Editing IPCC:AR6/SROCC/Chapter-5 (section)

== 5.2 Changing Ocean and Biodiversity ==

<span id="introduction-1"></span>
=== 5.2.1 Introduction ===

<div id="section-5-2-1introduction-block-1"></div>

This section assesses changes in the ocean. It includes the physical and chemical properties (Section 5.2.2), their impacts on the pelagic ecosystem (Section 5.2.3) and deep seafloor system (Section 5.2.4). In this assessment, the open ocean and deep seafloor includes areas where the water column is deeper than 200 m; it is the main subject of Section 5.2. Coastal and shelf seas are primarily discussed in Section 5.3.

<span id="changes-in-physical-and-biogeochemical-properties"></span>
=== 5.2.2 Changes in Physical and Biogeochemical Properties ===

<div id="section-5-2-2-1introduction-to-changing-open-ocean"></div>

<span id="introduction-to-changing-open-ocean"></span>
==== 5.2.2.1 Introduction to Changing Open Ocean ====

<div id="section-5-2-2-1introduction-to-changing-open-ocean-block-1"></div>

The ocean is getting progressively warmer, with parallel changes in ocean chemistry such as acidification and oxygen loss, as documented in the AR5 (Rhein et al., 2013 <sup>[[#fn:r5|5]]</sup> ). The global scale warming and acidification trends are readily detectable in oceanic observations, well understood scientifically, and consistently projected by ESMs. Each of these has been directly attributed to anthropogenic forcing from changing concentrations of greenhouse gases and aerosols (Bindoff et al., 2013 <sup>[[#fn:r6|6]]</sup> ). These trends in the global average ocean temperature will continue for centuries after the anthropogenic forcing is stabilised (Collins et al., 2013 <sup>[[#fn:r7|7]]</sup> ).

The impacts on ocean ecosystems and human societies are primarily driven by regional trends and by the local manifestation of the global-scale changes. At these smaller scales, the temperature, acidification, salinity, nutrient and oxygen concentrations in the ocean are also expected to exhibit basin and local-scale changes. However, the ocean also has significant natural variability at basin and local-scales with time scales from minutes to decades and longer (Rhein et al., 2013 <sup>[[#fn:r8|8]]</sup> ), which can mask the underlying observed and projected trends (see Box 5.1). The impact of multiple stressors on marine ecosystems is one of the main subjects of this chapter (Section 5.2.3, 5.2.4, 5.3), including new evidence and understanding since the last assessment report (e.g., Gunderson et al., 2016). The most severe impacts of a changing climate will typically be experienced when conditions are driven outside the range of previous experience at rates that are faster than human or ecological systems can adapt (Pörtner et al., 2014 <sup>[[#fn:r9|9]]</sup> ; Box 5.1).

This section summarises our emerging understanding of the primary changes to the ocean, along with an assessment of several key areas of scientific uncertainty about these changes. Because many of these long-term trends have already been extensively discussed in previous assessments (IPCC, 2013 <sup>[[#fn:r10|10]]</sup> ), much of this summary of the physical changes is brief except where there are significant new findings.

<div id="section-5-2-2-2changing-temperature-salinity-circulation"></div>

<span id="changing-temperature-salinity-circulation"></span>
==== 5.2.2.2 Changing Temperature, Salinity, Circulation ====

<div id="section-5-2-2-2changing-temperature-salinity-circulation-block-1"></div>

Historically, scientific research expeditions starting in the 19th century have provided occasional sections measuring deep ocean properties (Roemmich et al., 2012 <sup>[[#fn:r11|11]]</sup> ). Greater spatial and temporal coverage of temperatures down to about 700 m was obtained using expendable bathythermographs along commercial shipping tracks starting in the 1970s (Abraham et al., 2013 <sup>[[#fn:r12|12]]</sup> ). Since the early 2000s, thousands of autonomous profiling floats (Argo floats) have provided high-quality temperature and salinity profiles of the upper 2000 m in ice-free regions of the ocean (Abraham et al., 2013 <sup>[[#fn:r13|13]]</sup> ; Riser et al., 2016 <sup>[[#fn:r14|14]]</sup> ). Further advances in autonomous floats have been developed that now allow these floats to operate in seasonally ice covered oceans (Wong and Riser, 2011 <sup>[[#fn:r15|15]]</sup> ; Wong and Riser, 2013 <sup>[[#fn:r15|15]]</sup> ), and more recently to profile the entire depth of the water column down to 4000 or 6000 m (Johnson et al., 2015 <sup>[[#fn:r17|17]]</sup> ; Zilberman, 2017 <sup>[[#fn:r18|18]]</sup> ) and to include biogeochemical properties (Johnson et al., 2017 <sup>[[#fn:r19|19]]</sup> ). Autonomous floats have revolutionised our sampling and accuracy of the global ocean temperature and salinity records and increased certainty and confidence in global estimates of the earth heat (temperature) budget, particularly since 2004 (Von Schuckmann et al., 2014; Roemmich et al., 2015 <sup>[[#fn:r20|20]]</sup> ; Riser et al., 2016 <sup>[[#fn:r21|21]]</sup> ), as demonstrated by the convergence of observational estimates of the changes in the heat budget of the upper 2000 m (Figure 5.1). New findings using data collected from such observing platforms mark significant progress since AR5.

To understand the recent and future climate, we use ensembles of coupled ocean-atmosphere-cryosphere-ecosystem models (ESMs) with the full-time history of atmospheric forcing (greenhouse gases, aerosols, solar radiation and volcanic eruptions) for the historical period and projections of the concentrations or emissions of these forcings to 2100. For these projections the RCPs of atmospheric emissions scenarios are used as specified by the Coupled Model Intercomparison Project, Phase 5 (CMIP5) (see Section 1.8.2.3, Cross-Chapter Box 1, and also IPCC AR5) <sup>[[#fn:3|3]]</sup> . This chapter focuses on the low and high emissions scenarios RCP2.6 and RCP8.5, respectively. When these scenarios are used to drive ESMs, it is possible to simulate the recent and future patterns of changes in the ocean temperature, salinity and circulation (and other oceanic properties such as ocean oxygen concentration and acidification, Section 5.2.2.3 and 5.2.2.4). Finally, the projections of ocean changes also informs the detection, attribution and projection of risk and impacts on ecosystems (Sections 5.2.3, 5.2.4 and 5.3), ecosystem services (Section 5.4.1) and human well-being (Section 5.4.2) under climate change.

<div id="section-5-2-2-2changing-temperature-salinity-circulation-block-2"></div>

<span id="observed-and-projected-global-ocean-heat-uptake"></span>
===== 5.2.2.2.1 Observed and projected global ocean heat uptake =====

As AR5 concluded, the ocean is warming as a direct result of anthropogenic changes to the radiative properties of the atmosphere and the heat budget of the Earth ( ''very likely'' ) (Bindoff et al., 2013 <sup>[[#fn:r22|22]]</sup> ). Over the past few decades our ocean observing system has measured an increase in ocean temperature (Figure 5.1). This temperature increase corresponds to an uptake of over 90% of the excess heat accumulated in the Earth system over this period (Bindoff et al., 2013 <sup>[[#fn:r23|23]]</sup> ; Rhein et al., 2013 <sup>[[#fn:r24|24]]</sup> ). This heat in the ocean also causes it to expand and has contributed about 43% of the observed global mean SLR from 1970–2015 (Section 4.2.2.3.6).

Since AR5, there have been further improvements in our ability to understand and correct instrumental errors and new estimates also attempt to minimise biases in estimating temperature changes arising from traditional data-void filling strategies (Abraham et al., 2013 <sup>[[#fn:r25|25]]</sup> ; Durack, 2015 <sup>[[#fn:r26|26]]</sup> ; Cheng and Chen, 2017 <sup>[[#fn:r27|27]]</sup> ; Cheng et al., 2017 <sup>[[#fn:r28|28]]</sup> ). New estimates from ocean observations of ocean heat uptake in the top 2000 m between 1993 and 2017 ''very likely'' range from 9.2 ± 2.3 ZJ yr -1 to 12.1 ± 3.1 ZJ yr -1 (Johnson et al., 2018 <sup>[[#fn:r29|29]]</sup> ) <sup>[[#fn:4|4]]</sup> . Three recent independent estimates do a better job of accounting for instrumental biases and the sparseness of historical ocean temperature measurements than the older studies assessed in AR5, and provide larger and more consistent estimates of heat uptake rates for the 0-2000 m layer of 5.8 ± 1.0 ZJ yr -1 (Cheng and Chen, 2017 <sup>[[#fn:r30|30]]</sup> ; Cheng et al., 2017 <sup>[[#fn:r31|31]]</sup> ; Ishii et al., 2017 <sup>[[#fn:r32|32]]</sup> ), 6.0 ± 0.8 ZJ yr -1 (updated from Domingues et al. (2008)) and 6.3 ± 1.8 ZJ yr -1 (Cheng and Chen, 2017 <sup>[[#fn:r33|33]]</sup> ; Cheng et al., 2017 <sup>[[#fn:r34|34]]</sup> ; Ishii et al., 2017 <sup>[[#fn:r35|35]]</sup> ) for the 1971–2010 period assessed by AR5.

Based on these new published methods and revised atlases we update the estimates for ocean heat uptake (Table 5.1, and SM5.1). For all of the periods assessed in Table 5.1, it is ''virtually certain'' that the upper ocean (0–700 m) has warmed. These results are consistent with earlier research into the duration of record needed to detect a significant signal in global ocean heat content (Gleckler et al., 2012 <sup>[[#fn:r36|36]]</sup> ). Critically, the ''high confidence'' and ''high'' ''agreement'' in the ocean temperature data means we can detect discernable rates of increase in ocean heat uptake (Gleckler et al., 2012 <sup>[[#fn:r37|37]]</sup> ; Cheng et al., 2019 <sup>[[#fn:r38|38]]</sup> ). The rate of heat uptake in the upper ocean (0–700 m) is ''very likely'' higher in the 1993–2017 (or 2005–2017) period compared with the 1969–1993 period (see Table 5.1). The deeper layer (700–2000 m) heat uptake rate is ''likely'' to be higher in the 1993–2017 period compared with the 1969–1993 period.

<span id="table-5.1"></span>
<!-- START TABLE -->
'''Table 5.1'''

The assessed rate of increase in ocean heat content in the two depth layers 0–700 m and 700–2000 m and their ''very likely'' ranges. Fluxes in Wm -2 are averaged over the Earth’s entire surface area. The four periods cover earlier and more recent trends; the 2005–2017 period has the most complete interior ocean data coverage and the greatest consistency between estimates, while longer trends are better for distinguishing between forced changes and internal variability. These observationally-estimated rates come from an assessment of the recent research (see SM5.1), while the Coupled Model Intercomparison Project Phase 5 (CMIP5) Earth System Models (ESM) estimates are based on a combined 28-member ensemble of historical, Representative Concentration Pathway (RCP)2.6 and RCP8.5 simulations.

<!-- TABLE -->
{| class="wikitable"
|-
|
| colspan="4"| '''Ocean Heat Uptake Rate, ZJ yr''' '''-1'''
| colspan="4"| '''Ocean Heat Uptake as Average Fluxes, W m''' '''-2'''
|-
| '''Period'''
| '''1969–1993'''
| '''1993–2017'''
| '''1970–2017'''
| '''2005–2017'''
| '''1969–1993'''
| '''1993–2017'''
| '''1970–2017'''
| '''2005–2017'''
|-
| colspan="9"| '''Observationally Based Ocean Heat Uptake Estimates:'''
|-
| '''0–700 m'''
| 3.22 ± 1.61
| 6.28 ± 0.48
| 4.35 ± 0.80
| 5.31 ± 0.48
| 0.20 ± 0.1
| 0.39 ± 0.03
| 0.27 ± 0.05
| 0.33 ± 0.03
|-
| '''700–2000 m '''
| 0.97 ± 0.64
| 3.86 ± 2.09
| 2.25 ± 0.64
| 4.02 ± 0.97
| 0.06 ± 0.04
| 0.24 ± 0.13
| 0.14 ± 0.04
| 0.25 ± 0.06
|-
| colspan="9"| '''CMIP5 ESM Ensemble-mean Ocean Heat Uptake with 90% Certainty Range from Ensemble Spread:'''
|-
| '''0–700 m'''
| 3.60 ± 1.92
| 7.37 ± 2.09
| 5.64 ± 1.90
| 7.85 ± 2.71
| 0.22 ± 0.12
| 0.46 ± 0.13
| 0.35 ± 0.12
| 0.49 ± 0.17
|-
| '''700–2000 m '''
| 1.32 ± 1.49
| 2.72 ± 1.41
| 1.99 ± 1.51
| 3.33 ± 1.75
| 0.08 ± 0.09
| 0.17 ± 0.09
| 0.12 ± 0.09
| 0.21 ± 0.11
|}
<!-- END TABLE -->

The direct comparison of the observed changes in ocean heat content and the simulated historical changes is undertaken to detect climate change, to attribute the causes of climate change to the forcings in the system, and to evaluate the performance of ESMs. Attribution studies also reject competing hypotheses to explain the global ocean changes such as natural forcing from solar variability or volcanic eruptions (see Section 1.3) (Bindoff et al., 2013 <sup>[[#fn:r39|39]]</sup> ). Detection and attribution studies have since been used to detect changes in the rate of ocean heat uptake and to attribute these changes to human activity (Gleckler et al., 2016 <sup>[[#fn:r40|40]]</sup> ).

Updated observationally-based estimates of ocean heat uptake are consistent with simulations of equivalent time-periods from an ensemble of CMIP5 ESMs (Table 5.1 and the inset panel in Figure 5.1) ( ''high confidence'' ), once the limitations of the historical ocean observing network and the internally generated variability with a single realisation of the real world are taken into account (see Section 5.2.2.2). Following the CMIP5 protocol, the ESMs are radiatively forced with observationally derived estimates of greenhouse gas concentrations and aerosols, including natural forcing variations from volcanic eruptions and solar forcing, through 2005; after 2006 each of the ESMs uses either the RCP2.6 or RCP8.5 emissions scenarios. The ''very likely'' ranges of the observed trends of heat uptake for the four periods and two layers all fall within the ''very likely'' range of simulated heat uptake from the ESM ensemble (Table 5.1). The difference between observations and average of the simulations in the upper ocean is an overestimate of heat uptake by about 20% and for the deeper layer there an underestimate by a similar amount, but this difference is still well within the ''very likely'' range from the ensemble of simulations. The overall consistency between observationally-based estimates and ESM simulations of the historical period gives greater confidence in the projections; it is ''very likely'' that historical simulations agree with observations of the global ocean heat uptake (Table 5.1).

While the collection of the worlds’ ESMs have been criticised for having an ensemble mean that does not exhibit the observed ‘hiatus’ or ‘slowdown’ of global mean surface temperature increase in the early 21st century (Meehl et al., 2011 <sup>[[#fn:r57|57]]</sup> ; Trenberth et al., 2016 <sup>[[#fn:r58|58]]</sup> ) , it is increasingly clear that this is at least in part due to the redistribution of heat within the climate system from the surface into the interior ocean and between ocean basins. Individual realisations of ESMs do show decades with slow increases in mean surface temperature change comparable to what was observed, even though these cases exhibit continued interior ocean heat uptake, and every ensemble member exhibits surface warming closer to the ensemble-mean over multi-decadal timescales (Meehl et al., 2011 <sup>[[#fn:r59|59]]</sup> ; England et al., 2015 <sup>[[#fn:r60|60]]</sup> ; Knutson et al., 2016 <sup>[[#fn:r61|61]]</sup> )

<span id="figure-5.1"></span>
<!-- START IMG -->
<!-- IMG TITLE -->
'''Figure 5.1'''

<span id="figure-5.1-time-series-of-globally-integrated-upper-2000-m-ocean-heat-content-changes-in-zj-relative-to-the-20002010-period-average-as-inferred-from-observations-magenta-and-as-simulated-for-historical-tan-representative-concentration-pathway-rcp2.6-blue-and-rcp8.5-red-forcing-by-a-25-member-ensemble-of-coupled-model-intercomparison-project-phase-5-cmip5"></span>
<!-- IMG CAPTION -->
'''Figure 5.1 | Time series of globally integrated upper 2000 m ocean heat content changes in ZJ, relative to the 2000–2010 period average, as inferred from observations (magenta) and as simulated for historical (tan), Representative Concentration Pathway (RCP)2.6 (blue) and RCP8.5 (red) forcing by a 25-member ensemble of Coupled Model Intercomparison Project Phase 5 (CMIP5) […]'''
<!-- IMG FILE -->
[[File:e2518a5a96f60c4991d9797d8d3fea43 IPCC-SROCC-CH_5_1.jpg]]

Figure 5.1 | Time series of globally integrated upper 2000 m ocean heat content changes in ZJ, relative to the 2000–2010 period average, as inferred from observations (magenta) and as simulated for historical (tan), Representative Concentration Pathway (RCP)2.6 (blue) and RCP8.5 (red) forcing by a 25-member ensemble of Coupled Model Intercomparison Project Phase 5 (CMIP5) Earth System Models (ESMs) (Cheng et al. 2019 <sup>[[#fn:r41|41]]</sup> ). The shaded magenta in the outer panel is the very likely range determined by combining data from 4 long-term estimates (Palmer et al. 2007 <sup>[[#fn:r42|42]]</sup> ; Levitus et al. 2012 <sup>[[#fn:r43|43]]</sup> ; Lyman and Johnson, 2014 <sup>[[#fn:r44|44]]</sup> ; Cheng and Chen, 2017 <sup>[[#fn:r45|45]]</sup> ; Cheng et al. 2017 <sup>[[#fn:r46|46]]</sup> ; Ishii et al. 2017 <sup>[[#fn:r47|47]]</sup> ) processed as in Johnson et al. (2018) <sup>[[#fn:r48|48]]</sup> . The tan, blue and red lines are the ESM ensemble means, while shading shows each ensemble’s 5th to 95th percentile range. In the inset subpanel, the four different shaded magenta areas are the reported very likely range of heat content changes as inferred from observations by four independent groups (Magenta shading; Palmer et al. 2007 <sup>[[#fn:r49|49]]</sup> ; Lyman and Johnson, 2014 <sup>[[#fn:r50|50]]</sup> ; Cheng and Chen, 2017 <sup>[[#fn:r51|51]]</sup> ; Cheng et al. 2017 <sup>[[#fn:r52|52]]</sup> ; Ishii et al. 2017) <sup>[[#fn:r53|53]]</sup> processed as in Johnson et al. (2018) <sup>[[#fn:r54|54]]</sup> . In the inset subpanel the RCP2.6 and RCP8.5 projections after 2005 are combined into a single ensemble with the historical simulations.

<!-- END IMG -->
<span id="figure-5.2"></span>
<!-- START IMG -->
<!-- IMG TITLE -->
'''Figure 5.2'''

<span id="figure-5.2-heat-uptake-by-the-top-700-m-of-the-ocean-as-determined-by-differences-between-the-averages-over-two-5--or-20-year-intervals-converted-to-a-heat-flux-into-the-ocean-w-m2-either-from-observationally-based-analyses-or-a-38-member-ensemble-of-coupled-model-intercomparison-project-phase-5-cmip5-earth-system-models"></span>
<!-- IMG CAPTION -->
'''Figure 5.2 | Heat uptake by the top 700 m of the ocean, as determined by differences between the averages over two 5- or 20-year intervals converted to a heat flux into the ocean (W m–2), either from observationally-based analyses or a 38-member ensemble of Coupled Model Intercomparison Project Phase 5 (CMIP5) Earth System Models […]'''
<!-- IMG FILE -->
[[File:b7d41eb2445844a63fceb78ad98cdcb4 IPCC-SROCC-CH_5_2-1.jpg]]

Figure 5.2 | Heat uptake by the top 700 m of the ocean, as determined by differences between the averages over two 5- or 20-year intervals converted to a heat flux into the ocean (W m–2), either from observationally-based analyses or a 38-member ensemble of Coupled Model Intercomparison Project Phase 5 (CMIP5) Earth System Models (ESMs). (a) Change between (1971–1990) and (1998–2017) as inferred from observations (Good et al. 2013 <sup>[[#fn:r55|55]]</sup> ); (b) The ensemble mean change in CMIP5 ESMs for the same time periods as in (a); (c) Projected ensemble mean change in CMIP5 ESMs between (1986–2005) and (2081–2100) for the RCP8.5 forcing scenario. In panels (b) and (c), stippling indicates regions where the ensemble mean change is not significantly different from 0 at the 95% confidence level based on the models’ temporal variability. (d) Change between (2004–2008) and (2013–2017) as inferred from observations by the SODA 3.4.2 reanalysis product (Carton et al. 2018 <sup>[[#fn:r56|56]]</sup> ); (e) and (f) Estimates of change in heat uptake as in (d) but from two individual realisations of the CCSM ESM (Table SM5.2). These two realisations are identical apart from their initial conditions, which leads to different timing in their internal modes of variability; they were selected from the full CMIP5 ensemble as examples where one is reminiscent of the recent observed changes while the other has regional changes that have dissimilar timing.

<!-- END IMG -->
<div id="section-5-2-2-2changing-temperature-salinity-circulation-block-3"></div>

The ocean will continue to take up heat in the coming decades for all plausible scenarios. As depicted in Figure 5.1, the ensemble of CMIP5 ESMs used by Cheng et al. (2019) project that under RCP2.6, the top 2000 m of the ocean will take up 935 ZJ of heat between 2015 and 2100 (with a ''very likely'' range of 650–1340 ZJ based on the 5th and 95th percentiles of the 25 ESMs used here that have available data from the historical, scenario and control runs for RCP2.6). Under RCP8.5 this ensemble projects heat uptake of 2180 ZJ (with a ''very likely'' range of 1710–2790 ZJ, based on 35 ESMs) between 2015 and 2100. By 2100 the ocean is ''very likely'' to warm by 2 to 4 times as much for low emissions (RCP2.6) and 5 to 7 times as much for the high emissions scenario (RCP8.5) compared with the observed changes since 1970. With the RCP8.5 scenario, the ocean is ''very likely'' to take up about twice as much heat as RCP2.6 (Figure. 5.1). Even under RCP2.6 the ocean will continue to warm for several centuries to come (Collins et al., 2013 <sup>[[#fn:r52|52]]</sup> ). It is ''virtually certain'' that the ocean will continue to take up heat throughout the 21st century, and the rate of uptake will depend upon on the emissions scenario we collectively choose to follow.

<div id="section-5-2-2-2changing-temperature-salinity-circulation-block-4"></div>

<span id="figure-5.3"></span>
<!-- START IMG -->
<!-- IMG TITLE -->
'''Figure 5.3'''

<span id="figure-5.3-side-view-basin-averaged-zonal-mean-trends-change-per-century-in-water-mass-properties-in-the-top-2000-m-by-basin-a-as-inferred-from-observations-average-of-20132017-minus-average-of-20052009-and-b-coupled-model-intercomparison-project-phase-5-cmip5-model-projections-with-representative-concentration-pathway-rcp8.5-forcing-average-of-20812100-minus-average-of"></span>
<!-- IMG CAPTION -->
'''Figure 5.3 | Side-view basin-averaged zonal-mean trends (change per century) in water-mass properties in the top 2000 m by basin (a) as inferred from observations (average of 2013–2017 minus average of 2005–2009) and (b) Coupled Model Intercomparison Project Phase 5 (CMIP5) model projections with Representative Concentration Pathway (RCP)8.5 forcing (average of 2081–2100 minus average of […]'''
<!-- IMG FILE -->
[[File:e6f8060870293eb7dd25c79ffa06efb5 IPCC-SROCC-CH_5_3.jpg]]

Figure 5.3 | Side-view basin-averaged zonal-mean trends (change per century) in water-mass properties in the top 2000 m by basin (a) as inferred from observations (average of 2013–2017 minus average of 2005–2009) and (b) Coupled Model Intercomparison Project Phase 5 (CMIP5) model projections with Representative Concentration Pathway (RCP)8.5 forcing (average of 2081–2100 minus average of 1981–2000) trends in water-mass changes forcing. Subpanels within each group: top-to-bottom (Atlantic, combined Pacific and Indian, Global); left-to-right (Temperature, in situ Density, Salinity). Shaded areas show where the projected changes are not statistically significant at the 95% level. This figure uses the same observationally-derived reanalysis datasets and ensemble of Earth System Models (ESMs) as in Figure 5.2c and 5.2d. Solid lines show present contours of these fields; the notable structure in the northern hemisphere of the global-zonal mean contours of density and salinity are due to the relatively salty Mediterranean and fresh Black seas.

<!-- END IMG -->
<div id="section-5-2-2-2changing-temperature-salinity-circulation-block-5"></div>

<span id="structure-of-anthropogenic-climate-changes-in-the-ocean"></span>
===== 5.2.2.2.2 Structure of anthropogenic climate changes in the ocean =====

The ensemble average of the CMIP5 ESMs projects widespread ocean warming over the coming century, concentrated in the upper ocean (Figures 5.2c and 5.3) (Kuhlbrodt and Gregory, 2012). The anthropogenic heat will penetrate into the ocean following well-established circulation pathways (Jones et al., 2016a <sup>[[#fn:r63|63]]</sup> ). The greatest vertically integrated heat uptake occurs where there is already the formation of interior waters, such as Antarctic Intermediate Water along the Antarctic Circumpolar Current (Frölicher et al., 2015 <sup>[[#fn:r64|64]]</sup> ) or NADW precursors in the Nordic Seas (Figure 5.2c), but all water-masses <sup>[[#fn:5|5]]</sup> that are subducted over decades are

expected to experience significant warming (see Figure 5.3). The warming in the subtropical gyres penetrates deeper into the ocean than other gyres (roughly 15°N–45°N and 15°S–45°S in Figure 5.3), following the wind-driven bowing down of the density surfaces (the solid lines in Figure 5.3) in these gyres (Terada and Minobe, 2018 <sup>[[#fn:r65|65]]</sup> ). The greater warming at 700-2000 m in the Atlantic than the Pacific or Indian Oceans (Figure 5.3) reflects the strong southward transport of recently formed NADW at these depths by the AMOC. Two areas that commonly exhibit substantially reduced near-surface warming over the course of the 21st century are the northern north Atlantic, where a slowing AMOC (see Section 6.7.1.1) reduces the northward heat transport and brings the surface temperatures closer to what is found in other ocean basins at these latitudes (Collins et al., 2013 <sup>[[#fn:r66|66]]</sup> ), and the southern side of the Southern Ocean, where water upwells that has been submerged for so long that it has not yet experienced significant anthropogenic climate change (Armour et al., 2016 <sup>[[#fn:r67|67]]</sup> ). Most of these projected warming patterns are broadly consistent across the current and previous generations of climate models (Mitchell et al., 1995 <sup>[[#fn:r68|68]]</sup> ; Collins et al., 2014 <sup>[[#fn:r69|69]]</sup> ) as well as observations and theoretical understanding. These multiple lines of evidence give ''high confidence'' that the projections describe the changes in the real world ( ''high agreement, robust evidence'' ).

The near surface salinity of the ocean is both observed and projected to evolve in ways that reflect the increased intensity of the Earth’s hydrologic cycle (Durack, 2015 <sup>[[#fn:r70|70]]</sup> ) and the increasing near-surface ocean stratification (Zika et al., 2018 <sup>[[#fn:r71|71]]</sup> ). As described in WGI AR5, the ocean surface in areas that currently have net evaporation are expected to become saltier, while areas with net precipitation are expected to get fresher (Rhein et al., 2013 <sup>[[#fn:r72|72]]</sup> ), as the patterns of precipitation and evaporation are generally expected to be amplified (Held and Soden, 2006 <sup>[[#fn:r73|73]]</sup> ). At longer time-scales of decades, the larger scale changes in the ocean circulation and basin integrated freshwater imbalances emerge in the near-surface salinity changes, as shown in Figure 5.3b, with an increasingly salty tropical and subtropical Atlantic and Mediterranean contrasting with a freshening Pacific and polar Arctic emerging as robust signals across the suite of ESMs (Collins et al., 2013 <sup>[[#fn:r74|74]]</sup> ). The freshening of the high latitudes in the north Atlantic and Arctic basin is consistent with the widely expected weakening of the AMOC (also discussed in Section 6.7), hydrological cycle changes and a decline in the volume of sea ice (discussed in Section 3.2.2).

Projected salinity changes in the subsurface ocean reflect changes in the rates of formation of water masses or their newly formed properties (Purich et al., 2018 <sup>[[#fn:r75|75]]</sup> ). Thus, projected freshening of the Southern Ocean surface leads to a freshening of the Antarctic Intermediate Water that is subducted there, flowing northward from the Southern Ocean as a relatively fresh water-mass at depths of 500–1500 m (Figure 5.3b). Increased surface salinity in the Atlantic subtropical gyres are pumped into the interior by the winds, leading to an increased salinity of the interior subtropical gyres, along with contributions from increasingly salty Mediterranean water (Jordà et al., 2017 <sup>[[#fn:r76|76]]</sup> ). Conversely, freshwater capping of the northwestern north Atlantic is projected to inhibit deep convection in the Labrador Sea and the consequent production of Labrador Sea Water in some models (Collins et al., 2013 <sup>[[#fn:r77|77]]</sup> ), and contributes to the increased salinity of the north Atlantic between 1000–2000 m depths (Figure 5.3b).

Identifying the specific patterns of anthropogenic climate changes in oceanic observations is complicated by the presence of basin-scale natural variability with timescales ranging from tidal to multi-decadal, and due to the difficulties associated with maintaining high-precision observing systems spanning the ocean basins and limited observational coverage of the extratropical Southern Hemisphere before 2006 (Rhein et al., 2013 <sup>[[#fn:r78|78]]</sup> ). Inferences based on oceanographic observations from the 1970s onward show wide-spread warming of the upper 700 m (Figure 5.2a), in broad agreement with the ensemble of historical CMIP5 ESM simulations (Figure 5.2b). These ESMs indicate that anthropogenic regional warming over the past half-century should be discernable at the 95% confidence level in much of the upper oceans (un-stippled areas in Figure 5.2b). Most of the areas where observational analyses (Figure 5.2a) exhibit long-term cooling are either regions where the internally generated variability is large enough to mask the trends (e.g., the Eastern Tropical Pacific, Northwest Atlantic, and Kurushio extension east of Japan, which are stippled in Figure 5.2b), or where the observational coverage early in the record is limited and different analyses can disagree about trends (e.g., the Southern Ocean and extratropical South Pacific). When internal variability is taken into account, the broad consistency in the magnitude and regional distribution of observed and simulated 50-year trends gives confidence to the ESM projections of longer-term oceanic changes described previously.

Detailed regional patterns of trends in temperature and heat content at depths of 0–2000 m during the early 21st century are consistent in various analysis, owing to the improved observing network (Roemmich et al., 2015 <sup>[[#fn:r81|81]]</sup> ; Desbruyères et al., 2016a <sup>[[#fn:r82|82]]</sup> ) (Figure 5.2d). At depths of 700–2000 m, observations in all of the ocean basins show broadly warming trends in the well-observed Argo era (2006 to present), with particularly significant warming patterns in the Southern Hemisphere extratropics around 40 o S and the subpolar north Atlantic (Figure 5.3a). These observed changes support the notion that deep ocean heat content has been continuously increasing. As a result, regional climate change signatures emerge from confounding natural variability sooner in the 700–2000 m depth range than in upper 700 m of the ocean, where interannual modes of variability have a larger influence on the circulation (for a more complete discussion see Johnson et al. (2018). Despite regional patches of cooling water in the upper 700 m (Figure 5.2d), every one of the world’s ocean basins volume averaged over depths of 0–2000 m has experienced significant warming over the last decade (Figure 5.3, and also Desbruyères et al. (2016a) <sup>[[#fn:r82|82]]</sup> ). The greatest warming of the top 2000 m has been in the Southern Ocean (Roemmich et al., 2015 <sup>[[#fn:r83|83]]</sup> ; Trenberth et al., 2016 <sup>[[#fn:r84|84]]</sup> ), the tropical and subtropical Pacific Ocean (Roemmich et al., 2015 <sup>[[#fn:r85|85]]</sup> ), and the tropical and subtropical Atlantic Ocean (Cheng and Chen, 2017 <sup>[[#fn:r86|86]]</sup> ). The Southern Hemisphere extratropical oceans accounted for 67–98% of the total ocean heat increase in the uppermost 2000 m for the period of 2006–2013 (Roemmich et al., 2015 <sup>[[#fn:r87|87]]</sup> ). Shi et al. (2018) <sup>[[#fn:r88|88]]</sup> suggest that the dominant ocean heat uptake by the Southern Hemisphere in the early 21st century is expected to become more balanced between the hemispheres as the asymmetric cooling by aerosols decreases.

Large-scale patterns of natural variability at interannual to decadal time scales can mask the long-term warming trend in the upper 700 m, particularly in the tropical Pacific and Indian Oceans (England et al., 2014 <sup>[[#fn:r89|89]]</sup> ; Liu et al., 2016 <sup>[[#fn:r90|90]]</sup> ) and in the north Atlantic (Buckley and Marshall, 2015 <sup>[[#fn:r91|91]]</sup> ). The most significant upper 700 m warming between five-year averages centered on 2007–2015 occurred in a large extratropical band of the Southern Hemisphere between 30ºS–60ºS, and in the tropical Indian Ocean, the eastern North Pacific and western subtropical north Atlantic (Figure 5.2d). Warming of the southern hemisphere subtropical gyres is driven, in part, by an intensification of Southern Ocean winds in recent decades, facilitating the penetration of heat to deeper depths (Gao et al., 2018 <sup>[[#fn:r92|92]]</sup> ). Marginal seas, such as the Mediterranean and Red seas have also exhibited notable warming. Conversely, over this timeframe there were also regions of cooling in the upper 700 m, notably in the north Atlantic around 40 o N–60 o N and in the western tropical Pacific (Figure 5.2d). Recent relatively cold and fresh surface and subsurface conditions in the north Atlantic have been attributed to anomalous atmospheric forcing (Josey et al., 2018 <sup>[[#fn:r93|93]]</sup> ) or weakened transport by the north Atlantic Current and AMOC (Smeed et al., 2018 <sup>[[#fn:r94|94]]</sup> ), and in turn may have contributed to an intensification of deep convection in the Labrador Sea since 2012 (Yashayaev and Loder, 2017 <sup>[[#fn:r95|95]]</sup> ). All these observed decadal changes can be related to internal decadal variability (Robson et al., 2014 <sup>[[#fn:r96|96]]</sup> ; Yeager et al., 2015 <sup>[[#fn:r97|97]]</sup> ) even though they resemble expected longer-term anthropogenically forced trends. Substantial decadal-scale warming and cooling trends in the tropical Pacific and Indian oceans can arise from natural El Niño-Southern Oscillation (ENSO) and Indian Ocean Dipole variability (Han et al., 2014 <sup>[[#fn:r98|98]]</sup> ). Large ensembles of freely running CMIP5 ESM simulations also show that internal variability can dominate the regional manifestation of the anthropogenic climate signal on decadal timescales (Kay et al., 2014 <sup>[[#fn:r99|99]]</sup> ). This is illustrated by the differing warming trends in Figure 5.2e and 5.2f from two identical ESMs that differ only in the weather in their 1850 initial conditions, averaged over the whole 21st century, by contrast, the ensemble of CMIP5 models project statistically significant anthropogenic regional upper 700 m heat content trends almost everywhere (Figure 5.2c).

There are well documented changes in observed ocean temperatures and salinities (Abraham et al., 2013 <sup>[[#fn:r100|100]]</sup> ; Ishii et al., 2017 <sup>[[#fn:r101|101]]</sup> ). However, attributing these changes in the state of the ocean to anthropogenic causes can be challenging due to the presence of internally generated variability, which can swamp the underlying climate change signal in short records and on regional scales. As can be seen in Figure 5.2, the observed long-term trends (Figure 5.2a) exhibit a striking similarity to the CMIP5 ensemble mean in areas where the models suggest that anthropogenic changes should be statistically significant (Figure 5.2b). However, the trends in the shorter well-observed period covering 2005–2017 (Figure 5.2d) exhibits strong trends from internal variability, as illustrated by the differences of two ensemble members of the same ESM with the same forcing but initialised with different weather (Figure 5.2e and 5.2f). Detection and Attribution studies take the internal variability into account and separate the underlying climate signals with the same spatio-temporal sampling as the observations, and apply a range of statistical tests to determine the coherence of the observations with the co-sampled observations (Bindoff et al., 2013 <sup>[[#fn:r102|102]]</sup> ; AR5 WG1 Box 10.1).

Since AR5, the use of different and updated oceanographic data sets and increase in the number of ensembles of the CMIP5 simulations (Kay et al., 2014 <sup>[[#fn:r103|103]]</sup> ) has improved the overall detection and attribution of human influence. Together these measures increase the coherence of the simulations and reduce noise. For example, an isotherm approach used to reduce the noise from the displacement of isotherms in the upper water column allowing detection in each of the mid-latitude ocean basins was achieved on 60-year time series (Weller et al., 2016 <sup>[[#fn:r104|104]]</sup> ). Using all the available ocean temperature and salinity profiles from the Southern Ocean, Swart et al. (2018) show that the warming and freshening patterns were consistent primarily with increased human induced greenhouse gases and secondarily from ozone depletion in the stratosphere, but inconsistent with internal variability. Together the evidence from the AR5, and the discussion above with the new evidence on regional scales across the global oceans, we conclude that the observed long-term upper ocean temperature changes are ''very likely'' to have a substantial contribution from anthropogenic forcing.

The wind-driven ocean circulation at the end of the 21st century is expected to be qualitatively similar to that in the present day, even as important buoyancy-loss driven overturning circulations are expected to weaken. ESM projections suggest that some major ocean current transports will exhibit a modest increase (such as the Kuroshio Extension (Terada and Minobe, 2018) or a small decrease such as for the Indonesian Throughflow (Sen Gupta et al., 2016); many predominantly wind-driven current-system transports are expected to exhibit smaller than 20% changes by 2100 with RCP8.5 forcing. Climate-change induced changes of the circulation in other mid-latitude basins may be difficult to detect or reliably project because of significant natural variability at inter-annual (e.g., El Niño) to decadal (e.g., the Pacific Decadal Oscillation) timescales. The Antarctic Circumpolar Current is projected to be subject to strengthening westerly winds and substantially reduced rates of Antarctic Bottom Water (AABW) formation, as assessed in the Cross-Chapter Box 7 in Chapter 3. The heat transported by the buoyancy-loss driven AMOC, in particular, contributes to the relatively clement climate of northern Europe and the north Atlantic Basin as a whole, although the wind-driven ocean gyres also contribute to the meridional ocean heat transport (see the review by Buckley and Marshall (2015). As a result, there is a concern that significant changes in ocean circulation could lead to localised climate changes that are much larger than the global mean. Projected and observed changes in the AMOC and the rates of formation of deep water-masses in the north Atlantic are discussed in Chapter 6.7.1, along with the possibility of abrupt or enduring changes resulting from forcing by Greenlandic meltwater. A significant reduction in AMOC would, in turn, modestly weaken the Gulf Stream transport, which also has a substantial wind driven component (Frajka-Williams et al., 2016 <sup>[[#fn:r105|105]]</sup> ). Most aspects of the large-scale wind-driven ocean circulation are ''very likely'' to be qualitatively similar to the circulation in the present day, with only modest changes in transports and current location.

The global ocean below 2000 m has warmed significantly between the 1980s and 2010s (Figure 5.4), contributing to ocean heat uptake and through thermal expansion to SLR (Purkey and Johnson, 2010 <sup>[[#fn:r106|106]]</sup> ; Desbruyères et al., 2016b <sup>[[#fn:r107|107]]</sup> ). The observed deep warming rate varies regionally and by depth reflecting differences in the waters influencing particular regions. The deep and abyssal north Atlantic, fed by North Atlantic Deep Water (NADW), has reversed from warming to cooling over the past decade, possibly associated with the North Atlantic Oscillation (NAO) (e.g., Yashayaev, 2007; Desbruyères et al., 201 <sup>[[#fn:r108|108]]</sup> 4) or longer-term weakening in north Atlantic overturning circulation (Caesar et al., 2018 <sup>[[#fn:r109|109]]</sup> ; Thornalley et al., 2018 <sup>[[#fn:r110|110]]</sup> ). The strongest warming is observed in regions of the deep ocean AABW (Purkey et al., 2014 <sup>[[#fn:r111|111]]</sup> ). Regions of the ocean fed by AABW from the Weddell Sea have exhibited a possible slowdown in local AABW warming rates (Lyman and Johnson, 2014 <sup>[[#fn:r112|112]]</sup> ), while the Pacific, fed by AABW from the shelves along the Ross and Adelie Coast, has continued to warm at an accelerating rate between 1990 and 2018 (Desbruyères et al., 2016b <sup>[[#fn:r113|113]]</sup> ).

To date, assessment of deep ocean (below 2000 m) heat content has mostly been from ship-based data collected along decadal repeats of oceanographic transects (Figure 5.4b) (Talley et al., 2016 <sup>[[#fn:r114|114]]</sup> ). While relatively sparse in space and time compared to the upper ocean, these transects were positioned to optimise sampling of most deep ocean basins and provide the highest quality of salinity, temperature and pressure data. Argo floats capable of sampling to 6000 m have just started to populate select deep ocean basins; this Deep Argo data has just started providing regional deep ocean warming estimates (Johnson et al., 2019 <sup>[[#fn:r115|115]]</sup> ). Decadal monitoring by the full global Deep Argo array (Johnson et al., 2015 <sup>[[#fn:r116|116]]</sup> ), complemented by indirect estimates from space (Llovel et al., 2014 <sup>[[#fn:r117|117]]</sup> ; Von Schuckmann et al., 2014), will strongly reduce the currently large uncertainties of deep ocean heat content change estimates in the future.

The spatial and temporal sparseness of observations below 4000 m, along with significant differences between various ESMs, limits our understanding of the exact mechanisms driving the abyssal ocean variability. However, ESMs consistently predict an anthropogenic climate-change induced long-term abyssal warming trend originating in the Southern Ocean due to a reduction in the formation rates of cold AABW (Heuzé et al., 2015 <sup>[[#fn:r118|118]]</sup> ). Although the abyssal modes of natural variability are not as pronounced as closer to the surface, deep ocean heat content can vary on relatively short time scales through the communication of topographic and planetary waves driven by changes in the rate of deep water formation at high latitudes (Kawase, 1987 <sup>[[#fn:r119|119]]</sup> ; Masuda et al., 2010; Spence et al., 2017). AABW has shown variability in properties and production rates over the past half century (Purkey and Johnson, 2013 <sup>[[#fn:r121|121]]</sup> ; Menezes et al., 2017 <sup>[[#fn:r122|122]]</sup> ). A slowdown in AABW formation rates may arise from freshening of shelf waters, changes in local winds driving cross shelf mixing, or larger scale dynamics controlling the spin up or down of Southern Ocean gyres influencing the density of outflowing waters over deep sills. Large-scale circulation changes can also alter the properties of the ambient water that is entrained as dense water descends along the Antarctic continental slopes (Spence et al., 2017 <sup>[[#fn:r123|123]]</sup> ). Evolving AABW properties may also reflect changes in deep Southern Ocean convection. The Weddell Polynya is a large opening in the wintertime ice of the Weddell Sea that is kept ice-free despite intense heat loss to the atmosphere by convective mixing bringing up warm and salty water from the deep ocean. (See Box 3.2 for a more extensive discussion of polynyas and the Weddell Polynya in particular). The Weddell Polynya was present in three of the first years of infrared satellite observations of wintertime sea ice concentrations in the mid-1970s, but it has been closed since 1976, only to reopen in 2016 and 2017. The prominent Weddell Polynya in the mid-1970s greatly increased the volume of the coldest waters in the deep Weddell Sea. Weddell Polynyas are documented to drive abyssal cold and salty signals and can spread thermal signals as waves further and faster than could be explained by slow advective signals (Martin et al., 2015 <sup>[[#fn:r124|124]]</sup> ; Zanowski and Hallberg, 2017 <sup>[[#fn:r125|125]]</sup> ); these waves do not directly heat individual water parcels, but instead warm the ocean where they cause the coldest deep layers to spread laterally and thin. However, recovery from the large Weddell polynya of the early 1970s can only explain about 20% of the observed abyssal warming trend (Zanowski et al., 2015 <sup>[[#fn:r126|126]]</sup> ).

<span id="figure-5.4"></span>
<!-- START IMG -->
<!-- IMG TITLE -->
'''Figure 5.4'''

<span id="figure-5.4-observed-rates-of-warming-from-1981-to-2019-a-as-a-function-of-depth-globally-orange-and-south-of-the-sub-antarctic-front-the-purple-line-in-b-at-about-55s-purple-with-90-confidence-intervals-and-b-average-warming-rate-colours-in-the-abyss-below-4000-m-over-various-ocean-basins-whose"></span>
<!-- IMG CAPTION -->
'''Figure 5.4 | Observed rates of warming from 1981 to 2019 (a) as a function of depth globally (orange) and south of the Sub-Antarctic Front (the purple line in (b) at about 55°S) (purple) with 90% confidence intervals and (b) average warming rate (colours) in the abyss (below 4000 m) over various ocean basins (whose […]'''
<!-- IMG FILE -->
[[File:9f58ef729e7d153da43867c5b6e85cdf IPCC-SROCC-CH_5_4.jpg]]

Figure 5.4 | Observed rates of warming from 1981 to 2019 (a) as a function of depth globally (orange) and south of the Sub-Antarctic Front (the purple line in (b) at about 55°S) (purple) with 90% confidence intervals and (b) average warming rate (colours) in the abyss (below 4000 m) over various ocean basins (whose boundaries are shown in grey lines), with stippling indicating basins with no significant changes. The black lines show the repeat hydrographic sections used to make these estimates. These figures use updated GoShip data and the techniques of Purkey and Johnson (2010).

<!-- END IMG -->
<span id="figure-5.5"></span>
<!-- START IMG -->
<!-- IMG TITLE -->
'''Figure 5.5'''

<span id="figure-5.5-zonal-and-20-year-mean-stratification-averaged-over-the-top-200-m-of-the-ocean-for-the-coupled-model-intercomparison-project-phase-5-cmip5-ensemble-of-simulations-at-the-end-of-the-historical-runs-green-and-for-the-end-of-the-21st-century-for-representative-concentration-pathway-rcp2.6-blue-and-rcp8.5-red-scenarios."></span>
<!-- IMG CAPTION -->
'''Figure 5.5 | Zonal and 20-year mean stratification averaged over the top 200 m of the ocean for the Coupled Model Intercomparison Project Phase 5 (CMIP5) ensemble of simulations at the end of the historical runs (green), and for the end of the 21st century for Representative Concentration Pathway (RCP)2.6 (blue) and RCP8.5 (red) scenarios. […]'''
<!-- IMG FILE -->
[[File:d071ea667782879532641679e563ef15 IPCC-SROCC-CH_5_5.jpg]]

Figure 5.5 | Zonal and 20-year mean stratification averaged over the top 200 m of the ocean for the Coupled Model Intercomparison Project Phase 5 (CMIP5) ensemble of simulations at the end of the historical runs (green), and for the end of the 21st century for Representative Concentration Pathway (RCP)2.6 (blue) and RCP8.5 (red) scenarios. The values between the 5th and 95th percentiles of the ensembles are shaded, while the lines are the ensemble mean. These model results are not adjusted by the control-run, so the spread in the various estimates primarily reflect model formulation differences. The average squared buoyancy frequency shown here is nearly linearly proportional to the density difference between the surface and 200 m, and is a measure of the density stratification of the upper ocean.

The ocean’s properties are changing most rapidly in the near surface waters that are more immediately exposed to atmospheric forcing. As a result of the surface-intensified warming, the upper few hundred meters of the ocean are becoming more stably stratified (Helm et al., 2011 <sup>[[#fn:r126|126]]</sup> ; Talley et al., 2016 <sup>[[#fn:r128|128]]</sup> ). The combination of surface intensified warming and near-surface freshening at high latitudes leading to a projection of more intense near-surface stratification (the downward-increasing vertical gradient of density) across all ocean basins (Figures 5.3 and 5.5) is a robust result with a ''high'' ''agreement'' across successive generations of coupled climate models (Capotondi et al., 2012 <sup>[[#fn:r129|129]]</sup> ; Bopp et al., 2013 <sup>[[#fn:r130|130]]</sup> ). Based on the projected changes from individual models between 1986–2005 and 2081–2100, the mean stratification of the upper 200 m averaged between 60°S–60°N, normalised by the ensemble mean value from 1986–2005 will ''very likely'' increase by between 1.0–9.3% (with 95% confidence and a CMIP5 median change of 2.6%) for RCP2.6, and by between 12.2–30.0% (median value 21.2%) for RCP8.5. Inferences from oceanic observations (Good et al., 2013 <sup>[[#fn:r131|131]]</sup> ) suggest that the 20-year mean stratification averaged between 60°S–60°N and over the top 200 m ''very likely'' increased by between 2.18–2.42% from 1971–1990 to 1998–2017. By contrast, the bottom intensified warming in the abyss (see Figure 5.4) which is consistent with a slowing in the rate of AABW formation, is also associated with a reduction in the abyssal stratification of the ocean (Lyman and Johnson, 2014 <sup>[[#fn:r132|132]]</sup> ; Desbruyères et al., 2016b <sup>[[#fn:r133|133]]</sup> ). Both of these changes have consequences for the evolving turbulence and ocean water-mass structure. Based on observational evidence, theoretical understanding and robust ESM projections, it is ''very likely'' that stratification in the upper few hundred meters of the ocean below the mixed layer will increase significantly in the 21st century over most ocean basins as a result of climate change, and abyssal stratification will ''likely'' decrease.

Many dynamical consequences of increased stratification are understood with ''very high confidence'' (see, for instance, Gill (1982) and Vallis (2017)). For the same turbulent kinetic energy dissipation, locally increased stratification reduces the turbulent vertical diffusivity of heat, salinity, oxygen and nutrients (see Section 5.2.2.2.4). Increased stratification in the tropics and subtropical gyres will ''likely'' lead to a net reduction in the vertical diffusivities of nutrients and other gases within the main thermocline, reducing the flux of nutrients into the euphotic zone and increasing the gradient in oxygen concentrations between the near surface ocean and the interior. Increasing upper ocean stratification (Figure 5.5) acts to restrict the depth of the ocean’s surface mixed layer. Increasing stratification increases the buoyancy frequency and the lateral propagation speed of internal gravity waves and boundary waves by about half the percentage change of the stratification itself. Increasing stratification increases both the length of the internal deformation radius (a typical length scale in baroclinic eddy dynamics) and the horizontal scales of internal tides (see Section 5.2.2.2.3) proportionately with the changes in the internal gravity wave speeds. An increase in stratification will increase the lateral propagation of internal Rossby waves (which set up the basin-scale ocean density structure) proportionately. For the same forcing, increasing stratification reduces the geostrophically balanced slope of density surfaces, and hence the vertical extent of basin-scale wind-driven gyres or coastal upwelling circulations. The flattening of density surfaces by increased stratification inhibits advective exchange between the surface and interior ocean (Wang et al., 2015a <sup>[[#fn:r134|134]]</sup> ), with consequences for the uptake of anthropogenic carbon (Section 5.2.2.3), the evolving oxygen distribution (Section 5.2.2.4) and the supply of nutrients to support primary production (Section 5.2.2.5).

<!-- END IMG -->
<div id="section-5-2-2-2changing-temperature-salinity-circulation-block-6"></div>

<span id="tides-and-coastal-physical-changes-in-a-changing-climate"></span>
===== 5.2.2.2.3 Tides and coastal physical changes in a changing climate =====

Coastal systems are subject to the same large-scale warming trends as the open ocean, but the local response may be dominated by a complex of localised changes in factors such as circulation, mixing, river plumes or the seasonal upwelling of cold water. Using ESMs to project how these factors will interact often requires much finer resolution than is currently affordable in global models, however regional high-resolution models can be effective, especially in marginal seas like the Mediterranean with restricted interactions with the open ocean and that respond primarily to local forcing (Adloff et al., 2015 <sup>[[#fn:r135|135]]</sup> ). High resolution regional models have also been used to project robust localised ocean climate changes in wide shelf seas with more extensive interactions with the open ocean, like those in northwestern Europe (Tinker et al., 2016 <sup>[[#fn:r136|136]]</sup> ). The technical difficulties of using nested regional models are much greater in coasts adjacent to energetic large-scale currents like the Gulf Stream, Kuroshio, and Agulhas, and projecting detailed coastal climate change such places may require the use of expensive high resolution global models (Saba et al., 2016 <sup>[[#fn:r137|137]]</sup> ). These physical coastal changes have consequences that cascade through ecosystems to people, as is illustrated in detail for eastern boundary upwelling systems in Box 5.2.

Both human structures and ecological systems in the coastal zone are directly impacted by tidal amplitudes, which contribute to high-water levels and the tidal flushing rates of estuaries, embayments, marshes and mangroves. The tides are the response of a forced-damped-resonance system (Arbic et al., 2009 <sup>[[#fn:r138|138]]</sup> ). The M 2 tide is the dominant tidal constituent in most places, with a period of half a lunar day, or 12 hours, 25 minutes; the M 2 tides are created by the differential motion of the solid Earth and oceans in response to the gravitational attraction of the moon (Newton, 1687 <sup>[[#fn:r139|139]]</sup> ; Laplace, 1799 <sup>[[#fn:r140|140]]</sup> ). The astronomical forcing evolves only slowly, however the tidal damping and basin resonance at tidal frequencies can change in response to changes in sea level, stratification and coastal conditions (Müller, 2012 <sup>[[#fn:r141|141]]</sup> ; Schindelegger et al., 2018 <sup>[[#fn:r142|142]]</sup> ). Several recent studies have analysed historical coastal tide gauge data and found amplitude trends of order 1 – 4% per century (Ray, 2009; Woodworth, 2010; Müller et al., 2011). In some locations, the changes in the tides have been of comparable importance to changes in mean sea level for explaining changes in high water levels (Jay, 2009). For many individual tide gauges, the trends in tidal amplitude are strongly positively or negatively correlated with local time-mean sea level trends (Devlin et al., 2017 <sup>[[#fn:r145|145]]</sup> ). Another source of secular tidal changes, changes in oceanic stratification, modifies the rate of energy conversion from the barotropic tides to the internal tides (Jayne and St. Laurent, 2001 <sup>[[#fn:r146|146]]</sup> ), the vertical profile of turbulent viscosity on shelves (Müller, 2012 <sup>[[#fn:r147|147]]</sup> ), and the propagation speed of the internal tides (Zhao, 2016 <sup>[[#fn:r148|148]]</sup> ). For example, Colosi and Munk (2006) found an increase in the amplitude of the principal lunar semidiurnal tide M 2 in Honolulu of about 1 cm over the past 100 years, which they attributed primarily to changes in oceanic stratification bringing about local changes in relative phases of the internal and external M 2 tides, increasing constructive interference. Both sea level and stratification are expected to exhibit robust secular positive trends in the coming century due to climate change, at rates that are significantly larger than historical trends, and people may choose to replace natural beaches and marshes with sea-walls in response to rising sea levels. As a result, it is ''very likely'' that the majority of coastal regions will experience statistically significant changes in tidal amplitudes over the course of the 21st century.

Because coastal tides are near resonance in many locations, small changes in sea level and bay shape can change the local tides significantly. For example, the insertion of tidal power plants can have a significant impact on the local tides (Ward et al., 2012 <sup>[[#fn:r149|149]]</sup> ). Various observational and modeling studies demonstrate that SLR has spatial heterogeneous impacts on the tides, with some locations experiencing decreased tidal amplitudes and others experiencing increased tidal amplitudes (Pickering et al., 2012 <sup>[[#fn:r150|150]]</sup> ; Devlin et al., 2017 <sup>[[#fn:r151|151]]</sup> ; Pickering et al., 2017 <sup>[[#fn:r152|152]]</sup> ). Projections of tidal changes indicate that the patterns and even the sign of changes in tidal amplitudes depend on whether the coastlines are allowed to recede with rising sea levels or are held in place (Pickering et al., 2017 <sup>[[#fn:r153|153]]</sup> ; Schindelegger et al., 2018 <sup>[[#fn:r154|154]]</sup> ) . Pelling et al. (2013) <sup>[[#fn:r155|155]]</sup> and Hwang et al. (2014) <sup>[[#fn:r156|156]]</sup> demonstrate that the rapid coastline changes in China’s Bohai Sea have already altered the tides in that region and throughout the Yellow Sea (Hwang et al., 2014 <sup>[[#fn:r157|157]]</sup> ). Pelling and Green (2014) examine the impact of flood defenses as well as SLR on tides on the European Shelf. Such tidal changes have implications for designing flood defenses, for tidal renewable energy, for tidal flushing timescales of estuaries and embayments, and for navigational dredging requirements (Pickering et al., 2012 <sup>[[#fn:r158|158]]</sup> ) (Section 5.4.2). The sign and amplitude of local changes to tides are ''very likely'' to be impacted by both human coastal adaptation measures and climate drivers (listed above).

<div id="section-5-2-2-2changing-temperature-salinity-circulation-block-7"></div>

<span id="systematic-sources-of-uncertainty-in-projections-of-ocean-physical-changes"></span>
===== 5.2.2.2.4 5.2.2.2.4 Systematic sources of uncertainty in projections of ocean physical changes =====

ESMs are able to capture the dynamics of the climate system, but all numerical models have approximations and biases. The most commonly used type of ocean component in ESMs is known to exhibit numerically induced vertical mixing that can be a significant fraction of the physical mixing (Ilıcak et al., 2012 <sup>[[#fn:r159|159]]</sup> ; Megann, 2018 <sup>[[#fn:r160|160]]</sup> ). Because so many ocean models exhibit the same sign of bias, there is a systematic warming of the lower-main thermocline that is not cancelled out when taking the average over the ensemble of all the models in CMIP5. These biases are widely known within the ocean modelling community, and various groups are working to reduce these biases in future ESMs with better ocean model numerics and parameterisations. To correct for model biases, ESM projections are always taken as the difference from a control run without the anomalous forcing. However, some aspects of the ocean response to climate change are nonlinear, and model biases can introduce uncertainties into climate projections. In the case of heat uptake, this is of the order of 10% uncertainty, while for the rate of steric SLR (which depends on the nonlinear equation of state of seawater) the uncertainty in CMIP5 models is of the order of 20% (Hallberg et al., 2012 <sup>[[#fn:r161|161]]</sup> ).

Mesoscale eddies (geostrophic rotating vortices with spatial scales of 10–100 km that penetrate deeply into the water column, and are often described as the ocean’s weather) play an important role in regulating the changes to the larger scale ocean circulation, especially in the Antarctic Circumpolar current, as is discussed in Cross Chapter Box 7. In addition, sub-mesoscale eddies (rotationally influenced motions with smaller horizontal scales of hundreds of metres to about 10 km and intrinsic timescales of a few days that especially arise in association with fronts in the ocean’s surface properties) are known to be particularly important in the dynamics of the near-surface ocean boundary layer (see the review by Mahadevan (2016)). Sub-mesoscale instabilities are associated with re-stratifying overturning circulations that can limit the thickness of the well-mixed ocean surface boundary layer near fronts (Bachman et al., 2017 <sup>[[#fn:r162|162]]</sup> ). Moreover, sub-mesoscale motions generate strong vertical velocities that drive fluxes of nutrients from the interior ocean into the euphotic zone or create pockets of reduced mixing with increased phytoplankton residency time within the euphotic zone (Lévy et al., 2012 <sup>[[#fn:r163|163]]</sup> ). Intense mesoscale eddies are known to create favourable conditions for sub-mesoscale instabilities as shown in both observational (Bachman et al., 2017 <sup>[[#fn:r164|164]]</sup> ) and numerical studies (Brannigan et al., 2017 <sup>[[#fn:r165|165]]</sup> ). Intensifying Southern Ocean eddy fields will have a significant local impact on biological productivity, ecosystem structure, and carbon uptake, both directly and via sub-mesoscale processes. At typical CMIP5 ESM resolutions, it is only in the tropics that mesoscale eddies are adequately resolved to explicitly model their effects (Hallberg, 2013 <sup>[[#fn:r166|166]]</sup> ), while sub-mesoscale eddies are not resolved anywhere, so eddy effects need to be parameterised in ESMs. Despite great progress over the past 30 years in parameterising eddy effects, uncertainties in these parameterisations and how eddies will respond to novel conditions continue to contribute to uncertainties in projections of oceanic climate change ( ''medium confidence'' ).

Ocean turbulent mixing is a key process regulating the ocean circulation and climate. Turbulent mixing is important for the uptake and redistribution of heat, carbon, nutrients, oxygen and other tracers (properties that are carried along with the flow of water) in the ocean (Schmittner et al., 2009 <sup>[[#fn:r167|167]]</sup> ; MacKinnon et al., 2017 <sup>[[#fn:r168|168]]</sup> ). Both observations and theory indicate that turbulent mixing in the ocean is not constant in space or time. Global estimates of both the turbulent kinetic energy dissipation rate and the vertical diffusivity, two measures of ocean turbulence, vary over several orders of magnitude throughout the ocean (Figure 5.6) (Polzin et al., 1997 <sup>[[#fn:r170|170]]</sup> ; Waterman et al., 2012 <sup>[[#fn:r171|171]]</sup> ; Whalen et al., 2012 <sup>[[#fn:r172|172]]</sup> ; Alford et al., 2013 <sup>[[#fn:r173|173]]</sup> ; Hummels et al., 2013 <sup>[[#fn:r174|174]]</sup> ; Sheen et al., 2013 <sup>[[#fn:r175|175]]</sup> ; Waterhouse et al., 2014 <sup>[[#fn:r176|176]]</sup> ; Kunze, 2017 <sup>[[#fn:r177|177]]</sup> ). For a given energy dissipation rate, the turbulent diffusivities of heat, salinity, nutrients and other tracers tend to be smaller with stronger stratification. This dependency on stratification helps explain why the observationally inferred diffusivity in the heavily stratified main thermocline (250–1000 m depth) is of similar magnitude to those deeper in the water column, while the turbulent energy density and dissipation rate are much stronger at the shallower depths (Whalen et al., 2012 <sup>[[#fn:r178|178]]</sup> ). Oceanic turbulence also fluctuates in time, is modulated by tidal cycles (Klymak et al., 2008 <sup>[[#fn:r179|179]]</sup> ), the mesoscale eddy field and seasonal changes (Whalen et al., 2018 <sup>[[#fn:r180|180]]</sup> ). In the mixed layer and directly below, turbulence changes according to local conditions, such as the winds, heating rates and local stratification (Sloyan et al., 2010 <sup>[[#fn:r181|181]]</sup> ; Moum et al., 2013 <sup>[[#fn:r182|182]]</sup> ; D’Asaro, 2014 <sup>[[#fn:r183|183]]</sup> ; Tanaka et al., 2015 <sup>[[#fn:r184|184]]</sup> ) at diurnal to seasonal and longer timescales. These variations in near-surface turbulence need to be taken into account for ESMs to reproduce more accurately the observed seasonal cycle of surface properties and spatial structure of the depth of the thermally well-mixed near surface layer of the ocean. The spatial and temporal patterns of ocean turbulence help shape ocean tracer distributions (heat, dissolved greenhouse gases and nutrients) and how they will evolve in a changing climate ( ''high confidence'' ).

<span id="figure-5.6"></span>
<!-- START IMG -->
<!-- IMG TITLE -->
'''Figure 5.6'''

<span id="figure-5.6-estimate-of-the-average-vertical-turbulent-diffusivity-between-2501000-m-calculated-by-applying-fine-structure-techniques-to-argo-float-data-from-below-the-well-mixed-near-surface-boundary-layer.-only-bins-with-at-least-three-estimates-are-plotted-and-regions-with-insufficient-data-are-coloured-grey.-this-figure-was-created-using-updated-data-through"></span>
<!-- IMG CAPTION -->
'''Figure 5.6 | Estimate of the average vertical turbulent diffusivity between 250–1000 m calculated by applying fine structure techniques to Argo float data from below the well-mixed near-surface boundary layer. Only bins with at least three estimates are plotted and regions with insufficient data are coloured grey. This figure was created using updated data through […]'''
<!-- IMG FILE -->
[[File:53f4b4d15cc76ac8ab7c0b47a417f272 IPCC-SROCC-CH_5_6.jpg]]

Figure 5.6 | Estimate of the average vertical turbulent diffusivity between 250–1000 m calculated by applying fine structure techniques to Argo float data from below the well-mixed near-surface boundary layer. Only bins with at least three estimates are plotted and regions with insufficient data are coloured grey. This figure was created using updated data through April, 2018 with the techniques from Whalen et al. (2012).

Ocean turbulent mixing requires energy sources, many of which are expected to change with a changing climate. Surface wind and buoyancy forcing, the mean and eddying larger-scale ocean circulation itself, and the barotropic tides are all thought to be significant sources of the energy that drives mixing (Wunsch and Ferrari, 2004 <sup>[[#fn:r185|185]]</sup> ). Often this energy first passes through the ocean’s pervasive field of internal gravity waves that propagate and refract through the varying ocean circulation, often breaking into turbulent mixing far from their sources (Eden and Olbers, 2014 <sup>[[#fn:r186|186]]</sup> ; Alford et al., 2016 <sup>[[#fn:r187|187]]</sup> ; Melet et al., 2016 <sup>[[#fn:r188|188]]</sup> ; Meyer et al., 2016 <sup>[[#fn:r189|189]]</sup> ; Zhao et al., 2016b <sup>[[#fn:r190|190]]</sup> ). The energy contributing to the internal waves from the winds and the subsequent turbulence will be altered by changes in tropical storm activity or sea ice coverage. For example, the increasing extent of ice-free Arctic Ocean has already been observed to lead to increased wind-driven internal waves (Dosser and Rainville, 2016 <sup>[[#fn:r191|191]]</sup> ). The Southern Annular Mode is expected to intensify as a result of climate change (Young et al., 2011 <sup>[[#fn:r192|192]]</sup> ; Jones et al., 2016b <sup>[[#fn:r193|193]]</sup> ), bringing with it stronger winds, and more wind-energy input over most of the Southern Ocean and a more intense mesoscale eddy field (Hogg et al., 2015 <sup>[[#fn:r194|194]]</sup> ). Changes in the near-bottom stratification will alter the rate that the barotropic tides generate internal waves, thereby altering the strength and distribution of the tidally generated mixing. Some of the parameterisations of interior ocean mixing used in CMIP5 ESMs take some changing turbulent energy sources into account (Jayne and St. Laurent, 2001 <sup>[[#fn:r195|195]]</sup> ) , and more comprehensive mixing treatments are being developed for use in future generations of ESMs (Eden and Olbers, 2014 <sup>[[#fn:r196|196]]</sup> ). However, not all of the physical processes leading to the rich structure of mixing shown in Figure 5.6 are well understood or included in ESMs; the prospect of significant changes in the patterns and intensity of ocean turbulent mixing is a potential source of uncertainty (probably at the 10% level) in projections of physical and ecological changes in the ocean, including heat uptake, stratification changes, steric SLR, deoxygenisation and nutrient fluxes ( ''medium confidence'' ).

<!-- END IMG -->
<div id="section-5-2-2-3changes-in-ocean-carbon"></div>

<span id="changes-in-ocean-carbon"></span>
==== 5.2.2.3 Changes in Ocean Carbon ====

<div id="section-5-2-2-3changes-in-ocean-carbon-block-1"></div>

Since AR5, new global-scale data synthesis products, novel methods for their analyses, as well as progress in modeling have substantially increased our quantitative understanding of the role of the ocean in absorbing and storing CO 2 from the atmosphere. The most important progress concerns the data-based quantification of the temporal variability of the ocean carbon sink. While AR5 assessed primarily the climatological mean processes governing the ocean carbon cycle, the most recent work now permits us to assess how these processes have changed in recent decades in response to climate variability and change. Here we focus specifically on the open ocean carbon cycle.

<div id="section-5-2-2-3changes-in-ocean-carbon-block-2"></div>

<span id="ocean-carbon-fluxes-and-inventories"></span>
===== 5.2.2.3.1 Ocean carbon fluxes and inventories =====

The analyses of the steadily growing number of surface ocean CO 2 observations (now more than 20 million observations, SOCATv6 ( [http://www.socat.info/index.php/2018/06/19/v6-release/ www.socat.info/index.php/2018/06/19/v6-release] ) demonstrate that the net ocean uptake of CO 2 from the atmosphere has increased from around 1.2 ± 0.5 Pg C yr -1 in the early 1980s to 2.0 ± 0.5 Pg C yr -1 in the years 2010–2015 (Rödenbeck et al., 2014; Landschützer et al., 2016). Once new estimates of the outgassing flux stemming from river derived carbon of 0.8 Pg C yr -1 (Resplandy et al. 2018) are accounted for, these new observations imply that the rate of global ocean uptake of anthropogenic CO 2 increased from 2.0 ± 0.5 Pg C yr -1 to 2.8 ± 0.5 Pg C yr -1 between the early 1980s and 2010–2015 (Rödenbeck et al., 2014; Landschützer et al., 2016; Le Quéré et al., 2018). This increase is supported by the current generation of ocean carbon cycle models (Le Quéré et al., 2018), and commensurate with the increase in atmospheric CO 2 .

The continuing efforts to re-measure dissolved inorganic carbon (DIC) along many of the repeat hydrographic lines that were occupied during the 1980s and 1990 (Talley et al., 2016), alongside the preparation of a global quality controlled database of ocean interior observations (Olsen et al., 2016a), have led to progress since AR5 regarding to the oceanic interior storage of anthropogenic CO 2 . Several studies analysed the changes in the amount of anthropogenic CO 2 that have accumulated between different occupations in the different ocean basins (Wanninkhof et al., 2010; Pérez et al., 2013; Woosley et al., 2016; Carter et al., 2017), confirming that the anthropogenic CO 2 taken up from the atmosphere is transported to depth, where most of it is stored. Using a newly developed reconstruction method, Gruber et al. (2019) extended these results to the globe. They find that between 1994 and 2007, across two standard deviations, that the global ocean has accumulated an additional 30-38 Pg C of anthropogenic CO 2 , which is equivalent to an air-sea CO 2 flux of between 2.3–2.9 Pg C yr -1 (coherent with surface ocean CO 2 observations), bringing the total inventory for the year 2007 to 150 ± 20 Pg C. Extrapolating this estimate to the year 2010 gives an inventory of 158 ± 18 Pg C, which is statistically indistinguishable from the ‘best’ estimate provided by Khatiwala et al. (2013) of 155 ± 31 Pg C and more recently also found from a steady-state ocean model (DeVries, 2014) for this reference year. If the inventory-based estimates are adjusted for the loss of natural carbon, a ''very likely'' total increase in storage between 1994 and 2007 of 24–34 Pg C, or around 25% of total emissions, is found (Gruber, 2019).

Thus, there is ''very high confidence'' from surface ocean and ocean interior carbon data that the strength of the ocean sink for anthropogenic carbon has increased in the last two decades in response to the growth of atmospheric CO 2 . Multiple lines of evidence indicate that it is ''very likely'' that the ocean has taken up 20–30% of the global emissions of CO 2 from the burning of fossil fuels, cement production, and land-use change since the mid 1980s. The consistency between independent surface ocean observations and the ocean interior data-based reconstructions supports the assessment of ''very high confidence'' and provides ''robust evidence'' that fraction of emissions taken up by the ocean has not changed in a statistically significant manner in the last few decades and remains consistent with AR5.

Alongside a globally integrated perspective, these new surface ocean observations also reveal a substantial degree of variability at interannual and decadal scales (Rödenbeck et al., 2015; Landschützer et al., 2016; Le Quéré et al., 2018). Most notable are the air-sea CO 2 flux variations in the tropics linked to ENSO variations (Rödenbeck et al., 2015; Landschützer et al., 2016), as well as the strong decadal variations in the high latitudes, especially the Southern Ocean (Landschützer et al., 2015; Munro et al., 2015; Ritter et al., 2017), discussed further in Chapter 3 (Section 3.2.1.2.4). Fluctuations in the Southern Ocean CO 2 flux are important as they impart a substantial imprint also on the global uptake fluxes. For instance, reduced Southern Ocean uptake in the 1990–2000 period coincided with an exceptionally weak global net uptake of only about 0.8 ± 0.5 Pg C yr −1 .

Thus, there is growing evidence from multiple datasets that the ocean carbon sink exhibits decadal variability at regional scales that significantly alter the globally integrated sink ( ''medium confidence'' ).

Detailed analyses of the spatial structure of the change in storage of anthropogenic CO 2 confirm the variable nature of the ocean carbon sink suggested by the surface observations (Pérez et al., 2013), which are most likely a consequence of changes in ocean circulation (DeVries and Weber, 2017). The increase in anthropogenic CO 2 between 1994 and 2007 occurs throughout the upper 1000 m, but with very different penetration depths, reflecting largely differences in the efficiency, with which the anthropogenic CO 2 is transported from the surface to depth (Gruber et al., 2019) (Figure 5.7). This spatial distribution of how the amount of anthropogenic CO 2 has changed between 1994 and 2007 is similar to the distribution of anthropogenic CO 2 reconstructed for 1994 (Sabine et al., 2004), although the imprint of regional variations in ocean circulation and transport are discernible (Gruber, 2019).

<span id="figure-5.7"></span>
<!-- START IMG -->
<!-- IMG TITLE -->
'''Figure 5.7'''

<span id="figure-5.7-vertical-sections-of-the-change-in-anthropogenic-co2-from-1994-to-2007-represented-by-the-zonal-mean-sections-in-each-ocean-basin-organised-around-the-southern-ocean-in-the-centre.-the-upper-500-m-are-expanded.-contour-intervals-of-anthropogenic-co2-are-2-μmol-kg1-gruber-2019."></span>
<!-- IMG CAPTION -->
'''Figure 5.7 | Vertical sections of the change in anthropogenic CO2 from 1994 to 2007 represented by the zonal mean sections in each ocean basin, organised around the Southern Ocean in the centre. The upper 500 m are expanded. Contour intervals of anthropogenic CO2 are 2 μmol kg–1 (Gruber, 2019).'''
<!-- IMG FILE -->
[[File:61e952057d81b9dbc6f16f4911f31d5c IPCC-SROCC-CH_5_7.jpg]]

Figure 5.7 | Vertical sections of the change in anthropogenic CO2 from 1994 to 2007 represented by the zonal mean sections in each ocean basin, organised around the Southern Ocean in the centre. The upper 500 m are expanded. Contour intervals of anthropogenic CO2 are 2 μmol kg–1 (Gruber, 2019).

<!-- END IMG -->
<div id="section-5-2-2-3changes-in-ocean-carbon-block-3"></div>

<span id="ocean-carbon-chemistry"></span>
===== 5.2.2.3.2 Ocean carbon chemistry =====

Analyses of direct measurements of ocean chemistry from time series stations and merged shipboard studies show consistent decreases in surface-ocean pH over the past few decades. Reductions range between 0.013–0.03 pH units decade -1 over records that span up to 25 years (Table SM5.3). Focusing on the individual time series locations with records longer than 15 years, there is an overall decline of 0.017–0.027 (across 99% confidence intervals). Trends calculated from repeat measurements on ocean surveys show a consistent value of around –0.02 pH units decade -1 for diverse oceanic regions (Table SM5.3), with greater subsurface than surface trends reported in the subtropical oceans (Dore et al., 2009). At larger spatial scales, surface-ocean pH trends are assessed using shipboard observations of the fugacity of CO 2 and estimates of ocean alkalinity (Takahashi et al., 2014; Lauvset et al., 2015). Between 1991–2011, mean surface-ocean pH has declined by 0.018 ± 0.004 units decade –1 in 70% of ocean biomes, with the largest declines in the Indian Ocean (–0.027 units decade –1 ), eastern Equatorial Pacific (–0.026 units decade –1 ) and the South Pacific subtropical (–0.022 units decade –1 ) biomes (Lauvset et al., 2015). Due to the close link between carbonate ion concentrations and pH, mean trends in the stability of mineral forms of aragonite and calcite (known as the ‘saturation state’) that are important for organisms such as coccolithophorids, pteropods and corals follow those of pH, with high-latitude regions most vulnerable to under-saturation due to naturally lower mean values.

It is ''virtually certain'' that ocean pH is declining, and the ''very likely'' range of this decline is 0.017–0.027 pH units per decade for the 8 locations where individual time series observations longer than 15 years exist. This trend is lowering the chemical stability of mineral forms of calcium carbonate and can be attributed to rising atmospheric CO 2 levels.

CMIP5 models are in good agreement with historical observations of declining surface-ocean pH (Figure 5.8a). Models project global surface-ocean declines between 2006–2015 and 2081–2100 of 0.287–0.291 and 0.036–0.042 pH units (both across 99% confidence intervals) for the RCP2.6 and RCP8.5 scenarios, respectively, with higher reductions in the subsurface of subtropical oceans (Bopp et al., 2013; Gattuso et al., 2015). These changes in pH will be greatest in the Arctic Ocean and the high latitudes of the Atlantic and Pacific Oceans due to their lower buffer capacity and are lowest in contemporary upwelling systems (Figure 5.8b) and will also reduce the stability of calcite minerals (Bopp et al., 2013; Gattuso et al., 2015). The area of the surface ocean (0–10 m) characterised by undersaturated conditions in CMIP5 models by 2081–2100 reduces from a ''very likely'' range of 6.4–9.5 x 10 12 m 2 or 5.5–7.3 x 10 13 m 2 under RCP8.5 (as much as 16–20% of ocean surface area for aragonite), to just 0.01–0.2  x 10 12 m 2 or 0.01–0.13 x 10 13 m 2 under RCP2.6 for either calcite or aragonite minerals, respectively. Under RCP8.5, hotspots for undersaturated waters for calcite remain restricted to the Arctic Ocean, while for aragonite, much of the Southern Ocean and the North Pacific and Northwestern Atlantic Oceans are also projected to become undersaturated (Orr et al., 2005; Hauri et al., 2015; Sasse et al., 2015). These results arise from the very well understood reductions in carbonate ion concentrations at lower pH, the vulnerability of regions with naturally low mean values, and the greater overall sensitivity of aragonite solubility. Regional models, with higher resolution that ESMs, also project year-round corrosive conditions for aragonite in some eastern boundary upwelling systems (Franco et al., 2018a). In the ocean interior, the decline in pH and calcium carbonate saturation state is more uncertain across models (Steiner et al., 2014) as it is modulated by changes to ocean overturning and water mass subduction (Resplandy et al., 2013; Chen et al., 2017). Projected benthic changes in pH over the next century are highly localised and are linked to transport of surface anomalies to depth, with over 20% of the north Atlantic sea floor deeper than 500 m projected to experience pH reductions greater than 0.2 units by 2100 under the RCP8.5 scenario (Gehlen et al., 2014). Changes in pH in the abyssal ocean (&gt;3000 m deep) are greatest in the Atlantic and Arctic Oceans, with lesser impact in the Southern and Pacific Oceans by 2100, mainly due to the circulation timescales (Sweetman et al., 2017).

Overall, it is ''virtually certain'' that the future surface open ocean will experience pH drops of either 0.036–0.042 (RCP2.6) or 0.287–0.291 (RCP8.5) pH units by 2081–2100, relative to 2006–2105. These pH changes are ''very likely'' to cause 16–20% of the surface ocean, specifically the Arctic and Southern Oceans, as well as the northern Pacific and northwestern Atlantic Oceans, to experience year-round corrosive conditions for aragonite by 2081–2100. It is ''virtually certain'' these impacts will be avoided under the RCP2.6 scenario. There is ''medium confidence'' , due to the potential for parallel changes in ocean circulation, that the Arctic and north Atlantic seafloors will experience the largest pH changes over the next century.

Although ocean acidification results in long-term trends in mean ocean chemistry, it can also influence seasonal cycles. Observation-based products indicate that the seasonal cycle of global surface-ocean ''p'' CO 2 increased in amplitude by 2.2 ± 0.4 μatm between 1982 and 2014 (Landschützer et al., 2018). CMIP5 models and data-based products similarly project consistent future increases in the seasonal cycle of surface-ocean ''p'' CO 2 under the RCP8.5 emissions scenario, with enhanced amplification in high-latitude waters (McNeil and Sasse, 2016). The amplitude of the seasonal cycle of global surface-ocean free acidity ([H + ]) is projected to increase by 71–91% (across 90% confidence intervals) over the 21st century under RCP8.5, also with greater amplification in the high-latitudes (Kwiatkowski and Orr, 2018). Conversely, models project a 12–20% reduction (across 90% confidence intervals) in the seasonal amplitude of surface-ocean pH, as changes in pH represent relative changes in [H + ] due to their logarithmic relationship, and there are typically greater projected increases in annual mean state [H + ] than the seasonal amplitude of [H + ]. Models also project a 4–14% (across 90% confidence intervals) reduction in the seasonal amplitude of global mean surface-ocean aragonite saturation state under RCP8.5, with a slight amplification in the subtropics being outweighed by dampening elsewhere. The contrasting changes in the seasonal amplitudes of ocean carbonate chemistry variables derive from different sensitivities to atmospheric CO 2 and climate change and to diverging trends in the seasonal cycles of DIC, alkalinity and temperature. Model skill at simulating the seasonal cycles of carbonate chemistry is moderate, with persistent biases in the Southern Ocean, particularly for ''p'' CO 2 , [H + ] and pH (Kwiatkowski and Orr, 2018; Mongwe et al., 2018).

Overall, we assess that alongside the strong mean state changes, it is ''very likely'' that the amplitude of the seasonal cycle in free acidity will increase by 71–91%, while it is ''very likely'' that the seasonal cycles of pH and aragonite saturation will decrease by 12–20% and 4–14%, respectively.

<span id="figure-5.8"></span>
<!-- START IMG -->
<!-- IMG TITLE -->
'''Figure 5.8'''

<span id="figure-5.8-panels-a-d-g-and-j-display-simulated-global-changes-over-the-period-of-19002100-with-solid-lines-representing-the-multi-model-mean-and-the-envelope-representing-90-confidence-intervals-for-rcp8.5-and-rcp2.6-for-surface-ph-o2-concentration-averaged-over-100600-m-depth-upper-100-m-nitrate-concentrations-and-npp-integrated-over"></span>
<!-- IMG CAPTION -->
'''Figure 5.8 | Panels a, d, g and j display simulated global changes over the period of 1900–2100 (with solid lines representing the multi-model mean and the envelope representing 90% confidence intervals for RCP8.5 and RCP2.6), for surface pH, O2 concentration averaged over 100–600 m depth, upper 100 m nitrate concentrations and NPP integrated over […]'''
<!-- IMG FILE -->
[[File:dae0f5b925e76a8d0baa06593df5d103 IPCC-SROCC-CH_5_8-1.jpg]]

Figure 5.8 | Panels a, d, g and j display simulated global changes over the period of 1900–2100 (with solid lines representing the multi-model mean and the envelope representing 90% confidence intervals for RCP8.5 and RCP2.6), for surface pH, O2 concentration averaged over 100–600 m depth, upper 100 m nitrate concentrations and NPP integrated over the top 100 m. Differences are calculated relative to the 1850–1900 period. Panels b, e, h and k show spatial patterns of simulated change in surface pH, upper 100 m nitrate concentrations, O2 concentration averaged over 100 to 600 m depth, and NPP integrated over the top 100 m averaged over 2081–2100, relative to 1850–1900 for RCP8.5. Panels c, f, i and l display time series of the percentage of total uncertainty ascribed to internal variability uncertainty, model uncertainty, and scenario uncertainty in projections of global annual mean changes. Figure adapted after (Frölicher et al. 2016). Please note that confidence intervals can be affected by the different number of models available for the RCP8.5 and RCP2.6 scenarios and for different variables. See also Table SM5.4.

<!-- END IMG -->
<div id="section-5-2-2-4changing-ocean-oxygen"></div>

<span id="changing-ocean-oxygen"></span>
==== 5.2.2.4 Changing Ocean Oxygen ====

<div id="section-5-2-2-4changing-ocean-oxygen-block-1"></div>

Ocean oxygen (O 2 ) levels at the surface are controlled by the balance between oxygen production during photosynthesis, temperature-controlled solubility and air-sea exchange. Deeper in the water column, consumption of oxygen during respiration and redistribution by ocean circulation and mixing are dominant processes. In theory, a warmer more stratified ocean would have a reduced oxygen content, due to the combined influence of lowered gas solubility and a greater interior respiration of organic matter due to enhanced physical isolation of subsurface waters. In accord, global changes in ocean oxygen assessed from three different analyses of compiled global oxygen datasets going back to the 1960s agree that there is a net loss of oxygen from the ocean over all depths (see Table 5.2). For the 0–1000 m depth stratum that contains the most data and is common to all three analyses, oxygen is assessed to have declined by a ''very likely'' range of 0.5–3.3% between 1970 and 2010. For the surface ocean (0–100 m) and the thermocline later of 100–600 m the ''very likely'' range of oxygen declines are 0.2–2.1% and 0.7–3.5%, respectively (Table 5.2). Across two studies, global oxygen is assessed to have declined by a ''very likely'' range of 0.3–2.0%, with a similar range of decline for waters deeper than 600 m (Table 5.2). The regions of lowest oxygen, known as OMZs, with oxygen levels lower than 80 μ mol L -1 ), are observed to be expanding by a ''very likely'' range of 3.0–8.3% across the three studies.

Regionally, all studies agree that the north Pacific and Southern Oceans have shown the largest overall oxygen declines (Figure 5.9), but there is some disagreement regarding the magnitude of the oxygen change in the tropical ocean, with some studies suggesting significant declines (Schmidtko et al., 2017) and other reporting more modest reductions (Helm et al., 2011; Ito et al., 2017) and data coverage is still limited for some regions and deeper than 1000 m. Based on the available data, the strongest declines in deep ocean oxygen have occurred in the Equatorial Pacific, North Pacific, Southern Ocean and South Atlantic, with intermediate declines in the Arctic, South Pacific and Equatorial Atlantic, while the north Atlantic has experienced a moderate oxygen increase below 1200 m (Figure 5.9). A particular difference between parallel oxygen analyses concerns the means of integrating and mapping sparse data across the ocean, both horizontally and vertically, with different studies making specific decisions about averaging grids and integration methods. Moreover, data remains sparse for some ocean regions, depths and periods. Taken together, the challenges of data sparsity, regional differences and the relatively large uncertainties on the oxygen changes across different studies, but also recognising that oxygen declines are significantly different to zero, leads to ''medium confidence'' in the observed oxygen decline.

Syntheses of datasets from local time series tend to document stronger trends, with oxygen declines of over 20% at sites in the northeastern Pacific between 1956–2006 (Whitney et al., 2007), the Northwestern Pacific between 1954–2014 (Sasano et al., 2015) and the California Current between 1984–2011 (Bograd et al., 2015). Despite holding the highest inventory of oxygen in the ocean, oxygen levels in Southern Ocean contributed 25% to the global decline between 1970–1992 (Helm et al., 2011) and have fallen by over 150 Tmol per decade from the 1960s to present (Schmidtko et al., 2017). Observations along ocean cruises as part of the CLIVAR programme have also documented broad thermocline oxygen declines in the northern hemisphere oceans, accompanied by well understood oxygen increases in subtropical and southern hemispheres (Talley et al., 2016).

Overall there is ''medium confidence'' that the oxygen content of the upper 1000 m has declined with a ''very likely'' loss of 0.5–3.3% between 1970-2010. OMZ are expanding in volume, by a ''very likely'' range of 3.0–8.3%. There is ''medium confidence'' that the largest regional changes have occurred in the Southern Ocean, equatorial regions, North Pacific and South Atlantic due to ''medium agreement'' among studies.

The role of ocean warming alone in driving the oxygen changes can be appraised using solubility estimates, which vary between around 15–50% for the upper 1000 m oxygen trend between studies (Helm et al., 2011; Ito et al., 2017; Schmidtko et al., 2017). The role of other processes, linked to changing ocean ventilation and respiration are challenging to appraise directly, but tend to reinforce the impacts from warming and are probably predominant overall (Oschlies et al., 2018). Indeed, that the observed oxygen decline is negatively correlated with ocean heat content changes (Ito et al., 2017) reflects the overriding role of changing ocean ventilation and associated processes (see also Section 5.2.2). That the ratio of the associated oxygen to heat changes is larger than would be expected from thermal processes alone also highlights the role played by other processes (Oschlies et al., 2018). Local oxygen trends have emphasised the role of changes to ocean physics in western Northern Pacific (Whitney et al., 2013); Sasano et al. (2015), the southern California Current region (Goericke et al., 2015), and the Santa Barbara Basin (Goericke et al., 2015). In regions of high mesoscale activity, such as the tropical north Atlantic, low oxygen eddies can have a significant impact on oxygen dynamics (Karstensen et al., 2015; Grundle et al., 2017). Oxygen fluctuations in the deep ocean have been linked to changes in large scale ocean circulation (Watanabe et al., 2003; Stendardo and Gruber, 2012) and at the global scale, the observed oxygen decline is negatively correlated with ocean heat content changes (Ito et al., 2017). Changes to respiration rates, either due to temperature enhancement or in the amount/quality of organic material can also be important and the enhanced respiratory demand associated with an intensified monsoon has been invoked as a driver of the expansion of the Arabian Sea OMZ (Lachkar et al., 2018).

Ocean oxygen changes are also affected by climate variability on interannual and decadal timescales, especially for the tropical ocean OMZs (Deutsch et al., 2011). ENSO variability in particular affects the thermocline structure, which then alongside changes in circulation modulates oxygen solubility and respiratory demand in this region (Ito and Deutsch, 2013; Eddebbar et al., 2017). These drivers may then be combined with modifications to overturning and ventilation of OMZs by lateral jets and equatorial current intensity (Duteil et al., 2014). Centennial scale studies based on isotope proxies for low oxygen regions have demonstrated fluctuations in OMZ extent linked to decadal changes in tropical trade winds that affects interior ocean respiratory oxygen demand, which implies that it will be difficult to attribute recent changes in the Pacific OMZ to anthropogenic forcing alone (Deutsch et al., 2015). Parallel work based on oxygen observations (Llanillo et al., 2013), as well as modelling (Duteil et al., 2018) supports the importance of decadal scale variability in the eastern tropical Pacific OMZ. There is some evidence for the potential of a modulating impact on tropical Pacific oxygen at interannual timescales from atmospheric deposition of nitrogen and iron (Ito et al., 2016; Yang and Gruber, 2016).

<div id="section-5-2-2-4changing-ocean-oxygen-block-2"></div>

<span id="table-5.2"></span>
<!-- START TABLE -->
'''Table 5.2'''

Observed oxygen changes for the period 1970–2010 for 6 different layers within the ocean. The changes are shown as percentage change of global averages. The layers are depths 0–100, 100–600, 0–1000, and 600–bottom are in metres. The oxygen minimum zone (OMZ) is defined as the ocean volume change that is less than 80 μ mol L -1 . The estimates and confidence intervals are based published papers (Schmidtko et al. 2018, Ito et al. 2017 and Helm et al. 2011). The assessed change is the average of the available estimates and the 90% Confidence Interval (CI) combines the confidence as their standard deviation with two degrees of freedom.

<!-- TABLE -->
{| class="wikitable"
|-
|
| '''Schmidtko'''
|
| '''Ito'''
|
| '''Helm'''
|
| '''Assessed Change'''
|
|-
| '''Layer'''
| '''Period'''
| '''Change'''
| '''90 CI'''
| '''Change'''
| '''90 CI'''
| '''Change'''
| '''90 CI'''
| '''Change'''
| '''90 CI'''
|-
| '''0–100'''
| 1970–2010
| –0.38%
| ±1.06%
| –1.65%
| ±0.63%
| –1.30%
| ±0.54%
| –1.11%
| ±0.95%
|-
| '''100–600'''
| 1970–2010
| –1.06%
| ±1.36%
| –3.17%
| ±1.34%
| –2.04%
| ±0.60%
| –2.09%
| ±1.42%
|-
| '''0–1000'''
| 1970–2010
| –1.35%
| ±1.38%
| –2.70%
| ±1.30%
| –1.74%
| ±0.54%
| –1.93%
| ±1.39%
|-
| '''600–bottom'''
| 1970–2010
| –1.51%
| ±0.62%
| n.a.
| –0.81%
| ±0.57%
| –1.16%
| ±0.84%
|-
|
|-
| '''OMZ'''
| 1970–2010
| 6.33%
| ±2.52%
| 6.10%
| 1.2%
| 4.49%
| ±2.25%
| 5.64%
| ±2.66%
|-
|
|-
| '''Global'''
| 1970–2010
| –1.43%
| ±0.70%
| n.a.
| –0.87%
| ±0.53%
| –1.15%
| ±0.88%
|}
<!-- END TABLE -->

At the global scale, there is ''high confidence'' that the impact of a warmer ocean on oxygen levels is reinforced by other processes associated with ocean physics and biogeochemistry, which cause the majority of the observed oxygen decline. For the tropical Pacific OMZ, there is ''medium confidence'' arising from ''medium agreement'' from ''medium evidence'' that low frequency decadal changes in ocean physics have controlled past fluctuations in OMZ extent.

<div id="section-5-2-2-4changing-ocean-oxygen-block-3"></div>

<span id="figure-5.9"></span>
<!-- START IMG -->
<!-- IMG TITLE -->
'''Figure 5.9'''

<span id="figure-5.9-absolute-change-in-dissolved-oxygen-umol-kg1-per-decade-between-water-depths-of-a-0-and-1200-m-and-b-1200-m-and-the-sea-floor-over-the-period-19602010.-lines-indicate-boundaries-of-omzs-with-less-than-80-μ-mol-kg1-oxygen-anywhere-within-the-water-column-dasheddotted-less-than-40"></span>
<!-- IMG CAPTION -->
'''Figure 5.9 | Absolute change in dissolved oxygen (umol kg–1 per decade) between water depths of (a) 0 and 1200 m, and (b) 1200 m and the sea floor over the period 1960–2010. Lines indicate boundaries of OMZs with less than 80 μ mol kg–1 oxygen anywhere within the water column (dashed/dotted), less than 40 […]'''
<!-- IMG FILE -->
[[File:b75db0c50b18807042b37db95dc0403a IPCC-SROCC-CH_5_9.jpg]]

Figure 5.9 | Absolute change in dissolved oxygen (umol kg–1 per decade) between water depths of (a) 0 and 1200 m, and (b) 1200 m and the sea floor over the period 1960–2010. Lines indicate boundaries of OMZs with less than 80 μ mol kg–1 oxygen anywhere within the water column (dashed/dotted), less than 40 μ mol kg–1 (dashed) and less than 20 μ mol kg–1 (solid). Redrawn from Oschlies et al. (2018).

<!-- END IMG -->
<div id="section-5-2-2-4changing-ocean-oxygen-block-4"></div>

Future changes in oxygen can be appraised from ESMs that account for the combined effects of ocean physics and biogeochemistry. Globally, these models project that it is ''very likely'' oxygen will decline by 3.2–3.7% or 1.6–2.0% (both across 90% confidence limits) for RCP8.5 or RCP2.6, respectively, relative to 2000 (Bopp et al., 2013). Focussing on the 100–600 m depth stratum, O 2 changes by –4 to –3.1% for the RCP8.5 or by –0.5–0.1% for the RCP2.6 scenario (relative to 2006–2015, Figure 5.8d). It should be noted that ESMs appear to be underestimating the rate of oxygen change from available datasets from the historical period (Oschlies et al., 2018) . Increased tropical ocean stratification reduces interior ocean oxygen by diminishing pathways of ventilation in the subtropical gyres and by inhibiting turbulent mixing with the oxygen-rich surface ocean (see Section 5.2.2.2.4). This relatively robust global modelled trend (Figure 5.8d) however masks important uncertainties in the projection of regional trends (Figure 5.8e), particularly in the tropical ocean OMZs (Bopp et al., 2013; Cocco et al., 2013; Cabré et al., 2015). The uncertainty in the trends in tropical ocean OMZs arises due to the fact that oxygen depletion due to warming induced reductions in oxygen saturation are opposed by oxygen enrichment due to reduced oxygen consumption during respiration in response to predicted declines in marine export production, as well as biases due to model resolution in the tropics and the length of the model spin up (Bopp et al., 2017). The 80 μ mol L -1 threshold that may be used to define the volume of the oxygen minimum is projected to grow by a ''very likely'' range of 7.0 ± 5.6% by 2100 during the RCP8.5 scenario or show virtually no change during the RCP2.6 scenario, relative to a 1850–1900 reference period (Figure 5.10). At the seafloor, between 200–3000 m depth strata, the north Pacific, north Atlantic, Arctic and Southern Oceans may see oxygen declines by 0.3–3.7% by 2100 (relative to 2005), with abyssal ocean changes being lower and more localised around regions in the north Atlantic and Southern Ocean (Sweetman et al., 2017), but will be modulated by any future changes in overturning strength. There is ''high confidence'' that the largest changes in deep sea systems will occur after 2100 (Battaglia and Joos, 2018).

<div id="section-5-2-2-4changing-ocean-oxygen-block-5"></div>

<span id="figure-5.10"></span>
<!-- START IMG -->
<!-- IMG TITLE -->
'''Figure 5.10'''

<span id="figure-5.10-the-evolution-of-the-volume-of-the-100600-m-layer-of-the-ocean-with-oxygen-concentrations-less-than-80-mmol-l1-for-the-rcp8.5-red-line-and-the-rcp2.6-blue-line-normalised-to-the-volume-in-18501900.-dashed-lines-indicated-the-very-likely-range-90-confidence-intervals-across-the-cmip5-models-cnrm-cm5"></span>
<!-- IMG CAPTION -->
'''Figure 5.10 | The evolution of the volume of the 100–600 m layer of the ocean with oxygen concentrations less than 80 mmol L–1 for the RCP8.5 (red line) and the RCP2.6 (blue line), normalised to the volume in 1850–1900. Dashed lines indicated the very likely range (90% confidence intervals) across the CMIP5 models (CNRM-CM5, […]'''
<!-- IMG FILE -->
[[File:d0b3d4255f66d71da9ffd86d3c2fc212 IPCC-SROCC-CH_5_10.jpg]]

Figure 5.10 | The evolution of the volume of the 100–600 m layer of the ocean with oxygen concentrations less than 80 mmol L–1 for the RCP8.5 (red line) and the RCP2.6 (blue line), normalised to the volume in 1850–1900. Dashed lines indicated the very likely range (90% confidence intervals) across the CMIP5 models (CNRM-CM5, GFDL-ESM2M, GFDL-ESM2G, IPSL-CM5A-LR, IPSL-CM5A-MR, MPI-ESM-LR, MPI-ESM-MR and the NCAR-CESM1 models). Models are corrected for drift in O2 using their control simulations.

<!-- END IMG -->
<div id="section-5-2-2-4changing-ocean-oxygen-block-6"></div>

Simulations extended to 2300 suggest that by 2150 the trend of declining tropical ocean oxygen (both in terms of concentrations and volume of low oxygen waters) may reverse itself, mainly due to the effect of strong declines in primary production and organic matter fluxes to the ocean interior (Fu et al., 2018) or due to enhanced Antarctic ventilation (Yamamoto et al., 2015), but with ''low confidence'' due to ''limited evidence'' . At the global scale, 10,000 year intermediate complexity model simulations find that overall ocean oxygen loss shows near linear relationships to equilibrium temperature, itself linearly related to cumulative emissions, and any climate mitigation scenario will reduce peak oxygen loss by 4.4% per degree Celsius of avoided warming (Battaglia and Joos, 2018).

In summary, the total oxygen content of the ocean is ''very likely'' to decline by 3.2–3.7% by 2100, relative to 2000, for RCP8.5 or by between 1.6–2.0% for RCP2.6 with ''medium confidence'' . There is ''medium confidence'' that sea floor changes will be more localised in the north Atlantic and Southern Oceans by 2100, but ''high confidence'' that the largest deep sea floor changes in oxygen will occur after 2100.

<div id="section-5-2-2-5changing-ocean-nutrients"></div>

<span id="changing-ocean-nutrients"></span>
==== 5.2.2.5 Changing Ocean Nutrients ====

<div id="section-5-2-2-5changing-ocean-nutrients-block-1"></div>

Changes to ocean nutrient cycling are driven by modifications to ocean mixing and transport (Section 5.2.2.2.2), internal biogeochemical cycling and fluctuations in external supply, particularly from rivers and the atmosphere. This assessment will focus on the main nutrients important for driving microbial growth (Section 5.2.2.6), namely nitrogen, phosphorus and iron.

Diverse studies (including shipboard experiments and use of protein biomarkers) have highlighted nitrogen and phosphorus limitation in the stratified tropical ocean regions accompanied by widespread iron limitation at high latitudes and in upwelling regions that typically have elevated levels of productivity (Figure 5.11) (Moore et al., 2013 <sup>[[#fn:r288|288]]</sup> ; Saito et al., 2014 <sup>[[#fn:r289|289]]</sup> ; Browning et al., 2017 <sup>[[#fn:r290|290]]</sup> ; Tagliabue et al., 2017 <sup>[[#fn:r291|291]]</sup> ). Moreover, more extensive experimental work has demonstrated overlapping nitrogen-iron co-limitation at the boundaries between gyre and upwelling regimes (Browning et al., 2017 <sup>[[#fn:r292|292]]</sup> ). There is ''high confidence'' arising from ''robust evidence'' and ''high agreement'' across different types of studies that the main limiting nutrient is either iron (in most major upwelling regions and the Southern, north Atlantic and sub-Arctic Pacific Oceans) or nitrogen and phosphorus (in the low productivity tropical ocean gyres).

<div id="section-5-2-2-5changing-ocean-nutrients-block-2"></div>

<span id="figure-5.11"></span>
<!-- START IMG -->
<!-- IMG TITLE -->
'''Figure 5.11'''

<span id="figure-5.11-map-of-the-dominant-limiting-resource-moore-et-al.-2013-updated-to-include-new-experiments-from-the-north-pacific-tropical-atlantic-and-south-east-atlantic-browning-et-al.-2017-shilova-et-al.-2017.-the-background-is-depth-integrated-primary-productivity-using-the-vertically-generalized-production-model-algorithm.-colouring-of-the-circles-indicates"></span>
<!-- IMG CAPTION -->
'''Figure 5.11 | Map of the dominant limiting resource (Moore et al. 2013), updated to include new experiments from the north Pacific, tropical Atlantic and south east Atlantic (Browning et al. 2017; Shilova et al. 2017). The background is depth integrated primary productivity using the Vertically Generalized Production Model algorithm. Colouring of the circles indicates […]'''
<!-- IMG FILE -->
[[File:c72314062f49b48cb69732a24549a490 IPCC-SROCC-CH_5_11.jpg]]

Figure 5.11 | Map of the dominant limiting resource (Moore et al. 2013), updated to include new experiments from the north Pacific, tropical Atlantic and south east Atlantic (Browning et al. 2017; Shilova et al. 2017). The background is depth integrated primary productivity using the Vertically Generalized Production Model algorithm. Colouring of the circles indicates the primary limiting nutrients inferred from chlorophyll and/or primary productivity increases following artificial amendment of: N (blue), P (black), Fe (red), Co (yellow) and Zn (cyan). Divided circles indicate potentially co-limiting nutrients, for example, a red-blue divided circle indicates Fe-N co-limitation.

<!-- END IMG -->
<div id="section-5-2-2-5changing-ocean-nutrients-block-3"></div>

There is ''limited evidence'' on contemporary trends in nutrient levels, either from time series sites or broader meta-analyses. Increasing inputs of anthropogenic nitrogen from the atmosphere are perturbing ocean nutrient levels (Jickells et al., 2017 <sup>[[#fn:r293|293]]</sup> ). In the North Pacific in particular, additional atmospheric nitrogen input has raised the nitrogen to phosphorus ratio between 1988–2011 and induced a progressive shift towards phosphorus limitation in this region (Kim et al., 2011 <sup>[[#fn:r294|294]]</sup> ; Kim et al., 2014 <sup>[[#fn:r295|295]]</sup> ; Ren et al., 2017 <sup>[[#fn:r296|296]]</sup> ). This tendency is supported by modelling experiments that find enhanced atmospheric nitrogen input only has a small influence on productivity due to expanded phosphorus limitation (Yang and Gruber, 2016) and other nitrogen cycle feedbacks (Somes et al., 2016 <sup>[[#fn:r298|298]]</sup> ; Landolfi et al., 2017 <sup>[[#fn:r299|299]]</sup> ).

In general, future increases in stratification (Dave and Lozier, 2013 <sup>[[#fn:r300|300]]</sup> ; Talley et al., 2016 <sup>[[#fn:r301|301]]</sup> ; Kwiatkowski et al., 2017 <sup>[[#fn:r302|302]]</sup> ; and see also Section 5.2.2.2) will trap nutrients in the ocean interior and reduce upper ocean nutrient levels, alongside an additional local impact from changes to atmospheric delivery. However, no CMIP5 models accounted for changes in nutrient delivery from dust and anthropogenic aerosols during their experiments, which could be an important component of regional change (Wang et al., 2015b <sup>[[#fn:r303|303]]</sup> ; Somes et al., 2016 <sup>[[#fn:r304|304]]</sup> ; Yang and Gruber, 2016 <sup>[[#fn:r305|305]]</sup> ). ESMs project a decline in the nitrate content of the upper 100 m of 9–14% or 1.5–6% (across 90% confidence intervals) for the RCP8.5 or RCP2.6 scenario, respectively, by 2081–2100 relative to 2006–2015 (Figure 5.8g). The largest absolute declines in nitrate content is projected in the present day upwelling zones (Figure 5.8h). Projected changes to upper 100 m nitrate concentrations are significantly different to zero for both RCP8.5 and RCP2.6 at the 90% confidence level, but are overall lower for the RCP2.6. Scenario, internal variability and inter-model variability contribute roughly equally to the overall projection uncertainty in 2100 (Figure 5.8i) and there is no clear separation of nitrate trends between RCP8.5 and RCP2.6 outside the model uncertainty (Figure 5.8h).

Iron concentrations are projected to increase in the future from ESM simulations, due to enhanced lateral transport into high-latitude oceans and reduced biological consumption in regions of declining nitrate (Misumi et al., 2013 <sup>[[#fn:r314|314]]</sup> ). Other modelling efforts also suggest greater levels of the more biologically available Fe(II) species in a warmer and more acidic ocean (Tagliabue and Völker, 2011 <sup>[[#fn:r315|315]]</sup> ). These modelling studies tend to indicate greater ocean iron availability in the future overall, but the very limited skill of contemporary global ocean iron models in reproducing observations available from the new basin scale datasets from the international GEOTRACES program and neglect for parallel dust supply changes lower the confidence in the models’ projected changes (Tagliabue et al., 2016 <sup>[[#fn:r316|316]]</sup> ).

Overall, nitrate concentrations in the upper 100 m are ''very likely'' to decline by 9–14% by 2081–2100, relative to 2006–2015 for RCP8.5 or 1.5–6% for RCP2.6, in response to increased stratification, with ''medium confidence'' in these projections due to the ''limited evidence'' of past changes that can be robustly understood and reproduced by models. Surface ocean iron levels is projected to increase in the 21st century with ''low confidence'' due to systemic uncertainties in these models.

<div id="section-5-2-2-6changing-ocean-primary-and-export-production"></div>

<span id="changing-ocean-primary-and-export-production"></span>
==== 5.2.2.6 Changing Ocean Primary and Export Production ====

<div id="section-5-2-2-6changing-ocean-primary-and-export-production-block-1"></div>

Ocean primary productivity is a key process in the ocean carbon cycle (see Section 5.2.2.3), as well as for supporting pelagic ocean ecosystems (see Section 5.2.3). NPP is the product of phytoplankton growth rate and standing stock. Phytoplankton growth is controlled by the combination of temperature, light and nutrients, while the phytoplankton standing stock is modified by both gains from growth and losses due to grazing by zooplankton (Figure 5.12). Export production is here defined as the sinking flux of particulate organic carbon (produced by NPP) across a specified depth horizon. Otherwise known as the biological pump, export production is also a key component of the global carbon cycle (see Section 5.2.2.3) and an essential food supply to benthic organisms (see Section 5.2.3.2). Export production is regulated by the level of primary production and the transfer efficiency with depth, itself controlled by the type of sinking organic carbon, which is affected by the upper ocean food web structure (Boyd et al., 2019 <sup>[[#fn:r317|317]]</sup> ).

Satellite datasets that use mathematical algorithms to convert ocean colour, often alongside other remotely sensed information, into chlorophyll or other indexes of phytoplankton biomass and NPP provide the potential to deliver a global meta-analysis of changes in NPP. Since AR5, a variety of studies have reported relatively insignificant changes in overall open ocean chlorophyll levels of &lt;±1% yr –1 for individual time periods (Boyce et al., 2014 <sup>[[#fn:r318|318]]</sup> ; Gregg and Rousseaux, 2014 <sup>[[#fn:r319|319]]</sup> ; Boyce and Worm, 2015 <sup>[[#fn:r320|320]]</sup> ; Hammond et al., 2017 <sup>[[#fn:r321|321]]</sup> ). Regionally, trends of ±4% between 2002–2015 for different regions are found when different satellite products are merged, with increases at high latitudes and moderate decreases at low latitudes (Mélin et al., 2017 <sup>[[#fn:r322|322]]</sup> . While some studies report good comparability of merged products (Mélin et al., 2017 <sup>[[#fn:r323|323]]</sup> ), others highlight significant mismatches regarding absolute values and decadal trends in NPP between NPP algorithms (Gómez-Letona et al., 2017 <sup>[[#fn:r324|324]]</sup> ). Satellite derived NPP shows significant mismatches when compared to ''in situ'' data and reducing uncertainties in derived NPP is a high priority for the community (Lee et al., 2015 <sup>[[#fn:r325|325]]</sup> ), although there is a reasonable correlation in higher biomass coastal regions (Kahru et al., 2009 <sup>[[#fn:r326|326]]</sup> ). Importantly, satellite records are not yet long enough to unambiguously isolate long term climate related trends from natural variability (Beaulieu et al., 2013 <sup>[[#fn:r327|327]]</sup> ). Overall, there is ''low confidence'' in satellite-based trends in global ocean NPP due to the time series length and lack of corroborating ''in situ'' measurements or other validation time series. This is especially true at regional scales where distinct sets of poorly understood processes dominate.

Future changes in NPP will result from the changing influence from temperature, light, nutrients and grazing (Figure 5.12). Across CMIP models, NPP is predicted to broadly decline or remain constant by 2081–2100, with mean changes by 2100 of –3.8 to –10.6% and –1.1–0.8% across 90% confidence intervals for the RCP85 and RCP26 scenario, respectively (all relative to 2006–2015), with a strong degree of regional symmetry (Figure 5.8k). As seen for nitrate, changes are most marked in low-latitude upwelling regions, which are projected to show the largest absolute declines. As for nitrate, projected NPP changes are lower for the RCP26 scenario (Figure 5.8j), but the overall uncertainty is dominated by internal and inter-model variability in 2100 (Figure 5.8l) which results in no clear separation of NPP trends between the RCP85 and RCP26 (Figure 5.8j). Tropical ocean NPP is projected to show a large decline, but is underpinned by substantial intermodal uncertainty, with mean changes of 11 ± 24% across the suite of CMIP5 models by 2100, relative to 2000 under RCP8.5 (Laufkötter et al., 2015 <sup>[[#fn:r328|328]]</sup> ). However, if emergent constraints from the historical record that link the variability of tropical productivity to temperature anomalies then a four-fold decline in inter-model uncertainty results. This leads to a projected tropical ocean decline of 11 ± 6%, or from 6.8–16.2% across 90% confidence limits, depending on which historical constraint is used (Kwiatkowski et al., 2017 <sup>[[#fn:r329|329]]</sup> ). NPP is projected to increases for higher latitude regions, such as the Arctic and Southern Oceans.

Detailed analyses of the interplay between different drivers of NPP, including temperature, light, nutrient levels and grazing from a subset of CMIP5 models, reveals a complex interplay with a strong latitudinal dependence (Laufkötter et al., 2015 <sup>[[#fn:r330|330]]</sup> ) summarised in Figure 5.12. Warming acts to enhance growth, most notably at lower latitudes, while light conditions are also predicted to improve, mostly at the poles. Nutrient limitation shows a much more complex response across models, but tends to increase in the tropics and northern high latitudes, with little change in the Southern Ocean. Taken together there is a tendency for reduced growth rates across the entire ocean, but there is a large amount of inter-model variability. The changes in growth are allied to a consistent increase in the grazing loss of biomass to upper trophic levels. Since AR5, we have an increasing body of literature concerning role of biological feedbacks, especially due to interactions between organisms, specific physiological responses and from upper trophic levels on nutrient concentrations, linked to variable food quality (Kwiatkowski et al., 2018 <sup>[[#fn:r331|331]]</sup> ), resource recycling (Boyd et al., 2015a <sup>[[#fn:r332|332]]</sup> ; Tagliabue et al., 2017 <sup>[[#fn:r333|333]]</sup> ) and interactions between organisms (Lima-Mendez et al., 2015 <sup>[[#fn:r334|334]]</sup> ), but their role in shaping the response of NPP to climate change remains a major unknown. Lastly, modelling work suggests that the increasing deposition of anthropogenic aerosols (supplying N and Fe) stimulates biological activity (Wang et al., 2015b <sup>[[#fn:r335|335]]</sup> ) and may compensate for warming driven reductions in primary productivity (Wang et al., 2015b <sup>[[#fn:r336|336]]</sup> ), but these effects do not form part of the CMIP5 projections assessed here.

CMIP5 models show a strong negative relationship between changes in stratification that reduces net nutrient supply and integrated export production (Fu et al., 2016 <sup>[[#fn:r337|337]]</sup> ). Export production is projected to decline by 8.9–15.8% or 1.6–4.9% (across 90% confidence intervals) by 2100, relative to 2000 for the RCP8.5 or RCP2.6 scenario, respectively (Bopp et al., 2013 <sup>[[#fn:r338|338]]</sup> ; Fu et al., 2016 <sup>[[#fn:r339|339]]</sup> ; Laufkötter et al., 2016 <sup>[[#fn:r340|340]]</sup> ). The projected changes in export production can be larger than global primary production because they are affected by both the NPP changes, but also how shifts in food web structure modulates the ‘transfer efficiency’ of particulate organic material (Guidi et al., 2016 <sup>[[#fn:r341|341]]</sup> ; Tréguer et al., 2018 <sup>[[#fn:r342|342]]</sup> ), which then affects the sinking speed and lability of exported particles through the ocean interior to the sea floor (Bopp et al., 2013 <sup>[[#fn:r343|343]]</sup> ; Fu et al., 2016 <sup>[[#fn:r344|344]]</sup> ; Laufkötter et al., 2016 <sup>[[#fn:r345|345]]</sup> ). Declines in export production over much of the ocean mean that the flux arriving at the sea floor is also predicted to decline, while increases in export production are projected in the polar regions that see enhanced NPP (Sweetman et al., 2017 <sup>[[#fn:r346|346]]</sup> ).

The realism in model projections can be appraised via their ability to accurately simulate the limiting nutrient in specific ocean regions (Figure 5.11), with high model skill in reproducing surface distributions of nitrate and phosphate (Laufkötter et al., 2015 <sup>[[#fn:r347|347]]</sup> ), raising confidence in projections in nitrogen and phosphorus limited systems, but poor skill in reproducing iron distributions (Tagliabue et al., 2016 <sup>[[#fn:r348|348]]</sup> ) lowering confidence in iron limited regions (Figure 5.11). In addition to concentrations of specific nutrients, the response of NPP to environmental change is strongly controlled by accurate representation of the ratio of resources (Moreno et al., 2017 <sup>[[#fn:r349|349]]</sup> ). Overall CMIP5 models skill in reproducing patterns of NPP and export production from limited satellite derived estimates range from poor to average (correlation coefficients of 0.1–0.6 across different models (Laufkötter et al., 2016; Moreno et al., 2017 <sup>[[#fn:r350|350]]</sup> )), but it should be noted that complete comprehensive observational datasets do not exist for these metrics with very few ''in situ'' observations. As export production is a much better understood net integral of changing net nutrient supply (Sarmiento and Gruber, 2002 <sup>[[#fn:r351|351]]</sup> ) and can be constrained by interior ocean nutrient and oxygen levels, there is ''medium confidence'' in these projections for global changes. Improving the ability of models to reproduce historical NPP is crucial for more accurate projections as model biases in simulating contemporary ocean biogeochemistry play a key role in driving future projections (Fu et al., 2016 <sup>[[#fn:r352|352]]</sup> ).

Overall, these assessments balance the range of projections across models alongside the strength of different kinds of observational constraints available, as well as our theoretical or experimental understanding of the impact of a warmer, more stratified ocean on NPP and export production. As for AR5, net primary productivity is ''very likely'' to decline by 4–11% by 2081–2100, relative to 1850–1900, across CMIP5 models for RCP8.5, but there is ''low confidence'' for this estimate due to the ''medium agreement'' among models and the ''limited evidence'' from observations. It is ''very likely'' that tropical NPP will decline by 7–16% by 2100 for RCP8.5with ''medium confidence'' , as there are improved constraints from historical variability in this region. Globally, the increased stratification in the future is ''very likely'' to reduce export production by 9–16% in response to reduced nutrient supply, especially in tropical regions ( ''medium confidence'' ).

<div id="section-5-2-2-6changing-ocean-primary-and-export-production-block-2"></div>

<span id="figure-5.12"></span>
<!-- START IMG -->
<!-- IMG TITLE -->
'''Figure 5.12'''

<span id="figure-5.12-a-schematic-diagram-to-illustrate-how-net-primary-production-npp-is-a-combination-of-microbial-growth-and-biomass.-in-this-context-growth-is-controlled-by-three-limiting-factors-nutrients-light-and-temperature-while-biomass-is-affected-by-grazing.-the-grey-lines-in-the-plots-represent-results-from-different-coupled-model-intercomparison-project"></span>
<!-- IMG CAPTION -->
'''Figure 5.12 | A schematic diagram to illustrate how net primary production (NPP) is a combination of microbial growth and biomass. In this context, growth is controlled by three limiting factors (nutrients, light and temperature), while biomass is affected by grazing. The grey lines in the plots represent results from different Coupled Model Intercomparison Project […]'''
<!-- IMG FILE -->
[[File:a6fffbd204d9ba2ec8cc5d54b63d2bd3 IPCC-SROCC-CH_5_12-1.jpg]]

Figure 5.12 | A schematic diagram to illustrate how net primary production (NPP) is a combination of microbial growth and biomass. In this context, growth is controlled by three limiting factors (nutrients, light and temperature), while biomass is affected by grazing. The grey lines in the plots represent results from different Coupled Model Intercomparison Project Phase 5 (CMIP5) models as reported by Laufkötter et al. (2015) <sup>[[#fn:r353|353]]</sup> . Poorly understood feedbacks from upper trophic levels on autotroph biomass and nutrients are represented by dashed arrows.

<!-- END IMG -->
<div id="section-5-2-2-6changing-ocean-primary-and-export-production-block-3" class="box"></div>

<span id="box-5.1-time-of-emergence-and-exposure-to-climate-hazards"></span>