Product Information Document (PIDoc)

SeaDataCloud Mediterranean Heat Content estimate

SDC_MED_DP2

HORIZON 2020 2020 sdn-userdesk@.org – www.seadatanet.org

SeaDataCloud - Further developing the pan-European infrastructure for marine and ocean data management

Grant Agreement Number: 730960

Product Name

SDC_MED_DP2

Extended name

SeaDataCloud Mediterranean - Ocean Heat Content 1955-2018

Product DOI

https://doi.org/10.12770/504ea4ee-abee-4ebf-ab89-c4200e1cdad4

Short description

Ocean Heat Content estimates over the 0-700m and 0-2000m layers have been obtained from temperature sliding decadal fields computed with DIVAnd tool

Authors

Simona Simoncelli and Paolo Oliveri

Dissemination Copyright terms

Public

How to Cite

History

Version Authors Date Comments

V1 S. Simoncelli and P. Oliveri 07/12/2020 first draft

A. Iona 11/01/2021 review

V2 S. Simoncelli 21/01/2021 second version

V. Myroshnychenko 12/02/2021 second review

V3 S. Simoncelli 17/02/2021 final version

[email protected] – www.seadatanet.org

SeaDataCloud - Further developing the pan-European infrastructure for marine and ocean data management

Grant Agreement Number: 730960 2

Table of contents Abstract ...... 4 1. Input Data ...... 5 1.1. General description of the input SeaDataCloud data set ...... 5 1.2. Global CORA 5.2 REP data set ...... 7 1.3. Integration of SDC and CORA5.2 data sets ...... 10 2. Methodology ...... 14 2.1. Data QC ...... 14 2.2. DIVA implementation and settings ...... 14 2.2.1. DIVAnd settings ...... 15 2.3. Mediterranean Ocean Heat Content Anomaly ...... 16 3. Product Description ...... 19 4. Consistency analysis ...... 22 5. Technical Specifications ...... 25 5.1. Product Format ...... 25 5.2. Product Usability ...... 26 6. References ...... 27

[email protected] – www.seadatanet.org

SeaDataCloud - Further developing the pan-European infrastructure for marine and ocean data management

Grant Agreement Number: 730960 3

Abstract The SDC_MED_DP2 product contains 55 sliding decadal temperature fields (1955-1964, 1956-1965, 1957-1966, …, 2009-2018) at 1/8° horizontal resolution obtained in the 0-2000m layer and two derived OHC annual anomaly estimates for the 0-700m and the 0-2000m layers. Sliding decades of annual Temperature fields were obtained from an integrated Mediterranean Sea dataset covering the time period 1955-2018, which combines data extracted from SeaDataNet infrastructure at the end of July 2019 (SDC_MED_DATA_TS_V2, https://doi.org/10.12770/3f8eaace-9f9b-4b1b-a7a4-9c55270e205a) and the Coriolis Ocean Dataset for Reanalysis (CORA 5.2, accessed in July 2020, https://archimer.ifremer.fr/doc/00595/70726/). The resulting annual OHC anomaly time series span the 1960-2014 period. The analysis was performed with the DIVAnd (Data-Interpolating Variational Analysis in n dimensions), version 2.6.1.

Whenever SDC_MED_DP2 product is used, PIDoc should be cited in any publication. The data products providers normally possess insight on the quality and context of the product and tools not always shared with the SeaDataCloud team. Hence, inviting data providers and product leaders to collaborate in scientific investigations that depend on their data products is considered good and fair practice. Importantly, this will promote further sharing of products and will be beneficial to science.

[email protected] – www.seadatanet.org

SeaDataCloud - Further developing the pan-European infrastructure for marine and ocean data management

Grant Agreement Number: 730960 4

1. Input Data

1.1. General description of the input SeaDataCloud data set The SDC_MED_DP2 product contains Temperature climatological fields for the Mediterranean Sea at 1/8 of degree of horizontal resolution on 92 standard levels. Annual fields from sliding decades were produced over the time period 1955-2018. The climatological fields were obtained from an integrated dataset which combines data extracted from SeaDataNet infrastructure and the Coriolis Ocean Dataset for Reanalysis (CORA), version 5.2 (July 2020 update). Temperature profiles with Quality Flags (QF) 1 (good) and 2 (probably good) have been extracted from SDC_MED_DATA_TS_V2 unrestricted and restricted data collections to be used for the generation. The SDC_MED_DATA_TS_V2 collections have been obtained harvesting all measurements contained within SeaDataNet infrastructure at the end of July 2019. Simoncelli et al. (2020) describes the data set and the quality assessment procedures applied. SDC_MED_DATA_TS_V2 contains 1003258 stations, among which 997255 temperature profiles. Thermosalinograph (TSG hereafter) data (total 808364 stations) have been discarded due to their peculiar spatial and temporal distribution (see Fig.9 in Simoncelli et al., 2020) along trajectories at the near surface and their availability only after year 2000. The collection of restricted data consists of 22503 stations. Unrestricted and restricted data have been merged in ODV obtaining 216859 stations (without TSG). The duplicate check identified and removed 74 stations. The dataset contains in total 216759 temperature profiles spatially distributed as in Figure 1. Figure 2 presents the temporal data distribution, which is characterized by a data availability increase starting in late eighties and a reduction in the recent years, due to the time lag that occurs among sampling and data submission into the SeaDataNet infrastructure. Data availability is maximum during summertime thanks to the favorable sampling conditions.

[email protected] – www.seadatanet.org

SeaDataCloud - Further developing the pan-European infrastructure for marine and ocean data management

Grant Agreement Number: 730960 5

Figure 1 Distribution (top) and data density (bottom) maps of SeaDataCloud unrestricted and restricted Temperature stations used as input SDC_MED_DP2 product.

Figure 2 Temporal distribution of SeaDataCloud unrestricted and restricted Temperature stations: (left panel) annual distribution, and (right panel) seasonal distribution.

[email protected] – www.seadatanet.org

SeaDataCloud - Further developing the pan-European infrastructure for marine and ocean data management

Grant Agreement Number: 730960 6

1.2. Coriolis Global CORA 5.2 REP data set CORA 5.2 data set (Szekely et al., 2019), distributed through the Copernicus Marine Environment Monitoring Service (hereafter CMEMS), contains in-situ observations yearly delivery in delayed mode. The in situ delayed mode product integrates the best available version of in situ data for temperature and measurements. These data are collected from main global networks (, GOSUD, OceanSITES) completed by European data provided by EUROGOOS regional systems and national systems by the regional INS TAC components. The 5.2 version is a merged product between the previous version of CORA and EN4 distributed by the Met Office for the period 1950-1990. The data set covers the global ocean over the time period 1950-June 2019 (July 2020 access). The sub-set of data covering the Mediterranean Sea domain (Longitude > -5.5 degrees) has been downloaded and used to generate SDC_MED_CLIM_TS_V2 climatology. The CORA file system is based on several daily file types corresponding to the instrument type of the data provider. The following data types have been considered to be integrated with the SDC data set:  PR_PF files: data from Argo floats directly received from DACS (RT and DM if available). These data have a nominal accuracy of 0.01°C and 0.01 PSU and are transmitted with full resolution;  PR_CT files: contains CTD data from research vessels with accuracy on the order of 0.002° for temperature and 0.003 PSU for salinity after calibration;  PR_IC files: CTD from ICES dataset that completes the CTDs coverage on the period 1990-2011;  PR_SH files: profiles from the SHOM database, most of them cover the 1950-1990 period;  PR_ME files: CTD from SISMER database coming from French oceanographic campaigns;  PR_OS files: OceanSites data (mostly CTD);  PR_OC files, containing additional CTD and XCTD data from the high resolution CTD dataset of the World Ocean Database. Figure 3 shows the data time distribution of the selected CORA 5.2 data types, from which it appears a minimum of data availability in the early 2000s.

[email protected] – www.seadatanet.org

SeaDataCloud - Further developing the pan-European infrastructure for marine and ocean data management

Grant Agreement Number: 730960 7

Figure 3 CORA 5.2 data time distribution of the selected data types as input to the SDC_MED_DP2 climatology.

The extraction of the CORA 5.2 NetCDF files has been done through a common matlab procedure (shared by V. Myroshnychenko) which considers:  casts with QF =1 (good) for time and position  exported adjusted parameters when they exist. The data were reformatted to ODV spreadsheet (https://odv.awi.de/) and imported to ODV collection for QC and analysis. An additional quality check has been performed through visual inspection following the methodology detailed in Simoncelli et al. (2020). CORA 5.2 data presented anomalies, data flagged as good but visibly wrong. Moreover 235 internal duplicates in CORA data have been detected and eliminated from further analysis. This fundamental step aims at providing data consistency, since SeaDataNet and CMEMS apply different quality check strategies and procedures. As for SDC data, only measurements with associated QF=1,2 were considered for further analysis. The initial CORA 5.2 sub set consists of 355652 stations, 338114 profiles in the time span 1955- 2018, spatially distributed as in Figure 4. Figure 5 displays the temporal distribution of CORA 5.2 sub-set, which shows low data availability in the 2000s. Seasonal data distribution has maximum data availability in June-July and minimum in December-January.

[email protected] – www.seadatanet.org

SeaDataCloud - Further developing the pan-European infrastructure for marine and ocean data management

Grant Agreement Number: 730960 8

Figure 4 Distribution (top) and data density (bottom) maps of CORA5.2 stations (338114 temperature profiles) integrated with SDC_MED_DP2 product.

Figure 5 Temporal distribution of CORA 5.2 temperature stations integrated with the SDC_MED_DP2 product: (left panel) annual distribution, and (right panel) seasonal distribution.

[email protected] – www.seadatanet.org

SeaDataCloud - Further developing the pan-European infrastructure for marine and ocean data management

Grant Agreement Number: 730960 9

1.3. Integration of SDC and CORA5.2 data sets Data sets integration was performed merging SDC and CORA 5.2 data in ODV. The merged ODV collection contains 572160 temperature stations, 547830 in the time period 1955-2018. 338114 are from CORA 5.2 (Mediterranean only) and 209216 from SDC. The duplicate check has been performed through ODV looking for “quasi-perfect” duplicates within a coordinate interval of 0.0001deg (~11m) and a temporal interval of 0.0069 days (10 minutes). 92847 (16%) duplicates have been found over the 572160 stations, 86971 of which are from CORA 5.2. The remaining duplicates can be classified as two types of internal duplicates in SDC Med collection: 1) repeated samples on the same location but with assigned coincident time (62 groups of duplicates), or 2) profiles sampled with different instrument types at the same location (5543 profiles). The 86971 CORA 5.2 duplicates have been removed from the dataset, 3526 co-located T and S profiles over 5543 profiles from discrete samples have been eliminated, keeping the corresponding high resolution CTDs. The 62 groups of profiles having coincident time have been averaged adopting the superobbing approach (Palamarchuket al. 2017) used in data assimilation. The content of the final integrated and validated Mediterranean Sea Temperature and Salinity dataset of T profiles with QF equal to 1 (good) or 2 (probably good) is detailed per month in Table 1 and Figure 6. The input dataset consists in total of 451130 stations, 45% is from SDC and 55% is from CORA 5.2. Table 1 Content of the Mediterranean Sea integrated dataset per month in terms of number of temperature stations N and in percentage.

TOT SDC CORA5.2 N N % N % Jan 27237 11822 43% 15415 57% Feb 33812 14872 44% 18940 56% Mar 41020 19903 49% 21117 51% Apr 35753 15858 44% 19895 56% May 47738 22175 46% 25563 54% Jun 39229 18568 47% 20661 53% Jul 39346 17623 45% 21723 55% Aug 34896 15151 43% 19745 57% Sep 40941 18789 46% 22152 54% Oct 44625 18717 42% 25908 58% Nov 39422 16195 41% 23227 59% Dec 27111 11621 43% 15490 57% (TOT) 451130 (TOT) 201294 (MEAN) 45% (TOT) 249836 (MEAN) 55%

[email protected] – www.seadatanet.org

SeaDataCloud - Further developing the pan-European infrastructure for marine and ocean data management

Grant Agreement Number: 730960 10

Figure 6 Monthly data distribution over the time period 1955-2018 for the integrated data set in terms of number of stations.

The monthly spatial distribution of stations in the integrated dataset over the time period 1955-2018 is presented in Figure 7. CORA5.2 stations (in red) cover only the Mediterranean, while SDC ones cover also a buffer zone in the Atlantic Ocean in order to resolve the temperature distribution in the Gibraltar Strait. Major data gaps are apparent in the southern- central Mediterranean in all months.

Figure 7 Monthly spatial distribution of temperature profiles over the time period 1955-2018. In blue the SDC temperature stations and in red the CORA5.2 ones.

[email protected] – www.seadatanet.org

SeaDataCloud - Further developing the pan-European infrastructure for marine and ocean data management

Grant Agreement Number: 730960 11

The annual data distribution in

Figure 8 shows the data availability over the analysis time span.

Figure 8 Annual distribution of temperature stations in the integrated dataset over the time period 1955-2018.

The vertical data distribution in Figure 9 shows maximum data availability at about 10-15 m that rapidly decreases with depth. Below 2000m data are very sparse. At the surface (0m) data are sparse too thus requiring caution when using surface climatological fields

[email protected] – www.seadatanet.org

SeaDataCloud - Further developing the pan-European infrastructure for marine and ocean data management

Grant Agreement Number: 730960 12

Figure 9 Number of SDC (blue) and CORA5.2 (orange) stations per each standard depth level over the considered time period 1955-2018.

[email protected] – www.seadatanet.org

SeaDataCloud - Further developing the pan-European infrastructure for marine and ocean data management

Grant Agreement Number: 730960 13

2. Methodology Temperature sliding decadal fields (1955-1964, 1956-1965, 1957-1966, …, 2009-2018) have been computed considering all temperature measurements in each ten years window. 55 sliding decades have been produced covering the layer 0-2000m at 1/8° horizontal resolution on the WOA18 standard depths. The analysis was performed with the DIVAnd (Data- Interpolating Variational Analysis in n dimensions - Barth et al., 2014), version 2.6.1.

2.1. Data QC SeaDataCloud data have been quality checked according to the guidelines defined in the framework of SeaDataNet2 Project and further refined as described in Simoncelli et al. (2020b). Additional QC analysis in ODV has been performed on the CORA 5.2 data sub-set in order to discard data anomalies (outliers, stations on land, wrong spatial or temporal location) from further consideration during the analysis. This QC analysis has been performed also on the merged data set, once removed the duplicates, to guarantee data consistency. The SDC-CORA5.2 merged dataset has been exported in netCDF format and read with Matlab software to remove the list of detected duplicates and perform the linear interpolation of temperature profiles on standard levels. Extrapolation has been applied to improve the quality of surface and bottom fields:  measurements above 1.5m depth have been copied at the surface depth level (0m);  Near bottom observations at depth less than 700m have been copied to the next depth level if the difference between observed depth and standard depth is less than 10% of the layer thickness;  Near bottom observations deeper than 700m have been copied to the next depth level if the difference between observed depth and standard depth is less than 20% of the layer thickness. Linear interpolation has been applied in this first product version, but the impact of vertical interpolation and discretization on OHC estimates will be further investigated.

2.2. DIVA implementation and settings The Mediterranean Sea is a semi-enclosed basin connected to the Atlantic Ocean by the Gibraltar Strait and to the Marmara Sea by the Dardanelles. The Mediterranean Sea domain includes a small Atlantic box (Figure 10), in order to obtain a realistic representation of the Gibraltar Straight-Alboran Sea region. In this buffer zone only SDN data have been used since CORA5.2 sub set ends at Gibraltar as shown in Figure 4. The topography (Figure 10) has been generated using the new GEBCO_2019 Grid - the latest global bathymetric product released by the General of the (GEBCO, Olson et al., 2014). The GEBCO_2019 product provides global coverage on a 15 arc-second grid.

[email protected] – www.seadatanet.org

SeaDataCloud - Further developing the pan-European infrastructure for marine and ocean data management

Grant Agreement Number: 730960 14

Figure 10 Mediterranean Sea from GEBCO_2019 Grid product (Olson et al., 2014).

The land-sea mask (Figure 11) has the same dimensions (367x129x67) as the analysis grid and it was produced from the GEBCO_2019 bathymetry, masking the Sea of Marmara and some isolated grid cells in the analysis levels.

Figure 11 Land-sea mask at the surface.

2.2.1. DIVAnd settings Spatial coverage: Longitude: 9.25°W-36.5°E; Latitude: 30-46°N; Vertical: 0-2000m. Horizontal resolution: 1/8° Horizontal grid: 367x129 Vertical resolution: 67 depth levels, as in WOA18 (https://data.nodc.noaa.gov/woa/WOA18/DOC/woa18_vol1.pdf):

[email protected] – www.seadatanet.org

SeaDataCloud - Further developing the pan-European infrastructure for marine and ocean data management

Grant Agreement Number: 730960 15

 from 0 to 100 m by 5m steps;  from 125 to 500 m by 25m steps;  from 500 to 2000m by 50m steps; Temporal resolution: annual over 55 sliding decades spanning the time period 1955-2018. Background field Annual background field of Temperature has been defined with L=5.5 degrees and epsilon2 equal to 12. In our 32 Gb RAM setting the memtofit parameter is equal to 125. It controls how to cut the domain depending on the memory remaining available for inversion and it allows to utilize as much memory as possible, however this product has been computed in a Cluster with 384G of RAM solving memory problems. The tuning/optimization of correlation length (L) and epsilon2 parameters has been performed using synthetic profiles extracted from the Mediterranean Sea reanalysis data set (Simoncelli et al. 2019) at observed locations over the time period 1987-2018. Synthetic profiles have been used to compute Temperature and Salinity monthly (Simoncelli et al., 2020b) and compared with the reanalysis climatologies. The full methodology and results will be soon available in Simoncelli et al. (2021) in preparation. The correlation length (L) is isotropic and equal to 2° degrees for the analysis. The epsilon2 parameter is the error variance of the observations (normalized by the error variance of the background field). If epsilon2 is a scalar it is the inverse of the signal-to-noise ratio. In SDC_MED_DP2 climatology (Simoncelli et al., 2020b) epsilon2 parameter has been set to 6 to get smooth annual fields.

2.3. Mediterranean Ocean Heat Content Anomaly The aim of present SeaDataCloud product is to provide a new estimate of OHC anomaly for the Mediterranean Sea from an extensive data set which blends CORA 5.2 product and SDC historical data collection and a new DIVAnd mapping tool. Once produced the Temperature sliding decadal fields, the OHC has been computed for each decade in two layers: 1) 0-700m, and 2) 0-2000m. The OHC is considered an important Ocean Monitoring Indicator of warming due to climate change effect. Ocean Heat Content (OHC) has been defined within the Copernicus Marine Service (K. von Schuckmann et al., 2016 and 2018, Storto et al. 2019) as the deviation from a reference period and it is closely proportional to the average temperature change either from the surface to 700 m or 2000m depth, with a reference density of 1030 kgm-3 and a specific heat capacity (Cp) of 3980 J/kg°C (von Schuckmann et al., 2009). 푧2 푂퐻퐶 = 휌0퐶푝 ∫ (푇푖 − 푇̅)푑푧 푧1

[email protected] – www.seadatanet.org

SeaDataCloud - Further developing the pan-European infrastructure for marine and ocean data management

Grant Agreement Number: 730960 16

Figure 12 has been extracted from Storto et al. (2019) that introduced the ensemble mean approach of computing the OHC anomaly in the Mediterranean Sea from multiple products in order to get an uncertainty estimate from the ensemble spread.

Figure 12 Ocean heat content anomaly timeseries (0–700 m) in the Mediterranean Sea from the four GREP members and the ensemble mean (top panel); from GREP-EM, the CMEMS Mediterranean Sea Monitoring and Forecasting Center (MED-MFC) regional reanalysis and the CORA observation-only product (middle panel), with +/- one ensemble standard deviation represented by the grey shading and the CORA error estimated by the red bars; the ensemble spread from GREP (bottom panel) and the linearly fit line (dashed red line). Values are yearly means. From Storto et al. 2019.

[email protected] – www.seadatanet.org

SeaDataCloud - Further developing the pan-European infrastructure for marine and ocean data management

Grant Agreement Number: 730960 17

Figure 13 presents the latest released CMEMS OHC anomaly (0-700m) indicator (https://resources.marine.copernicus.eu/?option=com_csw&view=details&product_id=MED SEA_OMI_OHC_area_averaged_anomalies) computed as ensemble mean and relative spread from multiple products, considering as reference the climatological field 1993-2014. The Med OHC anomaly presents a progressive positive increase after year 2006.

Figure 13 Time series of area averaged ocean heat content anomaly in the Mediterranean Sea integrated over the 0-700m depth layer (MEDSEA_OMI_OHC_area_averaged_anomalies). Time series represents ensemble mean and the ensemble spread computed from different data products, i.e. Mediterranean Sea, global ocean reanalysis GLORYS, C-GLORS, ORAS5, FOAM, global observational based products CORA, ARMOR3D.

[email protected] – www.seadatanet.org

SeaDataCloud - Further developing the pan-European infrastructure for marine and ocean data management

Grant Agreement Number: 730960 18

3. Product Description The SDC_MED_DP2 product consists of sliding decadal temperature fields and annual OHC anomalies deriving from them over two layers 0-700m and 0-2000m. Figure 14 presents the hovmoller plot of basin average temperature profile and relative anomaly computed in the layer from surface to 2000m from the sliding decadal fields. Annual Temperature climatological values range from 13 degree, which characterize mainly the layer below 700m depth, and 16 degrees in the upper part of the . Temperature anomalies have been computed subtracting a mean field obtained from the decades: 1955- 1964, 1965-1974, 1975-1984, 1985-1994, 1995-2004, 2005-2014. Positive anomalies characterize the surface layer until the early seventies, then negative anomalies start dominate until mid-nineties, when positive anomalies prevail again. The intermediate layer behavior, up to about 600m, appear shifted in time with respect to the surface of about 10 years. Negative anomalies dominate below 700m until mid-nineties, positive anomalies prevail afterwards.

Figure 14 Hovmoller plot of Temperature (top) and Temperature anomaly (bottom) from surface to 2000m depth in the whole Mediterranean Sea. Basin averaged temperature profiles have been

[email protected] – www.seadatanet.org

SeaDataCloud - Further developing the pan-European infrastructure for marine and ocean data management

Grant Agreement Number: 730960 19 computed from sliding decadal fields. The anomaly field has been computed subtracting the temperature averaged field over the analysis time period.

Figure 15 shows time series of basin averaged ocean heat content anomaly integrated over the 0-700m depth layer computed from two V1 data sets (see below), the V2 data set (SDC_MED_DP2) and the reference product from Iona et al. (2018). The Iona et al (2018) OHC product used the SeaDataNet V2 (Simoncelli et al., 2015) dataset and the OHC anomaly has been computed for 5m and not from 0m. Experiments V1 A and V1 B are based on the dataset of collocated temperature and salinity profiles used to produce SDC_MED_CLIM_TS_V1 climatologies which span the time period 1955-2017. V1 A is based on sliding decades computed on IODE standard levels as in Iona et al. (2018), V1 B sliding decades are instead computed on the same 67 WOA18 standard levels as V2. V1 A and V1 B time series are similar indicating that the vertical resolution of the gridded fields does not substantially affect the OHC anomaly result. V2 time series is instead quite different from V1 A and B until mid- nineties and this can be ascribed to the input data set used which contains also all available MBT and XBT temperature profiles in SDC_MED_DATA_TS_V2 data set. None correction has been applied to SDC MBT and XBT profiles and this has to be considered in evaluating the results. This is another aspect that needs to be further investigated in the next product’s release. V1 and V2 match quite well after mid-nineties. V2 time series differs from V1 sensitivity tests’ results presenting positive anomalies from 1960 to mid-seventies, when they start to decrease. Negative anomalies characterize mainly the eighties until mid-nineties and after year 2000 the anomalies became positive maintaining a constant tendency to increase.

Figure 15 Time series of area averaged ocean heat content anomaly in the Mediterranean Sea integrated over the 0-700m depth layer.

Figure 16 shows time series of of basin averaged ocean heat content anomaly integrated over the 0-2000m layer. V2 time series differs from V1 sensitivity tests’ results presenting anomalies close to zero from 1960 to min-seventies, when they start to decrease. Negative anomalies characterize mainly the eighties and from mid-nineties the anomalies became positive with a constant increase tendency.

[email protected] – www.seadatanet.org

SeaDataCloud - Further developing the pan-European infrastructure for marine and ocean data management

Grant Agreement Number: 730960 20

Figure 16 Time series of area averaged ocean heat content anomaly in the Mediterranean Sea integrated over the 0-2000m depth layer.

[email protected] – www.seadatanet.org

SeaDataCloud - Further developing the pan-European infrastructure for marine and ocean data management

Grant Agreement Number: 730960 21

4. Consistency analysis A consistency analysis has been performed considering the work from Iona et al. (2018). Temperature fields, on which the computation is based, have been compared in Figure 17 and Figure 18. Temperature anomalies have all been computed subtracting a mean field obtained from the decades: 1955-1964, 1965-1974, 1975-1984, 1985-1994, 1995-2004, 2005-2014. The temperature evolution looks consistent, but with slightly more intense anomalies at the surface from Iona et al. (2018).

Figure 17 Hovmoller plot of Temperature from surface to 2000m depth in the whole Mediterranean Sea. Basin averaged temperature profiles have been computed from sliding decadal fields: (top) SDC_MED_DP2, and (bottom) Iona et al. (2018).

[email protected] – www.seadatanet.org

SeaDataCloud - Further developing the pan-European infrastructure for marine and ocean data management

Grant Agreement Number: 730960 22

Figure 18 Hovmoller plot of Temperature anomalies from surface to 2000m depth computed in the whole Mediterranean Sea. Basin averaged anomaly profiles have been computed from sliding decadal fields: (top) SDC_MED_DP2, and (bottom) Iona et al. (2018).

Figure 19 compares the two OHC anomalies in the 0-700m layer and their relative trend computed after 1993. SDC_MED_DP2 (labelled V2) displays a steeper trend than Iona et al. (2018), partially due to the extension of the analysis time period from 2015 to 2018. SDC_MED_DP2 behaviour appear consistent with the CMEMS time series in Figure 13, but its annual rate of OHC increase, 1.1Wm-2 is much higher than the CMEMS one 0.8Wm-2. This can be explained from the different reference period that in CMEMS indicator is 1993-2014, while in SDC_MED_DP2 is 1955-2014. The annual trend computed from Iona et al. (2018) is the smallest one, 0.5Wm-2.

[email protected] – www.seadatanet.org

SeaDataCloud - Further developing the pan-European infrastructure for marine and ocean data management

Grant Agreement Number: 730960 23

Figure 19 Time series of area averaged ocean heat content anomaly in the Mediterranean Sea integrated over the 0-700m depth layer from SDC_MED_DP2 (V2) and Iona et al. (2018), with the relative linear trends computed over the time period after 1993.

Figure 20 compares the two OHC anomalies in the 0-2000m layer and their relative trend computed after 1993. SDC_MED_DP2 (labelled V2) displays again a steeper trend than the one computed from Iona et al. (2018), partially due to the extension of the analysis time period from 2015 to 2018. The warming rates in the 0-2000m layer are larger than the ones computed up to 700m depth: 1.4Wm-2 for SDC_MED_DP2 (V2) and 1.0Wm-2 for Iona et al. (2018).

Figure 20 Time series of area averaged ocean heat content anomaly in the Mediterranean Sea integrated over the 0-2000m depth layer from SDC_MED_DP2 (V2) and Iona et al. (2018), with the relative linear trends computed over the time period after 1993.

[email protected] – www.seadatanet.org

SeaDataCloud - Further developing the pan-European infrastructure for marine and ocean data management

Grant Agreement Number: 730960 24

5. Technical Specifications

5.1. Product Format

Product files Description

SDC_MED_DP2_1_T_1960_2014_0125_annual.nc Temperature sliding decades

SDC_MED_DP2_2_O_1960_2014_annual.nc OHC and Temperature annual time series in the layers 0-700m and 0-2000m

The product is delivered in 2 files in netCDF format. SDC_MED_DP2_1 contains Temperature 4D matrixces (lon = 367; lat = 129 ; depth = 67 ; time = 55) named according to the following rule: • Parameter– 3D array for a parameter, • Parameter_L1 – parameter masked using relative error threshold 0.3, • Parameter_L2 – parameter masked using relative error threshold 0.5, • Parameter_relerr – relative error of parameter.

SDC_MED_DP2_2 netcdf file contains two (nv = 2) time series (depth = 2; time = 55) named according to the following rule: • Ocean_Heat_Content – 2D array containing OHC anomaly in the delivered layers 0-700 and 0-2000 m; • Temperature_anomaly – 2D array containing sea water temperature anomaly for the two layers 0-700m and 0-2000m. The netCDF file, along with the variable attributes, contains a set of global attributes describing the product: • Product code, version and abstract, • Name of the project, • EDMO code of the product developer, • Contact e-mail of developer, • Source of observations, • Keywords for the parameter and the area and their codes in SeaDataNet Vocabularies, Links to documentation, data and visualization tools.

[email protected] – www.seadatanet.org

SeaDataCloud - Further developing the pan-European infrastructure for marine and ocean data management

Grant Agreement Number: 730960 25

5.2. Product Usability While every effort is made to produce an error free grid, some artefacts may still appear in the data set. If you find any anomaly in the temperature fields or OHC calculations, please report them via email ([email protected]), giving the problem specifications, and we will investigate.

[email protected] – www.seadatanet.org

SeaDataCloud - Further developing the pan-European infrastructure for marine and ocean data management

Grant Agreement Number: 730960 26

6. References Barth, A., Beckers, J.-M., Troupin, C., Alvera-Azcárate, A., and Vandenbulcke, L.: DIVAnd-1.0: n- dimensional variational data analysis for , Geosci. Model Dev., 7, 225-241, doi:10.5194/gmd-7-225-2014, 2014. Cabanes, C., A. Grouazel, K. von Schuckmann, M. Hamon, V. Turpin, C. Coatanoan, F. Paris, S. Guinehut, C. Boone, N. Ferry, C. de Boyer Montégut, T. Carval, G. Reverdin, S. Pouliquen, and P. Y. Le Traon, 2013: The CORA dataset: validation and diagnostics of in-situ ocean temperature and salinity measurements. Ocean Science, 9, 1-18, http://www.ocean-sci.net/9/1/2013/os-9-1-2013.html, doi:10.5194/os-9-1- 2013 Iona, A., Theodorou, A., Sofianos, S., Watelet, S., Troupin, C., and Beckers, J.-M.: Mediterranean Sea climatic indices: monitoring long-term variability and climate changes, Earth Syst. Sci. Data, 10, 1829– 1842, https://doi.org/10.5194/essd-10-1829-2018, 2018. Olson, C J, Becker, J J and Sandwell, D T (2014). A new global bathymetry map at 15 arcsecond resolution for resolving seafloor fabric: SRTM15_PLUS, AGU Fall Meeting Abstracts 2014. Yuliia Palamarchuk, Serguei Ivanov, and Igor Ruban. Thinning and superobbing of radar data in precipitation simulations. Geophysical Research Abstracts, Vol. 19, EGU2017-3383-1, 2017EGU General Assembly 2017. https://meetingorganizer.copernicus.org/EGU2017/EGU2017-3383-1.pdf Simoncelli Simona, Coatanoan Christine, Myroshnychenko Volodymyr, Sagen Helge, Bäck Örjan, Scory Serge, Grandi Alessandro, Schlitzer Reiner, Fichaut Michele (2015). SeaDataNet. Second release of the aggregated data sets products. WP10 Fourth Year Report - DELIVERABLE D10.4. https://doi.org/10.13155/50382 Simoncelli S., Schaap D., Schlitzer R. (2018a). Mediterranean Sea - Temperature and salinity Historical Data Collection SeaDataCloud V1. https://doi.org/10.12770/2698a37e-c78b-4f78-be0b-ec536c4cb4b3 Simoncelli S., Myroshnychenko V., Coatanoan C. (2018b). SeaDataCloud Temperature and Salinity Historical Data Collection for the Mediterranean Sea (Version 1). Product Information Document (PIDoc). https://doi.org/10.13155/57036 Simoncelli, S., Fratianni, C., Pinardi, N., Grandi, A., Drudi, M., Oddo, P., & Dobricic, S. (2019). Mediterranean Sea Physical Reanalysis (CMEMS MED-Physics) [Data set]. Copernicus Monitoring Environment Marine Service (CMEMS). https://doi.org/10.25423/MEDSEA_REANALYSIS_PHYS_006_004 Simona Simoncelli, Paolo Oliveri (2019). SeaDataCloud Mediterranean Sea - Temperature and Salinity Climatology V1. https://doi.org/10.12770/ad07a55f-5de7-4abc-ba89-8899b16c4b59 Simoncelli Simona, Oliveri Paolo, Mattia Gelsomina (2020). SeaDataCloud Temperature and Salinity Climatology for the Mediterranean Sea (Version 1). Product Information Document (PIDoc). https://doi.org/10.13155/77506 Simona Simoncelli, Dick Schaap, Reiner Schlitzer (2020). Mediterranean Sea - Temperature and salinity Historical Data Collection SeaDataCloud V2. https://doi.org/10.12770/2a2aa0c5-4054-4a62-a18b- 3835b304fe64 Simoncelli Simona, Oliveri Paolo, Mattia Gelsomina, Myroshnychenko Volodymyr (2020a). SeaDataCloud Temperature and Salinity Historical Data Collection for the Mediterranean Sea (Version 2). Product Information Document (PIDoc). https://doi.org/10.13155/77059

[email protected] – www.seadatanet.org

SeaDataCloud - Further developing the pan-European infrastructure for marine and ocean data management

Grant Agreement Number: 730960 27

Simona Simoncelli, Paolo Oliveri, Gelsomina Mattia (2020b). SeaDataCloud Mediterranean Sea - V2 Temperature and Salinity Climatology. https://doi.org/10.12770/3f8eaace-9f9b-4b1b-a7a4- 9c55270e205a Storto, A., Masina, S., Simoncelli, S. et al. The added value of the multi-system spread information for ocean heat content and steric investigations in the CMEMS GREP ensemble reanalysis product. Clim Dyn 53, 287–312 (2019). https://doi.org/10.1007/s00382-018-4585-5 The GEBCO Digital Atlas published by the British Oceanographic Data Centre on behalf of IOC and IHO, 2003 Szekely Tanguy, Gourrion Jerome, Pouliquen Sylvie, Reverdin Gilles (2019). The CORA 5.2 dataset for global in situ temperature and salinity measurements: data description and validation. Ocean Science, 15(6), 1601-1614. Publisher's official version : https://doi.org/10.5194/os-15-1601-2019 , Open Access version : https://archimer.ifremer.fr/doc/00595/70726/ von Schuckmann, K., F. Gaillard and P.-Y. Le Traon, 2009: Global hydrographic variability patterns during 2003-2008, Journal of Geophysical Research, 114, C09007, doi:10.1029/2008JC005237. K. von Schuckmann, M. Balmaseda, S. Simoncelli 2016: Ocean heat content. In: The Copernicus Marine Environment Monitoring Service Ocean State Report, issue 1, Journal of Operational , Volume 9, 2016. http://dx.doi.org/10.1080/1755876X.2016.1273446. K. von Schuckmann, A.Storto, S. Simoncelli, R.P. Raj, A.Samuelsen, A. de Pascual Collar, M. Garcia Sotillo, T. Szerkely, 2018: Ocean heat content. In: The Copernicus Marine Environment Monitoring Service Ocean State Report, issue 2, Journal of Operational Oceanography, 11:sup1, s1-s142, DOI: 10.1080/1755876X.2018.1489208

[email protected] – www.seadatanet.org

SeaDataCloud - Further developing the pan-European infrastructure for marine and ocean data management

Grant Agreement Number: 730960 28