Rapid global warming over the past few decades has had consequences for weather, climate, ecosystems, human society and economy (IPCC, 2019). More heat available in the climate system is manifested in the oceans in many ways including increasing the ocean interior temperatures (Johnson et al., 2018; Cheng et al., 2019a), raising the sea level (Nerem et al., 2018), melting the ice sheets and permafrost (Shepherd et al., 2012; Meredith et al., 2019), altering the hydrological cycle (Durack et al., 2012), changing the atmospheric and oceanic circulation (Rahmstorf et al., 2015; Caesar et al., 2018), supporting stronger tropical cyclones with heavier rainfall (Trenberth et al., 2018), among others. Higher ocean heat content and sea surface temperatures invigorate tropical cyclones to make them more intense, bigger and longer lasting, and greatly increase their flooding rains.

In addition to global warming, rising concentrations of carbon dioxide (CO₂) in the atmosphere have a direct effect on the chemistry of the ocean through the absorption of CO₂ by surface waters. The oceans have absorbed about 25% of all CO₂ emissions since the pre-industrial period (Le Quéré et al., 2016; Gruber et al., 2019a,b; Friedlingstein et al., 2020). Increased CO₂ in the water lowers its pH, termed ocean acidification, making it harder for some marine organisms such as corals, oysters and pteropods (Hoegh-Guldberg et al., 2017; Lemasson et al., 2017) to form calcium carbonate shells and skeletons. In some cases, ocean acidification has also been shown to lower fitness in some species such as coccolithophores, crabs, sea urchins and early life stages of fishes (Baumann et al., 2012; Dodd et al., 2015; Campbell et al., 2016; Stiasny et al., 2016; Riebesell et al., 2017; Tasoff and Johnson, 2019). Research efforts over the past decade have built considerable understanding of how marine species, ecosystems, and biogeochemical cycles may be influenced by ocean acidification alone and in concert with other stressors including eutrophication, warming, and hypoxia (Breitburg et al., 2015; Baumann, 2019). Natural variability in carbonate chemistry, such as coastal upwelling and seasonal fluctuations in primary productivity, is also compounded by anthropogenic changes to create particularly extreme ocean acidification conditions in some regions of the global ocean (Feely et al., 2008; Cross et al., 2014).

Oxygen is the basis of life for the vast majority of all oceanic organisms and thus oceanic oxygen levels define habitat boundaries for marine life. Still oxygen can only be gained in the upper most waterlayers by photosynthesis or air sea gas exchange. Once a water mass has left the surface, oxygen is decreasing due to consumption. Global warming does reduce oxygen solubility at the surface, reducing the initial amount of subducted and convected oxygen. Furthermore, upper ocean warming has an impact on biological activity, oceanic stratification and overturning and other processes, which all in turn have the potential to decrease oceanic oxygen levels (Schmidtko et al., 2017; Stramma and Schmidtko, 2021).

Sea ice at the poles plays a critical role in maintaining global heat balance. Shortwave radiation from the sun bears down on the equator, while the global atmospheric and ocean circulations carry this heat to the relatively colder poles. The high albedo of sea ice and the cryosphere allows the global system to more effectively reflect insolation and radiate longwave heat to moderate the global heat balance. The loss of ice in the cryosphere (Meredith et al., 2019) lowers the planetary albedo, allows more heat from the ocean to flux to the atmosphere through the thinner sea ice and the more expansive areas of open water, and reduces Earth’s ability to maintain global heat balance. Sea ice also plays a role in the fresh water and salt budget of the global ocean (Polyakov et al., 2020). Salt is expelled in areas of sea ice growth; this ice drifts with the winds and ocean currents transporting fresh water to areas where it may melt during summer. Sea ice has also a significant impact on wildlife, many species depend on the sea ice for habitat, subsistence, and culture (e.g., Meier et al., 2014; Thoman et al., 2020).

Global mean sea level encompasses several processes and climatic systems. Global mean sea level rise is comprised of the change in the sea water volume due to global temperature rise (the thermosteric component) and the change in sea water mass (the barystatic component). The latter is the sum of the melting of ice sheets (Antarctica, Greenland), glaciers and of the input to the sea of terrestrial water storage (e.g., Gregory et al., 2019; Frederikse et al., 2020). Sea level rise poses a significant threat to low lying islands, coasts and communities around the world through inundations, the erosion of coastlines and the contamination of freshwater reserves and foodcrops (Oppenheimer et al., 2019).

The Atlantic Meridional Overturning Circulation (AMOC), a large system of ocean currents in the Atlantic, is an important factor in climate variability and change for several reasons. Changes in AMOC strength can have global impacts on the oceanic carbon sink (Zickfeld et al., 2008; Fontela et al., 2016), the position of the Intertropical Convergence Zone (Timmermann et al., 2007) and, as a consequence Sahel precipitation (Mulitza et al., 2008), the Asian monsoon regions (Fallah et al., 2016), and affect marine ecosystems (Schmittner, 2005). Despite its importance, the evolution of the AMOC since the beginning of the industrial era is poorly known and the question of whether the AMOC has already been weakening in response to global warming remains unknown.

The ocean is currently in a phase of significant climate change and evaluation of the rate of change is of upmost importance. We present here a review of seven key ocean climate change indicators: (i) Sea Surface Temperature, (ii) Ocean Heat Content, (iii) Global Mean Sea level, (iv) Ocean pH, (v) Dissolved oxygen concentration (vi) Arctic Sea Ice extent, thickness, and volume and (vii) the strength of the Atlantic Meridional Overturning Circulation (AMOC). In addition to reviewing the current state of the art, we discuss some research gaps and future developments and present a final visual summary of ocean climate change indicators with emphasis in recent changes (1993–2019/20).

Ocean acidification is the anthropogenic reduction in the pH of the ocean over an extended period of time, decades to centuries. The ocean has absorbed about 25% of all CO₂ emissions (1870–2015 period; Le Quéré et al., 2018; Gruber et al., 2019a,b; Friedlingstein et al., 2020) and the increased CO₂ in the water is lowering its pH through the formation of carbonic acid (Figure 3). Increased aqueous CO₂ is also leading to an increase in bicarbonate and decrease in carbonate ions.

Global surface ocean pH has declined on average by approximately 0.1 (from 8.2 to 8.1) since the Industrial Revolution (Caldeira and Wickett, 2003; Orr et al., 2005). Jiang et al. (2019) reports a similar global decrease of −0.11 ± 0.03 pH units from 1770 to 2000. The Arctic Ocean has experienced the largest pH decrease with a pH decline of −0.16 ± 0.04 pH units (1,770–2,000). There is natural variability of the ocean’s carbonate chemistry driven by a number of natural processes such as circulation, air-sea interchange, and remineralization for example, but carbonate chemistry at global scale is being driven by the increasing carbon dioxide in the atmosphere coming from emissions and land use change. The current changes can be observed in extended ocean time series and the rate of change is likely unparalleled in at least the past 66 million years (Hönisch et al., 2012; Zeebe et al., 2016). Ocean pH is projected to decline, approximately, by an additional 0.1–0.4 pH units by the end of century (2,100) depending on the future RCP (Representative Concentration Pathway) / SSP (Shared Socioeconomic Pathways) scenarios (Caldeira and Wickett, 2003; Feely et al., 2009; Jiang et al., 2019; Kwiatkowski et al., 2020).

Carbonate chemistry varies according to large-scale oceanic features including depth, distance from continents due to land influence, upwelling regime, freshwater/nutrient input and latitude (Jewett and Romanou, 2017). Due to this variability, as determined by these various characteristics, only longer term, observational time series can detect the predicted long-term increase in acidity at individual sites due to rising atmospheric CO₂ levels. Time of emergence of the signal varies from 8 to 15 years for open ocean sites, and 16 to 41 years for coastal sites (Sutton et al., 2019), making it necessary to commit to long-term observational records, especially in the coastal zone where most commercially and culturally important marine resources reside.

The ocean can only gain oxygen at the surface by air sea gas exchange and photosynthesis. Subsurface dissolved oceanic oxygen (DO) is advected along water mass distribution pathways and mixed into adjacent water masses while being consumed by respiration. Therefore, any change in solubility at the surface (due to warming), decrease of ventilation (due to stratification increase) and increase in deep ocean respiration (due to increased surface primary production and enhanced particle flux) can lead to oceanic deoxygenation. Thus, changes in deep ocean DO can be seen as an integrative long-term indicator for profound changes in physical or biogeochemical ocean dynamics.

Since oxygen is the marine biogeochemical parameter with little to no variations in analysis methods over time, long-term oceanic DO changes can be derived robustly with relatively high confidence (Carpenter, 1965; Wilcock et al., 1981; Knapp et al., 1991). Only limited data availability may compromise robust trends in all regions and depths. Winkler titration of water samples, established in 1903, has become the method of choice for DO measurements soon after its discovery and has since been used to calibrate DO measurements of all kind of platforms. More recent developments of computer-aided Winkler titration methods that provide higher accuracy, seem not to bias historical measurements (Schmidtko et al., 2017). Furthermore, a systematic relative bias due to reagent changes in the analysis was tested and determined as highly unlikely (Schmidtko et al., 2017). Therefore, any calibrated long-term DO observation can be used to derive long term trends and multi-decadal variability of timeseries spanning a century.

Coastal changes in DO are impacted on a very local scale by regional physical, biogeochemical and anthropogenic changes. These regional changes range from riverine run-off of nutrients, deposits of organic matter over heatwaves and tides, just to name a few. An observational study (Diaz and Rosenberg, 2008) reports an increased occurrence of coastal dead zones, with consequences for regional ecology and economy. While the occurrence of most of these coastal dead zones is locally driven, some low-oxygen events may have been affected by open ocean deoxygenation, making these events more likely by lowering background DO levels.

In the open ocean most long-term time series data from monitoring stations show decreasing DO levels despite temporal variations on annual to multidecadal time-cales (e.g., Keeling et al., 2010). Time stations with long-term increasing oxygen levels exist but are sporadic.

The long term monitoring stations support the findings of three major studies of global DO changes, covering the time period from the sixties to today (Helm et al., 2011; Ito et al., 2017; Schmidtko et al., 2017). Figure 4 shows an overview of the relevant changes. Despite diverging methods all studies agree that the global ocean is losing oxygen at a significant rate. The rate of decrease does vary over depth and method but is on the order of 2% over 50 years (Schmidtko et al., 2017). This accumulates to a loss of 4.8 ± 2.1 petamoles since 1960. This loss is not homogeneously distributed in the global ocean. Oceanic oxygen loss varies with depth and region, resembling the several processes involved in oxygen distribution and consumption. All the works generally agree on the large scale deoxygenation patterns with most pronounced deoxygenation in the north Pacific and Southern Oceans with some smaller disagreement regarding the intensity of deoxygenation in the tropical oceans.

From a global perspective, temperature-driven solubility decrease is dominating the oxygen loss in the upper most water layers. A warming ocean is gaining less oxygen from air sea gas exchange. Schmidtko et al. (2017) attribute 50% of the oxygen loss in the upper 1,000 m to solubility changes. This number drops to about 25% for the upper 2,000 m and only on the order of 13% of the overall full water column oxygen loss. This solubility driven deoxygenation is attributed to the time period 1960–2010, assuming linear warming. For an accelerating warming process these numbers will likely change. Since solubility is responsible for only part of the observed changes, other processes are similarly important. Nevertheless, we cannot disregard temperature as the key source of those changes as well, since processes other than solubility change are also largely driven by a warming upper ocean. These temperature driven processes are not limited to, but do include stratification increase, circulation changes and thermal impacts on biogeochemical cycles (e.g., Keeling and Garcia, 2002; Stendardo and Gruber, 2012; Bianchi et al., 2013).

More recent analysis of available regional studies (Stramma and Schmidtko, 2021) attribute various processes to observed regional changes (Figure 5). While solubility and stratification dominate high latitudes and the Atlantic Ocean, multi-decadal variability is dominant throughout most basins. The reason for this can be seen in atmosphere-ocean indices like the North Atlantic Oscillation (NAO), El Nino-Southern-Oscillation (ENSO), and Pacific Decadal Oscillation (PDO), among many others, impacting regional ocean dynamics with subsequent influences on dissolved oceanic oxygen. Biological and nutrient stimulation causes are mainly found near the coast and in particular upwelling regions. Source water changes and circulation driven changes point to physical parameters that have shifted in ocean dynamics. Many of these processes are linked to changing ocean ventilation and respiration and are therefore challenging to appraise directly. Still, all tend to reinforce the impacts from warming (Oschlies et al., 2018).

Along with globally decreasing oceanic DO, the volume of so-called oxygen minimum zones has grown significantly. Oxygen minimum zones (OMZ) are generally defined as oceanic volume with less than 80 mmol l^–1 DO, and thus not suited as habitat for many marine organisms that rely on continuous respiration although they may provide refuge for animals that can cope with low DO conditions. In areas where the OMZ DO levels are close to or completely depleted oxygen minimum zones have potential impacts on greenhouse gas driven climate warming, since they can emit large quantities of nitrous oxide, a potent greenhouse gas, owing to denitrification processes under anoxic conditions (e.g., Codispoti, 2010; Santoro et al., 2011). Such low OMZ DO levels can be found in the Pacific and Indian Ocean and have been found expanding.

With impacts on dissolved oxygen levels varying strongly on regional scales, predictions on future local oxygen can only be established knowing all local boundary conditions and predicted changes. At basin scale or even global scale, it can be stated that an increased warming of the upper ocean has impact on oxygen levels by solubility while simultaneously reinforcing other processes that are linked to ocean dynamics and biogeochemistry, and that are responsible for the majority of the observed deoxygenation.

The retreat of Arctic sea ice has been one of the most iconic indicators of climate change (Thoman et al., 2020). Arctic sea ice extent is declining by −13.1% per decade during summer (September 1979–2020), when it exhibits its seasonal minimum extent, and by −2.6% per decade during winter (March 1979–2018) (Fetterer et al., 2017; Perovich et al., 2020). Figures 6A,B present the mean sea ice concentration in summer and winter during the last 4 decades and Figures 6C,D the sea ice concentration trends during the same period and seasons. The decreasing trends during winter (Figure 6C) are observed in all the peripheral seas around the Arctic, with the greatest decreasing trends (−26% per decade) occurring in the Barents Sea. During summer the trends (Figure 6D) are almost twice as high in the Pacific and Asian sectors of the Arctic Ocean compared to the Atlantic Sector, with the greatest decreasing trends occurring in the Beaufort Sea (−29% per decade), which has been essentially ice free during summer over the past decade. The record minimum in summer sea ice extent was measured in September 2012, and the second lowest extent in the 42 years satellite record was measured in September 2020 (Perovich et al., 2020).

Similarly, the thickness of Arctic sea ice has also decreased. In one of the first studies to document this change, using measurements from upward looking sonars on submarines, Rothrock et al. (1999) showed that the average of thickness of sea ice decreased from 3.1 meters in 1958–1976, to just 1.8 meters in 1993–1997, with the largest decreases occurring in the central and eastern Arctic. In an updated study which includes estimates of sea ice thickness from satellites, Kwok (2018) showed that the thickness has now decreased by 2.0 meters, comparing 2011–2018 ICESat and CryoSat-2 data to the 1958–1976 and 1993–1997 submarine cruise measurements, or about 66% over the six decades.

Taken together, the observed trends in sea ice extent and sea ice thickness indicate that the volume of Arctic sea ice has decreased by over 75% since 1979. This estimate is coincident with many modeling studies, including the Pan-Arctic Ice Ocean Modeling and Assimilation System (Zhang and Rothrock, 2003; Schweiger et al., 2011, 2019) which estimates that the average volume of Arctic sea ice of 11.5 × 10³ km³ in September, 1979–2010, has decreased with a rate of −2.8 × 10³ km³ per decade. The current record minimum in total ice volume estimate using PIOMAS is 3.8 × 10³ km³ set in September 2012. The summers of 2019 and 2020 are tied for second minimum with 4.2 × 10³ km³.⁵

These trends in the decline of Arctic sea ice are the result of the complex interplay between the atmosphere, sea ice and the ocean. During the winter, the cold halocline layer protects sea ice from the underlying warm Atlantic water (e.g., Steele and Boyd, 1998), allowing sea ice to grow thermodynamically driven by air temperatures which historically were around −32°C (Rigor et al., 2000). Winds and ocean currents may also change the thickness of sea ice dynamically by ridging and rafting of sea ice during storms or against a coastline. This process also creates areas of open water which would rapidly freeze over and thicken during winter, quickly increasing the thickness distribution of sea ice. Most of the sea ice in the Arctic Ocean is exported through Fram Strait, and later melts in the warmer waters of the Greenland Sea and North Atlantic. Heat from the atmosphere also melts sea ice on the Arctic Ocean during summer.

The global trends in air temperature are more dramatic in the Arctic due to the ice-albedo feedback and Polar Amplification of global warming (e.g., Manabe and Stouffer, 1980). As warmer temperatures melt the ice, the lower albedo of the surface allows more heat to be absorbed by the surface leading to a positive feedback that amplifies the global warming signal. Changes in wind related to the Arctic Oscillation (AO, Thompson and Wallace, 1998) have also been linked to the decline of sea ice. For example, Rigor et al. (2002) have shown that during high AO winters the winds blow more sea ice away from the Eurasian coast which allows more heat from the ocean to warm the atmosphere over these areas. Rigor and Wallace (2004) found that the trends in sea ice extent during summer (e.g., Figure 6D) are a lagged response, specifically, the younger, thinner sea ice that develops along the Eurasian coast during high AO winters, is less likely to survive the summer melt. The younger, thinner sea ice pack is also blown faster by the winds and is more prone to fracturing and increasing amounts of open water and leads during all seasons, allowing more heat to be released by the ocean during winter, and more heat to be absorbed by the ocean during summer, which will delay freeze up. Thus the increased kinematics of sea ice provides a dynamic complexity that strengthens the ice-albedo feedback even more (Rampal et al., 2011).

The high Arctic Oscillation conditions may have also shifted the ocean currents so that river runoff from the Eurasian continent was diverted to the east, weakening the cold halocline later and allowing heat in the Atlantic waters to reach the surface and melt sea ice (Morison et al., 2012, 2021). Polyakov et al. (2020) show that the weakening of the cold halocline layer is also observed in in the Eurasian Basin through 2018 and estimate that the Atlantic waters heat increased to over 10 Wm^–2 in 2016–18 (from 3–4 Wm^–2 in 2007–2008), decreasing winter ice growth by twofold.

The IPCC Special Report on Oceans and Cryosphere (SROC, Oppenheimer et al., 2019) concluded that the rate of change of the Global Mean Sea Level (GMSL) was, respectively, 1.4 and 3.2 mm year^–1 for the periods 1901–1990 and 1993–2015. Several studies have advanced the analysis of the trend of GMSL by improving the accuracy of the observational data adding observations by different platforms, using different analysis methods, reconstructing time series, or using climate models to simulate the sea level evolution. By means of a novel hybrid sea-level reconstruction applied to the global tide gauge time series Dangendorf et al. (2019) estimated a GMSL trend of 1.6 mm year^–1 from 1900 to 2015, an increase of 0.19 m since 1900. From the analysis of altimetry for the period 1993–2017 Nerem et al. (2018) observed a GMSL rise of 2.9 ± 0.4 mm year^–1 and an acceleration of 0.084 ± 0.025 mm year^–2. Dangendorf et al. (2019) also found a notable increase in the GMSL rate from 1993 (2.1 mm year^–1) to 2015 (3.4 mm year^–1) and a persistent acceleration of 0.06 mm year^–2 since the 1960s. Cazenave et al. (2019) using the most recent time series (January 1993–February 2019) gives a mean rate of sea level rise of 3.15 ± 0.3 mm year^–1, with an acceleration of 0.10 ± 0.04 mm year^–2. This acceleration value agrees well with Nerem et al. (2018)’s estimate. GMSL is projected to rise by the end of the century (2100) between 0.43 m (0.29–0.59 m) under RCP2.6 and 0.84 m (0.61–1.10 m) under RCP8.5 and will continue to increase during centuries due to ocean heat uptake and the melting of the ice sheets and glaciers (Oppenheimer et al., 2019).

Sea level trend maps computed from satellite altimetry data reveal that even though a general GMSL rise has occurred during the altimetry era, the MSL change follows regional and local variability with diferences up to 8 mm year⁻¹ reflecting the pattern of ocean currents, the large-scale oceanic and atmospheric oscillations or the contribution of the melting ice-sheets among other factors (Figure 7). Basin-scale MSL trend variability is also observed from the longer tide gauges series (e.g., Slangen et al., 2017; Dangendorf et al., 2019). For the period 1900–2018, Frederikse et al. (2020) found a positive MSL trend in all the ocean basins, being the highest for the subtropical North Atlantic (2.49 mm year^–1) and south Atlantic (2.07 mm year^–1), the lowest for the subpolar North Atlantic (1.08 mm year^–1) and East Pacific (1.20 mm year^–1), while intermediate values were observed in the Indian-south Pacific (1.33 mm year^–1) and in the northwest Pacific (1.68 mm year^–1).

Separating the global drivers of the MSL variability at a regional scale and even more at a local scale, becomes complex. In addition to the anthropogenically forced sea-level signal, the internal variability is time and location-dependant (Stammer et al., 2013). Thus, large atmospheric and ocean oscillations (e.g., El Niño Southern Oscillation, Pacific Decadal Oscillation, North Atlantic Oscillation, Indian Ocean Dipole) have interannual and decadal signals in MSL time series being of different period and intensity depending on the ocean basin. Analyzing the MSL trend from a regional perspective additionally allows development of a more detailed characterization of the global sea level budget. For instance, from a hybrid MSL reconstruction from 1900 to 2015, Dangendorf et al. (2019) demonstrate that a great part (∼76%) of the GMSL acceleration from the 1960s has its origin in the Indo-Pacific (0.07 ± 0.01 mm year^–2) and South Atlantic sea (0.06 ± 0.01 mm year^–2) as a consequence of an intensification and a displacement of the southern hemispheric westerlies that contributed to increased heat uptake and consequently a more intense thermal expansion.

With regard to local MSL trends, besides the complex processes occurring near the coast (e.g., Benveniste et al., 2019), and the particular importance of land subsidence, the different scale processes contributing to the sea level variability has not only a local origin, but it can be generated far away (Woodworth et al., 2019). The IPCC Special Report on Oceans and Cryosphere (Oppenheimer et al., 2019) further indicates that the local Extreme Sea Level (ESL) events happening once in one hundred years will become annual events in many low-lying cities and small islands by 2050 under all RCP scenarios due to the projected Global Mean Sea Level Rise (GMSL).

Direct continuous measurements of the AMOC only started in 2004 with RAPID-MOCHA (Smeed et al., 2014), an array of moored instruments that spans the width of the Atlantic at latitude 26.5 degrees north and provides continuous monitoring of the AMOC. Before, there had only been five individual snapshots of the AMOC, computed from seawater density measurements taken at hydrographic sections in the years 1957, 1981, 1992, 1998, and 2004 (Frajka-Williams et al., 2019). A couple of other trans-basin observing arrays at different locations in the Atlantic followed, including SAMBA in the South Atlantic in 2009 (Meinen et al., 2018) and OSNAP in the subpolar North Atlantic in 2014 (Li et al., 2017).

The RAPID observations recorded a notable decrease of 2.7 Sverdrup (Sv; 1 Sv = 10⁶ m³ s^–1), about 15%, in AMOC strength between April 2004 until roughly April 2008, followed by a fairly stable period until the end of the recovered data in September 2018 (Smeed et al., 2018; Moat et al., 2020). Yet such a short record cannot distinguish between decadal variability and long-term slowdown. Various studies have attempted to reconstruct the AMOC for the time period before 2004 using other climatic variables, so-called proxies, like sea surface temperatures (Latif et al., 2006; Rahmstorf et al., 2015) and sea level heights (Frajka-Williams, 2015) as well as the available snapshots (Bryden et al., 2005; Kanzow et al., 2010). Using the latter Bryden et al. (2005) estimated a decrease in AMOC transport at 26°N of about 30% between 1957 and 2004. Two main criticisms were levied against this conclusion. One, the first 4 years of the observed overturning strength provided by the RAPID data (Kanzow et al., 2010; Smeed et al., 2014) suggested that the seasonal variability of the AMOC has an amplitude of several Sverdrup, significantly larger than previously thought, thus, the five snapshots used by Bryden et al. (2005) might have sub-sampled intense high-frequency variability, rather than a robust trend. Correcting the measurements for the seasonal cycle Kanzow et al. (2010) found a much smaller weakening of only 13%. A different approach, estimating the strength of the AMOC mainly from the more widely available measurements of CTD (Conductivity-Temperature-Depth) end stations, found a reduction of about 2–4 Sv between 1980 and 2005, but concluded that this trend cannot be statistically validated due to the large variability in the layer transports found in the data (Longworth et al., 2011). This is in direct opposition to the findings of Latif et al. (2006) who concluded from the observed linear trend in the sea surface temperatures in the North Atlantic that the AMOC has strengthened since 1980. Combining the observed density change in the region of the Denmark Straight with the results from ocean model simulations they estimated the increase to be about 1 Sv between 1970 and 2000. These seemingly contradictory results could be reconciled with the AMOC reconstruction by Caesar et al. (2018), based on the relative cooling in the subpolar North Atlantic, who see a decline of AMOC strength since the 1950s with a short-lived recovery that is evident in the 1980s and 90s before a return to decline from the mid-2000s (Figure 8). This short-lived recovery of the AMOC is also found by Jackson et al. (2016) by analyzing a global-ocean reanalysis product, the GloSea5 data, which covers the years 1989–2015 as well as by Frajka-Williams (2015) who combined sea surface height data from satellites with cable measurements to reconstruct the AMOC from 1993 to 2014. Caesar et al. (2018), found an overall decline of the AMOC of about 3 ± 1 Sv (about 15%) since the mid-twentieth century.

To put these changes into an even longer-term context, researchers rely on different paleo-proxies, including data from ice and marine sediment cores, to reconstruct the strength of the AMOC over the last more than 1,000 years. Using grains from cores of sediments from a key site off Cape Hatteras Thornalley et al. (2018) found that the AMOC is now at its weakest in at least 1,600 years. They confirmed this finding with foraminiferal-based temperature proxies which, when taken from specific locations in the North Atlantic, reflect the strength of the North Atlantic sea surface temperature dipole which has been repeatedly linked to AMOC changes (Thornalley et al., 2018). Similar conclusions were reached by Rahmstorf et al. (2015) who used a proxy compilation of tree-rings and ice cores that represent the relative temperature changes in the subpolar North Atlantic caused by AMOC changes. Sherwood et al. (2011) studied the δ¹⁵N concentration of deep-sea gorgonian corals and found a nutrient shift in the early 1970s that is unique in the context of the last approximately 1,800 years and indicates a decline in the presence of Labrador Slope Water associated with the AMOC. Thibodeau et al. (2018) found a similar decline in an AMOC record based on the δ¹⁸O in benthic foraminifera from sediment cores retrieved from the Laurentian Channel. Caesar et al. (2021) compared all these different proxy types and found that they provide a consistent picture of the evolution of the AMOC since AD 400 with a long and fairly stable period (that is intermitted with an initial decline during the nineteenth century) followed by another, more rapid weakening in the middle of the twentieth century. Together, these proxies indicate that the AMOC over the last decades is weaker than ever before in the last 1,600 years. Figure 9 shows collectively these observations since 1400 to the present.

Currently, while these findings provide strong evidence that the Atlantic Meridional Overturning Circulation has weakened relative to preindustrial times, there is insufficient data to quantify the exact magnitude of the weakening, or to properly attribute it to anthropogenic forcing (IPCC, 2019). This is also due to the fact that the ensemble means of the latest generation of climate models (CMIP6) show no trend in the strength of the overturning circulation over the historical period (Weijer et al., 2020). However, this might be due to an overestimation of the anthropogenic aerosol forcing in a majority of the models leading to a cooling of the subpolar North Atlantic and a subsequent AMOC strengthening. This is supported by the fact that those ensemble members that capture the North Atlantic cold blob, i.e., the SST fingerprint of a weaker AMOC, show a weakening of the AMOC over the historical period (Menary et al., 2020). For the future, the simulations of all CMIP6 models respond to the increasing greenhouse gas emissions with an AMOC weakening, showing on average a decline of 24–39%, depending on the emission scenario, over the course of the twenty-first century. When using the observed strength of the AMOC by RAPID/MOCHA as a constraint the mean decline increases, in particular for the low-emission scenarios, to 34–45% (Weijer et al., 2020).

The international ocean observing community has made persistent efforts in developing new technologies, observation networks and data sharing protocols to deliver credible climate change indicators and useful ocean information to a variety of users in a timely manner and at a global scale. Here we identify the progress of the global monitoring efforts and make some recommendations to fill some of the gaps in the coming years.

High quality and global coverage observations are essential to monitor the ocean temperature changes (Abraham et al., 2013). The primary instruments of the ocean subsurface observing system since the 1940s are MBTs, XBTs, Nansen/Niskin bottles, and CTDs (Figure 10A). MBTs typically go down to ∼125–250 m and were widely deployed from 1938 to the early 1960s. Shallow XBTs (e.g., T4/T6) reach 450 m, and were widely deployed during the 1970s∼1980s whereas deep XBTs (e.g., T7/DB) provide data to 800 m, and were widely used during the 1990s and early 2000s. The Argo Program, designed in 1998 achieved its initial goal of 3000 profiling floats in November 2007. Since 2007, the data coverage is > 80% of the global ocean area (3 by 3 degree box) from 0 to 1,200 m depth and > 70% for 1,200–2,000 m depth (Figure 10B; Meyssignac et al., 2019). Maintaining and improving the current ocean observation system are strongly recommended to ensure the accurate ocean climate monitoring. It is also essential to improve the historical record, for example, by recovering un-digitized temperature (OHC) and other observations.

Some limitations remain for the current ocean observation network, particularly for coastal regions, marginal seas, deep ocean regions below 2,000 m. It is important to establish a deep ocean system in the future to monitor ocean changes below 2,000 m, thus to provide a complete estimate of earth’s energy imbalance (Johnson et al., 2015; von Schuckmann et al., 2016). Currently, boundary currents are not fully represented by Argo as floats can swiftly pass through the energetic regions, e.g., western boundary current (WBC) regions which could induce an inverse cascade of kinetic energy and affect the large scale low-frequency variability (Wang et al., 2017; Llovel et al., 2018). Achieving adequate sampling will require an observing system design based on a mixture of observing technologies adopted to the different operating environments. There is a need to develop/maintain multiple platform observations for cross-validation and calibration purposes (Meyssignac et al., 2019).

Intensive national and international efforts focused on carbonate chemistry monitoring, biological observations and biogeochemical/ecological forecast modeling over the past decade have shed light on the status and impacts of ocean acidification on local to global scales. International observing networks deployed around the world which use moorings, repeat hydrography research cruise transects, ships of opportunity, and fixed ocean time-series to track ocean chemistry include the Global Ocean Ship-based Hydrographic Investigations Program (GO-SHIP) surveys, the Surface Ocean CO₂ Observing Network (SOCONET), the Ship of Opportunity Program (SOOP) volunteer observing ships, and the Ocean Sustained Interdisciplinary Time-series Environment Observation System (OceanSITES) time-series stations in the Atlantic, Pacific, and Indian oceans (Figure 11). These have provided essential, climate quality carbonate chemistry observations needed to understand ocean acidification in open waters. Biogeochemical Argo platforms, still under development, will increase the availability of pH profiles throughout the water column, along with other hydrographic parameters. In an effort to both coordinate with these international efforts and to coordinate with and expand national ocean acidification observing efforts, the Global Ocean Acidification Observing Network (GOA-ON) was launched in 2013. Through GOA-ON, organizations and scientists have established observation standards, enhanced data sharing, and quantified global and regional ocean acidification trends to identify areas of heightened vulnerability or resilience.

There have been many significant leaps in comprehending global ocean acidification trends and impacts and more research is needed to better inform models and improve predictions of the Earth system response to ocean acidification (Jewett et al., 2020). This includes the relationship between the impacts of ocean acidification and other stressors, such as warming, on organisms and communities. Understanding the direct and indirect impacts on marine populations and communities and the capacity of organisms to acclimate or adapt to the changes in ocean acidification-induced ocean chemistry will be extremely important in determining and predicting the economic, ecological, and societal impacts of ocean acidification. There remains a strong need for more extensive monitoring in coastal regions, including access to high quality, low-cost sensors to do this monitoring, and to satellite data and research into the long-term trends in ocean chemistry beyond the observational record (paleo-OA).

While dissolved oxygen data coverage has been proven to be sufficient to derive large scale and global trends, significant better long-term measurements are needed to address current questions. Measurements at greater spatial and temporal extent are needed in many regions in order to capture for example the high variability in oxygen content in coastal areas. In this context it is of significant importance that a variety of biogeochemical parameters are increasingly added to the automated monitoring of the global ocean, since they serve as vital indicators for changes in the biogeochemical dynamics, which cannot be analyzed detached from changes in small- and large-scale ocean dynamics.

The International Arctic Buoy Programme (IABP)⁶ maintains the fundamental Arctic Observing Network of drifting buoys which monitor ocean and sea ice circulation, as well as sea level pressure and surface temperature. While the IABP has been able to improve and maintain a denser network of drifting buoys, the prevailing winds and ocean circulation quickly carries these buoys away from the Eurasian coast of the Arctic Ocean, thus creating a recurring gap in the network during the winter that needs to be replenished since these gaps in the network hamper our ability to completely monitor and understand Arctic change (Thoman et al., 2020).

The best estimates of the long-term trends in Arctic sea ice volume are provided by models (e.g., Schweiger et al., 2019) given the paucity of in situ, pan-Arctic measurements of sea ice thickness. These models are compared to satellite retrievals of ice thickness and in situ measurements from field programs on the ice, aerial surveys (e.g., Haas et al., 2017), and submarine transits under the ice (e.g., Rothrock et al., 1999). The satellite retrievals are also compared to in situ measurements of ice thickness, and it has been shown that the primary source of uncertainty in these retrievals is the assumed depth of snow on top of the sea ice (Kwok and Cunningham, 2015; Kwok, 2018). Recently, the Multidisciplinary drifting Observatory for the Study of Arctic Climate (MOSAiC) Expedition completed a year-long drift across the Arctic Ocean (Shupe et al., 2020), and collected myriad of observations including measurements required to improve our understanding of Arctic climate processes, such as in situ measurements of snow and ice thickness, aerial and under ice surveys. Similar campaigns should be conducted routinely on a pan-Arctic scale (Haas et al., 2017; IPCC, 2019).

There are different programs dedicated to monitoring the different contributing factors of sea level change. The Argo program, mentioned before, is devoted to the monitoring of the temperature, salinity, currents and bio-optical properties of the global ocean reaching 2,000 m depths. From these globally distributed floats, changes in the density of the water column are estimated. These observations allow monitoring the contribution of the steric change of the GMSL of the ocean. With regard to the barystatic component, the GRACE space gravimetry mission (covering the period 2002–2017) and the subsequent GRACE Follow-On (from 2018 on) satellite mission are registering global anomalies of the Earth’s gravity field. To give continuity to the altimetry sea level record, Copernicus Sentinel-6 Michael Freilich satellite has been recently launched and has provided some first promising results. The launch of its twin, Sentinel 6B, is planned for 2025 (follow on of the Jason satellites). In the future the SWOT satellite will contribute to the improvement of the coastal data, due to its higher spatial resolution and in addition, this will measure river discharges and as such will be a valuable source of sea level budget from terrestrial contribution.

It is worth mentioning that when the footprint of the altimeter covers not only the sea but also the land the returning echoes are contaminated and consequently, it is not accurate enough at around 20–50 km from the coast. For retrieving altimetry data in those areas, the coastal altimetry community has investigated how to re-process these data by using different waveform retracking algorithms and applying different geophysical corrections (Vignudelli et al., 2019) and have provided different coastal altimetry databases (Birol et al., 2017; Cipollini et al., 2017; The Climate Change Initiative Coastal Sea Level Team, 2020) with more accurate data comparing to the conventional altimetry databases.

Information about the ocean circulation and its changes can be inferred from either direct measurements, proxies, model simulations and satellites. The main uncertainties regarding the trends in ocean circulation arise from the short time spans of the direct continuous measurements, the incompleteness when representing a circulation through proxies and the inherent uncertainties of the models. It is therefore essential that the existing observation programs like the Global Drifter Program (Dohan et al., 2010) and the Argo Program are sustained. This includes but is not limited to the main programs observing the AMOC, i.e., the RAPID programs (e.g., Smeed et al., 2014, 2018) that continuously measure the AMOC strength since 2004 at roughly 26°N, the SAMOC programs that measure AMOC strength in the South Atlantic and include the SAMBA array at about 34.5°S (Meinen et al., 2018; Kersale et al., 2020) and the OSNAP program (Lozier et al., 2017) measuring the overturning that feeds the AMOC since 2014.

International data programs like Copernicus and others can play a relevant role to integrate the different data sets and provide the signs for ocean climate change. With that objective Figures 12, 13 present here a final visual summary of key ocean climate change indicators and current trends based on the individual analyses of the Copernicus Marine Services (CMS), that emphasizes recent changes (1993–2019/20).

CG-S conceived and led the manuscript. All authors contributed to the writing of the manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.