OSOcean ScienceOSOcean Sci.1812-0792Copernicus PublicationsGöttingen, Germany10.5194/os-13-47-2017Changes in extreme regional sea level under global warmingBrunnabendS.-E.https://orcid.org/0000-0001-7837-5820DijkstraH. A.KliphuisM. A.BalH. E.SeinstraF.van WerkhovenB.https://orcid.org/0000-0002-7508-3272MaassenJ.van MeersbergenM.Institute of Marine and Atmospheric Research Utrecht, Utrecht University, Princetonplein 5, 3584 CC Utrecht, the NetherlandsLeibniz Institute for Baltic Sea Research Warnemünde, Seestrasse 15, 18119 Rostock, GermanyDepartment of Computer Science, VU University Amsterdam, 1081 HV Amsterdam, the NetherlandsNetherlands eScience Center, 1098 XG Amsterdam, the NetherlandsSandra-Esther Brunnabend (sandra.brunnabend@io-warnemuende.de)20January2017131476018July20162August20167December201617December2016This work is licensed under a Creative Commons Attribution 3.0 Unported License. To view a copy of this license, visit http://creativecommons.org/licenses/by/3.0/This article is available from https://os.copernicus.org/articles/13/47/2017/os-13-47-2017.htmlThe full text article is available as a PDF file from https://os.copernicus.org/articles/13/47/2017/os-13-47-2017.pdf
An important contribution to future changes in regional sea level extremes is
due to the changes in intrinsic ocean variability, in particular ocean
eddies. Here, we study a scenario of future dynamic sea level (DSL) extremes
using a high-resolution version of the Parallel Ocean Program and generalized
extreme value theory. This model is forced with atmospheric fluxes from a
coupled climate model which has been integrated under the IPCC-SRES-A1B
scenario over the period 2000–2100. Changes in 10-year return time DSL
extremes are very inhomogeneous over the globe and are related to changes in
ocean currents and corresponding regional shifts in ocean eddy pathways. In
this scenario, several regions in the North Atlantic experience an increase
in mean DSL of up to 0.4 m over the period 2000–2100. DSL extremes with a
10-year return time increase up to 0.2 m with largest values in the northern
and eastern Atlantic.
Introduction
From satellite measurements, it has been well established that global mean
sea level has increased by about 3 mm yr-1 over the period 1993–2010
. However, regional sea level
trends are very inhomogeneous over the oceans and range from about a
10 mm yr-1 increase in the western tropical Pacific to about
5 mm yr-1 decrease in the subtropical eastern Pacific
. Regional deviations from global mean sea level rise occur
due to ocean warming, global isostatic adjustment, land–ice mass loss and
changes in the ocean circulation. The dynamic sea level (DSL) component is
the sum of the contributions from local steric (thermal and saline) effects
and ocean mass redistribution.
Until 2100, global mean sea level is projected to rise up to roughly 1 m
depending on the climate change scenario considered
. For example, under the SRES-A1B scenario, the
global mean sea level is likely to rise between 0.42 and 0.80 m (compared to
1986–2005), with major contributions provided by thermal expansion of ocean
water and the mass loss of the major ice sheets and glaciers
. For the highest radiative forcing scenario
(RCP8.5), projected global sea level rise is between 0.52 and 0.98 m in 2100
. Regional sea level changes projected for the North
Atlantic show complex patterns that are partly caused by a weakening of the
Atlantic Meridional Overturning Circulation (AMOC), by a shift in the path of
the North Atlantic Current, and by changes in surface buoyancy fluxes
. Thermosteric
sea level evolves with a pattern that reflects the reduced heat transport to
the North Atlantic due to changes in ocean currents .
For example, DSL is rising near the North American continent because of a
reduction in the AMOC causing a redistribution of ocean mass
.
The spread in the projections of regional sea level change is largely
determined by internal ocean variability and model uncertainty. In
and , the spread due to
decadal-to-centennial variability is considered by looking at ensemble
simulations using CMIP3 and CMIP5 climate models, respectively. It was shown
that the (CMIP5) ensemble spread of the projected DSL is of the same order of
magnitude as the globally averaged sea level rise . Several
regions were identified where the forced sea level change signal is
relatively strong with respect to the internal variability, e.g., the
Indo-Pacific part of the Southern Ocean and the eastern equatorial Pacific,
and hence may be detected earlier .
However, in all these model studies the strongest component of oceanic
internal variability, i.e., that due to ocean meso-scale eddies, was not
represented. Rectification processes due to eddies can lead to strong changes
in mean ocean surface flows and their response to atmospheric forcing, in
particular in the Southern Ocean . In strongly eddying
ocean models even new modes of low-frequency variability may appear, such as
the multidecadal Southern Ocean Mode . Using the
eddy-permitting (about 1/4∘ horizontal resolution) version of the
MIROC3.2 model, showed that representing ocean eddies
provides a more detailed projection of regional sea level changes under the
IPCC SRES-A1B scenario and that eddies are strongly involved in regional sea
level extremes. In addition, as demonstrated by from
observational data, a high background sea level superposed on the sea level
change due to an arriving ocean eddy can lead to extreme local sea levels.
Eddies can also have a strong effect on the sensitivity of the AMOC to
freshwater forcing. The study of indicates that the AMOC in
the strongly eddying (about 0.1∘ horizontal resolution) version of
the Parallel Ocean Program (POP) model version is more sensitive to freshwater perturbations than the
non-eddying version of the same model. Climate model studies on the
projections of the AMOC with non-eddying ocean components show only an AMOC
decline of 22 to 40 % over the period 2000–2100, depending on the IPCC
scenario . Only 2 (out of 30) of these models project a
substantial decrease of the AMOC under the RCP8.5 scenario until year 2100,
and no model shows an abrupt transition after the 21st century
. However, in high-resolution ocean models strong variations
in the AMOC strength lead to changes in ocean currents and eddy pathways,
which induce an additional contribution to the variability in DSL and hence
affect extreme DSL values .
It is important to assess the role of eddies in projections of future
regional sea level changes, in particular on the DSL extremes. In this paper,
we study a scenario of future DSL change using the high-resolution version of
POP as in , but now forced with atmospheric fields from a
coupled climate model that evolved under the SRES-A1B scenario. We focus on
the changes in the probability density function of regional (and more local)
DSL values and 10-year return extreme values over the period 2000–2100,
computed using the generalized extreme value theory , and
compare these results to those obtained from a similarly forced non-eddying
version of POP.
Ocean model
The high-resolution version of the POP
(http://www.cesm.ucar.edu/models/ccsm4.0/pop/) used has a spatial
resolution of 0.1∘ horizontally and 42 depth levels of which the
thickness varies from 10 m near the surface to 250 m near the ocean bottom
. The high spatial resolution captures the processes leading
to meso-scale ocean eddies and provides a more detailed representation of the
western boundary currents. Specific details about the high-resolution model
setup, such as the treatment of the bottom topography, sea ice and river
runoff, are described in . The high-resolution model was
optimized for use on the Cartesius supercomputer in Amsterdam
(www.surfsara.nl) and about 3 model years are simulated per 24 h using
about 1000 cores.
The POP simulation was initialized from a 75-year spin-up simulation
under the CORE-I climatology dataset as
atmospheric forcing. This initial condition is indicated here as the year
1950. Under a freshwater flux which is diagnosed from the last 5 years of
this spin-up, the model displays only a very small drift over a 200-year
control simulation . Here, the model was forced with
monthly mean atmospheric forcing fluxes over the period 1950–2100, which were
derived from simulations with the ECHAM5-OM1 model within the ESSENCE
project (see www.knmi.nl/~sterl/Essence/). The used
forcing fields are 10 m wind speed, downward flux of short-wave and longwave
radiation, 2 m temperature, humidity, precipitation, runoff, and the surface
wind stress field. The atmospheric forcing fields are given on a global
1∘×1∘ grid and are interpolated to the curvilinear
POP model grid. The outgoing heat and freshwater fluxes are computed within
the model using bulk formulae. There is an initial adjustment after the
switch in forcing in 1950, for example measured by the change in the AMOC
strength, which lasts for about a decade.
Over the years 1950–2000, the POP model was forced by the ensemble mean
atmospheric fields from the ESSENCE project that take observed concentrations
of greenhouse gases and anthropogenic aerosols into account. Over the years
2001–2100, POP was forced with atmospheric forcing fields obtained from the
ECHAM5-OM1 model according to the SRES A1B scenario (from an arbitrarily
chosen ensemble member of the ESSENCE project; ). We focus
on ensemble member no. 021 for which the high-resolution simulation is
denoted by R021. Two
additional simulations are performed using the forcing from the ESSENCE
ensemble members nos. 029 and 033 to address the robustness of the change of
extreme DSL values.
In addition to this high-resolution simulation, a similarly forced simulation
is performed with a low-resolution POP version, indicated in the following by
R021low. This non-eddying version has an average 1.0∘
horizontal resolution and 40 vertical levels . The
scheme is used to represent eddy-driven tracer transports.
Such a scheme is not needed in the strongly eddying version as these tracer
transports are explicitly resolved.
The POP model directly computes the DSL, which can be decomposed into a mass
redistribution term and a steric contribution. Because the freshwater flux is
included into the model as a virtual salt flux and the global mean of
precipitation, evaporation and river runoff is zero, no mass-induced global
mean sea level changes can be represented. Due to the applied Boussinesq
approximation, global mean steric sea level variations are not accounted
for explicitly during this study, but this spatially independent contribution was
computed from the model output .
To demonstrate the performance of both versions of the POP model, we compare
the DSL over the years 1993–2012 (computed from monthly means) with
observations derived from altimetry. The altimeter products were produced by
Ssalto/Duacs and distributed by Aviso, with support from CNES
(http://www.aviso.altimetry.fr/duacs/). No salinity or heat restoring
is applied as even a weak salinity restoring artificially constrains the
AMOC. The model is also configured with no weak restoring of the global sea
surface salinity field. However, as the POP model does not include a
thermodynamic–dynamic sea-ice component, a prescribed climatological flux of
heat and salt is included in sea-ice regions. These fluxes are the same for
the control and hosing simulations and are an order of magnitude smaller than
the mean fluxes. The sea-ice regions are indicated by white areas and are not
considered in the analyses below. The mean DSL of simulation R021 over
the years 1993–2012 agrees well with observations, both for the mean
(compare Fig. a and c) and the standard deviation (compare
Fig. b and d). This shows that the model adequately determines
the mean ocean circulation, including the western boundary currents and also
represents the eddy-induced variability. Differences with respect to
observations appear due to the general overestimation of the modeled
variability during this time period, which may be due to the prescribed low
resolution of the atmospheric forcing and the lack of feedback of the
atmosphere on the ocean variability. Differences in variability may also
occur due to the higher horizontal resolution of the ocean model
(0.1∘) than the altimetry dataset used (0.25∘) as more
small-scale features can be resolved. Regional differences in variability in
the South Atlantic are caused by a too regular Agulhas ring formation rate in
POP compared to observations .
In contrast, results for the low-resolution simulation R021low
shown in Fig. e (mean) and Fig. f (standard
deviation) indicate that only the mean DSL change is reasonably well
captured. The variability captured in the model is mainly related to the
seasonal cycle and internal variability is weak, in particular in the regions
of the western boundary currents (similar to many other non-eddying ocean
model results; ). This weak variability also has
consequences for the mean flow through the lack of representation of
rectification processes causing, for example, too small DSL values in the
Agulhas and the Gulf Stream regions.
Mean sea surface height (SSH in meters) (a–c) and its
standard deviation (b–d) over the years 1993–2012. (a)
and (b) are derived from altimetry and (c) and
(d) of the high-resolution simulation R021. Panels (e)
and (f) show the mean SSH and the standard deviation for the
low-resolution simulation R021low, respectively.
Future dynamic sea level changes
In the results below, all long-term changes are computed by taking the
difference between values over the last 20 years (2081–2100) and the first
20 years (2001–2020) of the model simulations. Monthly mean data are used
for the analysis of changes in the mean and standard deviation (Sect. 3.1),
while daily data are used in the extreme value analyses (Sect. 3.2).
Mean and standard deviation
In the POP simulation R021, global mean steric height increases by about
2.2 mm yr-1 from year 2000 to 2100. As this signal is homogeneous over
the Earth, it is not considered in the results below. Largest changes in mean
DSL between the periods 2081–2100 and 2001–2020 occur in the North Atlantic
(Fig. a), in particular near the western part of this basin
(Fig. c). There is a mean DSL decrease in the Atlantic and
Pacific parts of the Southern Ocean, while mean DSL increases in the Indian
part of the Southern Ocean. The mean DSL increases in the eastern part of the
North Atlantic basin and decreases in the center of the subpolar gyre. Large
changes in DSL variability occur in the Agulhas retroflection region and near
Drake Passage (Fig. b). The DSL variability decreases in the
western North Atlantic, in the center of the subpolar gyre and slightly along
the western boundary of the North Atlantic while it substantially increases
in the eastern Atlantic (Fig. d). However, the separation of DSL
change in the North Atlantic into steric height change and change in regional
ocean mass show that the change is mainly caused by regional steric height
changes (Fig. a, b). These regional steric height changes
and the positive mass redistribution that increases DSL near the North
American coast (Fig. c, d) correspond well with the pattern
found by the study of .
Change in the (a) mean and (b) standard deviation
of modeled DSL in meters between the periods 2081–2100 and 2001–2020 for the
R021 simulation. The panels (c) and (d) are
magnifications of (a) and (b) for the North Atlantic
region. (e–h) Same as (a–d) but for the
R021low simulation.
In the POP simulation R021low, global mean steric height varies
only by a few centimeters over the period 2000 to 2100 and again is not considered
further. Regional steric height changes and the redistribution of ocean mass
towards the North American coast are also visible in the low-resolution
results. In addition, the small dipole pattern visible in the North Atlantic
is caused by the reduced strength and the shift in ocean currents, which are
discussed later in this section. Regarding mean DSL patterns and amplitudes,
the results of the low-resolution simulation (R021low), as shown
in Fig. e, agree well with many other model studies using
non-eddying ocean models . At first
sight, the results also look similar to those for the R021 simulation
(compare Fig. a and e). However, when regional details are
considered, the results are different. The Southern Ocean basin contrast
(Indian versus Atlantic/Pacific) is much stronger in the R021 results.
The DSL change in the Northern Atlantic is more dipolar in the North Atlantic
than in the R021low results, with a large area south of
Greenland with decreasing mean DSL. The change in DSL variability is, as
expected, different in both models (compare Fig. b and f), in
particular in western boundary current regions. In the North Atlantic,
(compare Fig. d and h), the changes in variability are less
coherent in the Gulf Stream region and have larger amplitudes in the eastern
part of the basin.
Change in (a) modeled mean steric height in meters, and
change in (b) modeled mean ocean bottom pressure change in meters of
equivalent water height between the periods 2081–2100 and
2001–2020 for the R021 simulation. The panels (c)
and (d) are the same as (a) and (b) but for the
R021low simulation.
(a, d) Maximum AMOC strength at 35∘ S (blue) and
26∘ N (red) over the period 2000–2100 of (a–c)R021
and (d–f)R021low; (b, e) AMOC streamfunction (mean of years 2001–2020); (c, f) same
as (b) and (e) but over the period 2081–2100.
To explain the changes in DSL in the North Atlantic for the R021
simulation (Fig. c–d), the behavior of the AMOC is shown in
Fig. . The maximum AMOC at 26∘ N decreases from about
20 Sv to about 5 Sv (red curve in Fig. a). The spatial pattern
of the AMOC does not change, but the North Atlantic Deep Water shallows by
about 1000 m (Fig. b–c). The maximum strength of the AMOC at
35∘ S decreases (blue curve in Fig. a) by more than
60 %. The decline in the AMOC causes a rise in mean DSL of up to 0.4 m
near the North American continent, mostly because of a redistribution of
ocean mass towards these regions (see Fig. a). The reduction of
the AMOC in the R021low simulation is only a few
sverdrups, as in the ESSENCE
ensemble . The strong variations at
26∘ N in the R021low simulation are very likely due to
an adjustment as a consequence of the change in forcing. At 26∘ N
the AMOC also measures the Gulf Stream in the model, which can intensify due
to a change in buoyancy gradient.
Difference of horizontal surface kinetic energy (energy flux per
unit area) in cm2 s-2 of the simulations (a)R021 and
(b)R021low in the North Atlantic (mean of years
2081–2100 minus mean of years 2001–2020). (c) shows the difference
in eddy kinetic energy (EKE) of the years 2090 and 2010 of R021. Before
computing EKE, the mean KE of the years 2080–2100 and 2000–2020 has been
subtracted. (d) is the same as (c) showing
only the North Atlantic.
The DSL change in the Southern Ocean between the periods 2081–2100 and
2001–2020 (Fig. a) is caused by a southward shift of the
westerly winds. In addition, the westerly wind stress strengthens by about
0.03 Pa (Fig. b). The increase in zonal momentum flux accelerates
the Antarctic Circumpolar Current and increases the northward Ekman transport
that changes the slope of the isopycnal surfaces in the South Atlantic
. These effects cause changes in the water mass properties
leading to steric contraction in the Southern Ocean and steric expansion in
the region of the Agulhas return current , explaining the
results in Fig. a.
The reduction of the AMOC is also associated with a northward shift of the
latitude separation of the Gulf Stream. This result has also been found in
the non-eddying model studies and previous strongly
eddying model studies . In addition, eastward shifts of
the path of the Gulf Stream and North Atlantic Current occur. This is shown
more clearly by the change in surface mean kinetic energy (Fig. a)
which has decreased over most of the Gulf Stream path in the R021
simulation. Figure c and d show the change of the eddy kinetic
energy (EKE) of year 2090 with respect to 2010. The changes in the mean
current path redirect eddies and lead to higher variability in the eastern
Atlantic while in the subpolar region the variability is reduced. In the
R021low simulation, similar shifts in the current system in the
North Atlantic occur (Fig. b). However, the amplitude of the
kinetic energy changes is much smaller compared to the R021 simulation,
in particular in the Labrador Sea and in the Caribbean Sea.
Change in (a) sea surface temperature (∘C),
(b) zonal wind stress (Pa), (c) surface heat flux
(W m-2), and (d) surface freshwater flux
(kg m-2 s-1) for the R021 simulation; again the mean over
the last 20 years (2081–2100) minus that over the first 20 years
(2001–2020) is shown. (positive values mean a flux into the ocean).
In the R021 simulation, the global mean sea surface temperature (SST)
rises by about 2 ∘C over the period 2000–2100 (Fig. a).
Almost all ocean regions experience a warming, and near the east coast of
North America there is a warming of up to 4 ∘C as also shown by
. However, in the Southern Ocean, SST remains almost unchanged
over large regions. This can be explained by the atmospheric forcing fields
associated with the SRES-A1B scenario as they lead to changes in the
radiative forcing between atmosphere and ocean. In addition, SST decreases by
more than 3 ∘C in the subpolar gyre region of the North Atlantic.
This cooling is related to changes in deepwater formation, as discussed by
, associated with a decrease of the AMOC strength and the
shift in the currents that reduce the heat transport to the northern polar
regions, which leads to thermal contraction and a negative DSL change
(Fig. a). The dipole pattern of SST changes and the corresponding
changes in DSL are robust fingerprints of AMOC weakening and are consistent
with most low-resolution coupled model projections (e.g.,
and others).
The reduction of the AMOC also decreases the ocean–atmosphere temperature
difference in the subpolar Atlantic region and hence leads to a reduction in
the net ocean–atmosphere surface heat flux, i.e., a reduced heat loss to the
atmosphere (Fig. c; positive values: flux into the ocean). However,
this heat gain is not strong enough to compensate for the cooling caused by
the reduced AMOC strength and the shift in current. The overall cooling in
the subpolar gyre region in the North Atlantic tends to strengthen the AMOC, but
it cannot compensate for the influence of the general warming in the upper
ocean. Furthermore, the cooling in this region leads to reduced evaporation
resulting in a further freshening of the upper ocean (Fig. d) in a
region where the AMOC is particularly sensitive to freshwater anomalies
. The reduced heat loss and the additional freshening
cause a further slowdown of the AMOC. The changes in surface fluxes for the
simulation R021low (not shown) are very similar as they are
derived from the same atmospheric forcing fields and are only slightly
differently affected by the ocean fields, compared to the R021
simulation. Because the mechanism of deepwater formation is very different
in the low-resolution model, the AMOC responds more mildly to changes in
surface forcing than that in the high-resolution model .
Regions in the North Atlantic (region of the subpolar gyre, near the
US east coast and near the European coast) and locations (near Lisbon,
Azores, and Bermuda) used for determining the PDFs and for the extreme value
analysis.
(a, c, e) Estimated probability density function (PDF) of
daily regional maximum DSL of simulation R021 and (b, d, f) of
the daily regional minimum DSL in the three different regions in the North
Atlantic shown in Fig. (a region of the subpolar gyre,
b near the US east coast, and c near the European coast).
In each plot, a maximum daily value over the region is identified after all
variability with frequencies lower than 550 days has been filtered out.
(g–l) Same but for the locations indicated in Fig. and
using (g, i, k) monthly maximum local DSL values and
(h, j, l) monthly minimum local DSL values derived from daily mean
time series. The green histogram is the PDF for the first 20 years
(2001–2020) and the blue histogram that for the last 20 years (2081–2100).
The green and blue lines are the GEV distribution function fitted to the
corresponding green and blue histogram, respectively.
Regional probability density function and extremes
To determine an estimate of the probability density function (PDF) of DSL we
show histograms of modeled daily-mean DSL data over two 20-year periods
(2001–2020 and 2081–2100). To remove variations on long timescales, all
signals with frequencies lower than 550 days are first filtered out of these
DSL time series. This leaves the seasonal and annual signals in the DSL time
series and hence changes on these timescales also lead to changes in the PDFs
and the DSL extremes. The PDFs are computed for three different regions in
the North Atlantic, i.e., in the region of the subpolar gyre, near the US
east coast and near the European coast (as shown in Fig. ) using
the daily-mean maximum value (over the region) in each of the regions from
the daily-mean time series. The PDFs for three specific locations near the
Azores, Bermuda, and Lisbon are also computed by using the monthly maximum
value (at that location) from the daily time series. A generalized extreme
value (GEV) distribution function has been fitted to the PDFs using the
maximum-likelihood method . It describes the behavior of the
extremes using the location, scale and shape parameter in and
is computed in the same way as in .
The changes in each PDF for the R021 simulation for the different
regions and locations are plotted in Fig. with the blue
histogram being the future PDF. The variance of DSL decreases in mid-Atlantic
region 1 (see Fig. b, d), which is seen by the shift of the PDF
to the left (Fig. a). This also leads to a reduction of the
highest DSL extremes by more than 10 cm. In region 2 (western North
Atlantic), DSL is mainly driven by mean changes due to steric effects and the
mass redistribution and hence the PDF shifts to the right
(Fig. c). In the eastern North Atlantic (region 3), the variance
of the DSL increases (see Fig. b, d) due to the changes in the
pathways of eddies causing the changes in EKE (Fig. c, d). This
leads to a rightward shift of the PDF by about 10 cm in this region
(Fig. e). The PDF of minimum DSL in region 2 and 3
(Fig. d, f) shifts left indicating an intensification of eddy
activity affecting the sea level change in these regions. In region 1
(Fig. a), the PDF of minimum DSL shifts right as the intensity of
the eddy activity decreases in this region.
(a, c, e) Estimated probability density function (PDF) of
daily regional maximum DSL of simulation R021low and
(b, d, f) of the daily regional minimum DSL in the three different
regions in the North Atlantic shown in Fig. (a region
of the subpolar gyre, b near the US east coast, and
c near the European coast). In each plot, a maximum daily value
over the region is identified after all variability with frequencies lower
than 550 days has been filtered out. (g–l) Same but for the
locations indicated in Fig. and using (g, i, k) monthly
maximum local DSL values and (h, j, l) monthly minimum local DSL
values derived from daily mean time series. The green histogram is the PDF
for the first 20 years (2001–2020) and the blue histogram that for the last
20 years (2081–2100). The green and blue lines are the GEV distribution
function fitted to the corresponding green and blue histogram, respectively.
Changes in the pathways of eddies are also important when considering local
DSL extremes. The Azores are located in a region of slightly decreased
variability (Fig. b, d) due to reduced eddy kinetic energy in
this region, shifting the PDF slightly to the left (Fig. g). Near
Bermuda the shift in the ocean currents leads to lower
probabilities of higher sea level extremes. (Fig. i). The most
interesting result, however, is shown in Fig. k for the coast
near Lisbon. Due to the shift in the Gulf Stream and North Atlantic Current
one would expect increased probabilities for high DSL values in this region.
However, because these currents are not only shifted but also reduced in
strength almost no changes in DSL extremes can be identified
(Fig. k). As the influence of ocean eddies decreases when
reaching coastal region, no clear signal in the eddy intensity change can be
identified at the three coastal locations (Fig. g–l).
The changes in the PDFs for the R021low simulation show quite a
different behavior than those in the R021 simulation for most regions
and locations. While the relative shift in the mean is comparable for both
models in the regions 1 and 2 (Fig. a, b), the amplitude is
much smaller for R021low. For region 3 (Fig. c),
the PDF has bimodal characteristics and hardly changes under climate change,
in contrast to the change in the R021 simulation (Fig. c).
The PDF change for the Azores is the opposite (Fig. d) in
both models due to the fact that the eastward shift in the Gulf Stream has no
influence on ocean eddy paths in the R021low simulation
(Fig. b). The PDFs of the other two locations
(Fig. e, f) show the same behavior as in the R021
simulation.
From the fit of parameters in generalized extreme value (GEV) distributions,
the extreme DSL values for a return time of 120 months (10 years) over the
period 2001–2020 and their changes over the different 20-year periods
(2081–2100 and 2001–2020) of the R021 simulation can be determined
(Fig. ). Over the period 2000–2020 higher extreme sea levels
occur in regions of high variability, i.e., in regions of the major current
systems such as the Gulf Stream and the Agulhas Current
(Fig. a, c). Therefore, the regional pattern of changes in
extreme sea levels for a return time of 10 years (Fig. b, d)
reflects the changes in sea level variability as shown in
Fig. b and d. Sea level extremes can increase by 50 cm near
Tasmania. Furthermore, in the northern and eastern North Atlantic, sea level
extremes with a 10-year return time will increase by up to 20 cm. A
comparison of the PDFs and the DSL extremes (for the 10-year return time)
using a 550-day filter and a 180-day filter (not shown) indicates that the
changes in DSL extremes are dominated by the change in short-term variability
caused mainly by the shift in the ocean currents changing the eddy pathways
(Fig. c, d).
Extreme DSL values in meters for a 10-year return time of simulation
R021 for (a) the first 20 years (2001–2020) and
(b) the differences between the period 2081–2100 and 2001–2020.
All signals with frequencies lower than 550 days have been filtered out. The
panels (c) and (d) are magnifications of (a)
and (b) for the North Atlantic region. (e, f) and
(g, h) are the differences between the period 2081–2100 and
2001–2020 for two additional simulations forced by ensemble members 029 and
033, respectively.
Extreme DSL values in meters for a 10-year return time of simulation
R021low for (a) the first 20 years (2001–2020) and
(b) the differences between the period 2081–2100 and 2001–2020.
All signals with frequencies lower than 550 days have been filtered out. The
panels (c) and (d) are magnifications of (a)
and (b) for the North Atlantic region.
To show that the mechanisms leading to extreme sea level change under the
SRES-A1B scenario are robust, Fig. e–h show the change of
extreme DSL values for a 10-year return time of two additional
high-resolution simulations forced by the ensemble members 029 and 033. The
similar pattern in the change of the extreme DSL values indicates similar
changes in behavior of the AMOC, ocean circulation, and DSL as in the
R021 simulation.
Changes in extreme sea level values are shown in Fig. for
the R021low simulation. The amplitude of these extremes is much
smaller, in particular in western boundary current regions
(Fig. a) and in the Gulf Stream region
(Fig. c). The low-resolution ocean model simulation leads
to different extreme sea level projections in the northern North Atlantic (in
particular, in the Labrador Sea and Barents Sea) than for the R021
simulation. The sign of the change in sea level extremes is also different in
the Caribbean Sea. This shows the importance of including an explicit
representation of eddy processes into an ocean model when looking at regional
projections of DSL.
Summary and discussion
In this paper, we considered future dynamic sea level (DSL) changes using a
strongly eddying ocean model forced by atmospheric fields according to an
SRES A1B scenario. The results show that changes in local and regional PDFs
(between the periods 2001–2020 and 2081–2100) are mainly due to changes in
DSL variability on short timescales and therefore related to changes in the
ocean eddy field. This can be deduced from both the changes in the eddy
kinetic energy of the ocean surface velocity field and from a comparison of
DSL changes in a non-eddying version of the same model. In the
high-resolution model simulation, the changes in eddy pathways are caused by
a strong decrease of the AMOC with simultaneous eastward shifts in the path
of the Gulf Stream and the North Atlantic Current.
Our main result is that the patterns of 10-year return time DSL extremes (as
shown in Fig. ) are determined by changes in the ocean eddy
field . In the POP model, eddies can come within
100 km of the coast and their maximum sea surface signal is often strongly
correlated with that at the coast. In some regions of the globe these extreme
DSL values can be up to 0.5 m, which are of the same order of magnitude as the
mean DSL change. This shows the importance of internal ocean variability for
regional extreme sea levels, not only on longer timescales
but also on shorter timescales .
These findings agree well with the study of , where it has been
shown that the influence of eddies on SSH variability is strongly reduced
near ocean boundaries but may still be several centimeters.
Low-resolution ocean–climate models are not capable of accurately
representing these changes in extreme sea levels. Some low-resolution model
studies do capture a shift in ocean currents in the case of a declining AMOC
. However, the model
resolution does not resolve DSL variability caused by ocean eddies, as the
parameterization of eddies in these models only affects the heat and salt
transport in the models. Although the use of an eddy-permitting ocean–climate
model (with a 0.25∘ horizontal resolution) already indicated the
importance of resolving ocean eddies to accurately estimate future sea level
variability , the western boundary currents usually do not
have a correct separation behavior in these models.
There are several caveats in this model study which may modify the results
quantitatively but which do not affect the main message of this paper that
strongly eddying models are important for regional future sea level change
projections. First, the AMOC in the R021 POP model simulation appears to
be quite sensitive to freshwater anomalies, and hence the scenario here may
be quite an extreme one. Second, the use of an ocean-only model with mixed
boundary conditions, restoring conditions below sea-ice regions, and
atmospheric forcing fields from a climate model restricts the capabilities of
the model in simulating the coupled ocean–atmosphere interactions occurring
in reality. However, it is expected that shifts in the ocean eddy fields
would also occur in coupled models with strongly eddying ocean model
components. Third, the model does not simulate many other processes causing
regional and coastal sea levels changes (e.g., glacial isostatic adjustment (GIA), gravity). Many of these
processes would only affect the mean DSL values and not their variability.
Hence, as a first approximation, these sea level changes can be added to the
mean DSL values determined here. Finally,
although we show robustness using a small ensemble it would be better to use
a larger ensemble of simulations to determine the effect of
ocean initial conditions and to have better statistics on the extreme DSL
values. The latter is still hardly feasible with the current computational
capabilities.
We conclude from the results that when developing plans for adapting to
future changing sea level, not only mean regional changes should be
considered, although they may be substantial. Also the changes in variability
should be accounted for, as with higher variability the probability of sea
level extremes may increase. This in particular holds for the North Atlantic
region, where many areas are vulnerable to sea level rise.
Data availability
For information about the POP model simulations and about how to get access
to the model output, please refer to
https://www.projects.science.uu.nl/oceanclimate/esalsa/overview_POP_runs_esalsa_project.pdf.
Acknowledgements
This study was supported by the Netherlands eScience Center (NLeSC) through
the eSALSA (An eScience Approach to determine future Sea-level chAnges)
project. The simulations were performed on the Cartesius supercomputer
at SURFsara (https://www.surfsara.nl) through the project SH-243-13.
This work was also partially funded by the Dutch national research program
COMMIT. The altimeter products were produced by Ssalto/Duacs and distributed
by Aviso, with support from CNES
(http://www.aviso.altimetry.fr/duacs/). We thank the two anonymous
reviewers for their constructive comments that improved the
manuscript. Edited by: M. Hecht
Reviewed by: two anonymous referees
ReferencesBöning, C. W., Dispert, A., Visbeck, M., Rintoul, S. R., and Schwarzkopf,
F. U.: The response of the Antarctic Circumpolar Current to recent
climate change, Nat. Geosci., 1, 864–869, 10.1038/ngeo362, 2008.Bordbar, M. H., Martin, T., Latif, M., and Park, W.: Effects of long-term
variability on projections of twenty-first-century dynamic sea level, Nat.
Clim. Change, 5, 343–347, 10.1038/NCLIMATE2569, 2015.Bouttes, N., Gregory, J. M., Kuhlbrodt, T., and Smith, R. S.: The drivers of
projected North Atlantic sea level change, Clim. Dynam., 43, 5, 1531–1544,
10.1007/s00382-013-1973-8, 2013.Brunnabend, S.-E., Dijkstra, H. A., Kliphuis, M. A., van Werkhoven, B., Bal,
H. E., Seinstra, F., Maassen, J., and van Meersbergen, M.: Changes in extreme
regional sea surface height due to an abrupt weakening of the Atlantic
meridional overturning circulation, Ocean Sci., 10, 881–891,
10.5194/os-10-881-2014, 2014.Church, J. A. and White, N. J.: Sea-Level Rise from the Late 19th to the
Early 21st Century, Surv. Geophys., 32, 585–602,
10.1007/s10712-011-9119-1, 2011.
Church, J. A., Clark, P. U., Cazenave, A., Gregory, J. M., Jevrejeva, S.,
Levermann, A., Merrifield, M. A., Milne, G. A., Nerem, R. S., Nunn, P. D.,
Payne, A. J., Pfeffer, W. T., Stammer, D., and Unnikrishnan, A. S.: Sea
Level Change, in: The Scientific Basis, Contribution of Working
Group I to the Fifth Assessment Report of the Intergovernmental
Panel on Climate Change, edited by: Stocker, T. F., Qin, D., Plattner,
G.-K., Tignor, M., Allen, S. K., Boschung, J., Nauels, A., Xia, Y., Bex, V.,
and Midgley, P. M., Cambridge University Press, Cambridge, New
York, 1137–1216, 2013.
Coles, S.: An Introduction to Statistical Modeling of Extreme
Values, Springer-Verlag London Ltd., ISBN: 1-85233-459-2, 2001.Danabasoglu, G., Yeager, S. G., Kwon, Y.-O., Tribbia, J. J., Phillips, A. S.,
and Hurrell, J. W.: Variability of the Atlantic Meridional
Overturning Circulation in CCSM4, J. climate, 25, 5153–5172,
10.1175/JCLI-D-11-00463.1, 2012.Drijfhout, S., Oldenborgh, G. J. V., and Cimatoribus, A.: Is a Decline of
AMOC Causing the Warming Hole above the North Atlantic in
Observed and Modeled Warming Patterns?, J. Climate, 25, 8373–8379,
10.1175/JCLI-D-12-00490.1, 2012.Drinkwater, K. F., Miles, M., Medhaug, I., Ottera, O. H., T.Kristiansen,
Sundby, S., and Gao, Y.: The Atlantic Multidecadal Oscillation: Its
manifestations and impacts with special emphasis on the Atlantic region north
of 60∘ N, J. Marine Syst., 133, 117–130,
10.1016/j.jmarsys.2013.11.001, 2014.Firing, Y. L. and Merrifield, M. A.: Extreme sea level events at Hawaii:
Influence of mesoscale eddies, Geophys. Res. Lett., 31, L24306
10.1029/2004GL021539, 2004.
Gent, P. R. and McWilliams, J. C.: Isopycnal mixing in ocean circulation
models, J. Phys. Oceanogr., 20, 150–155, 1990.
Greatbatch, R. J.: A note on the representation of steric sea level in models
that conserve volume rather than mass, J. Geophys. Res., 99, 12767–12771,
1994.Kanzow, T., Johnson, H. L., Marshall, D. P., Cunningham, S. A., Hirschi,
J.-M., Bryden, H. L., and Johns, W. E.: Besinwide Integration Volume
Transports in an Eddy-Filled Ocean, J. Phys. Oceanogr., 39, 3091–3110,
10.1175/2009JPO4185.1, 2009.Kienert, H. and Rahmstorf, S.: On the relation between Meridional Overturning
Circulation and sea-level gradients in the Atlantic, Earth Syst. Dynam., 3,
109–120, 10.5194/esd-3-109-2012, 2012.Landerer, F. W., Jungclaus, J. H., and Marotzke, J.: Regional Dynamic and
Steric Sea Level Change in Response to the IPCC-A1B Scenario,
J. Phys. Oceanogr., 37, 296–312, 10.1175/JPO3013.1, 2007.Large, W. G. and Yeager, S. G.: Diurnal to decadal global forcing for ocean
and sea-ice models; the datasets and flux climatologies, NCAR Technical Note
TN 460 STR, available at:
http://www.clivar.org/organization/wgomd/resources/core/core-i (last
access: 8 October 2008), 2004.Le Bars, D., Durgadoo, J. V., Dijkstra, H. A., Biastoch, A., and De Ruijter,
W. P. M.: An observed 20-year time series of Agulhas leakage, Ocean Sci., 10,
601–609, 10.5194/os-10-601-2014, 2014.Le Bars, D., Viebahn, J. P., and Dijkstra, H. A.: A Southern Ocean mode
of multidecadal variability, Geophys. Res. Lett., 43, 2102–2110,
10.1002/2016GL068177, 2016.Lorbacher, K., Dengg, J., Böning, C. W., and Biastoch, A.: Regional
Patterns of Sea Level Change Related to Interannual Variability
and Multidecadal Trends in the Atlantic Meridional Overturning
Circulation, J. Climate, 23, 4243–4254, 10.1175/2010JCLI3341.1, 2010.
Maltrud, M., Bryan, F., Hecht, M., Hunke, E., Ivanova, D., McClean, J., and
Peacock, S.: Global Ocean Modeling in the Eddying Regime Using
POP, CLIVAR Exchange, 44, 5–8, 2008.Pardaens, A. K., Gregory, J. M., and Lowe, J. A.: A model study of factors
influencing projected changes in regional sea level over the twenty-first
century, Clim. Dynam., 36, 2015–2033, 10.1007/s00382-009-0738-x, 2011.
Rhein, M., Rintoul, S., Aoki, S., Campos, E., Chambers, D., Feely, R. A.,
Gulev, S., Johnson, G. C., Josey, S., Kostianoy, A., Mauritzen, C., Roemmich,
D., Talley, L. D., and Wang, F.: Observations: Ocean, in: The Scientific
Basis, Contribution of Working Group I to the Fifth Assessment Report of the
Intergovernmental Panel on Climate Change, edited by: Stocker, T. F., Qin,
D., Plattner, G.-K., Tignor, M., Allen, S. K., Boschung, J., Nauels, A., Xia,
Y., Bex, V., and Midgley, P. M., Cambridge University Press, Cambridge, New
York, 255–316, 2013.Saba, V. S., Griffies, S. M., Anderson, W. G., Winton, M., Alexander, M. A.,
Delworth, T. L., Hare, J. A., Harrison, M. J., Rosati, A., Vecchi, G. A., and
Zhang, R.: Enhanced warming of the Northwest Atlantic Ocean under climate
change, J. Geophys. Res.-Oceans, 121, 118–132, 10.1002/2015JC011346,
2016.Slangen, A., Katsman, C. A., van de Wal, R. S. W., Vermeersen, L. L. A., and
Riva, R. E. M.: Towards regional projections of twenty-first century
sea-level change based on IPCC SRES scenarios, Clim. Dynam., 38,
1191–1209, 10.1007/s00382-011-1057-6, 2012.Slangen, A., Carson, M., Katsman, C. A., van de Wal, R. S. W., Koehl, A.,
Vermeersen, L. L. A., and Stammer, D.: Projecting twenty-first century
regional sea-level changes, Clim. Change, 124, 317–332,
10.1007/s10584-014-1080-9, 2014. Smith, R. S. and Gregory, J. M.: A study of the sensitivity of ocean
overturning circulation and climate to freshwater input in different regions
of the North Atlantic, Geophys. Res. Lett., 26, L15701
10.1029/2009GL038607, 2009.Sterl, A., Severijns, C., van Oldenborgh, G. J., Dijkstra, H., Hazeleger, W.,
van den Broeke, M., Burgers, G., van den Hurk, B., van Leeuwen, P., and van
Velthoven, P.: When can we expect extremely high surface temperatures?,
Geophys. Res. Lett., 35, L14703 10.1029/2008GL034071, 2008.Suzuki, T., Hasumi, H., Sakamoto, T. T., Nishimura, T., Abe-Ouchi, A.,
Segawa, T., Okada, N., Oka, A., and Emori, S.: Projection of future sea
level and its variability in a high-resolution climate model: Ocean
processes and Greenland and Antarctic ice-melt contributions, Geophys.
Res. Lett., 32, L19706 10.1029/2005GL023677, 2005.van Oldenborgh, G. J., te Raa, L. A., Dijkstra, H. A., and Philip, S. Y.:
Frequency- or amplitude-dependent effects of the Atlantic meridional
overturning on the tropical Pacific Ocean, Ocean Sci., 5, 293–301,
10.5194/os-5-293-2009, 2009.Watson, C. S., White, N. J., Church, J. A., King, M. A., and Burgette, R. J.:
Unabated global mean sea-level rise over the satellite altimeter era, Nat.
Clim. Change, 5, 565–569, 10.1038/NCLIMATE2635, 2015.Weaver, A. J., Sedlacek, J., Eby, M., Alexander, K., Crespin, E., Fichefet,
T., Philippon-Berthier, G., Joos, F., Kawamiya, M., Matsumonto, K., andK.
Tachiiri, M. S., Tokos, K., Yoshimori, M., and Zickfeld, K.: Stability of
the Atlantic meridional overturning circulation: A model intercomparison,
Geophys. Res. Lett., 39, L20709 10.1029/2012GL053763, 2013.Weijer, W., Maltrud, M. E., Hecht, M. W., Dijkstra, H. A., and Kliphuis,
M. A.: Response of the Atlantic Ocean circulation to Greenland Ice
Sheet melting in a strongly-eddying ocean model, Geophys. Res. Lett., 39,
L09606, 10.1029/2012GL051611, 2012.Yin, J., Schlesinger, M. E., and Stouffer, R. J.: Model projections of rapid
sea-level rise on the northeast coast of the United States, Nat. Geosci.,
2, 262–266, 10.1038/ngeo462, 2009.Yin, J., Griffies, S. M., and Stouffer, R. J.: Spatial Variability of
Sea Level Rise in Twenty-First Century Projections, J. Climate,
23, 4585–4607, 10.1175/2010JCLI3533.1, 2010.