D6.1: Implementation of CMEMS-MSCS Interfaces

D6.1: Implementation of CMEMS-MSCS Interfaces WP6: Implementation of new interfaces and methodologies in coastal systems 30-06-2025/v1.0 1 About this document Title D6.1: Implementation of CMEMS-MSCS Interfaces Work Package WP6, Implementation of new interfaces and methodologies in coastal systems Lead Partner Deltares (WP leads: Kai Christensen, MET Norway & Quentin Jamet, SHOM) Lead Author (Org) Lotta Beyaard (Deltares), Lorinc Meszaros (Deltares), Devanshi Pathak (Deltares) Contributing Author(s) Ann Kristin Sperrevik (MET.NO), Ina K. Kullmann (MET.NO), Kai Christensen (MET.NO), Quentin Jamet (SHOM), Franck Dumas (SHOM), Audrey Pasquet (SHOM), Fabien Brosse (SHOM), Denis Gourves (SHOM),Catarina Cecilio (+ATLANTIC),Cintia Bonanad (+ATLANTIC),Pedro Almeida (+ATLANTIC), Francisco Campuzano (+ATLANTIC), Soraia Romão (+ATLANTIC), Debora Bellafiore (CNR), Christian Ferrarin (CNR), Jun She (DMI), Vilnis Frishfelds (DMI), Johannes Schultz-Stellenfleth (HEREON), Kelli Johnson (HEREON), Wei Chen (HEREON), Johannes Pein (HEREON), Douglas Vieira da Silva (HEREON), Benjamin Jacob (HEREON), Carolina Gramcianinov (HEREON), Saheed Puthan Purayil (RBINS), Sebastien Legrand (RBINS), Lars Arneborg (SMHI), Sandra-Esther Brunnabend (SMHI), Emma Reyes (SOCIB), Máximo Garcia-Jove Navarro (SOCIB), Eric Jansen (CMCC), Ivan Federico (CMCC), Salvatore Causio (CMCC), Camilla Campanati (CMCC), Emanuela Mikhailov (MHD), Isabel Garcia Hermosa (MOI), Joanna Staneva (HEREON) Reviewers K. Johnson (Hereon), J. Staneva (Hereon) Due Date 30.06.2025, M18 Submission Date 29.06.2025 Version 1.0 Dissemination Level X PU: Public PP: Restricted to other programme participants (including the Commission) RE: Restricted to a group specified by the consortium (including the Commission) CO: Confidential, only for members of the consortium (including the Commission) FOCCUS: Forecasting and observing the open-to-coastal ocean for Copernicus users is a Research and Innovation action (RIA) funded by the Horizon Europe Work programme topics addressed: HORIZON-CL4-2023-SPACE-01: Strategic autonomy in developing, deploying and using global space based infrastructures, services, applications and data 2023. Start date: 01 January 2024. End date: 31 December 2026. Funded by the European Union (Grant Agreement No. 101133911). Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union or the European Health 2 / 52 and Digital Executive Agency (HaDEA). Neither the European Union nor the granting authority can be held responsible. Table of Contents Table of Contents............................................................................................................................ 3 Glossary and Abbreviations.............................................................................................................5 1. Executive summary..................................................................................................................... 6 2. Overview of Member State Coastal Systems (MSCS)....................................................................7 2.1. MSCS “3D DCSM-FM” (Dutch Continental Shelf).................................................................... 12 2.2. MSCS "Norkyst" (Norway)....................................................................................................... 13 2.3. MSCS “CROCO’’ (France)......................................................................................................... 14 2.4. MSCS “Tolosa-SW’’ (France).................................................................................................... 15 2.5. MSCS “GCOAST-GB” (German Bight)....................................................................................... 16 2.6. MSCS “GCOAST-BS” (Northwestern Black Sea)........................................................................17 2.7. MSCS "Lazio coastal areas" (Italy)............................................................................................18 2.8. MSCS "DKSS" (Denmark )........................................................................................................ 19 2.9. MSCS "OPTOS v3" (Belgian Coastal Zone ).............................................................................. 20 2.10. MSCS "SHYFEM-CNR" (Italy ).................................................................................................21 2.11. MSCS "MOHID-LisOcean" (Tagus and Sado Estuaries, Portugal)........................................... 22 2.12. MSCS "AdriFS" (Italy)............................................................................................................. 23 2.13. MSCS "WMOP" (Spain)..........................................................................................................24 2.14. MSCS "NEMO-ORUST50 + OpenDrift" (Orust-Tjörn fjord system).........................................25 3. Biogeochemical (BGC) mapping of MSCS for nesting in CMEMS................................................. 26 3.1 Use of CMEMS in BGC model - 3D DCSM-FM...........................................................................26 3.2 Use of CMEMS in BGC model - GCOAST................................................................................... 28 3.3 Use of CMEMS in BGC model - LisOcean WQ...........................................................................28 3.4 Conclusion and recommendations...........................................................................................29 4. Implementation of CMEMS-MCSC interface improvements....................................................... 30 4.1 3D DCSM-FM............................................................................................................................ 30 4.1.1 Thermobaricity effects in Equation of State [BETTER].................................................... 30 4.1.2 Suspended Particulate Matter (SPM) forcing from gap-filled CMEMS [FORC]................31 4.2 Norkyst..................................................................................................................................... 33 4.2.1 Adjustment of hydrography on external boundary condition [BETTER]......................... 33 4.3 CROCO-MANGA........................................................................................................................35 4.3.1 Open boundary perturbed ensemble [ENS]....................................................................35 4.4 Tolosa-SW.................................................................................................................................36 4.4.1 Filtering baroclinic dynamics [BETTER].......................................................................... 36 4.5 GCOAST-GB...............................................................................................................................38 4.5.1 Extended nesting via spongelayer relaxation [NUDG]....................................................38 4.6 GCOAST-BS................................................................................................................................39 3 / 52 4.6.1 2D Wave Spectra forcing computed from CMEMS parameters [WAVE]......................... 39 4.7 Lazio Coastal Areas...................................................................................................................40 4.7.1 Increasing temporal frequency for open lateral boundary [TECHN]...............................40 4.8 DKSS..........................................................................................................................................42 4.8.1 Nudging of T/S products [NUDG]................................................................................... 42 4.9 OPTOS v3.................................................................................................................................. 43 4.9.1 Baroclinic Nested Downscaling and SST Bias Correction through Nudging. [NEST] [NUDG].....................................................................................................................................43 4.10 SHYFEM-CNR.......................................................................................................................... 44 4.10.1 Spatially varying nudging of CMEMS T/S products [NUDG].......................................... 44 4.11 MOHID-LisOcean.................................................................................................................... 45 4.11.1 Circulation and waves coupling up to coastal and estuary structure level [TECHN]..... 45 4.12. AdriFS.................................................................................................................................... 47 4.12.1 Wave Spectra lateral boundary rebuild from CMEMS parameters [WAVE].................. 47 4.13 Conclusion and Next Steps..................................................................................................... 48 5. Annex........................................................................................................................................50 6. References.................................................................................................................................51 4 / 52 Glossary and Abbreviations BGC Biogeochemical CMEMS Copernicus Marine Environment Monitoring Service CNR National Research Council (Italy) CRMSE Centered Root Mean Square Error CRPS Continuous Rank Probability Score DMI Danish Meteorological Institute DCSM Dutch Continental Shelf Model EMODnet European Marine Observation and Data Network FOCCUS Forecasting and observing the open-to-coastal ocean for Copernicus users HABs Harmful Algal Blooms HEREON Helmholtz-Zentrum Hereon MAE Mean Absolute Error MET.NO Norwegian Meteorological Institute MHD Marine Hydrographic Directorate (Romania) ML Machine Learning MSCS Member State Coastal System (s) NetCDF Network Common Data Form PP Restricted to other programme participants (including the Commission) PU Public PUM Product User Manual RE Restricted to a group specified by the consortium (including the Commission) RBINS Royal Belgian Institute of Natural Sciences RMSE Root Mean Square Error S Salinity SCHISM SHOM Semi-implicit Cross-scale Hydroscience Integrated System Model Service Hydrographique et Océanographique de la Marine (France) SD Standard Deviation SHYFEM System of HydrodYnamic Finite Element Modules SOCIB Balearic Islands Coastal Observing and Forecasting System SPM Suspended Particulate Matter SST Sea Surface Temperature SWAN Simulating WAves Nearshore T Temperature WP Work Package WGS84 World Geodetic System 1984 WQ Water quality WW3 WaveWatch III® XBEACH Cross-shore Beach Dynamics Model 5 / 52 1. Executive summary The FOCCUS project (https://foccus-project.eu/) aims to improve and advance the coastal dimension of CMEMS and demonstrate a convincing leverage and coupling of CMEMS and Member State Coastal System (MSCS) for advancing knowledge about the coastal environment and associated applications. Work Package 6 (WP6) focuses on strengthening the connections between MSCS and CMEMS through the application of innovative nesting methods and new data fusion approaches using stochastic simulations, ensemble approaches and AI technologies. This report provides an overview of all MSCS, assesses the current status of biogeochemical (BGC) mapping and describes the newly implemented nesting methods per MSCS. Thereby, this report summarises the work on: - Subtask 6.1.1 Parameter mapping: Mapping of MSCS with respect to nesting requirements with focus on BGC models, where integration is the most challenging. - Subtask 6.1.2 Implementation of new nesting capabilities: Better utilisation of CMEMS products in MSCS. Out of the status assessment of biogeochemical mapping, it is recommended to: - Extend the distribution of ammonium to both hindcast and multi-year products for more regions. - Harmonise efforts between different products, with the ultimate aim to have identical variables in all released products. Fourteen new interfacing techniques have been used to successfully improve interfacing of CMEMS into the MSCS. Out of the improvements, four have implemented nudging techniques, three have improved the model numerics or physics, two use new wave modelling techniques and two have changed their technical implementation. The four other improvements use bias corrections, update model surface forcings, use ensemble techniques and initialise nesting in CMEMS to improve the interfacing. In the next steps of the FOCCUS project, as part of Work Package 7, these nesting techniques will be tested and the improvements will be validated to quantify the added value of the improved nesting techniques. The document is structured to provide an overview of the wide array of newly implemented nesting techniques and the MSCS in which they are applied. Section 2 provides an overview of all MSCS used throughout the FOCCUS project, both in a summary table and through individual fact sheets. Section 3 provides the outcomes of subtask 6.1.1, where the frequent users of biogeochemical models describe how CMEMS BGC products are currently implemented and share recommendations on how to improve this. Section 4, which is based off MS6.1 (see Annex), describes the wide variety of newly implemented ways in which the connection between MSCS and CMEMS is strengthened through new nesting techniques. 6 / 52 2. Overview of Member State Coastal Systems (MSCS) This chapter serves as an overview of all Member State Coastal Systems that are being worked on. Until system number 12, the MSCS worked on in this deliverable is described (in blue/white). Afterwards, MSCS are described that are used for other tasks in Work Package 6, not related to nesting improvements (in yellow). Factsheets of all MSCS are presented as reference material. Further details on these MSCS are also available in this Inventory of Coastal Systems (see Annex). Table 2. Overview table of MSCS used in FOCCUS. # MSCS name Lead Coverage (area of interest) CMEMS interface based off MS6.1 (see Annex) Short description of nesting work and expected improvements based off MS6.1 (see Annex) 1. 3D DCSM-FM DELTARES North Sea/Dutch Continental Shelf Global Ocean Physics Reanalysis Global Ocean Biogeochemistry Hindcast - Improving consistency in physics between CMEMS NEMO and 3D DCSM-FM on thermobaricity: taking pressure into account in the Equation of State. This will result in improved stability in the oceanic areas of 3D DCSM-FM. [BETTER] - Replace current SPM forcing with weekly SPM CMEMS satellite, gap-filled with 4DVarNet. This will result in a more accurate weekly SPM field that can be applied to 3D DCSM-FM. [FORC] 2. Norkyst MET Norway/N ERSC From the Kattegat along the Norwegian coast to the Barents Sea Arctic Ocean Physics Analysis and Forecast Baltic Sea Physics Analysis and Forecast - Improving lateral boundary conditions using analysis increments from 4D-Var to assess inconsistencies between Norkyst and the CMEMS models. [BETTER] 3. CROCO-MANG A SHOM French Atlantic coast Global Ocean Physics Analysis and Forecast Global Ocean Physics Ensemble Reanalysis - Use of oceanic ensemble simulations (GLO4ens) at the open boundaries, to quantify the sensitivity of our regional system to uncertainties from remote regions. [ENS] 7 / 52 (not distributed yet) 4. Tolosa-SW SHOM French Atlantic coast Global Ocean Physics Analysis and Forecast Atlantic-Iberian Biscay IrishOcean Physics Analysis and Forecast - Filtering of the non-barotropic dynamics of IBI / GLO products. It is expected that taking into account physical processes not currently represented in Tolosa-SW will improve water level forecasts. [BETTER] - In parallel, development of ML-assisted calibration, trained via IBI datasets in particular, to assess the impact of non-barotropic processes on ssh bias. [BIAIS] 5. GCOAST-GB HEREON German Bight Atlantic - European NWS Physics Analysis and Forecast - The Nesting approach of the GCOAST-GB into the (IBI_MFC) was improved by in addition to apply 3d time series at the boundary line of elements, also to apply relaxation to daily temperature and salinity fields to gain an improved thermohaline field. [NUDG] 6. GCOAST-BS HEREON / MHD Northwestern Black Sea Black Sea Physics Analysis and Forecast Black Sea Waves Analysis and Forecast Black Sea, Bio-Geo-Chemical , L3, daily Satellite Observations - Improved operability using parametric 2D wave spectra computed from CMEMS wave analysis and forecasting products. This approach avoids running a basin-scale wave model (e.g., WW3) to generate spectral forcing for SCHISM. [WAVE] 7. Lazio Coastal Areas CMCC Tyrrhenian Sea CMEMS Mediterranean Sea Analysis and Forecasts -PHY - Addressing the effect of different frequency for lateral boundary forcings to improving the simulation accuracy of the MSCS. [TECHN] 8. DKSS DMI Baltic-North Sea CMEMS NWS MFC Multi-Year - Addressing effects of non-local static sea level impacts at the lateral boundary condition - Reducing initial error in subsurface T/S fields by nudging CMEMS products. [NUDG] 8 / 52 Product-PHY; CMEMS BAL MFC Multi-year Product-PHY; CMEMS NWS MFC Analysis and Forecast - WAV; BAL MFC Analysis and Forecast - WAV; WAV TAC NRT and MYP products; SST TAC NRT and MYP products - Aggregating CMEMS forecasts, national forecasts and satellite observations to provide multi-forecasts based ensemble forecast (SST and Waves). 9. OPTOS v3 RBINS North Sea/ Belgian Coastal Zone CMEMS Atlantic-European North West Shelf-Ocean Physics Reanalysis (NWSHELF_MULTI YEAR_PHY_004_0 09) OSTIA SST - Lateral Boundary condition of salinity/temperature to the first 3D domain in the nested chain of domains, NoS (North Sea). [NEST] [NUDG] - Accurate downscaled salinity and temperature fields at the lateral boundaries of the NoS domain, ensuring smooth and realistic transitions from external forcing data into the nested model domain. 10. SHYFEM-CNR CNR-ISMA R Adriatic Sea CMEMS MED-MFC physical multiyear product - Nudging of T/S MED-REA over the whole Adriatic basin using spatially variable relaxation coefficient. [NUDG] 11. MOHID-LisOcean +Atlantic Tagus and Sado estuarine areas, Portugal Global Ocean Physics Analysis and Forecast Atlantic-Iberian Biscay Irish- - Downscaling wave model using CMEMS IBI Ocean Wave Analysis and Forecast. [WAVE] - Improve the forecast capability of coastal circulation and wave propagation through one-way coupling between a circulation model (MOHID) and a wave model (SWAN). [TECHN] 9 / 52 Projection WGS84 Data assimilation No Update frequency Daily Format CSV (1D) · Binary, VTK (2D) Provider Shom, France External Link https://data.shom.fr/ 2.5. MSCS “GCOAST-GB” (German Bight) INFORMATION Full Name GCOAST-GB Description Based on SCHISM, covers German Coastal Waters Source Numerical models Spatial extent Lat [53.0º to 55.6º] · Lon [5.1º to 10.4º] Spatial resolution Cross-scale 1.5 km - 100 m horizontal resolution Temporal extent From 01/12/2017 Temporal resolution Hourly, Daily Depth levels 21 vertical layers using sigma coordinates Variables Temperature (T) Salinity (S) Sea surface height (SSH) Velocity (UV) Vertical velocity (W) Wave significant height (SWH) Wave mean period (MWT) Wave Peak Period (VTPK) Wave mean direction (VMDR) Suspended matter (SPM) 16 / 52 Feature type Triangular unstructured grid (native, distributed as gridded 0.005 degree) Projection WGS84 Data assimilation No Update frequency Daily Format NetCDF-4 Provider Helmholtz Zentrum Hereon GmbH External Link HCDC TDS Catalog 2.6. MSCS “GCOAST-BS” (Northwestern Black Sea) INFORMATION Full Name GCOAST-BS Description Based on SCHISM, this setup covers the Northwestern Black Sea shelf. The SCHISM is coupled with wave (WWM) and sediment (SED3D) models, providing physical, sea state, and sediment-related variables. Source Numerical models Spatial extent Lat [41.2º to 46.7º] · Lon [27.4º to 31.6º] Spatial resolution Cross-scale 3 km - 100 m horizontal resolution Temporal extent TBD – not operational yet Temporal resolution Hourly, Daily Depth levels LSC² up to 18 vertical coordinates Variables Temperature (T) Salinity (S) Sea surface height (SSH) Velocity (UV) 17 / 52 Vertical velocity (W) Wave significant height (SWH) Wave mean period (MWT) Wave Peak Period (VTPK) Wave mean direction (VMDR) Suspended matter (SPM) Feature type grid/mesh Projection WGS84 Data assimilation No Update frequency TBD – not operational yet Format NetCDF-4 Provider Helmholtz Zentrum Hereon GmbH / MHD External Link Will be made available at The Helmholtz Coastal Data Center (HCDC - https://thredds.hcdc.hereon.de/thredds/catalog/catalog.html ) 2.7. MSCS "Lazio coastal areas" (Italy) INFORMATION Full Name Lazio coastal areas forecasting system Description Coastal ocean forecasting system based on SHYFEM-MPI and WW3 covering the coastal areas of Western Mediterranean. Source Numerical models: SHYFEM-MPI WW3 Spatial extent ~ Lat [41.7º to 42.5º] · Lon [11.2º to 12.2º] Spatial resolution from 2 km off-shore to 30 m along the coast Temporal extent From 2022 18 / 52 Temporal resolution Hourly Depth levels 56 Variables Sea water potential temperature (T) Sea water salinity (S) Sea surface height above geoid (SSH) Velocity (UV) Wave Significant Height (SWH) Wave MeanPeriod (MWT) Wave Mean Direction (VMDR) Feature type Triangular unstructured grid Projection Mercator Data assimilation No Update frequency Daily Format NetCDF-4 Provider CMCC External Link https://civitavecchia.cmcc.it/ 2.8. MSCS "DKSS" (Denmark ) INFORMATION Full Name Danish Storm Surge Model in the Baltic-North Sea Description DKSS provides operational 3D ocean forecasts, with 7 two-way nested domains in the Baltic-North Sea region, with focus on Danish waters. It uses a three-dimensional, two-way nested, coupled ocean-ice model HBM. Source Numerical models: HBM (HIROMB-BOOS Model) Spatial extent ~ Lat [48.5ºN to 65.9ºN] · Lon [4.1ºW to 30.3ºE] Spatial resolution from 5 km off-shore to 185 m in fjørd areas Temporal extent From 2010 Temporal resolution Hourly Depth levels 10-52 z-levels, varying with sub-domains Variables Sea water potential temperature (T) Sea water salinity (S) 19 / 52 Sea surface height above geoid (SSH) Eastward sea water velocity (U) Northward sea water velocity (V) Sea ice concentration (SIC) Sea ice thickness (SIT) Feature type Regular latitude-longitude grid Projection WGS84 Data assimilation No Update frequency 4 times per day Format NetCDF-4 Provider DMI External Link https://ocean.dmi.dk/models/hbm.uk.php 2.9. MSCS "OPTOS v3" (Belgian Coastal Zone ) INFORMATION Full Name Operational Forecasting System for Belgian Coastal zone, OPTOS-BeC Description A multi nested domain coastal forecasting System based on the model COHERENS. Source Numerical Model COHERENS Spatial extent Longitude :- 0.3000° E to 4.6333° E Latitude :- 49.7333° N to 52.7000° N Spatial resolution ~145 m / ~116 m Temporal extent From 2019 20 / 52 Temporal resolution Hourly Depth levels 10 Sigma Layers (BeC) Variables Temperature (T), Salinity (S), Barotropic velocity (UV), Vertical velocity (W), Sea surface height (SSH), Horizontal and Vertical diffusivity Feature type Structured Lat-Long grid Projection WGS-84 Data assimilation No Update frequency daily Format netcdf Provider RBINS External Link None 2.10. MSCS "SHYFEM-CNR" (Italy ) INFORMATION Full Name Shyfem-CNR Adriatic Modelling System Description Shyfem-CNR Adriatic Modelling System is the application of the unstructured SHYFEM model to a computational domain that represents the whole Adriatic Sea, the lagoon of Marano-Grado, the lagoon of Venice and the Po River Delta (including the Scardovari and Goro lagoons). The spatial resolution of the triangular elements varies from 4 km in the 21 / 52 open sea to a few hundred meters along the coast and tens of meters in the inner lagoon and river channels. Source Numerical models: SHYFEMcm-ISMAR Spatial extent Lat [40.0º to 45.8º] · Lon [12.1º to 19.9º] Spatial resolution Cross-scale 4 km - 10m horizontal resolution Temporal extent 2016 Temporal resolution Hourly Depth levels 77 Variables Sea water temperature (T) Sea water salinity (S) Sea surface height (SSH) Eastward sea water velocity (U) Northward sea water velocity (V) Feature type Triangular unstructured grid Projection WGS84 Data assimilation No Update frequency Not operational, used for hindcast simulations Format NetCDF Provider CNR-ISMAR External Link None 2.11. MSCS "MOHID-LisOcean" (Tagus and Sado Estuaries, Portugal) INFORMATION Full Name LisOcean Operational Model Description The 3D LisOcean Operational Model system is based on the MOHID and SWAN numerical models applied for the coastal area covering the Tagus and Sado estuarine areas, in Portugal. The operational models provide a 3-days forecast of hydrodynamic and water properties and 5-days forecast of wave conditions. 22 / 52 Source Numerical models: MOHIDWater and SWAN Spatial extent Lat: 38.16° to 38.96°N Lon: 8.66° to 9.65°W Spatial resolution 280 m Temporal extent Hydrodynamics: 15/10/2023 – up to date Waves: 01/01/2025 – up to date Temporal resolution Hourly 2D fields, 3-Hourly 3D fields; Daily Depth levels 7 Sigma Layers (0-8.68 m) + 37 Cartesian Layers (8.68-3037.72 m) Variables Salinity (S), Temperature (T) Velocity (UV), vertical velocity (W), Sea Surface Height (SSH), Wave Significant Height (SWH), Wave Mean Period (MWT), Wave Mean Direction (VMDR). Feature type Structured regular grid Projection WGS84 Data assimilation None Update frequency Daily Format NetCDF, HDF5 Provider +Atlantic External Link https://thredds.atlanticsense.com/ 2.12. MSCS "AdriFS" (Italy) INFORMATION Full Name Adriatic Sea forecasting system 23 / 52 Description Coastal ocean forecasting system based on SHYFEM-MPI and WW3 covering the coastal areas of Western Mediterranean. Source Numerical models: SHYFEM-MPI WW3 Spatial extent ~ Lat [39º to 45.8º] · Lon [12.1º to 19.8º] Spatial resolution from 2.5 km off-shore to 100 m along the Italian coast Temporal extent From 2022 Temporal resolution Hourly Depth levels 102 Variables Sea water potential temperature (T) Sea water salinity (S) Sea surface height above geoid (SSH) Velocity (UV) Wave Significant Height (SWH) Wave MeanPeriod (MWT) Wave Mean Direction (VMDR) Feature type Triangular unstructured grid Projection Mercator Data assimilation No Update frequency Daily Format NetCDF-4 Provider CMCC External Link https://adri.cmcc.it/ 2.13. MSCS "WMOP" (Spain) INFORMATION Full Name WMOP 24 / 52 Description Assimilative coastal ocean forecasting system based on ROMS with EnOI covering the coastal areas of Western Mediterranean. Source Numerical models: ROMS - The Regional Ocean Modeling System Spatial extent Coastal area · Lat [34.9º to 44.7º] · Lon [-5.7º to 9.1º] Spatial resolution 2 km Temporal extent From 2014 Temporal resolution 3 Hourly · Daily Depth levels 32 Variables Temperature (T) Salinity (S) Sea surface height (SSH) Velocity (UV) Vertical velocity (W) Feature type Grid Projection Mercator Data assimilation Argo profiles, Satellite altimetry, Satellite SST, HF Radar data, Buoy data. Satellite altimetry and SST data from CMEMS are utilized for both data assimilation and model validation. Update frequency Daily Format NetCDF-4 Provider SOCIB [email protected] External Link https://www.socib.es/en/what-we-do/ocean-forecasting/wmo p 25 / 52 based off MS6.1 (see Annex) These changes have been made computationally efficient and are publicly released in the software source code. Expected Results based off MS6.1 (see Annex) The expected results are vertical temperature and salinity profiles closer to the GLORYS12V1 CMEMS ocean model and in situ observations, thereby making the ocean areas within the model domain more stable. Metrics RMSE (Vertical temperature and salinity profiles of GLORYS12V1 data and model data at locations near boundary) Link to material https://doi.org/10.5281/zenodo.15690055 Additional Information: The functionality has been added to D-Flow FM, through incorporation of a new formulation of the Equation of State for the density. The existing EOS-80 formulation has been extended to EOS-83, which includes pressure dependence and thermobaric effects, and the addition has been made computationally efficient. The EOS-80 formulation provides a standardised expression of seawater density, as a function of temperature and salinity at atmospheric pressure. This equation works with the assumption that the pressure is zero, interpreting the output as surface-referenced potential density. This assumptions are valid in shallow waters. By updating the EOS-80 to EOS-83 formulation, the formulation becomes valid when pressure is non-zero (in deeper oceanic areas). This is consistent with modern oceanographic standards and ensures physically realistic behavior. For the full equation breakdown, see the D-Flow Flexible Mesh User manual, linked in the table above. 4.1.2 Suspended Particulate Matter (SPM) forcing from gap-filled CMEMS [FORC] Suspended Particulate Matter (SPM) forcing from gap-filled CMEMS Existing nesting based off MS6.1 (see Annex) The current forcing file to initialise suspended particulate matter is a yearly average of MODIS satellite observations with a cosine function imposed on it to resemble seasonal variations (elevated SPM values in winter and lower SPM values in summer). This forcing resembles a 7-day averaged spatially static sediment field, where the values fluctuate, but the spatial distribution of the sediment plumes remains constant. Improved nesting based off MS6.1 (see Annex) A machine learning algorithm (4DVarNet—double LSTM) has been used to gap-fill satellite observations, resulting in spatially and temporally fluctuating SPM forcing fields based on weekly average CMEMS observations. Expected Results based off MS6.1 (see Annex) The new forcing field is expected to be a more accurate and realistic SPM field, compared to the current forcing. As a result, it is expected to improve chlorophyll-a and primary production 32 / 52 modelling in the MSCS. Metrics RMSE, R2 (Existing forcing and new forcing compared to CMEMS observations) Link to material https://doi.org/10.5281/zenodo.15683195 Additional Information: The gap-filling machine learning algorithm was trained on (a twice repeated) one year of daily modelled Total Inorganic Matter (TIM) data, retrieved from output of the 3D DCSM-FM, when sediment transport was actively modelled. This was considered the ground truth data. The patch data was created by masking the ground truth with clouds from the CMEMS dataset (OCEANCOLOUR_ATL_BGC_L3_MY_009_113). This was done for two years of cloud data, resulting in two years of ground truth and patch data for training, which was then used to train and validate the model on (60% training, 40% validation). The model was tested on a weekly mean of the patch data. To deal with the large domain and resolution of the 3D DCSM-FM, the 4DVarNet model was trained on the ground truth and patch data at a coarse resolution (~ 10 km x ~ 10 km ). To apply the model to high resolution data (~1 km x ~1 km), the data was split into 25 different slices. The coarsely trained model was then applied to each of these slices, and they were merged together. This results in a gapfilled dataset of 1km resolution, with weekly SPM fields. The full workflow, with data details, is described in Figure 4.1.2. Figure 4.1.2: Workflow overview of 4DVarNet Double LSTM gap-filling to create the observation-based SPM forcing. 33 / 52 4.2 Norkyst 4.2.1 Adjustment of hydrography on external boundary condition [BETTER] Adjustment of hydrography on external boundary condition Existing nesting based off MS6.1 (see Annex) Currently we use daily means from the BAL-MFC and the ARC-MFC at the boundaries of Norkyst, adding tides and using the inverted barometer effect in post-processing and online during simulations to adjust the mean sea level. Improved nesting based off MS6.1 (see Annex) We use 4D-Var for data assimilation, and the lateral boundary conditions are part of the control vector. Hence we have an archive of analysis increments at the boundaries that reflect the inconsistencies between the CMEMS and the Norkyst physics. There is no obvious mean bias over time, but a slowly changing difference and often a mismatch in vertical structure. We will apply a correction scheme based on the latest analysis increments to be applied in the forecast cycle. Expected Results based off MS6.1 (see Annex) We expect an improvement in the description of the hydrography, particularly in the open ocean areas. Metrics Insitu salinity and temperature bias and RMSE. Link to material https://github.com/metno/FOCCUS/coastal_system/boundary_co nditions https://thredds.met.no/thredds/projects/foccus.html (see also D6.2 report) https://doi.org/10.5281/zenodo.15656700 Additional Information: Norkyst-DA is the assimilative version of the family of Norkyst models, producing forecasts daily for the Norwegian shelf seas. It uses ROMS-4DVAR (e.g. Iversen et al., 2023) for data assimilation, ingesting satellite SST, insitu salinity/temperature, HF-radar radials during 48h analysis cycles. Forecasts from the ARC-MFC and BAL-MFC are used for lateral boundary conditions, with the latter only used for the portion of the boundary in the Kattegat. ROMS uses sigma type coordinates in contrast to the isopycnal coordinates of the ARC-MFC model HYCOM. ROMS-4DVAR has the ability to adjust the lateral boundary conditions and the surface forcing as part of the assimilation procedure. An assessment of lateral boundary conditions analysis increments shows a slowly varying discrepancy between the internal Norkyst salinity and temperature fields and the imposed boundary conditions. A time varying bias correction will be applied to CMEMS boundary data used in the forecast cycle. This correction will be based on the latest analysis increments, using a prespecified timescale to limit memory in the procedure. An evaluation of the correction scheme will be done for a range of timescales from one day up to a month. 34 / 52 Figure 4.2.1 : Difference between external boundary condition values before and after the 4D-Var analysis, snapshot example on the model grid coordinates along the western boundary. 4.3 CROCO-MANGA 4.3.1 Open boundary perturbed ensemble [ENS] Open boundary perturbed ensemble Existing nesting based off MS6.1 (see Annex) Following the current strategy of the HYCOM-based coastal system, the new CROCO-based coastal system is being set up with GLO-MFC products. Daily averaged 3D horizontal velocities and tracers are imposed with adaptive radiation conditions, and daily averaged barotropic velocities and sea surface height are imposed with Flather-type conditions. Inverse barometer effect from atmospheric pressure (ECMWF-ERA5) is used to correct free surface elevation. Improved nesting based off MS6.1 (see Annex) Use of Global Ocean Physics Ensemble Reanalysis (GLO4ens; provided by MOi but not distributed yet at CMEMS) to produce regional ocean ensembles with consistent open boundary perturbations. Expected Results based off MS6.1 (see Annex) Capability to quantify the sensitivity of our regional system to uncertainties from remote regions. 35 / 52 Metrics Surface currents magnitude and direction with probabilistic approaches (e.g. ensemble mean and spread, continuous rank probability score (CRPS)). Link to material https://doi.org/10.5281/zenodo.15526387 Additional Information: Our interfacing improvements aim at leveraging ensemble estimates of ocean state at global scale to force our coastal system at the boundaries. This is done with the Global Ocean Physics Ensemble Reanalysis (GLO4ens) currently developed at Mercator Ocean (MOi). It is a NEMO-based, ¼o ocean ensemble reanalysis providing 17-member ensemble-based probabilistic daily estimates of ocean state through model grid, sea-ice lateral melt and air temperature perturbations based on Brankart et al. (2015) and Leroux et al. (2022) methods. From native NEMO model outputs provided by MOi, we converted this ensemble dataset into a CMEMS-like dataset in order to fit with the data format expected from our preprocessing tools used to derive open boundary conditions (i.e. https://croco-ocean.gitlabpages.inria.fr/croco_pytools/prepro/). Our coastal system being currently interfaced with GLO12 products, a first consistency check consisted in comparing tracer transports across the boundaries (Fig. 1). It turns out GLO4ens systematically deviates from GLO12, the latter being only marginally included in the ensemble envelope of GLO4ens and showing a less pronounced variability over the considered period (Marsh 2019) . We are currently investigating solutions to leverage these ensemble estimates in a consistent way regarding these discrepancies. This effort in producing boundary perturbed ensemble simulation complement the other ensemble approaches reported in Deliverable 6.2 “Processing Techniques”. 36 / 52 Figure 4.3.1: Integrated temperature (left) and salinity (right) transport across the three open boundaries of our regional system as derived from GLO12 (black) and GLO4ens (green). 4.4 Tolosa-SW 4.4.1 Filtering baroclinic dynamics [BETTER] Filtering baroclinic dynamic Existing nesting based off MS6.1 (see Annex) The latest version of the French surge forecasting system is based on the shallow-water barotropic configuration of the Tolosa toolbox, computationally efficient to meet operational constraints. This implies that physical baroclinic processes such as waves and height/currents interactions (including wave setup) are not represented, and their impact on the simulated water levels is neglected. Improved nesting based off MS6.1 (see Annex) The first approach consists of filtering the dynamics of IBI-ANFC and GLO-ANFC to extract baroclinic processes, such as dynamic variations in sea level or steric variations in sea level, excluding tides and atmospheric forcings. Expected Results based off MS6.1 (see Annex) Assessing the impact of unrepresented physical processes on simulated water levels will improve the representation of physics within the system, thereby increasing the accuracy of the forecasting system. Corrections are expected for the 2D fields, but 37 / 52 the incorrect representation of the wave setup will not be integrated. Metrics BIAS, MAE, RMSE, SD Link to material https://doi.org/10.5281/zenodo.15537127 4.4.2 ML-assisted calibration method [BIAIS] ML-assisted calibration method Existing nesting based off MS6.1 (see Annex) The latest version of the French surge forecasting system is based on the shallow-water barotropic configuration of the Tolosa toolbox, computationally efficient to meet operational constraints. This implies that the physical baroclinic processes such as waves and height/currents interactions (including wave setup) are not represented, and their impact on the simulated water levels is neglected. Improved nesting based off MS6.1 (see Annex) The second approach is based on the development of a machine learning-assisted calibration method. Several algorithms are considered, such as Gradient Boosting or Long Short-Term Memory. These models use predictors related to 3D dynamics of CMEMS IBI-MFC outputs (such as sea surface temperature, eastwards / northwards velocity at the surface and sea surface height), together with sea state (Wave significant height and wave mean period) and tide gauges data (sea surface height). Expected Results based off MS6.1 (see Annex) Estimating the impact of unrepresented physical processes on simulated water levels through Machine Learning will reduce the bias on the predicted storm surge and improve the accuracy of the forecasting system. The corrective method will be applied on 1D time series only, and will take account of all 3D effects and wave/heights/currents interactions effects. Metrics BIAS, MAE, RMSE, SD Link to material https://doi.org/10.5281/zenodo.15537127 Additional Information: The French storm surge forecasting system, operated by Météo-France, uses the shallow water version of the Tolosa toolbox. It solves the non-linear shallow-water equations with source terms corresponding to bottom dissipation, tidal forcing, atmospheric pressure and surface wind. These equations are solved using the finite volume method, which enables the use of unstructured grids to achieve much finer resolutions. Consequently, the resolution varies spatially, ranging from 100 m in the English Channel and along the Atlantic coast to approximately 10-20 km in the northern part of the computational domain, off the Norwegian coast. The model uses low-dissipation 38 / 52 numerical schemes, which have been demonstrated to effectively calculate non-linear flows in the low-Mach (i.e. low-speed) regime (Couderc et al. 2017). The open boundaries of the domain are forced with water levels from the FES2014b tidal model, and the tidal potential is also included in all simulations. The other domain boundaries are treated as vertical walls (zero flow), except for those corresponding to rivers for which flows are imposed, whether constant or not. Wind stress is represented in the model by a Charnock formulation with a constant drag coefficient. Finally, the effects of atmospheric pressure on water levels are represented by the inverse barometer. 4.5 GCOAST-GB 4.5.1 Extended nesting via spongelayer relaxation [NUDG] Extended nesting via spongelayer relaxation Existing nesting based off MS6.1 (see Annex) The nesting approach is based on the downscaling from CMEMS products for the North West Shelf. The existing methodology uses NWS-MFC sea level, temperature, salinity and 3D currents at hourly frequency as lateral boundary conditions. Improved nesting based off MS6.1 (see Annex) In addition to the lateral boundary forcing larger domain nudging is deployed relaxing to T/S fields over a larger area. Also A similar approach for the nudging towards Sea surface SPM will be explored based on satellite Observations making a step towards data assimilation. Expected Results based off MS6.1 (see Annex) The T/S nudging is expected to improve hydrodynamics and coast-to-open-sea dynamics while the SPM nudging should help increase accuracy of matter simulations. Metrics bias, rmse, correlation Link to material Existing nesting (updated for FOCCUS): https://doi.org/10.5281/zenodo.15689870 Nudging: https://doi.org/10.5281/zenodo.15689988 Additional Information: Classical boundary condition forcing is applied along the outermost row of grid elements for temperature, salinity, sea surface elevation, and horizontal currents. These boundary conditions typically involve time-varying data from a larger-scale model or observational dataset. In extended simulations, internal model dynamics can lead to systematic drift, especially in tracer fields such as temperature and salinity. To counteract this, a relaxation (or nudging) scheme is implemented over a buffer zone extending approximately 20 km into the model domain from the open boundaries. This buffer typically corresponds to about 10 grid elements (indicated by the reddish area in Fig. 4.5.1). 39 / 52 Within this zone, the model solution is gently relaxed toward the external (parent) model values The temporal evolution of any scalar tracer variable (e.g., temperature or salinity) is modified according to the relaxation equation: where is the simulated value within the nested model, is the corresponding value from the outer (CMEMS NWS-MFC) model, Δt is the model time step, γ is a dimensionless relaxation coefficient, scaled such that the maximum nudging timescale is limited to 1 day (i.e., γ=1 corresponds to full relaxation within one day). This formulation introduces a source term that reduces discrepancies with the parent model over time. By tapering the nudging strength inward from the boundary, it ensures numerical stability and maintains large-scale consistency while preserving interior small-scale dynamics. Figure 4.5.1: Distribution of nudging strength factor γ. 4.6 GCOAST-BS 4.6.1 2D Wave Spectra forcing computed from CMEMS parameters [WAVE] 2D Wave Spectra forcing computed from CMEMS parameters Existing nesting based off MS6.1 (see Annex) Wave spectra are unavailable through CMEMS, and other data sources that provide wave spectra do not present enough resolution in the Black Sea (e.g., IFREMER WW3 hindcast). In this way, we need to run a basin-scale wave model only to produce spectral forcing at a few points to force our model. This required step before the target run demands time and computational resources, hampering the operability of GCOAST-BS. 40 / 52 Improved nesting based off MS6.1 (see Annex) We use wave parameters from CMEMS wave products to compute parametric 2D wave spectra for each time step at the model open boundary. The spectrum is calculated using the JONSWAP and Cartwright approach. Expected Results based off MS6.1 (see Annex) The parametric wave spectra will improve the GCOAST-BS workflow and allow a better interface with CMEMS. Additionally, this approach enables the use of the high-quality CMEMS Black Sea wave products, which assimilate data and are extensively validated. Despite its simplifications compared to the full wave spectrum, the parametric option has performed well when validated with observations in coastal areas. Metrics bias, RMSE, Pearson correlation Link to material The wave spectra is computed using wavespectra Python package, a Jupyter Notebook for demonstration is available at: https://doi.org/10.5281/zenodo.15690198 Additional Information: The 2D Wave Spectra is computed based on the CMEMS wave analysis and forecast parameters to generate spectral boundary conditions for the wave module of GCOAST-BS (WWM). The method gets the wave parameters available at CMEMS (wave significant wave height - SWH, wave mean direction - VMDR, and wave peak period - VTPK) on the open boundary points of the computational mesh. The directional wave spectra are then reconstructed at those points by applying the JONSWAP spectrum for frequency shape and the Cartwright method for directional spreading. The resulting 2D spectra (frequency and direction) at each boundary point and time step are stored and used as boundary conditions. 4.7 Lazio Coastal Areas 4.7.1 Increasing temporal frequency for open lateral boundary [TECHN] Increasing temporal frequency for open lateral boundary Existing nesting based off MS6.1 (see Annex) The nesting approach is based on the downscaling from CMEMS products for the Mediterranean Sea. The existing methodology uses MED-MFC sea level at hourly frequency, and 3D daily current velocity, salinity and water temperature are used as lateral boundary conditions. Improved nesting based off MS6.1 (see Annex) The new methodology for nesting the MSCS into the MED-PHY framework uses three-dimensional hourly data as lateral boundary conditions for all the variables. This includes current velocity, salinity, temperature, and sea level, thereby enhancing the temporal consistency and dynamical coherence of the nested simulations with respect to the large-scale circulation patterns 41 / 52 4.12. AdriFS 4.12.1 Wave Spectra lateral boundary rebuild from CMEMS parameters [WAVE] Wave Spectra lateral boundary rebuild from CMEMS parameters Existing nesting based off MS6.1 (see Annex) Wave spectra are not directly provided by CMEMS products, primarily due to the substantial disk space required for their production and storage. As a result, a spectral approximation strategy is adopted for nesting the AdriFS model within the MED-WAV CMEMS product. In this approach, wave spectra are reconstructed from standard integrated wave parameters—such as significant wave height, peak period, and mean direction—by assuming a JONSWAP spectral shape and applying the directional spreading formulation proposed by Yamaguchi (1984). This method provides a reasonable approximation of the incoming wave energy. However, it does not account for spectral partitioning, meaning that the representation of complex sea states composed of multiple wave systems (e.g., wind sea and swell) remains limited. Improved nesting based off MS6.1 (see Annex) The new method allows for the reconstruction of multiple spectral components. This enables a more realistic representation of complex sea states, including the coexistence of wind sea and swell systems. As a result, the input wave conditions used to drive the AdriFS model better reflect the true variability and structure of the wave field. Expected Results based off MS6.1 (see Annex) The use of multi-partitioned wave spectra is expected to improve the accuracy of wave simulations within the AdriFS forecasting system, particularly under conditions where the wave field exhibits multiple spectral peaks. This leads to more reliable forecasts, especially during complex sea state events, such as storms or rapidly changing wind conditions. Metrics BIAS, RMSE, Pearson correlation Link to material https://doi.org/10.5281/zenodo.15637958 Additional Information: The approach we present for downscaling the WW3 wave model in the MSCS focuses on a multi-partitioned method for reconstructing wave spectral boundary conditions. This method is evaluated against two alternative approaches: full spectral forcing and a standard single-peak reconstruction. The three methods are outlined below: · Full Spectra Forcing (Reference Case): This method provides the most accurate representation of wave conditions, as it retains the full spectral energy distribution. However, it requires a parent large-scale model to supply spectral boundary data, leading to high storage demands and limited practical availability. 48 / 52 · Rebuilding Spectra from Standard Mean Wave Parameters (Single-Peak Approach): In this simplified downscaling method, the directional wave spectrum is reconstructed using only standard wave statistics — significant wave height (Hs), peak period (Tp), and mean wave direction (mDir). A single spectral peak is assumed, using an approximation method that provides a reasonable estimate for the zeroth spectral moment. While computationally efficient and easy to implement, this approach does not capture the complexity of multi-modal sea states. · Rebuilding Spectra from Partitioning (Multi-Partition Approach): This method improves the spectral representation by reconstructing several distinct wave systems (e.g., wind-sea and 2–3 swell components). Each partition is modelled using the Yamaguchi (1984) approximation and combined to form a complete, multi-peaked spectrum. This reflects the physical nature of ocean waves as a superposition of monochromatic wave components traveling at different frequencies and directions, as described by the dispersion relation. An example of the resulting spectra from each method is shown in figure 4.12.1, using polar plots. These plots highlight the differences in directional energy distribution between the reference and the singleand multi-partition approaches. Figure 4.12.1 Comparison of the reference spectrum with spectra reconstructed using the multi-partition and single-peak methods for 2 random time steps (A and B). This MSCS is also part of the evolution related to ensemble methods. Link for the software is in the table above, in the link to material. 4.13 Conclusion and Next Steps 49 / 52 To conclude, the numerous improvements described above have been implemented to better integrate CMEMS products into the various MSCS. Out of all improvements, four have implemented nudging techniques, three have improved the model numerics or physics, two use new wave modelling techniques and two have changed their technical implementation. The four other improvements use bias correction, update model surface forcings, use ensemble techniques or initialise nesting in CMEMS to improve the interfacing. Proof of implementation can be found in the corresponding sections and in D6.2. Also described here are the metrics, with which further steps will be taken. The next steps are to test the new nesting techniques and validate the improvement to quantify the added value of the improved nesting techniques. This will be done using the metrics described in each improvement, over the duration of Work Package 7. 50 / 52 5. Annex MSCS Inventory: A preliminary inventory of coastal systems was agreed upon and created by the Consortium during the first year of the project, in order to better organize efforts to achieve the goals laid out in the DoA to improve them. A summary of this document can be found below, as it is not available on the EU Portal or for public dissemination. Additionally, the Commission has the right to request access to this document directly using the project’s shared drive here: https://docs.google.com/spreadsheets/d/1JxFKYBhboVHDFwAgmMikijJ_9N2UFUu6xNmFOAGB21 M/edit?usp=sharing This inventory includes a list of project partners, their relevant WPs and regions of interest, in addition to details about their coastal systems which are being used and developed in this project per the DoA. This includes the name of the systems, objectives for improvement, models used, current observations and forcing data, and how these systems interface with Copernicus Marine. Additionally, cross-WP connections for each coastal system are listed here, specifically how each has links to hydrology. Links to initial performance metrics for each coastal system are included here, in addition to roadmaps for improving each system’s interfacing with Copernicus (detailed further below). Milestone MS6.1: Interface Roadmaps for Improved Interfacing with CMEMS: Roadmaps for improving each coastal system’s interfacing with Copernicus was created in M6 of the project. A summary of this milestone can be found below, as it is not available on the EU Portal or for public dissemination. Additionally, the Commission has the right to request access to the documents that make up this milestone directly using the project’s shared drive here: https://drive.google.com/drive/folders/1tlmc7NZxheF69ls86eIyiv3eLRsgggPH?usp=drive_link These roadmaps include a summary of each member state coastal system, the Copernicus Marine model products used, as well as a description of the existing nesting method used. Crucially, these roadmaps detail for each coastal system the proposed improvements in this nesting, and will be used for the work outlined in the DoA in WP6/7. 51 / 52 6. References Agatova, A. I., Lapina, N. M., & Torgunova, N. I. (2008). Organic matter of the North Atlantic. Oceanology, 48(2), 182–195. https://doi.org/10.1134/S0001437008020045 Brankart, J. M., Candille, G., Garnier, F., Calone, C., Melet, A., Bouttier, P. A., ... & Verron, J. (2015). A generic approach to explicit simulation of uncertainty in the NEMO ocean model. Geoscientific Model Development, 8(5), 1285-1297. https://doi.org/10.5194/gmd-8-1285-2015 Brzezinski, M. A. (1985). THE Si:C:N RATIO OF MARINE DIATOMS: INTERSPECIFIC VARIABILITY AND THE EFFECT OF SOME ENVIRONMENTAL VARIABLES. Journal of Phycology, 21(3), 347–357. https://doi.org/10.1111/j.0022-3646.1985.00347.x Couderc, F., Duran, A., Vila, J.-P. (2017). An explicit asymptotic preserving low Froude scheme for the multilayer shallow water model with density stratification. Journal of Computational Physics 343 (Supplement C) (2017) 235 – 270. https://doi.org/10.1016/j.jcp.2017.04.018 Iversen, S. C., Sperrevik, A. K., & Goux, O. (2023). Improving sea surface temperature in a regional ocean model through refined sea surface temperature assimilation. Ocean Science, 19(3), 729-744. https://doi.org/10.5194/os-19-729-2023 Leroux, S., Brankart, J. M., Albert, A., Brodeau, L., Molines, J. M., Jamet, Q., ... & Brasseur, P. (2022). Ensemble quantification of short-term predictability of the ocean dynamics at a kilometric-scale resolution: a Western Mediterranean test case. Ocean Science, 18(6), 1619-1644. https://doi.org/10.5194/os-18-1619-2022 Letscher, R. T., Moore, J. K., Teng, Y.-C., & Primeau, F. (2015). Variable C : N : P stoichiometry of dissolved organic matter cycling in the Community Earth System Model. Biogeosciences, 12(1), 209–221. https://doi.org/10.5194/bg-12-209-2015 Moore, Andrew M., Hernan G. Arango, Gregoire Broquet, Brian S. Powell, Anthony T. Weaver, and Javier Zavala-Garay. (2011), "The Regional Ocean Modeling System (ROMS) 4-dimensional variational data assimilation systems: Part I–System overview and formulation." Progress in Oceanography 91(1), 34-49. https://doi.org/10.1016/j.pocean.2011.05.004 Redfield, A. C. (1934). On the proportions of organic derivatives in sea water and their relation to the composition of plankton. In James Johnston Memorial Volume (pp. 176–192). University Press of Liverpool. 52 / 52