Advanced search
CrossRef citations to date
Research Article

Copernicus Marine Service Ocean State Report, Issue 5

(Editor), (Editor), (Review Editor), (Review Editor), (Review Editor), (Review Editor), (Review Editor), , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , & show all

Chapter 1: CMEMS OSR5     1

 1.1 Introduction

Karina von Schuckmann and Pierre-Yves Le Traon     1

 1.2 Knowledge and data for international Ocean governance

Paula Kellett, Brittany E. Alexander and Johanna J. Heymans     6

CMEMS OSR5, Chapter 2     10

 2.1 Modelled sea-ice volume and area transport from the Arctic Ocean to the Nordic and Barents Seas

Vidar S. Lien, Roshin P. Raj and Sourav Chatterjee     10

 2.2 Ocean heat content in the High North

Michael Mayer, Vidar S. Lien, Kjell Arne Mork, Karina von Schuckmann, Maeva Monier and Eric Greiner     17

 2.3 Declining silicate and nitrate concentrations in the northern North Atlantic

Kjell Gundersen, Vidar S. Lien, Jane S. Møgster, Jan Even Øie Nilsen and Håvard Vindenes (IMR)     23

 2.4 Eutrophic and oligotrophic indicators for the North Atlantic Ocean

Silvia Pardo, Shubha Sathyendranath and Trevor Platt     31

 2.5 Nitrate, ammonium and phosphate pools in the Baltic Sea

Mariliis Kõuts, Ilja Maljutenko, Ye Liu and Urmas Raudsepp     37

 2.6 Long term changes monitored in two Mediterranean Channels

Sana Ben Ismail, Katrin Schroeder, Jacopo Chiggiato, Stefania Sparnocchia and Mireno Borghini     48

 2.7 Interannual variations of the Black Sea Rim Current

Elisaveta Peneva, Emil Stanev, Stefania Ciliberti, Leonardo Lima, Ali Aydogdu, Veselka Marinova and Nadejda Valcheva     53

 2.8 Climatology and 2019 anomaly of maximum waves in the Mediterranean and Black Seas

Alvise Benetazzo, Francesco Barbariol, Joanna Staneva, Silvio Davison, Antonio Ricchi, Arno Behrens, Gerhard Gayer and Paolo Pezzutto     59

 2.9 Strong positive Indian Ocean Dipole events over the period 1993–2019

Simon Good, Jean-François Legeais and Richard Graham     65

CMEMS OSR5, Chapter 3     82

 3.1 The chlorophyll-a gradient as primary Earth observation index of marine ecosystem feeding capacity

Jean-Noël Druon, Antoine Mangin, Pierre Hélaouët and Andreas Palialexis     82

 3.2 Marine heatwaves and cold-spells, and their impact on fisheries in the North Sea

Sarah Wakelin, Bryony Townhill, Georg Engelhard, Jason Holt and Richard Renshaw     91

 3.3 Massive occurrence of the jellyfish Portuguese Man-of-War in the Mediterranean Sea: Implication for coastal management

Laura Prieto, Diego Macías, José L. Oviedo, Mélanie Juza and Javier Ruiz     96

 3.4 Recent changes of the salinity distribution and zooplankton community in the South Adriatic Pit

Elena Mauri, Milena Menna, Rade Garić, Mirna Batistić, Simone Libralato, Giulio Notarstefano, Riccardo Martellucci, Riccardo Gerin, Annunziata Pirro, Marijana Hure and Pierre-Marie Poulain     102

 3.5 Delivering high quality sea-ice information around the Svalbard archipelago to marine end-users

Cyril Palerme, Panagiotis Kountouris, Alexandra Stocker, Maaike Knol-Kauffman, Paul Cochrane, Malte Müller and Lasse Rabenstein     109

 3.6 Developing spatial distribution models for demersal species by the integration of trawl surveys data and relevant ocean variables

Panzeri D., Bitetto I., Carlucci R., Cipriano G., Cossarini G., D'Andrea L., Masnadi F., Querin S., Reale M., Russo T., Scarcella G., Spedicato M.T., Teruzzi A., Vrgoč N., Zupa W. and Libralato S.     114

 3.7 A benthic hypoxia index (BHindex) for assessing the Good Environmental Status of the Black Sea's north-western shelf waters

Arthur Capet, Luc Vandenbulcke, Catherine Meulders and Marilaure Grégoire     123

CMEMS OSR5, Chapter 4     140

 4.1 Sea-ice and ocean conditions surprisingly normal in the Svalbard-Barents Sea region after large sea-ice inflows in 2019

Signe Aaboe, Sigrid Lind, Stefan Hendricks, Emily Down, Thomas Lavergne and Robert Ricker     140

 4.2 Monitoring storms by merged data sources for the Malta shelf area in 2019

Aldo Drago, Adam Gauci, Joel Azzopardi, Jorge Sanchez and Andrés Alonso-Martirena     148

 4.3 The November 2019 record high water levels in Venice, Italy

Rianne Giesen, Emanuela Clementi, Marco Bajo, Ivan Federico, Ad Stoffelen and Rosalia Santoleri     156

 4.4 Extreme waves and low sea level during the storm in the Gulf of Bothnia, Baltic Sea

Urmas Raudsepp, Aarne Männik, Ilja Maljutenko, Priidik Lagemaa, Sander Rikka, Victor Alari and Rivo Uiboupin     162

 4.5 Establishment of Pterois miles (Bennett, 1828) in the Ionian Sea

Laura Bray and Dimitris Kassis     173

CMEMS OSR5 – Chapter 1

1.1. Introduction

Authors: Karina von Schuckmann, Pierre-Yves Le Traon

1.1.1. Copernicus Marine Service status and achievements

The first operational phase 2014–2021 of the Copernicus Marine Service has successfully implemented a unique European Union ocean monitoring and forecasting service ( Thirty thousand expert downstream services and users are now connected to the service that responds to public and private user needs and policies related to all marine and maritime sectors: maritime safety, coastal environment monitoring, trade and marine navigation, fishery, aquaculture, marine renewable energy, marine conservation and biodiversity, ocean health, climate and climate adaptation, recreation, education, science and innovation. The Copernicus Marine Service organises the value chain that goes from observation to information (Le Traon et al. ) and is an essential tool for Ocean Governance (Section 1.4) and sustainable management of the ocean based on comprehensive ocean monitoring and forecasting capabilities.

The Copernicus Marine Service is unique by its coverage and comprehensiveness, its balance between state-of-the-art science and operational commitments, and the consistency of its portfolio where satellite observations, in situ observations, and model simulations are used coherently to describe the physical (blue), biogeochemical (green) ocean and sea-ice (white) state the European regional seas and the global ocean. The Copernicus Marine Service gathers a strong network of European ocean information producers. Thanks to a well-established and organised evolutions of the Copernicus Marine Service system of systems, the capabilities to operate a marine service responsive to user needs and scientific/technological advances have fully been demonstrated (CMEMS General Assembly ). The product and service portfolios have evolved from 2015 to 2021 with, in particular, the integration of new parameters (e.g. waves, carbon, turbidity, sea ice thickness and icebergs), the improvement of resolution and quality, longer time series, the full uptake of Sentinel missions (S1, S3 and recently S2) and the development of new means to access and visualise the data.

The Copernicus Marine Service provides invaluable observation and model products to assess and report on past and present marine environmental conditions and to analyse and interpret changes and trends in the marine environment as for example discussed in previous Ocean State Reports,, and as part of the Ocean Monitoring Indicator framework, The CMEMS Ocean State Reports and Ocean Monitoring Indicators provide, in particular, a unique ocean monitoring dashboard for policy and decision makers as well as for the general public to support actions and assess progresses in policy implementation. There have been excellent feedbacks on the annual Ocean State Reports and their high level summaries that are now part of the EU ocean state assessment landscape and provide a high visibility of the Copernicus Marine Service. They have also federated a unique pooling of EU scientific expertise to assess the state of the ocean based on the Copernicus Marine Service ocean monitoring products.

A remaining challege is to establish a comprehensive monitoring of the ocean, a challenge that demands international cooperation. In response, the Copernicus Marine Service has set up important partnerships with GOOS, OceanPredict, GEO and GEO Blue Planet. The UN Decade of Ocean Science for Sustainable Development ( will be a unique opportunity to develop further the required international cooperation to support delivery of the information, action and solutions needed to achieve the 2030 Agenda for Sustainable Development.

1.1.2. Plans for Copernicus 2

The Copernicus Marine Service has developed an ambitious plan for the next phase of the Copernicus programme (Copernicus 2). The objective is to further establish CMEMS products as a worldwide reference, foster further the service uptake and respond to increasing and pressing user and policy needs (in particular the EU Green Deal – see Chapter 1 in the 4th issue of the CMEMS Ocean State Report, Peterlin et al. ) for improved ocean monitoring and prediction capabilities.

The plan identifies three levels of implementation for the evolution of the Copernicus Marine Service product and service portfolio over the period 2021–2027: baseline (continuity of service with incremental evolutions), enhanced continuity (major product improvements) and new services.

Baseline will be implemented from the start of Copernicus 2 to ensure the continuity of the present service and maintain a consistent blue, white and green offer. This includes incremental evolutions to improve product quality, integrate future Sentinel missions and new in-situ observations (e.g. Biogeochemical Argo, Claustre et al. ), to improve the estimation of product uncertainty and benefit from new capabilities of digital services through the WEkEO DIAS platform ( User interaction and user engagement will be strengthened by developing dedicated sectorial offers per applications and policies and enhancing the training and capacity building offer. The objective is also to re-enforce the Copernicus programmme consistency by producing marine data for the other Copernicus services and developing sectorial approaches (thematic hubs) with the other Copernicus Services (e.g. Coastal, Arctic).

The enhanced continuity and new services streams will build from present and future H2020 and Horizon Europe R&D projects and will be developed depending on budget and priorities. Improved digital services, ensemble forecasts, higher resolution, step change in Arctic monitoring, air/sea CO2 fluxes, twentieth century reanalyses are proposed under the enhanced continuity scenario. Coastal, marine biology, climate projection (coastal, ecosystem) are proposed under the new services scenario. A strong priority is, in particular, to offer new services for the coastal ocean through a co-design and co-development approach between the Copernicus Marine Service and coastal marine services operated by EU Member States.

The Copernicus Marine Service Ocean State Reports and Ocean Monitoring Indicators will continue to be an essential component of the Copernicus Marine Service in Copernicus 2. The plan is to foster further the integration with other Copernicus Services and, in particular, with the Climate Change Service ( to provide an integrated assessment of the state of ocean and climate. Links with international activities and organisations related to ocean and climate assessments (Intergovernmental Panel on Climate Change, Intergovernmental Oceanographic Commission; World Meteorological Organisation; UN World Ocean Assessment, Global Ocean Observing System, Global Climate Observing System, G7 Future of the Seas and Oceans Initiative, UN Decade of Ocean Science) will be strengthened. The Copernicus Marine Service could contribute, in particular, to the development of an international framework for ocean monitoring indicators. The delivery of meaningful Ocean Monitoring Indicators requires an adequate ocean observing system. This dependency will be documented. This is essential to advocate for the sustainability of the ocean observing system.

1.1.3. Major outcomes of the 5th issue of the Ocean State Report

The 5th issue of the CMEMS OSR incorporates a large range of topics for the blue, white and green ocean for all European regional seas, and the global scale over 1993–2019 with a special focus on 2019. As previous reports, this report is organised within four principal chapters:

  • Chapter 1 provides the introduction and a synthesised overview, together with an informative section on ocean governance written in collaboration with the European Marine Board (

  • Chapter 2 includes various novel scientific analyses of the ocean state variability at subseasonal, seasonal and multi-annual scales.

  • Chapter 3 connects science and policy by reporting science cases of (potential) socio-economic relavance.

  • Chapter 4 highlights unusual events during the year 2019.

The reporting and indicators are focused on the seven Copernicus Marine Service regions, i.e. the global ocean, the Arctic, the North-West-Shelf, the Iberia-Biscay-Ireland, the Baltic Sea, the Mediterranean Sea and the Black Sea. The uncertainty assessment based on a ‘multi-product-approach’ is also used here (see von Schuckmann et al., 2018 for more details). The OSR is predominantly based on CMEMS products, and many analyses are complemented by additional datasets. CMEMS includes both satellite and in-situ high level products prepared by the Thematic Assembly Centres (TACs) – including reprocessed products – and modelling and data assimilation products prepared by Monitoring and Forecasting Centres (MFCs). Products are described in Product User Manuals (PUMs) and their quality in the Quality Information Documents (QUID; CMEMS ). Within this report, all CMEMS products used are cited by their product name, and download links to corresponding QUID and PUM documents are provided. The use of other products has also been documented to provide further links to their product information, and data source.

The major outcomes of the fifth issue of the Copernicus Marine Service Ocean State Report are synthesised in Figure 1.1.1, and are summarised below.

Figure 1.1.1. Overview on major outcomes of this 5th issue of the CMEMS Ocean State Report.

Global / Large scale:

Investigations in OSR5 address three topics at large to global ocean scale: the development of an indicator to monitor the eutrophic and oligotrophic state in the ocean; a chlorophyll-a based indicator linked to a plankton-to-fish index as well as an improved indicator for an air–sea coupled mode of variability in the Indian Ocean.

Eutrophication is the process by which an excess of nutrients (mainly phosphorus and nitrogen) leads to increased growth of plant material in an aquatic body – an issue particularly of relevance in coastal regions and areas with restricted water flow. Eutrophication can be linked to anthropogenic activities, such as farming, agriculture, aquaculture, industry and sewage, and results in decreased water quality through enhanced plant growth (e.g. algal blooms) causing death by hypoxia of aquatic organisms. Oligotrophication is the opposite of eutrophication, where reduction in some limiting resource leads to a decrease in photosynthesis by aquatic plants, which might in turn reduce the capacity of the ecosystem to sustain the higher organisms in it. A new indicator of eutrophic and oligotrophic waters proposed in OSR5 derived from satellite chlorophyll-a data (Section 2.4) showed hardly any localities in the North Atlantic where the eutrophic flag was positive in 2019 (i.e. above the 1993–2017 P90 climatological reference). Oligotrophic flags were positive mostly along coastal waters, but also along scattered points within the 30–40°N latitudes. Waters flagged as eutrophic can be then classified as eutrophication or oligotrophication when the eutrophic state is sustained over several years, such as a significant trend over time. This indicator methodology has been distributed to EuroStat in the context of SDG14.1a eutrophication reporting over the period 1998–2019 for all European Seas.

The horizontal gradient of chlorophyll-a derived from remote sensing chl-a data and linked to a plankton-to-fish index (Section 3.1) has been shown to be highly valuable to marine biologists and ecosystem modellers and, in turn, to regional fisheries management and authorities facing overexploitation and the effects of climate change. Marine policies will ultimately be efficiently supported by the use of chlorophyll-a gradient as a direct, observation-based, biological variable monitoring the marine ecosystem productivity across a wide range of spatial and temporal scales.

The Indian Ocean Dipole (IOD) is an air–sea coupled mode of variability in the Indian Ocean exacerbating moderate to extreme variations at the air–sea interface such as precipitation and ocean hydrography changes. OSR5 has analysed the classical IOD index – the Dipole Mode Index – based on Sea Surface Temperature (SST), and complemented the analysis with a sea level-based indicator demonstrating the increased performance of IOD monitoring based on a combined use of both indexes (Section 2.9). Results report on two particularly strong events in 1997 and 2019, inducing drought periods in the land areas bordering the eastern Indian Ocean, and extreme precipitation in the western part of the basin.

The North-Atlantic / Arctic gateway

Sea-ice conditions in the North-Atlantic / Artic gateway strongly impact ecosystem, weather, and economic activities, such as tourism, fisheries, and shipping. OSR5 has particularly emphasised blue, white and green oceanographic conditions at Svalbard, the Barent Sea and other areas of the Nordic Seas. The Fram Strait represents the major gateway for sea ice transport between the Arctic and North Atlantic Ocean, affecting the mass balance of the perennial ice cover in the Arctic. Contrary to previous results, OSR5 finds a significant, negative trend in the sea-ice area export through this strait over the last two and a half decades (1993–2019) as affected by a strong reduction of nearly 90% in average sea-ice thickness in the Barents Sea (Section 2.1).

In 2019 however, sea-ice conditions had been surprisingly normal around Svalbard and parts of the Barents Sea, albeit concurrent unusual low sea-ice extent in summer and autumn in the Arctic ocean. The OSR5 results (Section 4.1) have shown that sea-ice (old, and thick) redistribution from the Artic into this area have acted to recover the sea-ice cover and ocean stratification through adding sea ice and freshwater to the region. This supports that large sea-ice inflows act to maintain an Arctic-type ocean climate with a cold, stratified, and sea-ice covered water column and is a key player among others in the Arctic climate system. OSR5 also introduces a new tool (IcySea) providing near-real time monitoring (satellite images from the Sentinel 1 satellites) and sea-ice drift forecasts in the Svalbard area to inform operational planning and safety for the transport and navigation economic sector (Section 3.6).

OSR5 has also addressed Artic ocean warming over the past decades (1993–2019) – a critical missing piece of knowledge for global scale ocean warming linked to the current positive Earth energy imbalance: Currently, global ocean warming estimates are limited between 60°S and 60°N and less is known on the role of ocean warming in areas polewards 60°N latitude. In OSR5, the relative contribution of polar ocean warming north of 60°N to global ocean warming rates accounts for nearly 4% – a comparable value to its area fraction of the global ocean (Section 2.2). The ice-free ocean area warms substantially faster as compared to the ice-covered ocean, and this compensating effect leads to a warming trend of 0.6 (0.7) ± 0.2 Wm−2 for the upper 700 m (full-depth) ocean layer of the pan-Arctic region north of 60°N as consistently derived from the combined analysis of observations and reanalyses data over the period 1993–2019.

The molar nitrate:silicate ratio is an important indicator of nutrient availability related to the requirement of diatoms – a major group of microscopic algae, which underpin ocean biological productivity and transfer carbon from the surface to the deep layers of the ocean when they die. Globally, they are responsible for 40% of marine primary production and 40% of the particulate organic carbon exported to the deep ocean. Consequently, changes in diatom concentration can greatly influence global climate, atmospheric carbon dioxide concentration, and the function of marine ecosystems. OSR5 investigated for the first time a 30-year record (1990–2019) of water column silicate and nitrate in the Nordic Seas (Section 2.3), showing a steady increase of the nitrate:silicate ratio throughout the thirty-year period and linked to concurrent statistically significant decline in surface silicate. For this specific region, less access to silicate and other macronutrients in the Nordic Seas may shorten the spring diatom bloom period and hamper zooplankton growth, which in turn may have consequences for growth and development of commercially important fish stocks in these waters.

Baltic Sea:

A specific evaluation of eutrophication is also proposed for the Baltic Sea based on reanalysis results (Section 2.5) with particular focus on the nitrogen to phosphorus ratio, which has been reported to decrease in the water column across the entire Baltic Sea over the period 1993–2017, particularly in the Baltic Proper, the Gulf of Finland and the Gulf of Riga. An exception is the Gulf of Bothnia, which shows a relatively good environmental state over this period. Decrease in the nitrogen to phosphorus ratio affects phytoplankton blooms, supports nitrogen-fixing cyanobacteria growth, which leads to increased production of organic matter – and its decomposition consequently decreases the oxygen content – and enables eutrophication to endure. In addition to changes in the nutrient ratio, cyanobacteria blooms are also facilitated by ocean warming from climate change. Based on the results obtained in OSR5, it is very likely that the Baltic Sea will continue to experience frequent cyanobacterial blooms in the future.

Even a single passage of an extreme storm associated with high waves forces high sea level at the coast and a rough sea outside the sheltered areas. In the Baltic Sea (Section 4.4), simultaneous high sea level and waves lead to significant coastal erosion and flooding of low-lying areas. Low sea levels may complicate operations of heavily loaded cargo ships at the ports. In Januray 2019 however, a very unusual situation was documented: high waves coincided with a low coastal sea level. In January 2019, the Bothanian Sea had been hit by a severe storm and record-breaking significant wave height of 8.1 m was recorded. Surprisingly, exceptionally low sea levels were concurrently recorded in many coastal stations (as low as −1.1 m), both on the Finnish and the Swedish side of the Gulf of Bothania. As discussed in Section 4.4, the interplay of the extreme event, local sea ice extent and ocean circulation changes have triggered these unusual conditions.

North Sea:

OSR5 also draws a linkage between extreme variability such as marine heat waves and cold spells and key fish and shellfish stocks in the North Sea (Section 3.2). Catches of sole and sea bass increased in years with cold-spells (1994, 1996, 1997, 2010, 2011, 2013 and 2018), while catches of red mullet and edible crabs decreased. For heatwaves (1998, 2002, 2003, 2006, 2007 and 2014–2019), the impact on fisheries catch data lagged the temperature events by five years: sole, European lobster and sea bass catches increased whilst red mullet catches reduced.

Mediterranean Sea:

OSR5 also focuses on the Mediterranean Sea which has been recognised to be a climatic hotspot. During the past decade, its water masses have experienced strong and fast increases in temperature and salinity, responding very rapidly to global warming and to changes in the regional freshwater budget – an outcome that is envisaged to be important for climate science, environmental agencies, concerned citizens as well as regional policy-makers (Section 2.6).

The monitoring of the spatial and temporal evolution of storms is crucial to provide an information service to responsible emergency entities like coastguards and civil protection units which might need to intervene even under harsh weather and sea conditions (Section 4.2). In such circumstances, accurate nowcasts and short-term predictions are essential to prepare interventions that are timely, effective and with minimal risk. This kind of service is also essential to provide information and guidance for safer navigation by avoiding the higher impacted sea areas or delaying transits. OSR5 also discussed a new approach for the Maltese shelf area to predict, monitor, and assess extreme meteo-marine conditions, to verify the evolution of the storms in real time, and to provide improved services to users such as for civil protection, marine safety, and risks to essential assets. The method (Section 4.2) is based on merging complement data from observation and modelling systems (CMEMS products, CALYPSO HF radar network), aided by the support of artificial intelligence techniques in-cooperating knowledge from past extreme events. The methods performance is demonstrated on two extreme events in January and December 2019.

The city of Venice in Italy experienced four exceptionally high tidal peaks in the week from 11 to 18 November 2019, flooding large parts of the city. Venice had not suffered from four successive extreme events within one single week before. The OSR5 results (Section 4.3) show that spring tides coincided with a very high mean November sea level during this week. Additionally, and concurrent with the tidal maximum, strong Sirocco winds pushed Adriatic Sea water towards Venice during three of the four exceptional water level events. For the most extreme event on 12 November, a storm passed over Venice just at the time of the maximum tide. The official forecast underestimated maximum water levels for this event as the model forcing did not resolve the local storm. Higher-resolution atmospheric model fields and the use of satellite wind observations for nowcasting may further improve water level forecasts under extreme conditions.

OSR5 also covers a study on the intensity and geographical distribution of maximum wave height in the Mediterranean and Black Seas over 27 years (1993–2019) using CMEMS wave model hindcasts and the wave model WAVEWATCH III® (Section 2.8). Results show that in 2019 maximum wave heights were smaller than usual in the Black Sea (up to −1.5 m), while in the South Mediterranean Sea higher-than-average wave heights (+2.5 m) are reported linked to atmospheric depressions that rapidly passed over this area.

OSR5 also demonstrated the strong linkage between changes in South Adriatic hydrography as triggered by the ocean circulation – particularly for salinity – and the biodiversity in this area with potential effects on fish species of commercial interest (Section 3.5). These changes in the ecosystem can strongly impact economies and coastal communities that might need to adapt to the declining abundance of traditional target species and/or to the increasing abundance of other species, which previously were secondary to the local market.

The ecological and socio-economic consequences (e.g. on tourism, aquaculture, fisheries) of jellyfish outbreaks on the shorelines are relevant worldwide, and the development of prediction tools is critical to anticipate and mitigate the arrival of the jellyfish blooms (Section 3.4). While the Portuguese Man-of-War is not native to the Mediterranean Sea, their appearance had been reported several times during the past decade in the Gulf of Cadiz and in the Western Mediterranean. OSR5 presents a new forecasting system for the spread of this jellyfish which shows good skills during a strong event in 2018. The main potential benefits of this new forecasting system are to support coastal managers, and to minimise associated socio-economic losses.

Section 3.6 discusses the benefits of integrating the CMEMS variables in combination with trawl surveys into the modelling of fishery independent data for predicting fish species distribution in the Adriatic and Ionian basins. An integrated ecosystem approach is discussed, which incorporates anthropogenic and other environmental stressors into the advice for fisheries management. The results robustly demonstrate that the combined use of data improves the species distribution in the models.

The presence of invasive species in the Mediterranean Sea is much higher than in other European seas, and understanding the reasons behind the range expansion of this invasive species is important for minimising any possible impacts to the already highly pressurised Mediterranean marine ecosystem. OSR5 describes in Section 4.5 sightings of the invasive lionfish Pterois miles in the Ionian Sea, together with an analysis of ocean temperature in this region, and in 2019, warm water conditions have favoured the northward spread of this thermophilic species along the coast of the Mani Peninsula and the Greek mainland. These results are critical for ecological modellers and regional stakeholders involved aiming to monitor the spread of this generalist predator in their waters.

Black Sea:

In this 5th issue of the CMEMS OSR, topics tackled for the Black Sea include aspects of the basin-scale circulation, as well as discussing hypoxia monitoring in the northwestern part of the basin. The general circulation in the Black Sea features a cyclonic gyre encompassing the entire basin (Rim Current). OSR5 provides a new method for the Black Sea Rim Current ocean monitoring indicator (Section 2.7). Results over the period 1993–2019 show Rim current speed variations of 30% in close relation to the atmospheric circulation (e.g. wind) and an increase in Rim Current speed of ∼0.1 m/s/decade.

During the 1970s to 1990s, large areas of the Black Sea, particularly along the Romanian and Ukrainian coasts, had been hit by severe hypoxia predominantly driven by eutrophication, and this dead zone reached up to 40,000 km2 at its extreme in the 1990s. OSR5 (Section 3.8) analyses a Benthic Hypoxia index in this area over the period 1992–2019 depicting general recovery from the preceding eutrophication period (1980s), but also a re-increase in the severity of benthic hypoxia for the years 2016–2019 which is attributed to warming atmospheric conditions. Results demonstrate that a joint consideration of oceanographic and climate conditions and riverine and coastal nutrient discharge, incorperated into an operational indicator such as presented in this study could be a critical tool in support of coastal management and marine protection strategies.

1.2. Knowledge and data for international Ocean governance

Authors: Paula Kellett, Brittany E. Alexander, Johanna J. Heymans

1.2.1. What is international Ocean governance?

Covering 71% of the Earth’s surface and holding 99% of the area that can be inhabited by life, the Ocean plays a pivotal role in sustaining life on Earth, including through the provision of climate regulation, food, energy, and many other resources. The Ocean, or ‘blue’, economy in Europe alone was estimated to have a turnover of €750 billion in 2018 (European Commission ) and there is significant interest in developing this further through increased jobs and by supporting innovation. However, over-exploitation of the Ocean as a result of human activities is a very real challenge, and coupled with increasing pressures from climate change impacts and pollution, its ability to continue supporting life on Earth is threatened. There is hence a balance to be achieved: in order to continue supporting life on Earth and to achieve the Sustainable Development Goals (SDGs), the Ocean must be productive, clean, healthy, and resilient. For this, we must ensure that human impacts on the Ocean and its resources are managed sustainably. Given the interconnected nature of the Ocean, and that the majority of its volume lies outside of nation’s Exclusive Economic Zones (EEZs), strong international cooperation is needed for the sustainable management of the Ocean as a global common through international Ocean governance. This includes rules, agreements, processes and institutions, which need to be organised in a way that ensures that the human use of the Ocean will be sustainable into the future.

At the core of the international Ocean governance system lies the United Nations Convention on the Law of the Sea (UNCLOS – United Nations ). This is an international agreement that defines both the rights and responsibilities that nations have when using and managing the Ocean and its resources. Building on this foundation, laws, frameworks, institutions and jurisdictional rights have been established at different regulatory levels (local, national, regional, international) and for different marine sectors (e.g. shipping, fishing, and research). However, this has made the Ocean governance system very fragmented, and there is often a lack of coordination between different organisations and governance systems. Coupled with this are the challenges of ratifying and enforcing laws and regulations, especially in areas beyond national jurisdiction (ABNJ), of gaining international agreements for governance in a timely manner, and of gaps in the legal framework, especially linked to emerging sectors (e.g. seabed mining).

More coordination is therefore needed across regulatory systems and marine sectors, and all stakeholders should be involved in the process of developing and implementing governance regulations for the sustainable management of the Ocean and its resources.

1.2.2. The role of Ocean observation and data in international Ocean governance

Knowledge underpins Ocean governance, and provides the means to understand the Ocean and its functioning, and develop appropriate measures for its sustainable management and use. Ocean observing allows the collection of data to monitor and report on the state of the Ocean, make predictions about its future, and to assess the impact of governance regulations and success towards achieving the intended sustainability goals. Observational data also help to ensure that the development of economic activities in the Ocean are indeed sustainable.

In order to plan for and assess the sustainable use of the Ocean, a wide range of different types of data, from different sources and different providers are required. Significant investment in Ocean observing systems and personnel, along with appropriate maintenance and support, is required in order to collect these data, and efficiency and coordination across Ocean observing systems are critical (EMB ). These are challenges that the Global Ocean Observing System1 (GOOS) and the Group on Earth Observations2 (GEO) are working to improve at the international level. Within this context the European Global Ocean Observing System3 (EuroGOOS), the European Ocean Observing System4 (EOOS), and the European Commission’s Foreign Policy Instrument Action on international Ocean governance: EU component to global observations5 (EU4OceanObs), are tackling these challenges at the European level. Data sharing and interoperability between marine data infrastructure are also key enablers for effective international Ocean governance.

1.2.3. The European state of play

In 2016, the European Commission published its Joint Communication on International Ocean Governance (IOG): An agenda for the future of our oceans (European Commission ), which aligned strongly with the UN 2030 Agenda for Sustainable Development (United Nations ) and specifically the targets of Sustainable Development Goal 14 (Life Below Water). The Communication outlined the European Commission’s 50 planned actions for developing Ocean governance, not only in Europe but also internationally. The Communication was structured around three priority areas, one of which was ‘strengthening international Ocean research and data’: recognising the critical role that data and knowledge play in supporting the Ocean governance system.

The Joint Communication was followed by a report published in 2019, which examined the progress made towards implementing the 50 actions (European Commission ). The report highlighted the annual publication of the CMEMS Ocean State Report as one of the direct actions taken by the European Commission towards developing international Ocean governance. This report was initiated as a means to ‘promote ocean research, data and science with the aim of developing comprehensive, reliable, comparable and accessible Ocean knowledge to improve policy-making, drive innovation and facilitate a sustainable “blue” economy’.

In addition, in 2019 the European Commission, together with the European External Action Service, established the IOG Forum.6 The IOG Forum provided a platform for stakeholders within and beyond Europe to engage in interactive cross-sectoral and cross-boundary dialogue on Ocean challenges and governance solutions in support of the follow-up of the EU’s IOG Agenda. A diversity of stakeholders from across the globe have engaged with this initiative through a series of expert workshops, consultations and events, and the final recommendations7 were launched during a high-Level event on 20 April 2021. As well as continuing to align with the UN 2030 Agenda for Sustainable Development, the future of the EU’s IOG Agenda will also align closely with the European Green Deal,8 the EU 2030 Biodiversity Strategy9 and the aims of the UN Decade of Ocean Science for Sustainable Development.10

The activities of the IOG Forum have elicited a number of priority areas for action that are recommended to be addressed in the future EU IOG Agenda to ensure a clean, healthy, productive, resilient, and understood Ocean. The priority area on improving the Ocean knowledge system focuses on ensuring that future Ocean governance is knowledge-based and driven by inclusive and effective knowledge-policy interfaces.11 This priority area emphasises the need to intensify transdisciplinary co-designed research to address key knowledge gaps, integrate knowledge from relevant stakeholders and knowledge sources, and ensure strong observations and data capacity. The EU has many strong initiatives that it can build on in order to take a leading role in addressing these recommendations including the European Data Strategy, 8th Environment Action Programme, Horizon Europe, and the Destination Earth/Digital Twin Ocean initiative.

1.2.4. The international state of play

In recent years there has been focus on the development of a new legal instrument under UNCLOS, which covers the conservation and sustainable use of marine biological diversity in areas beyond national jurisdiction (BBNJ).12 This process formally began in 2015 and a revised draft text was published by the UN in early 2020.13 Negotiations are ongoing, and data and observations will be key in supporting its implementation, as was highlighted in the Intergovernmental Oceanographic Commission of UNESCO (IOC) Report of the Executive Secretary14 at its meeting on 3 February 2021, where the IOC proposed a State of the Ocean Report, to provide annual information about Ocean variables, and the status of Ocean observations (among other information). Strong data and observations will also be required to monitor progress towards commitments made as part of the Paris Agreement and the Convention on Biological Diversity. The upcoming UN Climate Change Conference (COP26) and the Convention on Biological Diversity’s COP15, both to be held in 2021, offer critical opportunities to increase ambition and cooperation towards Ocean sustainability at the international level. Several nations, and the EU, have also committed to designating three new marine protected areas in the Antarctic, the successful management of which will rely on international cooperation for data sharing and observational infrastructure in these areas.

Ocean observations and data have also gained increased attention in other international fora. In the 2016 Tsukuba Communiqué, the Science and Technology ministers of the G7 group Member States (G7 Science and Technology Ministers ) recognised the importance of developing stronger scientific knowledge in order to develop ‘appropriate policies to ensure the sustainable use of the seas and Ocean’. In order to achieve this, they stated their support for a number of actions linked to enhancing Ocean observations globally, promoting Ocean science and improving data sharing infrastructures, and strengthening collaboration to encourage regional developments in observing capabilities and knowledge networks. Subsequently, as presented in the G7 Future of the Seas and Oceans Working Group Statement to the OceanObs’19 Conference,15 the G7 established a dedicated Coordination Centre for Ocean observation platforms, which will be interlinked with other G7 priority areas16 and will interface with GOOS.

The OceanObs conferences, held every 10 years, are an opportunity for the Ocean observing community to discuss progress and define goals for the coming decade. At OceanObs’19 the importance of establishing effective collaborations with multiple stakeholders to advance effective Ocean governance was specifically recognised (Speich et al. ).

1.2.5. Where do we need to go from here?

Through all of the Ocean governance initiatives presented above, there is a clear message calling for improved co-ordination and increased stakeholder engagement in the co-design of Ocean research, observations and data. There is also a call for Ocean governance to be based on a sound foundation of knowledge that is effectively translated and available for use in policy-making so that sustainable and resilient management practices can be implemented. The Ocean observations and data communities have an integral part to play in both of these aims, including through actions such as the Ocean State Report. Knowledge cannot be developed without Ocean observations and data, and to support the co-ordination of Ocean governance at a global level, data need to be open, interoperable and guided by principles such as FAIR (Finadable, Accessible, Interoperable, Resuable). The observations and data communities are key stakeholders in Ocean governance, and should be engaged in dialogues around Ocean governance.

CMEMS OSR5 – Chapter 2

Table of content


Modelled sea-ice volume and area transport from the Arctic Ocean to the Nordic and Barents Seas


Ocean heat content in the High North


Declining silicate and nitrate concentrations in the northern North Atlantic


Eutrophic and oligotrophic indicators for the North Atlantic Ocean


Nitrate, ammonium and phosphate pools in the Baltic Sea


Long term changes monitored in two Mediterranean Channels


Interannual variations of the Black Sea Rim Current


Climatology and 2019 anomaly of maximum waves in the Mediterranean and Black Seas


Strong positive Indian Ocean Dipole events over the period 1993–2019

Section 2.1 Modelled sea-ice volume and area transport from the Arctic Ocean to the Nordic and Barents Seas

Authors: Vidar S. Lien, Roshin P. Raj, Sourav Chatterjee

Statement of main outcome: The Fram Strait represents the major gateway for sea ice transport from the Arctic Ocean, affecting the mass balance of the perennial ice cover in the Arctic. Our model results show a distinct seasonal cycle in both sea ice area and volume transport with a maximum in winter, in agreement with observations and other model-based studies. Contrary to several previously published studies, we find a significant, negative trend in the sea-ice area export through Fram Strait over the last two and a half decades. Possible explanations for this discrepancy are discussed. The reduction in area transport translated into a reduction also in the volume transport. In the Barents Sea, a strong reduction of nearly 90% in average sea-ice thickness has diminished the sea-ice import from the Polar Basin.

Product used:

2.1.1. Introduction

The Arctic Ocean contains a large amount of freshwater, and the freshwater export from the Arctic to the North Atlantic influence the stratification, and, hence, the Atlantic meridional overturning circulation (e.g. Aagaard et al. ; Aagaard and Carmack ; Holland et al. ). The Fram Strait represents a major gateway for freshwater transport, both as liquid freshwater and as sea ice, from the Arctic Ocean to the North Atlantic through the East Greenland Current (e.g. Vinje et al. ; Fahrbach et al. ; Lique et al. ; Smedsrud et al. ; Rabe et al. ; Figure 2.1.1). Two main factors contributing to this are the vicinity to the area north of Greenland where the thickest multi-year pack ice in the Arctic resides, and the fact that the Fram Strait is the largest (and deepest) gateway between the Arctic Ocean and the World Ocean. The transport of sea ice through the Fram Strait is therefore important for the mass balance of the perennial sea-ice cover in the Arctic and represents an annual loss corresponding to about 10% of the total Arctic perennial sea-ice volume (e.g. Vinje et al. ; Rampal et al. ). Indeed, sea ice export through the Fram Strait has been found to explain 54% of the variations in Arctic winter multi-year sea ice volume changes (Ricker et al. ). However, while satellites provide high-quality data on the Arctic sea-ice extent, estimates of the sea-ice thickness, and, hence, total volume of the Arctic sea ice has up until recently been rather limited (e.g. Ricker et al. ), and consequently model re-analyses are often used for assessing sea-ice volume and transport (e.g. Zhang et al. ).

A large part of the interannual variability in the sea-ice export through the Fram Strait is explained by the large-scale synoptic forcing, while thermal wind forcing across the Fram Strait represents a persistent, although declining, atmospheric forcing (van Angelen et al. ). Moreover, the geostrophic winds across Fram Strait have been stronger in the 2000s compared with the period 1960s–1990s, which has caused an increase in the sea-ice transport, more than compensating for the decrease in the sea-ice concentration in the Fram Strait over the same period (Widell et al. ; Smedsrud et al. ). As an example, in 2007 anomalous wind conditions over the Arctic contributed to a strengthened transpolar drift and increased sea-ice export through the Fram Strait, reducing the amount of multi-year sea ice within the Arctic (Smedsrud et al. ; Zhang et al. ). However, Bi et al. () found a decrease in the sea-ice volume export of 600 km3/year in the period 2011–2014 compared to the periods 1990–1994 and 2003–2008. They attributed this decline first and foremost to changes in sea-ice drift, and to a lesser extent to changes in sea-ice thickness and even less to a decrease in the sea-ice concentration.

Another gateway of major exchange between the Arctic and the North Atlantic oceans is the Barents Sea (e.g. Smedsrud et al. ). Although there is a considerable net sea-ice transport from the Nansen Basin of the Arctic Ocean to the Barents Sea shelf both from the north and from the east (e.g. Sorteberg and Kvingedal 2006; Lind et al. ; Figure 2.1.1), this sea ice melts locally within the Barents Sea. However, the freshwater input from the melting sea ice in the central Barents Sea is reported to be an important factor in maintaining the pool of Arctic Water, and, thus, the Arctic conditions and strong stratification of the northern Barents Sea (Lind et al. ). Moreover, the presence of Arctic Water maintains the Polar Front towards the southern Barents Sea dominated by warmer and more saline Atlantic Water (e.g. Loeng ). A borealization (e.g. Polyakov et al. ) of the Barents Sea, where the influence of Atlantic Water increases at the expense of the influence from Arctic Water, has been found to be ongoing due to an increase in the temperature of the Atlantic Water in the southwest (e.g. Årthun et al. ; Fossheim et al. ; Onarheim et al. ), and further due to increasing temperature in the atmosphere, and, hence, weaker heat loss from the ocean (Skagseth et al. ). A declining sea-ice import from the Arctic Ocean and a subsequent weakening of the Arctic Water reservoir further adds to the borealization (Lind et al. ).

Here, we present net modelled sea-ice area and volume transport through the Fram Strait and into the Barents Sea and compare the results with observation-base estimates reported in literature. Furthermore, we provide a scientific rationale for including these transport time series as Ocean Monitoring Indicators for the Copernicus Marine Environmental Monitoring Service.

2.1.2. Material and methods

The Arctic Monitoring and Forecasting Center model (product 2.1.1) is the TOPAZ4 system based on a North Atlantic and Arctic configuration of the HYCOM ocean model coupled to a modified version of the CICE3 sea ice model at a horizontal resolution of 12 km, assimilating various observations once a week, including sea-ice concentrations from OSI SAF and thin-ice thickness (from SMOS, since 2014) with an Ensemble Kalman Filter (Xie et al. 2018). The best estimate output is the average of the 100-members ensemble. For further information about the model system and data assimilation, see Sakov et al. (). For composite analysis shown in Figure 4, time periods of high/low sea ice area export are determined as those above/below 1 STD of the mean for the full monthly time series during the time period 1993–2019.

We have chosen to calculate the sea-ice area and volume transport through the Fram Strait across the 79°N latitude. In some other studies the section is chosen differently. Generally, Spreen et al. () found a convergence between a northern (80°N) and southern (76°N) section in summer (i.e. the transport at 80°N is larger than further south/downstream at 76°N), while in winter they found a divergence (i.e. the transport at 80°N was smaller than further south due to local ice production in the Greenland Sea). Thus, our results may not be directly comparable to results reported from the literature.

2.1.3. Results and discussion

The average monthly modelled sea ice export through the Fram Strait, both volume and area, follows the observations in terms of seasonal cycle, with a maximum in winter/spring (Oct–Apr) and a minimum in summer (Jul–Aug; Figure 2.1.2a,c). The maximum area export is found in March, with an average export of 133 × 103 km2, and with a standard deviation of 36 × 103 km2. However, because the maximum is governed both by the maximum sea-ice extent and favourable wind conditions, the actual month of the maximum area transport may vary between years. The minimum area export decreases to close to zero in July (9 × 103 km2) but with a standard deviation of 20 × 103 km2. For the full year, the average annual modelled sea-ice area export is 989 × 103 km2. This is about 10% larger than the estimate of 880 × 103 km2 reported by Smedsrud et al. () for the period 1935–2014, based on observations of satellite radar images and surface pressure observations and estimated across 79°N. Moreover, they reported some tendency of a positive trend during recent decades. Similar results were reported by Zamani et al. (), although they used a different section located further upstream at 82°N that likely increases the transport estimate in summer and decreases the transport estimate in winter (Spreen et al. ). Zamani et al. () estimated an annual sea-ice area transport of 860 × 103 km2 for the period 1990–2010. Moreover, they found a positive trend of +10% per decade (i.e. an increase of around 90 × 103 km2/year per decade). On the contrary, our model results indicate a negative, statistically significant (p < 0.05) trend of – 1.0 × 103 km2/month per year (i.e. on the order of – 100 × 103 km2/year per decade) for the period 1993–2019. However, we also find that the trend varies with the season, with the largest trend seen in winter while being close to zero in late spring and summer (Figure 2.1.2a). For the full period 1993–2019, this trend represents a 28% reduction in area transport. Other observation-based estimates based on ice motion and concentration fields are lower (Bi et al. ). They reported an average maximum sea-ice area transport of 78 × 103 km2 in March, and an annual average area export totalling 644 × 103 km2 (Tables 2.1.1 and 2.1.2).

Table 2.1.1. Average sea-ice area transports through the Fram Strait.

Table 2.1.2. Average sea-ice volume transports through the Fram Strait. * Annual average calculated from 12 times monthly average.

Average modelled sea-ice volume export through the Fram Strait exhibits a seasonal cycle comparable to that of the sea-ice area export (Figure 2.1.2c), with a maximum of 167 km3 in March, and a minimum of 13 km3 in July and August. However, the variability is large, with a maximum standard deviation in February (88 km3) and a minimum standard deviation in August of 22 km3 implying that there are also months where the net sea-ice transport is directed northwards. Indeed, the maximum modelled monthly sea-ice volume export is 353 km3 (April 1995), while the minimum modelled monthly sea-ice volume export is −34 km3 (July 2004; Figure 2.1.3c). Moreover, as for the sea-ice area transport, the actual month of the maximum area transport may vary between years. Our modelled estimates are comparable with observation-based estimates. Spreen et al. () reported an average monthly sea-ice volume export estimate of 217 km3 for the years 2003–2008, varying between 92 and 420 km3, which is higher than our modelled estimates. Note, however, that Spreen et al. () used a different section at 80°N which tended to overestimate the transport compared to further south in summer, and slightly underestimate in winter. Moreover, our results suggest that there was indeed an elevated export during the years 2003–2008 (Figure 2.1.3b,c). On the other hand, for the period 2010–2017, Ricker et al. () using a section located at 82°N reported that monthly export estimates varied between 21 and 540 km3. Annual export varied between 1250 km3 for the year 2012–2013 and 1910 km3 for the year 2011–2012 (Ricker et al. ), which is close to the average annual transport estimate of 1267 km3 found in our model results. Based on both satellite and numerical model data Zhang et al. () estimated an annual average sea-ice volume transport of 1132 km3, ranging from 10 km3/month in August to 145 km3/month in March. On the higher end, Spreen et al. () reported an average annual sea-ice volume transport of 2400 ± 640 km3, which they considered to be a conservative estimate, based on satellite measurement of sea-ice drift and upward-looking sonar for measuring sea-ice thickness. Based on linear regression, we find a statistically significant (p < 0.05) trend of −2.2 km3/month per year during the period 1993–2019. This trend represents a decrease in the sea-ice volume transport of 43% over the period, which is larger than the relative reduction in sea-ice area transport, implying also a reduction in the average sea-ice thickness during the period. As for the sea-ice area transport, the trend in sea-ice volume transport through the Fram Strait varies with season. A strong, negative trend is found in winter and partly autumn (October), while a weak, positive trend is found in early summer (Figure 2.1.2c).

Figure 2.1.1. Map showing the location of the sections across the Fram Strait between Greenland and Spitsbergen, the Barents Sea north between Svalbard and Franz Josef Land, and the Barents Sea east between Franz Josef Land and Novaya Zemlya, used to estimate the area and volume of sea ice transport. Mean sea ice drift speed during the time period 1993–2019 is overlaid in arrows. Black isobaths are drawn at 500-meter intervals.

The model results do not show any statistically significant trend in the sea-ice thickness in the Fram Strait. However, there are some indications of decreasing thickness from the 1990s to the 2000s, a period during which Spreen et al. , observed a negative trend in thickness of 15–21% per decade. This period was followed by an increase in the average modelled thickness from 2015 and onwards. Moreover, the first half of the period corresponds to the period when Hansen et al. () reported a decline in the average sea-ice thickness in the Fram Strait from 3.0 m during the 1990s to 2.2 m during the period 2008–2011.

The Barents Sea ice area and volume transport seasonal cycle differ somewhat from that in the Fram Strait. A maximum is seen in late winter/early spring (April–May), along with a secondary maximum in December (Figure 2.1.2b,d). The modelled seasonal sea-ice area transport maximum is 69 × 103 km2 in April, while the maximum sea-ice volume transport is 64 km3 in May. A seasonal minimum in both area and volume transport is found in summer (Jul-Sep), with minimum values of 2 × 103 km2 and 5 km3, respectively. A secondary minimum is found in January (and also Feb-Mar for volume).

Figure 2.1.2. Seasonal cycle of modelled sea-ice area and volume transport and corresponding trend. (a) Black line shows monthly average sea-ice area transport through the Fram Strait for the period 1993–2019 (positive values southward). Grey shading shows the corresponding ±1 standard deviation. Red, broken line shows the linear trend per year for each month. (b) Similar to (a), but for sea-ice area transport into the Barents Sea (positive values into the Barents Sea). Note that the scale on the y axes are similar in (a) and (b). (c) Similar to (a), but showing sea-ice volume transport through the Fram Strait. (d) Similar to (c), but showing for the Barents Sea. Note that the scale on the y axes are similar in (c) and (d).

Figure 2.1.3. Time series showing modelled sea-ice thickness, area and volume transport. (a) Average monthly sea-ice thickness in the Fram Strait (black line). The red line shows the linear trend for the period 1993–2019. (b) Twelve-month cumulative sea-ice area transport through the Fram Strait (black line; positive southward). The red line shows the linear trend for the period 1993–2019. (c) Similar to (b), but showing sea-ice volume transport through the Fram Strait. (d) Similar to (a) but showing for the Barents Sea. The broken red lines show the linear trend for the periods 1993–2006 and 2007–2019, respectively. (e) Similar to (b), but showing sea-ice area transport into the Barents Sea. The broken red lines show the linear trend for the periods 1993–2006 and 2007–2019, respectively. (f) similar to (e), but showing sea-ice volume transport into the Barents Sea.

As for the Fram Strait, the time series of both sea-ice area and volume transport show large variability, although the sea-ice transport into the Barents Sea shows smaller seasonal variability than the sea-ice transport through the Fram Strait (Figures 2.1.2 and 2.1.3). Moreover, we find no significant trend in the modelled sea-ice area transport into the Barents Sea, while the volume transport exhibit a statistically significant (p < 0.05) linear trend of −1.5 km3/year for the period 1993–2019, representing a total reduction of 84% over the period, using a linear regression analysis. However, there are some indications of a minimum during the 2000s, when the seasonal maximum seems to be more or less absent, while large seasonal maxima have ocurred in recent years leading to an increase in total volume transport (2014, 2017, and 2019; Figure 2.1.3f). While we did not find a significant trend in the sea-ice area transport overall, the linear regression indicated a reduction in the area transport of nearly 40% from 1993 to 2019. Moreover, there was an apparent minimum in the sea-ice area transport around 2007, followed by an apparent increase, especially after 2015 (Figure 2.1.3e). This contrasts with the findings of Lind et al. (), who reported a strong decline in the sea-ice area import to the Barents Sea from the 2000s until 2016. Our results show that the average thickness of the sea ice entering the Barents Sea decreased strongly between 1993 and 2005 (Figure 2.1.3d). From 2005 and onwards, however, there is no trend in the modelled thickness of the sea-ice entering the Barents Sea. Indeed, while we did not find a statistically significant change in the drift speed of the sea ice entering the Barents Sea (not shown), there is a statistically significant (p < 0.05), negative trend in the thickness of the sea ice that enters the Barents Sea, with a total reduction of 89% from 1993 to 2019 when using linear regression, but with the major part taking place prior to 2005. Similar to the Fram Strait, there was a strong increase in both the sea-ice area and volume transport in 2019 compared to the preceding years. This finding is in line with the reporting of the sea ice conditions around the Svalbard archipelago returning to ‘normal’ in 2019 through a combination of preconditioning from anomalous advection of thick multi-year sea ice from the Arctic Ocean and strong, northerly winds in the region (see Section 4.1 in this issue).

A major driver for the interannual variability of the Fram Strait sea-ice transport is the synoptic atmospheric circulation (e.g. van Angelen et al. ). Using ± 1 standard deviation to define high/low monthly transports, we find dipole patterns in the large-scale mean sea-level pressure and corresponding cross-sectional winds associated with anomalous sea-ice transport through all the three sections investigated (Figure 2.1.4). This result indicates that atmospheric forcing is a major driver also for the variability in the modelled sea-ice transport, and in agreement with findings based on observations (e.g. Tsukernik et al. ). However, our model results indicate a statistically significant, negative trend in the sea-ice area transport through the Fram Strait, opposite of the reported positive trend from increasing winds in recent decades in observations (e.g. Smedsrud et al. ) and model simulations (e.g. Zamani et al. ). Moreover, the main driver of the trend found in our model results is a negative trend in the sea-ice drift (not shown). On the other hand, the average modelled sea-ice concentration does not show any significant trend. Note, that although a negative trend in sea-ice volume transport was also found based on observations by Spreen et al. (), they concluded that a decrease in sea-ice thickness was the main driver of the negative trend. Possible explanations for this discrepancy include differences in the calculation of the area transport, uncertainties in the estimation of sea-ice drift (e.g. Sumata et al. ), and errors relating to the parameterisation of wind drag on sea ice in the model (e.g. Chikhar et al. ). Moreover, different studies refer to different areas for defining the Fram Strait. However, results presented by Spreen et al. () indicate that the transport estimates are not sensitive to the exact flux gate location. A thorough investigation of these matters is beyond the scope of this study, but we will point out some aspects that could be further elaborated in future studies. Our model results show larger sea-ice drift on the Greenland shelf in the western Fram Strait (Figure 2.1.1) than observed (Smedsrud et al. , see their Figure 2). This discrepancy points to differences in the treatment of fast ice and its impact on sea-ice drift in the model and the observation-based datasets. The current model configuration utilised in this study include a sea-ice model with an elastic-viscous-plastic rheology (Hunke and Dukowicz ), while an improved sea-ice model using a Maxwell elastic-brittle rheology (Dansereau et al. ), which takes into account ice deformation, will be implemented in the near future.

Figure 2.1.4. Composite sea-level pressure anomalies during periods of high sea-ice transport (left) and low sea-ice transport (right) through the Fram Strait (top; a,b), the Barents Sea north section (middle; c,d) and through the Barents Sea east section (bottom; e,f).

Differences in the position of Fram Strait sea-ice export estimation may affect the transport estimates in several ways. Complex recirculation patterns within the Fram Strait area (e.g. Beszczynzka-Möller et al. ) cause uncertainties in the northward and southward sea-ice drift estimations dependent on the position of the transect used. Depending on the season, sea ice freezing or melting occurs along-stream from north to south in the Fram Strait (e.g. Spreen et al. ). Additionally, the West Spitsbergen Current flowing northward in the eastern parts of the Fram Strait is more or less ice free all year round at 79°N, while further north there is varying presence of sea ice, also depending on the northward flow of Atlantic Water (e.g. Ivanov et al. ; Polyakov et al. ).

The relation between the sea-ice thickness and drift to the total sea-ice volume import in the Barents Sea points toward a change in the properties of the sea ice upstream of the Barents Sea, that is, in the Nansen Basin of the Arctic Ocean (in agreement with the findings of a thinning sea ice cover by Polyakov et al. ), rather than a change in atmospheric conditions affecting the general circulation, as the main driver of the changes in the sea-ice volume import to the Barents Sea during the period 1993–2019. However, our results indicate that the preconditioning through decreasing sea-ice thickness from less multi-year sea ice present in the Nansen Basin was a main driver of the decline in the sea-ice import to the Barents Sea in the 1990s and the first half of the 2000s, whereas after the mid-2000s, the area transport has increased (while the thickness has remained more or less constant), which indicates that changes to the general atmospheric circulation may have played an increasingly important role in the most recent decade.

Section 2.2. Ocean heat content in the High North

Authors: Michael Mayer, Vidar S. Lien, Kjell Arne Mork, Karina von Schuckmann, Maeva Monier, Eric Greiner

Statement of main outcome: This section presents an analysis of ocean heat content (OHC) north of 60°N. We use a range of data sources including in-situ data and ocean reanalyses, which are in remarkably good agreement despite the relatively sparse observational coverage. For the 1993–2019 period, we find a warming trend of 0.6 ± 0.2 Wm−2 for the upper 700 m and 0.7 ± 0.2 Wm−2 for the full-depth ocean north of 60°N. This suggests that neglecting the Arctic ocean in quasi-global OHC evaluations leads to an underestimate of ∼4%, which is similar to its area fraction of the global ocean. The strongest warming below 700 m is found in the Norwegian and Greenland Seas. The ice-free ocean area warms substantially faster at a rate of 1.4 ± 0.5 (1.9 ± 0.7) Wm−2 in the upper 700 m (full-depth) ocean, compared to the ice-covered ocean which shows a non-significant warming rate of 0.2 ± 0.2 Wm−2. The slow warming of the ice-covered Arctic Ocean thus masks the rapid and above-global-average warming of the ice-free ocean when considering the pan-Arctic region north of 60°N as a whole.

Product used:

2.2.1. Introduction

Ocean heat content (OHC) is an important quantitative variable in the Earth’s climate system, and provides a unique measure of the current status and prospects of global warming (von Schuckmann et al. ; Cheng et al. ). In the upper 700 m of the near-global (60°N-60°S) ocean, area averaged OHC has increased at a rate of 0.3 ± 0.1 Wm−2 during 1960–2018 and the rate increased to 0.9 ± 0.1 Wm−2 over the period 2005–2018 (von Schuckmann et al. ). The pan-Arctic (here defined as north of 60°N) – a region where near-surface warming has been reported to occur faster than the global average (e.g. Serreze and Barry ) – is usually not represented in the near-global global analysis of OHC, predominantly due to the fact of low subsurface temperature data coverage in this area. Recent studies have shown that the Arctic OHC is warming at a rate close to the global warming rate (e.g. Mayer et al. ; von Schuckmann et al. ; Mayer et al. ). Under anthropogenic pressure, Arctic sea-ice cover is declining, thereby altering the ocean-atmosphere heat exchange through larger areas of open water. In addition, increased ocean heat transport to the Arctic is pushing oceanic conditions typical of sub-polar regions (e.g. a comparatively weak stratification) downstream into the northern polar basin, a process termed ‘Atlantification’ (e.g. Polyakov et al. ) affecting also ocean-ice fluxes (Polyakov et al. ). Thus, the OHC and its spatial distribution within the Arctic is a relevant metric at the interface of oceanic transports, surface fluxes, sea ice melt, and the global energy inventory.

The Nordic Seas (Norwegian, Greenland, and Iceland Seas) is a region of major water mass transformation in the northern loop of the global thermohaline circulation (e.g. Aagaard et al. ; Mauritzen ). Here, saline and relatively warm Atlantic Water flows through the Norwegian Sea to the east en-route to the Arctic Ocean, while colder and fresher water masses of Arctic origin flows southward through the Greenland Sea to the west. Especially the northern part of the Norwegian Sea, the Lofoten Basin, is a reservoir of Atlantic Water and represents a major heat sink, with strong and persistent heat loss to the atmosphere, and as will be shown in this contribution, above-average regional warming. The OHC as an integrated measure of the hydrographic conditions in the Norwegian Sea represent a robust indicator and precursor of the ocean heat transport to the Arctic Ocean (e.g. Furevik ). The variability in the OHC locally in the Nordic Seas is mainly a product of local air–sea fluxes and advection. Several studies have indicated that variation in ocean heat loss to the atmosphere can explain about half of the year-to-year OHC changes in the Norwegian Sea (e.g. Mork et al. ). Other studies have revealed that advection is the primary cause of interannual to decadal heat content variability in this area (e.g. Carton et al. 2011; Asbjørnsen et al. 2019). Thus, advection tends to play a dominating role in the OHC variability in the Norwegian Sea on time scales longer than one year (Mork et al. ).

Quantification of ocean warming in the high north is crucial for a better understanding of Arctic change. However, earlier assessments of Arctic OHC examined neither the spatial details of the trends nor the robustness of the results, which were obtained largely from reanalyses. Here we assess Arctic OHC in more detail than before and also examine the agreement of reanalysis products and in-situ observations for a better understanding of uncertainties.

2.2.2. Data & methods

OHC integrated over the upper 1000 m of the Norwegian Sea is calculated from Argo data (CORA-GLOBAL-5.1, Ref No 2.2.1), and is presented as anomalies (OHCA), relative to the World Ocean Atlas 2018 (WOA18; Locarnini et al. ) climatology using the latest averaging period, 2005–2017. For each Argo profile, the temperature climatology is interpolated horizontally and vertically to match the location and vertical resolution of the Argo profile (Mork et al. ). The advantage of using anomalies is that they are independent of the reference value. For each Argo profile, OHC anomalies are calculated above the reference depth h as (1) OHC=cpρ0h0(TTclim)dz,(1) where cp is the heat capacity (4.0 × 103 Jkg−1 K−1), ρ0 is a reference density (1030 kgm−3), and z is the vertical axis with z = 0 at the sea surface. Subscript ‘clim’ indicates climatological data from WOA18, and is a function of month, depth and location. The integrated depth, h, is set to 1000 m, as the upper 1000 m of the Norwegian Sea comprises the bulk of the Atlantic Water flow towards the Arctic. Monthly means with uncertainties (standard deviation) are calculated from all OHC anomalies for the Norwegian and Lofoten basins within each month.

To cover the full Arctic, we compute OHC from the Global Ocean Reanalysis Ensemble product (GREP; product ref. 2.2.2) at a ¼° spatial and monthly temporal resolution. We present results for the upper 300, 700 m, upper 1000 m and full-depth (average depth north of 60 N to 1240 m, maximum depth > 4600 m) OHC, covering 1993–2019. In addition to the GREP, we also present results from CMEMS ARMOR3D (Ref No 2.2.3; Guinehut et al. ). ARMOR3D is a weekly global product with a 1/4° spatial resolution and 33 vertical levels. It provides temperature, salinity, geostrophic currents and mixed layer gridded fields from 1993 to present. Satellite altimetry and sea surface temperature are combined with in situ data by optimal interpolation. In situ data include vertical profiles from moorings, scientific campaigns, autonomous profilers, gliders, ships of opportunity, sea mammals, as well as surface data from various buoys and ferry boxes. Satellite data cover most of the Arctic in summer (altimetry goes up to 82°N), and there is some data near Svalbard even in winter.

Anomalies of the GREP and ARMOR3D data are calculated with respect to their 1993–2014 climatology. For the comparison with Argo data, we use the same climatology but additionally remove the long-term-means of the period for which Argo data is available (2002–2019 for the Norwegian Basin and 2005–2019 for the Lofoten Basin). Additionally, we provide regional averages for the ice-free ocean and ice-covered ocean. We use 30% annual mean sea ice concentration based on the GREP ensemble mean as a threshold for sea ice cover.

Trend estimates are based on the ordinary least squares method and are provided for the full 1993–2019 period. The uncertainty estimate used for significance testing takes into account random and structural errors as follows. Random errors are estimated from the standard errors of the linear regression coefficients of the GREP ensemble mean and the Argo OHC evolutions, respectively, taking temporal auto-correlation into account. Structural uncertainty for the GREP is estimated from the spread in the four OHC trends computed for the four reanalyses separately, divided by n1 (3). For the Argo, 100 synthetic time series were constructed and the structural uncertainty was estimated by the standard deviation of the linear regression coefficients of these 100 time series. The synthetic time series were constructed by creating 100 random numbers for each month that were normally distributed around the monthly OHC estimate with a standard deviation equalling the uncertainty during the respective month. Random and structural errors are added in quadrature to obtain the uncertainty estimate used for significance testing.

2.2.3. Results Reanalysis – observation comparison

Before examining the OHC in the pan-Arctic region from the gridded products (GREP and ARMORD3D), we perform a comparison between the estimated OHC from the reanalyses and observations based on Argo. The region is limited to the Lofoten and Norwegian basins of the Norwegian Sea, and the depth range is from the surface to 1000 m depth. The OHC is estimated according to equation (1). Due to limitations in the availability of Argo observations, the observation-based estimates are limited to the period June 2002 to July 2018 in the Norwegian basin and March 2005 to September 2018 in the Lofoten basin, respectively. In the comparison, the analysis of both the reanalysis-based and observation-based time series include the period covered by observations only. The time series based on the reanalyses and the observations are shown in Figure 2.2.1. Generally, OHCA estimates from both the GREP and ARMOR3D follow the observations in terms of both magnitude and variability. However, the observation-based estimates display larger month-to-month variability (Figure 2.2.1), which could be related to sampling noise in the observations that is smoothed by data assimilation. Based on monthly averages, the correlation, R, between the de-trended GREP and observation-based estimates is 0.59 (p < 0.001) in the Norwegian basin and 0.56 (p < 0.001) in the Lofoten basin. However, when we filter the time series using a 12-month moving boxcar window (i.e. annual averages), the correlation coefficients increase to R = 0.85 (p < 0.001) and R = 0.90 (p < 0.001) in the Norwegian and Lofoten basins, respectively. Another noteworthy feature of Figure 2.2.1 is that the agreement between the GREP and ARMOR3D is degraded prior to the availability of Argo data, which demonstrates the strong constraint that Argo provides on ocean state estimates.

Figure 2.2.1. Time Series of 0–1000 m ocean heat content anomalies in the Lofoten (top) and Norwegian (bottom) basins estimated from Argo (Ref. No. 2.2.1) and the Global Ocean Reanalysis Ensemble product (Ref. No. 2.2.2). Note that GREP and Argo data represent three-monthly averaged OHCA, while ARMOR3D data represents annual OHCA.

The trends in the reanalysis-based OHC estimates are comparable to the observation-based estimates in both basins (1.22 ± 0.85 Wm−2 and 1.69 ± 0.92 Wm−2 in the Norwegian basin, 3.75 ± 0.96 Wm−2 and 3.16 ± 1.35 Wm−2 in the Lofoten basin, with 95% confidence intervals computed as described in the Data & Methods section). All trends are significant at the 95% level. Moreover, the individual trend estimates fall within the uncertainty range (1σ) of the associated estimate from the reanalyses/observations. Thus, there is a close agreement between the reanalysed and observed OHC estimates in both the Norwegian Sea basins, both in terms of trend during the 2000s and inter-annual variability. While there is a statistically significant correlation also on monthly timescales, this correlation is rather weak. Ocean heat content in the high North from reanalyses

Figure 2.2.2a presents linear trends of upper 700 m OHC over 1993–2019. We focus on the spatial features that are statistically significant on the 95% level. Widespread and significant warming is present in the upper 700 m in basically all ice-free ocean north of 60N. Maximum warming is found in the Barents and Norwegian seas, with regional trends in the Norwegian Sea exceeding 6 Wm−2, which is almost twice as high compared to the upper 300 m OHC trends (not shown). Strongest warming in the Barents Sea occurs in the southeastern part, which is in agreement with the findings of Skagseth et al. (). Weak but significant warming is present also in the shelf seas along the Siberian coast. Moderate but still significant warming is found also in the Beaufort Gyre, an area that is ice-covered most of the year. Weak and largely insignificant cooling is present around the Amundsen Basin. It is important to note that observational data in the sea-ice-covered areas is scarce, but at least the moderate warming in the Beaufort Gyre seems consistent with observation-based estimates. For example, Timmermans et al. () found a halocline warming of ∼0.3 Wm−2 during 1992–2015 in this region (based on the numbers in their Figure 1).

Figure 2.2.2. Linear OHC trends (converted to Wm−2) (a) of the upper 700 m and (b) below 700 m for 1993–2019 from the GREP (Ref. No. 2.2.2). Stippling denotes regions where trends are significant at the 95% confidence level. Figure (b) additionally indicates the location of water bodies referred to in the text: Iceland Sea (IS), Norwegian Sea (NS) with two basins Lofoten Basin (LB) and Norwegian Basin (NB), Greenland Sea (GS), and Barents Sea (BS).

Figure 2.2.2b shows linear trends of OHC below 700 m. In the Norwegian and Greenland Seas, substantial warming is present at these depths, attaining values greater than 5 Wm−2. Thus, full-depth OHC trends in the Nordic Seas exceed 10 Wm−2 regionally. In contrast, penetration of warming OHC trends below 700 m into the central Arctic is progressing relatively slowly according to the results from the GREP.

We now turn to regionally averaged OHC evolution during 1993–2019. Figure 2.2.3 presents a time series of upper 700 m OHC anomalies for the pan-Arctic, including results for the two sub-regions (ice-covered and ice-free). Pan-Arctic upper 700 m OHC has increased since 1993, with a much larger contribution from the ice-free Arctic Ocean. One should note that the ice-free ocean covers only ∼37% of the total ocean area north of 60°N. This is consistent with Figs 2.2.2a and b showing particularly strong warming in the Greenland and Norwegian Seas. Remarkably rapid warming occurred in the ice-free Arctic ocean during 2002–2004, which can also be seen in the pan-Arctic OHC. More in-depth diagnostics reveal that this rapid warming signal persists when removing Argo data in an observing system experiment (not shown), suggesting that this is not an artefact from increased data coverage and is indeed a real climate signal. The long-term OHC increase and interannual signals, including the rapid warming in 2002–2004, are consistently shown also by ARMOR3D. Even in the data-sparse regions under sea ice, ARMOR3D yields similar results as the GREP.

Figure 2.2.3. OHC anomaly (OHCA) of the upper 700 m in 1021 J relative to the 1993–2014 climatology for the Arctic Ocean north of 60 N and the partition into ice-covered regions (under_seaice) and ice-free regions (no_seaice). The partition is based on the annual mean sea ice concentration using a 30% threshold. Curves represent the GREP (Ref. No. 2.2.2) ensemble mean, and the shading represents the intra-ensemble spread. The dashed lines represent results from ARMOR3D (Ref. No. 2.2.3; based on annual mean values).

Finally, we present area-averaged OHC trends for the pan-Arctic as well as the contributions from ice-free and ice-covered ocean in Figure 2.2.4. The 1993–2019 warming rate for the pan-Arctic was 0.4 ± 0.1 Wm−2 (0.44 ± 0.07 Wm−2 when allowing two decimals) in the 0–300 m layer and 0.6 ± 0.2 Wm−2 (0.57 ± 0.16 Wm−2) in the 0–700 m layer. The trend is enhanced by a factor of ∼1.3 when integrating down to 700 m, while the uncertainty increases by a factor of ∼2.1. This indicates that the signal-to-noise ratio, defined as the ratio of the trend and its uncertainty, is smaller for the 0–700 m layer compared to the 0–300 m layer. For the full-depth ocean, warming is further enhanced to 0.7 ± 0.2 Wm−2 (0.74 ± 0.19 Wm−2).

Figure 2.2.4. Linear OHC trends 1993–2019 for Arctic Ocean north of 60N and decomposition into ice-free and ice-covered ocean north of 60N (based on the 1993–2019 sea ice concentration climatology with a threshold of 30% concentration) estimated from the GREP (Ref. No. 2.2.2). Trends are converted to warming rates given in Wm−2. The conversion factor from Wm−2 to ZJ/yr is 0.51 for the pan-Arctic Ocean, 0.17 for the ice-free ocean, and 0.34 for the ice-covered ocean. Uncertainties are the 1σ-uncertainties.

Warming is weak in the ice-covered ocean (both in area-specific and area-integrated units), probably because sea ice melt takes up a substantial amount of extra energy and the ocean below is protected from the atmosphere above (Mayer et al. ), which is elaborated on in the discussion. Upper 300 m trends are 0.2 ± 0.1 Wm−2. Extending to the full-depth ocean does not alter the obtained trend, but uncertainty is much increased (0.2 ± 0.2 Wm−2).

Strong warming is found in the ice-free ocean, with a warming trend of 1.0 ± 0.2 Wm−2 in the upper 300 m, 1.4 ± 0.5 Wm−2 in the upper 700 m layer, and 1.9 ± 0.7 Wm−2 in the full-depth ocean. We note the much higher signal-to-noise ratio of the trends in the ice-free ocean compared to the ice-covered ocean, which likely is a result of the much better observational coverage.

2.2.4. Discussion

We have shown that the GREP represents realistically the observed trend and temporal variability of OHC in the Norwegian Sea, which is a robust indicator of heat transport toward the Arctic Ocean. This finding indicates that the mechanisms that control the variability in Atlantic Water heat content en route to the Arctic Ocean are realistically represented in the GREP. We note that the GREP is not independent of the Argo data, as these observations are assimilated into ocean reanalyses. Moreover, the found agreement does not necessarily guarantee a correct balance of oceanic transports (e.g. Carton et al. ; Asbjørnsen et al. ), local air–sea heat fluxes (e.g. Segtnan et al. ; Mork et al. ), and shelf slope – basin exchange (e.g. Isachsen et al. ; Mork and Skagseth ) within the Norwegian Sea in the GREP, as a potentially strong contribution to regional OHC stems from data assimilation. However, we note that earlier studies found good agreement between oceanic transports obtained from oceanic reanalyses and mooring-derived data in the High North (e.g. Pietschnig et al. ; Uotila et al. ; Mayer et al. ), which provides additional confidence in our results. Nevertheless, more validation work beyond OHC is needed to further corroborate our findings.

A rapid increase in 0–700 m pan-Arctic OHC is reported during the period 2002–2004, with the open-water area accounting for most of the increase (Figure 2.2.3). However, while the increase in OHC persists for the pan-Arctic region as a whole, it represents a transient signal within the Norwegian Sea (Figure 2.2.1). This suggests an advective nature of the warming, where the signal propagates from the northern North Atlantic to the entrance of the Arctic Ocean in a few years (Furevik ; Årthun and Eldevik ). Thus, it may be a delayed effect of the rapid warming of the North Atlantic subpolar gyre in the mid-1990s as documented by Robson et al. (), in conjunction with observed changes to atmospheric circulation as discussed by Proshutinsky et al. (). Indeed, a temperature increase was observed in the Faroe-Shetland Channel in 2002, preceding the OHC increase observed in the Norwegian Sea in 2002–2003 (Figure 2.2.1) and subsequent temperature maxima in the Barents Sea Opening and the Fram Strait in 2006 (e.g. González-Pola et al. ).

The large difference in the rate of increase in OHC between open ocean and ice-covered areas is worthy of further discussion. The rapid warming of the ice-free ocean appears to be driven by a combination of increased ocean heat advection and decreased surface net heat loss. For example, Tsubouchi et al. () found a general increase of ocean heat transport into the Nordic Seas after 2001, and Mork et al. () found that both changes to oceanic transports and air–sea exchange play a role in the Norwegian Sea. Furthermore, Skagseth et al. () showed that within the Barents Sea – a main area for oceanic heat loss in the Arctic region – a reduction in the turbulent heat loss from the Atlantic Water to the atmosphere due to increased air temperatures more than offset the increase in surface heat loss due to increased open water area and enhanced shortwave radiation absorption arising from the ice-albedo feedback. The contrastingly slow warming of the ice-covered Arctic Ocean indicates that a large fraction of the increased OHC in the upstream, ice-free areas is lost in the northward-moving seasonal and marginal ice zones either directly to the atmosphere, especially with trends towards larger areas of open waters (e.g. Ivanov et al. ; Skagseth et al. ), or through melting sea ice, as suggested by, e.g. Dmitrenko et al. (). However, a smaller residual fraction of the increase in OHC is advected into the ice-covered part of the Arctic, seen as a subsequent warming along the path of the Atlantic Water flow through the Fram Strait and also the St. Anna Trough and to the interior Polar Basin (Figure 2.2.2a; Lien and Trofimov ; Skagseth et al. ). In addition, the positive OHC trends in the Beaufort Gyre (Figure 2.2.2) can be linked with its acceleration in association with recent trends towards more anticyclonic atmospheric circulation in this region (Proshutinsky et al. ).

The distinction between OHC trends in ice-free and ice-covered regions helps to explain the surprising result of earlier studies (Mayer et al. ; von Schuckmann et al. ; Mayer et al. ) that pan-Arctic OHC increase is not stronger than its global average despite Arctic amplification in the near-surface climate. OHC increase in the ice-free ocean north of 60N is indeed above global average values, but relatively weak trends under sea ice mask this when considering pan-Arctic trends. However, increased temperature in the Atlantic Water flowing into (formerly) ice-covered regions is also projected to continue in the future (Årthun et al. ). Thus, propagation of ocean heat downstream to the interior Polar Basin through advection may play an increasingly important role in Arctic Ocean warming in the coming decades. One important caveat of the results for the sea-ice-covered regions is the low number of observations in presence of sea ice. However, we do find good agreement between the GREP-based results and those from ARMOR3D, which use very different data assimilation approaches and use different observational data. Moreover, there is quantitative agreement with an observation-based study in the Beaufort Gyre (Timmermans et al. ), and the weak OHC increase under sea ice is physically plausible. Hence, we consider our results for sea-ice-covered regions reasonable, but our confidence intervals likely underestimate true uncertainty.

To put these results into a global context, it is useful to consider area-integrated quantities. The pan-Arctic linear full-depth OHC trend 1993–2019 converts to 0.38 ± 0.10 ZJ/yr. This can be compared to the 60S–60N 0–2000 m OHC trend of 9.20 ± 0.92 ZJ/yr (1993–2018) as estimated by von Schuckmann et al. (). Thus, neglect of the Arctic ocean in quasi-global OHC assessments leads to an underestimate of ∼4%, which is similar to its area fraction of the global ocean.

Section 2.3. Declining silicate and nitrate concentrations in the northern North Atlantic

Authors: Kjell Gundersen, Vidar S. Lien, Jane S. Møgster, Jan Even Øie Nilsen, Håvard Vindenes (IMR)

Statement of main outcome: A comprehensive analysis of nutrient data in three regions of the Nordic Seas between 1990 and 2019, shows a statistically significant decline in surface silicate. This finding is in agreement with previous reports on silicate from the northern regions of the North Atlantic, but this is the first look at a 30-year record of water column silicate and nitrate in the Nordic Seas. Surface nutrient concentrations in the Nordic Seas appear to be regulated by the Subpolar Gyre situated south of Greenland and Iceland. The Subpolar Gyre Index has been in a decline for most of the period investigated, which means that the gyre has moved westward allowing more subtropical (and more nutrient depleted) water into the Nordic Seas. The largest decline in silicate occurred during the first ten years investigated and is still on a downward slope. We also found a statistically significant decline of silicate in Arctic Water in the Greenland Sea, but with a time-lag relative to the decline in Atlantic Water. Nitrate on the other hand, did not decline as uniformly and only had a significant drop midway through the time-series (2005–2009) and only in the Norwegian Sea, before it again increased to previous levels. The molar nitrate:silicate ratio however, showed a steady increase throughout the thirty year period investigated. Less access to silicate and other macronutrients in the Nordic Seas may shorten the spring diatom bloom period and hamper zooplankton growth, which in turn may have consequences for growth and development of commercially important fish stocks in these waters.

Product used:

2.3.1. Introduction

Periodic changes in the eastward extent of the Subpolar Gyre (SPG), an important regulator of the North Atlantic thermohaline circulation, determines the inflow of high-saline water to the Nordic Seas (Hátún et al. ). Rey () first noticed a bidecadal drop in dissolved silicic acid (silicate) concentrations in the Nordic Seas, along with an increase in salinities, and suggested that periods of relative weak North Atlantic Oscillation (NAO) caused warmer, high-saline Atlantic Water to enter the Norwegian Sea. The strength of the NAO determines the east–west positioning and frontal shifts of the SPG and hence, the magnitude of Atlantic Water influx to the Norwegian Sea (Sarafanov ). More recently, Hátún et al. () were able to demonstrate that the shifting fronts of the SPG to a large extent determine surface hydrography, and silicate content, in several areas of the subpolar North Atlantic region (Norwegian Sea, Irminger Sea, Labrador Sea). Water entering the Greenland Sea does not only originate from Atlantic Water, but has another pathway that may take years to decades through the Arctic and the Fram Strait (e.g. Schlosser et al. ). Given this time-lapse, diminishing silicate concentrations in the Norwegian and Barents Seas may not immediately appear in the Greenland Sea, as this region may be a mia mixture of Atlantic Water and Arctic waters.

Extended periods of reduced access to silicate, and possibly other nutrient elements, may have severe implications for the pelagic community composition in the Nordic Seas. The four nutrients most commonly measured in the Nordic Seas are nitrate, nitrite, phosphate and silicate. Of these, nitrite is a product of microbiological activity and appears in very low concentrations relative to nitrate, and phosphate is readily recyclable within the euphotic zone. Therefore, we expect to see more immediate effects from diminishing nutrient concentrations in the pool determining annual new production (nitrate) and in the single most important element for sustained diatom growth (silicate). North Atlantic spring blooms are often dominated by diatoms (Savidge et al. ) as these primary consumers of dissolved silicate rapidly respond to the high nutrient concentrations. Toward the end of the bloom however, silicate and other macronutrients are approaching depletion and the phytoplankton community becomes less buoyant. As the remnants of the spring bloom exit surface waters it becomes a major component of the export of particulate organic matter to the deep ocean (Honjo and Manganini ). The latter has led Pollock () and others to suggest that available silicate in surface waters may determine atmospheric CO2 drawdown via the ‘biological pump’ (Volk and Hoffert ). However, an overall decline in accessible silicate in surface waters, the end product of diminishing silicate concentrations in deeper waters, may reduce the length and magnitude of the annual phytoplankton spring bloom.

Diatoms are only competitive with other phytoplankton species at silicate concentrations higher than 2 μmol/L (Furnas ; Egge and Aksnes ). Therefore, reduced annual influx of silicate to the euphotic zone, relative to other macronutrients, may hasten the switch towards a flagellate dominated community and possibly regenerated production (Smayda et al. ). As of today, spring blooms in the Nordic Seas are mainly composed of diatoms (Rey ) and the magnitude and timing of each bloom is considered a deciding factor in both growth and recruitment of Calanus finmarchicus and C. glacialis, two of the most important calanoid copepods in the region (Melle et al. ). Calanoid copepods, such as the most abundant species, C. finmarchicus, appears to have a preference for diatoms (Meyer-Harms et al. ) and a number of studies have concluded that food quality affects copepod production and ultimately, growth at higher trophic levels (Barofsky et al. , and references therein). Calanoid copepods in turn, are considered one of the main staples for several species of commercially important pelagic fish, both in the Norwegian Sea and the Barents Sea (Skjoldal et al. ; Gjøsæther ). Therefore, changes in the magnitude or composition of the spring bloom, and less significance of diatoms, may ultimately have direct implications for commercial fisheries in these regions.

Here we present an updated, 30 year time-series of silicate measurements from the Nordic Seas, as we also consider the fate of nitrate concentrations in these waters during the same period of time. The relative change between the two macronutrient concentrations (the nitrate:silicate-ratio) and its implications for phytoplankton growth and community composition, is also discussed.

2.3.2. Data and methods

The nutrient data used in this study, which include silicate and nitrate collected from the Nordic Seas between 1990 and 2019, are available from the Copernicus Marine database (product ref. 2.3.1; 2.3.2) and the majority of data (>99%) originate from the Institute of Marine Research (IMR). Therefore, we provide a detailed overview of the seawater sampling and rigorous data quality control at the IMR, that precedes the quality control procedures applied by the Copernicus Marine Environment Monitoring Service (CMEMS; Jaccard et al. ), where all data are further scrutinised by automated tests and all outliers are checked manually (product ref. 2.3.1; Jaccard et al. ). Only data flagged as ‘good data’ (flag = 1) were included in this study.

Seawater samples were collected from Niskin-type water bottles at predetermined depths triggered by a Conductivity Temperature Depth (CTD) instrument package mounted on a rosette. Up until the last millenium, the majority of nutrient samples were analysed in real time onboard the ships. As the research fleet expanded at IMR, the number of autoanalyzers could no longer match the number of ships operating simultaneously, and nutrient samples (20 mL) were poisoned with chloroform (200 µL) and stored in the fridge at +4°C for analysis at the home laboratory within 1–6 weeks after collection. The samples were allowed to acclimatise to room temperature as the chloroform was evacuated by vacuum, prior to analysis on an Automated Analyzer (AA) system. With one exception, all nutrients were run on homemade AA system assemblies (Skalar and Alpkem hybrids) up until recently. The latest upgrade was the first complete AA system purchased from Skalar Analytical B.V. (Breda, The Netherlands). Although analytical instruments have changed over the years the nutrient chemistry has stayed the same. Colorimetric determinations of dissolved inorganic nutrients are based on the methods first described by Bendschneider and Robinson () and Grasshoff () with a number of minor adjustments suggested by the manufacturers (Alpkem, Skalar Analytical). The AA system measures nitrate (NO3), nitrite (NO2), phosphate (PO4) and silicate (SiO4) but only nitrate and silicate are reported in this study. Briefly, nitrate in seawater is reduced to nitrite coupled to a diazonium ion and, in the presence of aromatic amines, the resulting blue azo-dye is determined spectrophotometrically at 540 nm. The nitrate concentration is corrected for ambient nitrite (same analytical method as for nitrate, but without cadmium reduction) measured concurrently. Silicate (silicic acid) reacts to molybdate at low pH and the resulting silicomolybdate is reduced by ascorbic acid to a blue dye measured spectrophotometrically at 810 nm.

The Plankton Chemistry Laboratory at IMR maintains quality control of precision and accuracy by daily assessments of analytical standard curves and internal standards. Both nitrate and silicate concentrations are measured with a precision <0.2% and the accuracy deviate <1% from the internal standard. The laboratory is an active participant in the biannual intercomparisons initiated by the QUASIMEME Laboratory Performance Studies (, as well as other inter-laboratory comparisons, such as the International Ocean Carbon Coordination Project (IOCCP) intercalibration of nutrient analysis in 2017 (

In this study we focus on the deep water reservoir of nutrients in order to further document the decadal decline already reported in literature (Rey ; Hátún et al. ). Because the main source of nutrients in the Nordic Seas is the North Atlantic mode water, we used silicate and nitrate data from high-saline Atlantic Water collected in winter (January-March) in order to avoid effects of biological uptake by primary production. Therefore, we limit our study to silicate and nitrate data from two Atlantic Water regions (Figure 2.3.1) including Atlantic Water only (here defined by S > 34.9) in order to avoid continued supply of nutrients to surface waters through the year, as is the case with river runoff in coastal waters. In the Greenland Sea region we focus on water from the Arctic as a source of nutrients. Here Atlantic Water influence is mainly from the subsurface Return Atlantic Water from the Fram Strait in the north which can be traced by temperature (T > 0°C). Therefore, we limit our study of the Greenland Sea to only include water with temperature below 0°C. We divided the data into two depth segments; the euphotic zone (0–50 m), and the core of Atlantic Water flow through the Nordic Seas below the euphotic zone (100–200 m). We included the euphotic zone (0–50 m) in this study in order to detect potential, temporal discrepancies between surface water nutrients and the large midwater reservoirs (100–200 m). Due to poor cruise coverage in winter in the Greenland Sea, we used deep water data only (100–200 m range) from the whole year in that region. Data from the whole year, and below 100 m depth in the Norwegian and Barents Seas, produced qualitatively similar results as when using winter data only (data not shown).

Figure 2.3.1. The Nordic Seas and all stations visited (blue dots) where nutrient data have been collected (product ref. 2.3.1; 2.3.2) and analysed in the period investigated (1990–2019). Black rectangles show the regions with data used in this study: NS (Norwegian Sea), BS (Barents Sea), GS (Greenland Sea).

2.3.3. Results

Surface and deep water silicate concentrations from the three selected regions were binned in 5-year periods (Figure 2.3.2). To check the significance of any changes between two consecutive pentads (i.e. 5-year periods), we have used a two-tailed student’s t-test with n-2 degrees of freedom, where n is the number of observations in the pentad with least observations among the two. We used higher than 95 % confidence (p < 0.05) to imply statistical significance (where p is the probability). No available data from the Barents Sea in the second period (1995–1999) met our selection criteria (region and salinity range) and is therefore not shown. There is a significant drop in silicate concentrations in both surface waters (0–50 m) and in the Atlantic Water (100–200 m), between the first two periods (1990–1999) and the subsequent periods in the Norwegian Sea, in both depth layers, while in the Barents Sea the decline continues through the first half of the entire time-series in both layers (Figure 2.3.2). The largest decrease appeared in the deeper layer (100–200 m) in the Barents Sea, where silicate concentrations dropped by 19 % (p < 0.05) from 1990–1994 to 2015–2019. A statistically significant (p < 0.05) decrease of comparable magnitude (17 %) was also observed in the euphotic zone (0–50 m) between the same periods. The decline in the Norwegian Sea, during the same two periods of time, were 17% in surface waters and 16% in deeper waters. A larger part of the decrease occurred during the 1990–2004 period, followed by an apparent levelling off after 2005. The last period (2015–2019) showed a statistically significant decline compared to 2010–2014 in the Barents Sea (Figure 2.3.2).

Figure 2.3.2. Box-whisker plots of silicate concentrations (in micromol per litre; obtained from product ref. 1.4.1) bin-averaged in 5-year periods. Blue box shows 25th–75th percentile, red bar shows median, red cross shows mean, and the whiskers show min/max values. (a) Norwegian Sea, 0–50 m. (b) Barents Sea, 0–50 m. (c) Norwegian Sea, 100–200 m. (d) Barents Sea, 100–200 m. (e) Greenland Sea, 100–200 m. Note the different scales on the y-axis. Sample size (n) is provided above each top whisker and shown in boldface where the change from the previous period is statistically significant (>95%).

Initially, silicate in the Greenland Sea showed no or only a little decline, except for a statistically significant decrease (p < 0.05) towards the end of the period investigated (Figure 2.3.2). It is possible that silicate concentrations in the Greenland Sea not only reflect the conditions in Arctic waters, as this region can also be directly influenced by Atlantic water masses from adjacent regions and that this influence may change over the study period. Therefore, we wanted to investigate the influence of Atlantic Water in the Greenland Sea. In order to do this, we used temperature instead of salinity as a proxy for Atlantic Water influence (due to the relatively larger differences in temperature than what we see in salinity, between Atlantic and Arctic water masses in the region) and compared silicate concentration and the associated in-situ temperature. We omitted all observations where T > 0°C, and our analysis revealed no correlation between silicate concentrations and in-situ temperature at each individual station (R = −0.02, p = 0.69; n = 519). We concluded that the observed changes in average silicate concentrations in the Greenland Sea were not caused by sampling in different water masses in different time-periods. Average temperature increased during the 5-year periods investigated (−0.86°C, −0.85°C, −0.60°C, −0.40°C, 0.41°C and −0.25°C, respectively). Average deeper layer silicate concentrations were overall, higher in the Greenland Sea for the entire period (6.0 μmol/L) compared to the Norwegian Sea (4.8 μmol/L) and the Barents Sea (4.5 μmol/L).

We observed a less dramatic decline in nitrate concentrations in all three regions (Figure 2.3.3). The decline in nitrate appeared more pronounced (significant at the p < 0.05 level) in the Norwegian and Barents Seas (Atlantic Water) during the first half of the periods investigated, while the decrease in the Greenland Sea (Arctic Water) appeared mostly in the latter half of the investigated period. In the Norwegian Sea, the average nitrate concentration decreased by 8 % in the euphotic zone and 9 % in the deeper waters, respectively, between 1990–1994 and 2005–2009. Subsequently, nitrate concentrations in the Norwegian Sea approached initial levels found in 1990–1994 (Figure 2.3.3). In the Barents Sea, maximum nitrate concentrations were observed in 2000–2004, followed by a decrease of 6 % in both surface and deeper waters until 2010–2014. The overall decline in nitrate concentrations in the Greenland Sea (9 %) was comparable to the other two regions in magnitude, but the drop was mostly confined to the latter half of the investigated period. Average concentrations of deep water nitrate were overall, higher in the Greenland Sea for the entire period (12.4 μmol/L) compared to the Norwegian Sea (11.3 μmol/L) and the Barents Sea (10.9 μmol/L).

Figure 2.3.3. Same as Figure 1.4.2, but for nitrate (obtained from product ref. 1.4.1).

As silicate showed an overall decline and nitrate appeared to recover during the latter half of the time-series (Figures 2.3.2 and 2.3.3), the relative proportion between the two macronutrients (the molar nitrate:silicate-ratio) showed an overall increase during the same period of time (Figure 2.3.4). There was an initial decline to 2.2 in the Norwegian Sea (1995–1999) prior to an increase and a subsequent levelling off around 2.5, both in the euphotic zone and in deeper waters. In the Barents Sea the ratio was 2.1 in both the euphotic zone and the deeper layer in the first period (1990–1994) before it increased to approximately 2.5. Note that we only have 19 and 14 data points (deep waters and surface layers, respectively) in the Barents Sea during the first 5-year period. Yet, the increase during this period (1990–1994) was significant at the 95% confidence level. The observed nitrate:silicate-ratios were >2.5 during the last 5-year period in both the Norwegian and Barents Seas. In the Greenland Sea, the nitrate:silicate-ratio remained stable around 2 during the entire period investigated (Figure 2.3.4).

Figure 2.3.4. Same as Figure 2.3.2, but for the nitrate:silicate-ratio (obtained from product ref. 2.3.1).

2.3.4. Discussion Decadal decline in silicate and nitrate, and the role of the SPG

Our compilation of surface and midwater nutrient concentrations over the last 30 years (1990–2019) show an abrupt and significant decline in silicate during the first decade (Figure 2.3.2). The decline is slower for nitrate and reaches a minimum after 15–20 years, followed by a subsequent increase (Figure 2.3.3). Rey () noticed the same initial drop as in this study, as he used the same Nordic Seas data sets up until 2010. Hátún et al. () focused much of their study on Arctic waters surrounding the southern tip of Greenland, but also observed the initial drop in the Nordic Seas as they used the same data sets as this study, but only up until 2015. Only one other study (Johnson et al. ) has reported a decline in nitrate from the Rockall Trough, a marginal region off Scotland in the Atlantic Water inflow entering the southern parts of the Nordic Seas.

Winter deep water mixing in open oceans will to a large extent determine nutrient concentrations in surface waters at the onset of the spring bloom each year, and our data also show that the reduction in silicate and nitrate is reflected in surface waters (0–50 m) in winter (Figures 2.3.2 and 2.3.3). The largest decline appears during the first half of the period investigated, coinciding with a major weakening of the Subpolar Gyre Index (Figure 2.3.5), as the index decreased strongly from 1994 to 1998. The SPG continued to decrease from 2000 and throughout the period, and this is also evident in the nutrient concentrations in the Norwegian and the Barents Seas (Figures 2.3.2 and 2.3.3). We observed concurrent increases in nitrate concentrations during periods of minor strengthening of the SPG (2005–2009 and 2012–2015), while there was no uniform and significant reversals in our 5-year bin-averaging of deep water and surface silicate concentrations during these periods.

Figure 2.3.5. Time series of silicate concentrations (top panel, blue lines) and nitrate concentrations (bottom panel, blue lines) in the Norwegian Sea (solid line), the Barents Sea (dashed line) and the Greenland Sea (dash-dotted line) in the 100–200 m depth range. The data were obtained from product ref. 2.3.1; 2.3.2. Black lines show the annual Subpolar Gyre Index (data from Berx and Paye ). Values for the Norwegian and Barents seas represent January-March averages, and values from the Greenland Sea are annual deep water averages. Note that data from the Barents Sea only contain values from 1992 prior to the year 2000.

We acknowledge that differences in the spatial sampling distribution between 5-year periods may affect our results by imposing artificial biases. However, most of the periods have similar spatial sampling distribution as depicted in Figure 2.3.1, with a few exceptions: In the Norwegian Sea, there are only data available from the Ocean Weather Station Mike (located at 66° N, 2° E) during the second period (1995–1999) and there are no observations from the Barents Sea during that period. Moreover, in the Greenland Sea there are only two stations (located in the southeastern corner) present in the fifth pentad (2010–2014). Thus, the higher silicate concentrations and lower nitrate:silicate-ratios seen in the Norwegian Sea in the second period is likely to be partly due to a spatial sampling bias compared with the other periods. Since we are using samples from Atlantic Water only (in the Norwegian and Barents seas), we argue that any spatial sampling bias is expected to be small as the samples nevertheless represent waters of similar origin. However, there could still be some effects of a time-lagged response to changes in the source waters when we are sampling the core of the inflow along the shelf slope versus the ocean basins, due to shelf-basin exchange processes and residence times in the basins, but the use of 5-year bins is chosen in order to dampen such effects.

There is a statistically significant decline in Arctic deep water silicate in the Greenland Sea appearing approximately 15 years after the initial lowering of silicate and nitrate in Atlantic Water in the Norwegian and Barents Seas. Water entering the Greenland Sea can come from the North-Atlantic, the Arctic through the Fram Strait, or as a mixture of the two regions. The lack of correlation between silicate concentrations and sea temperature do suggest that the source water in the Greenland Sea gyre is of Arctic origin and entering the area through the Fram Strait. Also, since we observed a delay in decline of silicate in the Greenland Sea, relative to the adjacent Norwegian and Barent Seas (Figure 2.3.2), we may rule out the North-Atlantic as a source. Water travelling from the Norwegian and Barents Seas, through the Polar Basin and the East Greenland Current, to the Greenland Sea, has an approximate periodicity of several years to a few decades, depending on the depth (Schlosser et al. ). Therefore, the decadal-scale time lag of the decrease in nutrients in the Greenland Sea (Figure 2.3.2) suggests that nutrient concentrations in Arctic waters (indirectly spurred by changes in the Norwegian and Barents Seas) are also determining bioavailable silicate and nitrate in the Greenland Sea.

The decline in biologically available silicate throughout the northern North Atlantic in surface waters, has been attributed to changes in the SPG circulation (Hátún et al. ). This appears to concur with the decline in silicate, and nitrate, concentrations reported from the Norwegian and Barents Seas in this study, and is similar in scope to the observations made by Johnson et al. () for nitrate and phosphate at the southern entrance to the Nordic Seas, in the Rockall Trough area. The drop in nutrient concentrations have already been attributed to a weakening and westward movement of the Subpolar Gyre Index (Rey ; Johnsen et al. ; Hatun et al. ) as less nutrient rich North Atlantic water has entered the Nordic Seas. Therefore, our reported decline in silicate is similar to other findings in the same region and concur with an almost unidirectional weakening of the Subpolar Gyre Index in the period investigated (Figure 2.3.5). The rebound of nitrate during the latter half of the time-series, most prominent in the Norwegian Sea (Figure 2.3.4), coincides with a levelling off in the SPG in the 2005–2009 period. Since a weakening of the SPG appears to increase the inflow of southern, nutrient poor water, we assume that the adverse may diminish and perhaps alter the Nordic Seas inflow from the south. McCartney and Mauritzen () concluded that the Nordic Seas inflow originate from upper ocean (0–800 m) subtropical waters, the subtropical mode water (STMW). Although silicate concentrations are similar in the gyre, measured nitrate concentrations in the eastern STMW are low (1–2 µmol/L) whereas they appear five times higher in the western part of the subtropical gyre (Garcia et al. ). Therefore, Nordic Seas inflow originating from the eastern STMW will carry a very different nutrient signature (i.e. the nitrate:silicate-ratio) to the Norwegian Sea, than water from the western side. If this difference in nitrate concentrations (and hence nitrate:silicate-ratios) is reflected in the Nordic Seas inflow, our results may suggest that a weakening of the SPG and increased inflow from the south to the Norwegian Sea, may carry proportionately more STMW of eastern origin. Adversely, we suggest that a levelling off or strengthening of the SPG will carry less STMW of eastern origin. The initial drop in Norwegian Sea nitrate levels took place in 2005–2009 (Figure 2.3.3) coinciding with a levelling off, but not strengthening, of the SPG index (Figure 2.3.3). Therefore, we suggest that this halt in the weakening of the SPG could have altered the subtropical mode water source and hence, caused a Nordic Seas inflow with a lower nitrate:silicate-ratio during that period of time. The drop in the SPG index is not as dramatic and unidirectional after 2009 as prior to 2005 (Figure 2.3.5), and this ubiquity may have retained existing inflow patterns leading to a continued rise in the nitrate:silicate-ratios in the Norwegian and Barents Seas during the remainder of the time investigated. Phytoplankton community composition and zooplankton grazing

During an annual growth season, Officer and Ryther () suggested two basic phytoplankton communities; the diatom-dominated ones and flagellated (non-diatom) primary producers. Due to their rapid growth at high nutrient concentrations (a typical scenario for early spring bloom conditions in these waters), the former community is outcompeting other phytoplankton at silicate concentrations >2 µmol/L, whereas flagellated cells only appear to flourish at silicate concentration <2 µmol/L (Furnas ; Egge and Aksnes ). However, more recent studies of the inflow waters to the Norwegian Sea (e.g. Daniels et al. ), may suggest that large diatom spring blooms at times are preceded by smaller phytoplankton with higher nutrient affinity (picoeukaryotes and nanoplankton, including nanosized diatoms) in the North Atlantic. Daniels et al. () suggested that not only physicochemical factors may determine regional spring blooms dynamics, as biological factors (e.g. microzooplankton grazing pressure, presence–absence of larger diatom seed populations) also may control bloom-formations in this region. Large diatom spring blooms in the Nordic Seas are often followed by flagellated, non-diatom phytoplankton, only interrupted by smaller episodes of minor diatom blooms towards the end of the growth season in the fall (McQuatters-Gollop et al. ). Therefore, as the spring bloom is depleting nutrients and silicate concentrations goes below 2 µmol/L, diatoms are not competitive with other phytoplankton until new nutrient intrusions (and silicate conc. >2 µmol/L) in surface waters at the end of the growth season in the fall. Zooplankton growth and development is closely tied in with feeding efficiency (Pond et al. ) and measured clearance and ingestion rates for Calanus finmarchicus are at a peak during diatom spring blooms in the Norwegian Sea (Meyer-Harms et al. ). Growth development and reproduction of zooplankton, such as C. finmarchicus in the Norwegian and Barents Seas, is highly dependent of the timing and magnitude of the spring diatom blooms in these oceans (Melle et al. ; Eiane and Tande ). Juvenile and to some extent adult fish, such as commercial stocks of herring, cod and capelin in the Norwegian and the Barents Seas, are in turn strongly dependent on successful zooplankton growth and reproduction each year (Skjoldal et al. ; Gjøsæther ). Therefore, lowered initial silicate concentrations may lead to less extensive spring diatom blooms and possibly, less successful growth and survival of zooplankton and, ultimately, fish larvae from commercial stocks in these oceans.

Chemosensory detection of phytoplankton means that zooplankton can discriminate between toxic and non-toxic cells, even within the same species expressing different levels of toxicity (e.g. Selander et al. ). The chemosensory capabilities in zooplankton are also used to assess food particle quality in order to optimise the intake of high quality foods (Mayzaud et al. ; Olsen et al. ). We still do not have a complete understanding of all the ques that determine ‘high quality foods’ but copepods are also known to be able to discriminate different growth phases (and hence, a potentially different nutritional quality) in the same species of phytoplankton (Barofsky et al. ). This change in the ‘phycosphere’ (Moore et al. ) surrounding each phytoplankton cell will, amongst others, depend on nutrient availability and associated physical/chemical stress factors, that can create different reproductive outcomes and growth in zooplankton. Changes in the relative proportion of available nutrients (e.g. the nitrate:silicate-ratio) may also create stress in phytoplankton (diatoms) who are depending on silicate for cellular growth. Calculated nitrate:silicate-ratios in this study (Figure 2.3.3) were well above a molar ratio of 2 and hence, proportionately more nitrate than silicate was available at the onset of the spring bloom in all three regions investigated. This may indicate however, that if the decline in silicate are continuing, and on a larger scale than e.g. the decline in nitrate, the requisite for successful diatom spring blooms may also diminish. Successful reproductive outcomes for zooplankton (egg production and survival of nauplii) is highly dependent on access to an abundance of diatoms of optimal nutritional value (Jonasdottir ; Irigoien et al. ). Therefore, limited access to silicate for diatoms growth and development, may hamper zooplankton egg production and survival of zooplankton nauplii. If the negative Subpolar Gyre Index persist, the Nordic Seas may see longer periods of surface waters with silicate levels below 2 µmol/L (critical for successful diatom growth) and reduced growth and recruitment in zooplankton (e.g. Calanus sp.) may in turn have consequences for growth and development of fish larvae of commercially important stocks in these waters

Section 2.4. Eutrophic and oligotrophic indicators for the North Atlantic Ocean

Authors: Silvia Pardo, Shubha Sathyendranath, Trevor Platt

Products used:

Statement of main outcome: We have presented a satellite-based map of eutrophic/oligotrophic flags for the North Atlantic, for the year 2019. The flags were generated on the basis of comparison with the chlorophyll climatology for the area based on data from 20 previous years (1998–2017). The results showed hardly any localities where the eutrophic flag was positive, but some locations were positive for oligotrophic flag. Oligotrophic flags were positive mostly along coastal waters, but also along scattered points within the 30–40°N latitudes. They point to localities that should be on a watch to determine whether the trend is sustained into the future.

2.4.1. Introduction

Eutrophication is the process by which an excess of nutrients – mainly phosphorus and nitrogen – leads to increased growth of plant material in an aquatic body. Anthropogenic activities, such as farming, agriculture, aquaculture, industry and sewage, are the main source of nutrient input in problem areas (Jickells ; Schindler ; Galloway et al. ). Eutrophication is an issue particularly in coastal regions (Malone and Newton ) and areas with restricted water flow, such as lakes and rivers (Howarth and Marino ; Smith ). The impact of eutrophication on aquatic ecosystems is well known: nutrient availability boosts plant growth – particularly algal blooms – resulting in a decrease in water quality (Anderson et al. ; Howarth et al. ). This can, in turn, cause death by hypoxia of aquatic organisms (Breitburg et al. ), ultimately driving changes in community composition (Van Meerssche and Pinckney ). Eutrophication has also been linked to changes in the pH (Cai et al. ; Wallace et al. ) and depletion of inorganic carbon in the aquatic environment (Balmer and Downing ). Oligotrophication is the opposite of eutrophication, where reduction in some limiting resource leads to a decrease in photosynthesis by aquatic plants, which might in turn reduce the capacity of the ecosystem to sustain the higher organisms in it.

Eutrophication is one of the more long-lasting water quality problems in Europe (OSPAR ICG-EUT, ), and is on the forefront of most European Directives on water-protection. Efforts to reduce anthropogenically-induced pollution resulted in the implementation of the Water Framework Directive (WFD) in 2000 (Carvalho et al. ). In a similar way, the more recent Marine Strategy Framework Directive (MSFD) established in 2008 a requirement for EU member states to report on eutrophication and other water quality parameters for their regional seas for directive review purposes every 6 years. Various international conventions (e.g. OSPAR, Helsinki and Barcelona Conventions) and commissions (e.g. HELCOM) promote ecological status monitoring in order to enforce said water directives.

As a proxy for phytoplankton biomass, chlorophyll concentration is frequently used in ecological status assessments as an indicator for eutrophication, either on its own (Ferreira et al. ; Van der Zande et al. ) or as part of multi-metric ensembles (Murray et al. ; Papathanasopoulou et al. ). While in situ sampling is a powerful tool for the acquisition of reference chlorophyll concentration baseline values, the resulting datasets frequently lack the temporal and spatial resolution needed for effective monitoring and subsequent management of the problem. This is particularly true for traditionally undersampled open-ocean regions, where the risk of eutrophication is considered very low.

The use of remotely-sensed ocean colour for eutrophication monitoring provides some advantages in temporal and spatial coverage, and can help to reduce the cost of implementation of water directives (NOWPAP ; Ferreira et al. ). Several studies have exploited satellite-derived chlorophyll concentration for the study of the eutrophication status of European regional seas, e.g. Cristina et al. () for the Iberian Seas, Harvey et al. () and Attila et al. () for the Baltic Sea, Novoa et al. () for the Bay of Biscay, Lefebvre et al. () and Gohin et al. (, ) for the English Channel, and Coppini et al. () for the global region, among many others.

Many of the studies cited above use percentile-derived thresholds to classify water bodies according to their ecological status. Recent efforts such as the Joint Monitoring Programme for the North Sea and the Celtic Sea (JMP NS/CS) and the Joint Monitoring Programme of the Eutrophication of the North Sea with Satellite data (JMP-EUNOSAT) have highlighted the need for a coherent and unified method to derive eutrophication indicators, as well as the importance of using a well-validated, high-quality satellite chlorophyll product (Baretta-Bekker et al. ; Blauw et al. ; Van der Zande et al. ) to compute them. In this work we follow these recommendations to develop an indicator suite based on chlorophyll P90 and P10 percentiles, as described in the following methods section. As an illustration, we use these indicators to report on the status of the CMEMS North Atlantic region during 2019.

2.4.2. Method

We have derived a suite of annual eutrophic and eutrophic indicator maps for the North Atlantic Ocean using satellite-derived chlorophyll concentration provided in the CMEMS North Atlantic OC-CCI REP product (product reference 2.4.1). The chlorophyll 90 percentile (P90) and chlorophyll 10 percentile (P10) were used to derive, respectively, the eutrophic and oligotrophic indicators. P90 and P10 are typically defined as dynamic thresholds such as 90% of the chlorophyll values are below the P90 value, and 10% of the chlorophyll values are below the P10 value. P90 is considered a good indicator of high chlorophyll episodes (Park et al. ), and is a standard metric in eutrophication status assessments in the region (Gohin et al. ). While P10 is not widely used in the context of eutrophication studies, it has been exploited in conjunction with P90 to provide the baselines of the annual chlorophyll cycle (Gohin et al. ).

Using the satellite-derived chlorophyll products distributed in the regional North Atlantic CMEMS REP Ocean Colour dataset (OC-CCI), we computed a set of daily P90 and P10 climatologies on a pixel-by pixel basis for the region of interest, as done in Gohin et al. (). The period selected for the climatology was 1998–2017. Most existing regional studies compute these percentile climatologies over a locally-defined productive season, i.e. March to October the Bay of Biscay and eastern English Channel in Gohin et al. (). In order to apply the method to a broader area such as the CMEMS North Atlantic region, in this paper we avoided the definition of a regional productive season and used the whole year in the calculation of the P90 and P10 climatologies instead.

The region covered by the CMEMS North Atlantic OC-CCI REP product (product reference 2.4.1) is characterised by strong seasonality in both concentration values and cloud cover, which might lead to irregular sampling (Sathyendranath et al. ). These effects have been shown to be the cause of relative errors of up to 30% in the estimation of P90 (Van der Zande et al. ). To minimise the effect of gaps in the data in the computation of these P90 and P10 climatological values, we imposed a threshold of 25% valid data for the daily climatology. For the 20-year 1998–2017 climatology this means that, for a given pixel and day of the year, at least 5 years must contain valid data for the resulting climatological value to be considered significant. Pixels where the P90 and P10 climatological values were non-significant were not considered in further calculations.

To assess the average regional status during 2019 when compared with these P90 and P10 percentile baselines, we computed time series of the 2019 daily area-averaged chlorophyll concentration, daily P90 climatology and daily P10 climatology. For consistency with the method employed to derive chlorophyll time series in previous Ocean State Reports (Sathyendranath et al. ), pixel-area weighted averages were used to calculate the daily values.

Finally, to generate the eutrophic and oligotrophic indicator maps for 2019, we compared every valid daily observation over the year with the corresponding daily climatology on a pixel-by-pixel basis, to determine if values were above the P90 threshold, below the P10 threshold or within the [P10, P90] range. Values above the P90 threshold or below the P10 were flagged as anomalous. The number of anomalous and total valid observations were stored during this process. We then calculated the percentage of valid anomalous observations (above/below the P90/P10 thresholds) for each pixel, to create percentile anomaly maps in terms of % days per year. Lastly, we derived an annual eutrophic/oligotrophic flag map: if 25% of the valid observations for a given pixel and year were above the P90 threshold, the pixel was flagged as eutrophic. Similarly, if 25% of the observations for a given pixel were below the P10 threshold, the pixel was flagged as oligotrophic.

2.4.3. Results

Figure 2.4.1 shows the 2019 North Atlantic daily average time series for chlorophyll concentration (black), together with the P90 (red) and P10 (blue) time series of daily average climatologies, obtained using the CMEMS North Atlantic OC-CCI REP product (product reference 2.4.1). The daily average chlorophyll time series showcased primary (spring) and secondary (early winter) peaks consistent with previous assessments of the North Atlantic chlorophyll seasonal cycle (Sathyendranath et al. ; CMEMS OMI ). Changes in cloud cover and satellite coverage affect data availability and can cause considerable day-to-day variability in the time series, as shown in Figure 2.4.1. While the mean variability in daily coverage was less than 3% during 2019, it is worth mentioning that the average number of valid pixels during the winter months (November-January) was almost 10% lower than the average number of valid pixels during the rest of the year.

Figure 2.4.1. 2019 daily chlorophyll average for the North Atlantic Ocean (black), daily P90 (red) and P10 (blue) 1998–2017 climatological values, calculated using the CMEMS Ocean Colour ATL REP dataset (OC-CCI, product reference 2.4.1).

The percentile parameters also presented a distinct seasonality that matches the chlorophyll concentration; in particular P90 exhibited a strong spring bloom signal peaking in May. The average chlorophyll concentration in the area for 2019 was 0.24 mg m−3, against values of 0.37 mg m−3 average P90 and 0.17 mg m−3 average P10. During 2019, the average daily chlorophyll concentration was well below the P90 climatological values during the first and second quarters of the year, and closer to the percentile climatologies during the second half of 2019. The analysis of the mean anomaly values revealed that chlorophyll was on average 31% lower than the P90 climatology and 48% higher than the P10 climatology during 2019.

While the time series comparison provides a comparison of 2019 with the average regional P90 and P10 climatologies, it offers very little information regarding the spatial location and significance of potential eutrophication and oligotrophication episodes. As we detailed in the methods section, to identify eutrophication and oligotrophication levels for the year we assessed the P90 and P10 anomalies on a daily pixels-by-pixel basis, and generated eutrophic and oligotrophic indicator maps based on the percentage of anomalous observations. The 2019 eutrophic indicator map for the North Atlantic is shown in Figure 2.4.2. On average, 6.4% of the valid observations acquired in the region were above the P90 climatological value (i.e. in a eutrophic state). For the North Sea, Bay of Biscay and coastal waters of Portugal and Spain, this value increased to 15–20% of the valid observations being above the P90 climatological value. Waters around the Canary Islands, west coast of Morocco, and some hotspots in the open ocean showcased values above the 25% threshold, reaching 50% of the valid observations in some cases (i.e. around 182 days for regions with continuous coverage). Note that a considerable portion of the region above 40 degrees North did not provide a measure of P90 anomaly. This was due to gaps in data lowering the significance of the P90 below the 5-year threshold we imposed in the climatology calculation. The eutrophic indicator distribution in this annual average map was consistent with the 2019 anomaly (Sathyendranath et al. ; CMEMS OMI ), particularly with the positive anomalies reported for the Bay of Biscay and southern coast of Ireland.

Figure 2.4.2. 2019 eutrophic indicator map for the North Atlantic Ocean, in terms of percentage of days with chlorophyll values above the 1998–2017 P90 climatological reference, calculated using the CMEMS Ocean Colour ATL REP dataset (OC-CCI, product reference 2.4.1).

The 2019 oligotrophic indicator map for the North Atlantic is shown in Figure 2.4.3. On average, 12.7% of the valid observations acquired in the region were below the P10 climatological value (i.e. in an oligotrophic state). Values for this indicator appeared to be polarised, with regions showing either no decrease in chlorophyll – less than 5% of the valid observations are below the P10 reference value – or values above the 25% threshold. Areas in the latter category included the coastal waters of Norway, Sweden and Denmark, the Iberian Shelf, the Celtic and Irish Seas, and the English Channel. Maximum values in hotspots for these areas were around 70% of the valid observations (i.e. around 255 days for regions with continuous coverage).

Figure 2.4.3. 2019 oligotrophic indicator map for the North Atlantic Ocean, in terms of percentage of days with chlorophyll values below the 1998–2017 P10 climatological reference, calculated using the CMEMS Ocean Colour ATL REP dataset (OC-CCI, product reference 2.4.1).

The oligotrophic indicator distribution in the annual average map was consistent with the negative anomalies found in the 2019 anomaly map for the same product and region (Sathyendranath et al. ; CMEMS OMI ), particularly with the strong decrease in chlorophyll concentration reported for the Iberian Shelf seas. The higher percentage values for this second indicator might be due to a non-proportional distribution of the observation between bloom and non-bloom periods, i.e. higher frequency of non-bloom observations.

Using a 25% threshold, we derived the final 2019 annual eutrophic/oligotrophic flag indicator map for the North Atlantic, shown in Figure 2.4.4. Extensive coastal and shelf waters showed active oligotrophic flags for 2019, with areas including but not restricted to the Iberian Shelf waters, North Sea, Celtic Sea and Irish Sea. The results for the English Channel and the French Atlantic continental shelf are consistent with the decrease in eutrophication risk recently reported by Gohin et al. () using both in situ and satellite chlorophyll data.

Figure 2.4.4. 2019 annual eutrophic (red) and oligotrophic(blue) flag indicator map calculated using the CMEMS Ocean Colour ATL REP dataset (OC-CCI, product reference 2.4.1). Active eutrophic flags indicate that more than 25% of the valid observations were above the 1998–2017 P90 climatological reference. Active oligotrophic flags indicate that more than 25% of the valid observations were below the 1998–2017 P10 climatological reference.

The flag indicator map showed very few areas with active eutrophic flags for 2019. The Third Integrated Report on the Eutrophication Status of the OSPAR Maritime Area (OSPAR ICG-EUT, ) reported an improvement from 2008 to 2017 in eutrophication status across offshore and outer coastal waters of the Greater North Sea, with a decrease in the size of coastal problem areas in Denmark, France, Germany, Ireland, Norway and the United Kingdom. The absence of active eutrophic flags in the North Sea coastal regions might reflect these trends and a low occurrence of eutrophic episodes during 2019.

2.4.4. Conclusions

The indicators developed here provide a tool to monitor spatial and temporal variations in chlorophyll concentration. While various regions have been flagged as eutrophic/oligotrophic in the 2019 indicator map (Figure 2.4.4), we can only talk about eutrophication/oligotrophication when this state is sustained and showcases a significant trend across a longer period of time. Following studies will only benefit from progressive temporal extensions of the dataset (improving the significance of the climatologies) and from the incorporation of high-resolution satellite ocean colour datasets – such as Sentinel3 OLCI – into the CMEMS OC-CCI REP dataset.

From the results obtained in Figure 2.4.4, we can conclude that the occurrence of eutrophic episodes (chlorophyll concentration higher than the P90 climatological value for more than 25% valid observations) was very low during 2019. Nonetheless, the eutrophic indicator provided in Figure 2.4.2 included some significant P90 anomalies in the 15% to 25% range, and lowering the threshold from 25% to 15% would result in an increase in the areas flagged as eutrophic in Figure 2.4.4. This implies that the 2019 eutrophic episodes in those regions were shorter in duration than their typical phytoplankton growth season. While most regional studies identify the growth season as the standard period for heightened eutrophication risk, and employ it to compute P90 climatologies (Gohin et al. ), this paper employed whole-year datasets for the calculation of the P90 and P10 climatologies. This approach was motivated by the need to report on the broad CMEMS North Atlantic region, which includes both coastal and open waters characterised by different productive periods.

It is worth noting that the absence of flags in the top left quarter of Figure 2.4.4 was due to the low significance of the P90/P10 indicators. The impact of satellite data quality and availability in the accuracy of eutrophication indicators has been highlighted for various ocean colour datasets (Gohin et al. ; Ha et al. ), with Van der Zande et al. () demonstrating that the mean relative error in the estimation of the P90 values can reach 25.4% for MERIS. While the use of the climate-grade CMEMS OC-CCI REP dataset ameliorates these issues by merging various bias-corrected ocean colour streams (increasing the coverage and sampling frequency), some well-known eutrophication problem areas seem to be below our threshold of detection using the percentile climatology method presented here. This is the case for the coastal waters of the Skagerrak and Kattegat straits, where the significance of the percentile climatologies was too small to confirm previous trends reported for these problem areas (Andersen et al. ).

Section 2.5. Nitrate, ammonium and phosphate pools in the Baltic Sea

Authors: Mariliis Kõuts, Ilja Maljutenko, Ye Liu, Urmas Raudsepp

Statement of outcome: Eutrophication is a challenge in the Baltic Sea, with estimates of annual total input of about 830 kT of nitrogen and 31 kT of phosphorus in 2014. Nutrient and oxygen pools were estimated using model reanalysis data. During the period of 1993–2017, the pelagic nitrate pool decreased from ∼2400 to 1700 kT in the Baltic Sea. The reduction of the nitrate pool was uniform over the Baltic Sea, with the exception of the Gulf of Bothnia, where decrease in the intermediate layer was small or even turned into an increase in the deep layer. Pelagic ammonium pools increased in the deep layer of the Baltic Proper due to decreasing oxygen concentrations there. The pelagic phosphate pool increased from ∼600 to 750 kT, with most of the increase taking place in the surface and intermediate layers. The pool in the deep layer remained on a more stable level, with only a slight increasing trend. The increase of the phosphate pool covers the Baltic Proper area. In the Gulf of Bothnia, a decrease of phosphate has occurred in the upper layers and a slight increase in the deep layer. Only slight changes in the oxygen pool can bring about significant changes in the nutrient pools. Decreasing dissolved inorganic nitrogen pools and increasing phosphorus pools in the water column across the entire Baltic Sea, with the exception of the Bothnian Bay, results in a decreasing nitrogen to phosphorus ratio.

2.5.1. Introduction

Eutrophication is a major challenge in the Baltic Sea, with at least 97% of the region assessed as eutrophied in 2011–2016 according to the HELCOM thematic assessment of eutrophication (HELCOM ). However, the validity of this statement is questionable, as the least eutrophied basin – the Gulf of Bothnia – constitutes as much as 25.16% of the Baltic Sea volume. The HELCOM Eutrophication Assessment (HELCOM ) featured nutrient levels, specifically winter dissolved inorganic phosphate and nitrogen, in the upper 10 m of the water column. According to the assessment, dissolved inorganic nitrogen varied from 3–5 μmol l−1 in the Bothnian Bay to ∼10 μmol l−1 in the Gulf of Riga. A similar pattern occurred with dissolved inorganic phosphorus levels, which varied from 0.04 μmol l−1 in the Bothnian Bay to ∼1.3 μmol l−1 in the Gulf of Riga. The hypotheses about the Gulf of Bothnia showing a tendency to become eutrophied are currently emerging (Lundberg et al. ; Fleming-Lehtinen et al. ; Andersen et al. ).

Eutrophication is caused by the addition of nutrients from land that accumulate in the marine system. In 2014, the Baltic Sea received an annual total input of about 826,000 tons of nitrogen and 30,900 tons of phosphorus (HELCOM ). Atmospheric inputs account for about 30% of total nitrogen inputs (HELCOM ). The input has decreased since the 1980s to the level of the 1960s for nitrogen and to the level of the 1950s for phosphorus by 2014 (HELCOM ). The importance of nutrients is illustrated by different trophic states the ecosystem can evolve into, depending on the scale of nutrient inputs (Meier et al. ; Saraiva et al. ; Murray et al. ). History has shown that a massive input of nutrients into any lake or sea, including the Baltic, results in a disrupted ecosystem that eventually brings about eutrophication (Gustafsson et al. ; Murray et al. ). The aforementioned reductions in nutrient inputs are expected to be reflected in the time series of the pools.

The aim of this study is to assess the nutrient status of the Baltic Sea using model reanalysis data for the period 1993–2017. Based on ur results, we can evaluate the possible problem areas as well as the success of mitigation measures so far. We can evaluate the state-of-the-art quality of the reanalysis in the Baltic Sea. We do not claim that our analyses are unique, but rather, we intend to provide another set of estimates of total nutrient pools to the family of already existing estimates. Having an ‘ensemble’ of estimates obtained with different tools narrows the interval of uncertainties around the ‘true’ value.

2.5.2. Materials and methods

The BALMFC CMEMS biochemistry reanalysis product (product reference 2.5.1) is calculated using the Nemo-Nordic physical model (Hordoir et al. ; Pemberton et al. ) coupled with the Swedish Coastal and Ocean Biogeochemical model (SCOBI) (Eilola et al. ; Almroth-Rosell et al. ). The model system uses the Localised Singular Evolutive Interpolated Kalman filter data assimilation method (LSEIK, Nerger et al. ).

The model domain consists of the North Sea and the Baltic Sea (Hordoir et al. ). The horizontal resolution of the Nemo-Nordic model is approximately 2 nautical miles, and there are 56 vertical levels. Vertical resolution varies from 3 m at the surface up to 10 m below the depth of 100 m. The numerical model simulation data for this research covers the period 1993–2017 (product reference 2.5.1). At the lateral boundaries in the western English Channel and along the Scotland-Norway boundary, the sea levels are prescribed using 24 nautical mile resolution storm-surge North Atlantic Model. Climatological monthly mean values of temperature, salinity and biogeochemical variables are used at the open boundary. Atmospheric deposition of biogeochemical substances is supplied as spatial maps of a monthly climatology.

The river runoff is specified as daily means for the whole reanalysis period, 1993–2017, from the output of the HYdrological Predictions for the Environment (HYPE) model (Donnelly et al. ). The nutrient loads are specified as monthly mean values. River runoff as well as nutrient load are spread over more than 250 rivers located in the Baltic Sea (Hordoir et al. ).

The meteorological forcing is from HIRLAM (High-Resolution Limited Area Model) with a 22 km resolution, from project Euro4M (Dahlgren et al. ; Landelius et al. ), and covers most of the reanalysis period, 1993–2011. From 2012 onwards, the meteorological forcing is from the UERRA reanalysis product (European Regional analysis) with an 11 km resolution.

The sea surface temperature from the Swedish Ice Service and in-situ measured temperature and salinity profiles from the ICES database ( are assimilated into the physical model. In-situ profiles of nitrate, ammonium, phosphate and dissolved oxygen from the Swedish Ocean Archive database (SHARK; are assimilated into the biogeochemical model. In total, 3200 nitrate, 3500 ammonium, 4000 phosphate and 7500 dissolved oxygen profiles were assimilated into the model. Spatial data coverage was strongly inhomogeneous, with high density data at the Swedish coast and in the western Baltic Sea, less so in the eastern Baltic sea, and in the Gulf of Riga and Gulf of Finland there were only a few profiles. The reanalysis has been produced using 72-hour cycling, which implies that every 72 h all available observations are assimilated into the model before a 72-hour forecast is made. When there was more than one observation per model layer, an average value of the observations in the same layer was used.

In addition to model reanalysis data validation provided in the QUID, we have compared model reanalysis data with the observation data at the central Gotland basin and in the Gulf of Riga for the period 1993–2017. The nutrients concentrations have been extracted at central Gotland basin (BMPJ1 / BY15) and from the Gulf of Riga (BMPG1/ G1) from the model simulation and compared with observational data extracted from two different databases. The observational data has been extracted from The Marine distributed databases in the NEST system (Wulff et al. ), which provides monthly mean aggregate data at reference depths, and from the EMODnet database (Buga et al. ), which provides data at measured time and depth levels.

Main statistics of the comparison at BY15 for surface and bottom nutrients are summarised in Table 2.5.1. We have used the data from the EMODnet database. The differences in the statistics for the NEST database and the model reanalysis were marginal compared to the EmodNet data. Cost function (CF) was formulated as CF   =   |(M–D)/SD|, where the bias (M–D) of the model mean (M) relative to the mean of observations (D) is normalised to the standard deviation (SD) of the observations (Eilola et al. ). Cost function values 0–1 indicate a good match between the model results and measurements, values 1–2 indicate a reasonable match and values above 2 indicate a poor match. Cost function shows that model reanalysis data are good (Table 2.5.1). In general, the model underestimates all nutrients except phosphates at the surface.

Table 2.5.1. Mean and standard deviation (SD) of the model reanalysis data and measurements, model bias (model minus observations), root mean square difference (RMSD) and cost function (CF) at the central Gotland basin (station BY15). Units are mmol/m3.

In order to assess the nutrient status of the Baltic Sea, we use the combination of time-series analyses of nitrogen and phosphorus pools in three different layers of the water column and spatial maps of vertically integrated mean concentrations for two periods: 1993–1999 and 2000–2017. The selection of these two periods is motivated by the temporal variations of the hypoxic area of the Baltic Sea. After the stagnation period, which was terminated by the Major Baltic Inflow in 1993 (e.g. Mohrholz ), hypoxia development has shown two regimes. The first period, from 1993 to 1999, represents an increase of hypoxic area from 20,000 km2 to a level of about 60,000 km2 (Savchuk ; Meier et al. ; Carstensen and Conley ; Carstensen et al. ). The second period, from 2000 to 2017, can be characterised as variations of hypoxic area around a mean level between 60,000 and 80,000 km2 (Savchuk ).

The pools of nitrate, ammonium, phosphate and oxygen have been calculated from the biochemistry reanalysis product (product reference 2.5.1). We have separated the water column into 3 layers: upper layer 0–15 m, intermediate layer 15–80 m and deep layer 80 m and below. The time-series of winter (February) nitrate and phosphate in the 0–15 m layer of the upper water column integrated over the Baltic Sea is an estimate of nutrient availability for phytoplankton and the concurrent transformation into biomass during the productive seasons. The upper layer, down to 15 m, characterises the summer mixed layer depth where the nutrients are depleted by plankton production. The intermediate layer characterises nutrient pools down to the average wintertime mixed layer depth, i.e. the depth to which the water column is seasonally mixed. The deep layer characterises the nutrient pool which is within or below the permanent halocline and, therefore, not easily mixed into the euphotic layer and accessible to phytoplankton production. Time-series of nitrate and phosphate pools under the halocline indicate the intensity of removal processes and storage of nutrients in the deep parts of the Baltic Sea.

Time series of pelagic pools are calculated as spatial integrals of nitrate, ammonium, phosphate and oxygen concentrations from the daily mean fields of product reference 2.5.1 for the whole Baltic Sea and for the three different layers separately. The dissolved inorganic nitrogen pool is taken as the sum of nitrate and ammonia. The spatial distributions of nutrient pools (g/m2) are calculated as vertical integrals over different layers: 0–15 m, 15–80 m, 80 m to the bottom.

The sediment processes are included in the model (Almroth-Rosell et al. ), but the nutrient pools in the sediments are not included in the product reference 2.5.1 output list. Thus, the dynamics of benthic pools of nutrients is not analysed in this study.

2.5.3. Results

Over the period of 1993–2017, the nitrate pool decreased monotonically from about 2500 kT to 1400 kT in the Baltic Sea (Figure 2.5.1a). In the upper layer, peak values in winter fell from 800 kT to 400 kT. In the intermediate layer, which constitutes the largest part of the nitrate reserve in the water column, the drop was from ∼1600 to 800 kT. The nitrate pool in the deep layer decreased from ∼600 to ∼400 kT.

Figure 2.5.1. Stacked plot of time series of summarised dissolved nutrient pools: nitrate on (a), ammonium on (b), phosphate on (c) and dissolved oxygen on (d) in three different layers (Upper layer 0–15 m: blue; Intermediate layer 15–80 m: yellow; and Deep layer starting at 80 m and extending to the bottom: red) across the entire Baltic Sea. Time period is 1993–2017 (Product reference 2.5.1).