19 0.23 0.21

lev 0.13 0.24 0.16 0.12 0.15 0.19 0.16 0.28 0.23 0.19 0.21 0.25

aeg 0.16 0.22

Table IV.4. Averaged RMS between model and Bio-Argo chlorophyll profiles. Averages are computed for matched profiles considering each sub-basin and a monthly temporal resolution.

IV. Chlorophyll and Oxygen time series validation using in-situ data from moorings

At the present stage, the comparison of CMEMS version V2 model system with the in-situ NRT data from moorings and vessel is at a very preliminary stage, due to the very scarce availability of data offered through CMEMS product INSITU_MED_NRT_OBSERVATIONS_013_035. Indeed, only two moorings, E1M3A and Pylos, have some available data of chlorophyll and oxygen for the period corresponding to the pre-operational task (April 2014 – June 2015). Although planned quantitative metrics (see list of metrics of Table III.1) cannot be computed, some highlights can be evidenced for the modelled variables at these specific points. V2 version model system reproduces quite well the different temporal evolutions of chlorophyll at the different depths of the two moorings. Chlorophyll values at E1M3A are more consistent with observations than in Pylos, for which an underestimation is evident for the first two layers (Figure IV.9). Results of oxygen (Figure IV.10) show that model reproduces quite well the seasonal cycle but with a lower variability.

It is worth to remind that this is the first attempt to compare model results with high-frequency time series derived from in-situ platforms. Therefore, these highlights will be definitely useful to the future quality product assessments. Indeed, developments are under progression in order to check the consistency of the observations, the appropriateness of the model-observation matching procedure (i.e. representative of the observation within model domain) and to select the appropriate metrics for this kind of data.

Figure IV.9. Chlorophyll timeseries at different depths at the E1M3A (left) and Pylos (right) in-situ moorings for the period of the pre-operational qualification run (April 2014-June 2015). Model (red) and in-situ observation (blue) are reported.



Date : 16-05-2017

Issue : 1.4


Figure IV.10. Oxygen timeseries at surface layer of the E1M3A in-situ moorings for the period of the pre-operational qualification run (April 2014-June 2015).. Model (red) and in-situ observation (blue) are reported.

V. Carbonate system variables and CARB dataset variables validation

The new component of the 3DVAR-OGSTM-BFM model system (the carbonate system module) is able to reproduce the principal spatial patterns of the carbonate system variables in the Mediterranean Sea. Further, the validation of the new model variables (DIC and alkalinity) serves to estimate the accuracy of the new CARB variables of the MEDSEA_ANALYSIS_FORECAST_BIO_006_006 product: pH and pCO2 at surface.

In particular, DIC and alkalinity constitute the prognostic (state variables) of the carbonate system, and are those with the largest number of observations (up to 90% of the datasets of Tab.II.2). Further, pH and pCO2 constitute diagnostic variables of the carbonate system module: they are very seldom collected (less than 30% of the datasets of Table II.2 contain pH) but can be easily computed off-line when the values of DIC, alkalinity and other ancillary variables such as T, S and concentration of phosphate and silicate are known. The validation task of the V2 system was performed on the 2 master variables of the carbonate system (DIC and alkalinity). Then, the accuracy of pCO2 and pH is eventually estimated starting from the accuracy of DIC and alkalinity using a propagation error approach.

Results of the pre-operational task run (July 2014 – June 2015) is used to build the comparison of mean annual maps with the 1°x1° climatology (Figure IV.11). The results indicate a strong surface west-to-east gradient of both alkalinity and DIC. The eastern marginal seas (Adriatic and Aegean seas) are characterized by the highest values. The west-to-east gradient is a permanent structure recognizable at all depths, but less marked in the maps of the intermediate and deep layers (see for example the layer 200-500, Figure IV.12). At the surface, DIC and alkalinity dynamics are driven by three major factors: the input in the eastern marginal seas (the terrestrial input from the Po and other Italian rivers and the input from the Dardanelles), the effect of evaporation in the eastern basin and the influx of the low-alkaline ad low-DIC Atlantic waters. The thermohaline basin-wide circulation modulates the intensity and the patterns of the spatial gradients. Intermediate and deep layers show lower dynamics and less variability.



Date : 16-05-2017

Issue : 1.4


Figure IV.11. Mean annual map of Alkalinity (left) and DIC (right) simulated by the OGSTM-BFM model (upper panels) and reconstructed by the 1°x1° climatology (lower panels) for the layer 0-50m.

Figure IV.12. Mean annual map of Alkalinity (left) and DIC (right) simulated by the OGSTM-BFM model (upper panels) and reconstructed by the 1°x1° climatology (lower panels) for the layer 200-500m.

The skill performance was analyzed by computing the correlation, the bias and the root mean square of differences (RMSD) between the model map and the 1°x1° climatology maps for the selected layers. Model results have been transformed on the observation space. The results of the skill metrics, reported in Table IV.5, show that correlations are significant and very high for both variables at the surface layer. The correlations of DIC for the layers between 100 and 500 are lower than 0.5, because of spatial patterns characterized by very low gradients from east to west, with the climatological maps characterized by rather fuzzy patterns. The mean BIAS error is less than 10 μmol/kg and less than 15 μmol/kg for alkalinity and DIC, respectively, while RMSD, which is used as an estimation of the accuracy, is of about 20-30 μmol/kg for both variables.



Date : 16-05-2017

Issue : 1.4


Alkalinity DIC

Layer depth

Corr Bias RMSD Corr Bias RMSD

0-50 0.91 4.8 27.1 0.85 -0.1 20.6

50-100 0.87 -1.2 27.9 0.57 -4.8 23.3

100-150 0.89 -2.6 21.0 0.33 -12.2 26.8

150-200 0.90 -3.9 16.3 0.37 -14.6 26.8

200-500 0.89 -7.2 14.0 0.47 -16.7 19.9

500-1500 0.83 -7.5 13.8 0.58 -7.3 10.4

1500-4000 0.78 -10.9 15.4 0.57 -6.8 9.2

Table IV.5. Skill metrics for the comparison of alkalinity and DIC on the 1x1 grid domain and selected layers.

The performance of the new component of the model system (i.e. the carbonate system) is also evaluated by computing the skill metrics on vertical profiles of 18 selected areas. Examples of the comparison between mean model profiles and the climatological profiles are given in Figure IV.13 and IV.14 while the skill metric values are given in Table IV.6. The analysis pretty well confirms the capability of the model to reproduce the different shape of the profiles and the distribution of vertical values along the west to east gradient for both variables. It is worth to conclude that, given the available information and data, the CMEMS-MED-MFC-Biogeochemistry model system presents a rather good capability to reproduce the average conditions of the carbonate system variables at the basin wide scale.

Figure IV.13. Alkalinity profiles: mean monthly model profiles (colored lines; from July 2014 to June 2015) and climatological (±standard deviation) profiles (black lines) for the areas 2, 6, 15 of Fig. III.2.



Date : 16-05-2017

Issue : 1.4


Figure IV.14. DIC profiles: mean monthly model profiles (colored lines; from July 2014 to June 2015) and climatological (±standard deviation) profiles (black lines) for the areas 1, 5, 18 of Fig. III.2.

DIC Alkalinity

Areas Corr Bias RMSD Corr Bias RMSD

1 0.95 -12.1 22.2 0.99 2.5 13.2

2 0.94 -20.8 27.1 0.98 -15.2 18.0

3 0.98 -7.1 10.9 0.98 -1.1 8.3

4 0.95 -7.1 10.7 0.91 -10.1 11.8

5 0.95 -8.6 14.9 0.98 -2.3 8.4

6 0.91 -12.5 15.7 0.97 -5.8 6.9

7 0.99 1.2 20.3 0.99 6.2 22.4

8 0.94 -5.2 11.2 0.93 -9.8 13.9

9 n.a. n.a. n.a. 0.13 -12.6 17.1

10 0.98 -1.5 7.1 0.91 -2.8 12.7

11 0.76 -8.3 12.0 0.76 -14.9 19.7

12 0.92 -36.0 37.1 -0.41 -23.8 26.1

13 0.78 -7.5 11.5 0.06 -20.3 31.2

14 0.88 -7.9 10.0 0.29 -10.3 13.5

15 0.88 -18.6 19.4 -0.01 -17.3 25.4

16 -0.63 -23.5 35.1 -0.06 -28.7 38.6

17 0.86 -7.0 10.3 0.02 -2.8 8.6

18 0.85 -8.0 11.0 0.62 -3.3 6.8

Table IV.6. Skill metrics of the comparison of alkalinity and DIC on the 18 selected areas 4°x4° wide.



Date : 16-05-2017

Issue : 1.4


Two new variables of the product MEDSEA_ANALYSIS_FORECAST_BIO_006_006 are released in V2 version: ocean pH and ocean pCO2, which are essential climate variables (ECVs) and are commonly used in the UNFCCC and IPCC reports as indicators of the climate change and the ocean acidification processes. Examples of mean monthly values of pH and pCO2 for the Mediterranean Sea at surface are given in Figure IV.15 and IV.16 for given months: pH and pCO2 exhibit spatial patterns partly overlapped to those of DIC and alkalinity and partly connected to the temperature dynamics since the important role played by the thermal effect on the activity of [H

+] and on the CO2 solubility.

Indeed, the monthly maps depict a variability of the annual cycle that is as high as 0.15-0.20 and 150-200 ppm for pH and pCO2, respectively.

Figure IV.15. Mean monthly values of pH at in situ condition and reported in Sea Water Scale (SWS) at the 0-50m layer.

Figure IV.16. Mean monthly values of pCO2 [ppm] for the 0-50m layer.



Date : 16-05-2017

Issue : 1.4


No validation with observations is possible for pCO2 and pH since available data are too scarce and it results not possible to compute a reliable climatology for these variables. Therefore the quality assessment of the new products is based on an indirect estimate. The accuracy of pCO2 (UpCO2) and pH (UpH) is estimated starting from the accuracy of DIC (UDIC) and alkalinity (Ualk) adopting the propagation error approach:

UpCO2 = ((δpCO2

δDIC⋅ UDIC)

2

+ (δpCO2

δalk⋅ Ualk)

2

)0.5

UpH = ((δpH

δDIC⋅ UDIC)

2

+ (δpH

δalk⋅ Ualk)

2

)0.5

Using the results of the comparison on the 1°x1° maps, the RMS of DIC and alkalinity, reported in Table IV.5, is assumed as the measure of their uncertainty. The derivative of pCO2 and pH respect to DIC and alkalinity are approximated by their local linear deviation values. Then, the resulting uncertainty of the mean values of pCO2 and pH in the surface layer are:

UpCO2=50.8 ppm

UpH= 0.05

These are conservative estimates of the uncertainty of pH and pCO2. First, it is worth to note that DIC and alkalinity are, to some degree, correlated variables (Figures IV.11 and IV.12) since some processes have similar impacts on both variables. Indeed, an underestimation of DIC is likely to occur with an underestimation of alkalinity (e.g. an error on the dilution – concentration process at surface has the same impact on both variables). Then, DIC and alkalinity have opposite effect on pH and pCO2 (Zeebe and Wolf-Gladrow, 2001). Therefore, considering a concomitant overestimation / underestimation of DIC and alkalinity, it is presumable to assume that the error estimate of pH and pCO2 does not increase with the sum of the errors of the two variables. Further, the choice of using the basin wide RMS estimate as the measure of the uncertainty might have given an overestimation of the total uncertainty. Indeed, the highest model-climatology errors are computed in transitional areas between sub-basins and in marginal areas. DIC and alkalinity uncertainties computed on sub-basin-wide scale would have had lower values. However, since the low availability of data, the estimates would also be less reliable.



Date : 16-05-2017

Issue : 1.4


V RECENT EVENTS HAVING AN IMPACT ON DATA AVAILABILITY AND/OR QUALITY

VI.1 Update on 11 July 2016: replacement of data file.

A technical error occurred on 15 April 2016: it did not involve the model and did not affect its performance with respect to what is described in the present QUID. The archived files from April 15th to July 11th had been updated and the operational product restarted from the corrected data. The quality of these archived and current operational products is the same of that one described in the QUID. The quality of the operational analysis & forecast products is described in the Quarterly Validation Report. The QVR for Apr-May-Jun 2016 was built using the products corrected.

VI.2 Update on 22 November 2016: impact on Biogeochemistry due to update in Med-Currents for GLO-LOBC.

The product MEDSEA_ANALYSIS_FORECAST_BIO_006_006 is updated from 23rd November: biogeochemistry is now forced by the updated MEDSEA_ANALYSIS_FORECAST_PHYS_006_001 product.

Validation activities showed that the new physical forcing has negligible impacts on the quality of biogeochemical products. However, users have to take into account that modest discontinuities in horizontal spatial patterns of biogeochemical variables at surface might occur.

VI.3 Update on 18 May 2017: impact on Biogeochemistry due to the update in V3 surface chlorophyll OCTAC product. From V3 chlorophyll products are released as multisensory Near Real Time (NRT) within few days and substituted with the Delay Time (DT) version after 25-30 days. Due to the difference between the V2 MODIS-based surface chlorophyll and the V3 multisensor-based chlorophyll, the product MEDSEA_ANALYSIS_FORECAST_BIO_006_006 has a small negative bias of the chlorophyll forecasts (<0.005 mg/m3) with respect to the previous version. The impact on the quality of chlorophyll (which will be able to use the NRT data) and on other biogeochemical variables is negligible. However it must be considered that there could be inconsistencies between V3 DT chlorophyll OCTAC product and the chlorophyll provided by the biogeochemical analysis and forecasts Med-MFC that used the NRT OCTAC product.



Date : 16-05-2017

Issue : 1.4


VI QUALITY CHANGES SINCE PREVIOUS VERSION

V2 version is basically equal to the previous version of the CMEMS operational system: there are no changes in the datasets NUTR, BIOL and PFTC of the MEDSEA_ANALYSIS_FORECAST_BIO_006_006 product. The improvements of the V2 version are due to the implementation of two new product variables included in the CARB dataset (ocean pH and ocean pCO2) and to the validation of the new components. Further, improvement of the new system are related to the implementation of new data streams for model validation that will produce an amelioration of the product quality activities and therefore of the product reliability.

Date post:	21-May-2020
Category:	Documents
Upload:	others
View:	7 times
Download:	0 times