Climate and marine biogeochemistry changes over the Holocene are investigated
based on transient global climate and biogeochemistry model simulations over
the last 9500 years. The simulations are forced by accelerated and
non-accelerated orbital parameters, respectively, and atmospheric
Numerical models that combine the ocean circulation and marine
biogeochemistry have been developed since the 1980s
Of the many features that characterize the biogeochemical system in the
ocean, here we will concentrate on oxygen minimum zones (OMZs),
atmosphere–ocean carbon fluxes, and the marine ecosystem. OMZs have received
particular attention in the recent past. This is in large part due to the
observation that in the last 5 decades, a general deoxygenation of the
world's ocean and an intensification of the ocean's main OMZs have occurred
A few studies that investigate past oxygen variations have already been
performed: based on a model study with an intermediate complexity model to
investigate glacial–interglacial variations of oxygen,
Although the focus of this paper is on marine
biogeochemistry, it is mainly the changes in climate that are driving the
changes in marine biogeochemistry. Hence, some characteristics of the
Holocene climate variability need to be addressed. Model-based investigations
of Holocene climate are performed under the auspices of the Paleo Model
Intercomparison Project
A second source of information about climate variability during the Holocene
comes from proxy data. A concerted effort to synthesize these estimates by
the PAGES2K project has resulted in a temperature reconstruction over the
last 2000 years at a fairly high temporal resolution
A continuous reconstruction of temperatures for the entire Holocene, i.e. the
past 11 300 years, albeit with lower temporal resolution before the PAGES2K
period, has been assembled by
Model simulations and proxy-based estimates of past climate variability
apparently show some disagreement
In this paper we aim at closing the gap between glacial–interglacial and
future greenhouse gas (GHG) driven simulations of climate and the marine
carbon cycle and earlier time-slice experiments of the Holocene. Given the
differences in simulated and proxy-derived climate evolution over the
Holocene, this study should be regarded as a sensitivity study to orbital and
GHG forcing. Following earlier time-slice experiments with a coupled
atmosphere–ocean–sea-ice climate model and a marine biogeochemistry model
In addition we want to address the more technical question to what extent
simulations with accelerated orbital forcing are suitable for Holocene marine
biogeochemistry simulations. In the accelerated-forcing experiments, the
change in orbital parameters between 2 model years corresponds to a 10-year
step in the real orbital forcing (see Sect.
We will first describe the numerical models, the experiment set-up, and
characteristics of the time-varying forcing in Sect.
Oceanic physical conditions are obtained from the KCM
KCM has previously been used to conduct and analyse time-slice simulations of
the pre-industrial and mid-Holocene climate and hydrological cycle
Monthly mean fields of temperature, salinity, and the velocity from the KCM
experiment were used in offline mode to force a global model of the marine
biogeochemistry
Since the description of PISCES in
Primary production is simulated by two phytoplankton groups representing nanophytoplankton and diatoms. Growth rates are based on temperature, the availability of light, the nutrients P and N (both as nitrate and ammonium), Si (for diatoms), and the micronutrient Fe. The elemental ratios of iron, chlorophyll, and silicate within diatoms are computed prognostically based on the surrounding water's concentration of nutrients. Otherwise elemental ratios are constant following the Redfield ratios. Photosynthetically available radiation (PAR) is computed from the short-wave radiation passed from ECHAM to NEMO. Sea ice is assumed to reflect all incoming radiation, so there is no biological production in areas that are completely sea-ice-covered (i.e. where the sea-ice fraction is equal to 1).
There are three non-living components of organic carbon in PISCES:
semi-labile DOC, as well as large and small POC, which are fuelled by
mortality, aggregation, fecal pellet production, and grazing. In the standard
version of PISCES, large and small POC sinks to the sea floor at respective
settling velocities of 2 and 50 m d
We also added an age tracer to PISCES. The age tracer is set to zero at surface grid points, and then the age increases with model time elsewhere. Advection and mixing are also applied to the age tracer.
As GHG and orbital forcing are the boundary conditions driving the forced variations in the KCM experiments, we describe this forcing in a little more detail. We do not take into account changes in total solar irradiance (TSI), sea level, changes in ice sheets (neither topography nor albedo), freshwater input into the North Atlantic, or volcanic aerosols.
Greenhouse gas concentrations were obtained from the PMIP database
(
Eccentricity remained fairly constant at a value of 0.02 over the entire
Holocene. The precessional index increased from
We note that the total annual radiation driven by precession changes remains
fairly constant at each latitude and globally, whereas obliquity changes
cause changes also in the annual mean insolation
Forcing for the KCM-HOL and BGC-HOL experiments:
For our analyses that focus on ocean physical conditions and marine
biogeochemistry, however, we need to consider the TOA forcing as filtered by
the atmosphere, i.e. at the sea surface. In Fig.
The basis for the KCM experiments is a 1000-year KCM experiment with
9.5 kyr BP orbital parameters, 286.6 ppm
Experiment names and characteristics. See also
Fig.
n/a: not applicable
To spin up the biogeochemical model, monthly mean ocean model output from experiment KCM-CTL was used as forcing. This then available 2000-year long forcing (first 2000 years of KCM-CTL) was repeated three times to spin up PISCES for 6000 years, after which period the model drift as defined by air–sea carbon flux and age of water masses was negligible. It was in particular the age tracer in the deep North Pacific that required the long spin-up time. Note that this BGC spin-up simulation does not achieve a “classical” time-invariant steady state but reflects the internal variability of the first 2000 years from experiment KCM-CTL and any remaining drift. After repeating the KCM-CTL forcing three times for the spin-up, PISCES was integrated for a further 8700 years with the available KCM-CTL forcing as a control experiment for the marine biogeochemistry (BGC-CTL).
Similarly to the set-up of the Holocene KCM experiments, we performed two
transient experiments with PISCES in offline mode. Both transient experiments
are also started from year 6000 of the PISCES spin-up experiment. In the
accelerated experiment BGC-HOLx10, oceanic fields of KCM-HOLx10 and
the same atmospheric
Note that the approach here differs from earlier work to investigate Holocene
OMZ changes with a KCM/PISCES model set-up, where PISCES was forced by
PMIP-protocol time-averaged oceanic conditions for specific time slices
All plots in the results section are based on model output interpolated to a
regular 1
For all time series the time axis represents the forcing years. This
corresponds to model years for the non-accelerated experiments but not for
the accelerated experiments, so any variation caused by long-term internal
variability of the model would be spread out in time in the accelerated
experiment compared to the non-accelerated experiment. For all time series,
the long-term changes are indicated by the fourth-order polynomial fits from
the Grace software package
(
Since the biogeochemical variations depend to a large extent on the changes in ocean physics, we will first examine the relevant aspects of the simulated climate variations over the Holocene.
As a first indicator of simulated changes in ocean physics, we present time
series of the global and annual mean SST (Fig.
Also, the KCM-CTL control experiment displays a decrease in global mean SST
of about 0.1
The SST evolution in KCM-CTL implies that the simulated early Holocene
decrease in SST in KCM-HOL and KCM-HOLx10 is the combined result
of a remaining model drift, and the orbital and
A Hovmöller diagram of zonal mean SST anomalies of KCM-HOL
(Fig.
As in Fig.
The seasonal cycle of global mean SST in KCM-HOL doubles its amplitude from
around 0.35
The Atlantic meridional overturning circulation (AMOC) serves as an indicator
of the intensity of deep water formation in the source region of the global
conveyor belt. From the NEMO-package output, maximum AMOC at 30
In KCM-HOLx10 the mean AMOC and its temporal evolution are similar to KCM-HOL, with a slightly higher mean value. The KCM-CTL control experiment, however, also displays changes in AMOC, similar to the changes in KCM-HOL. Overall, the long-term changes in AMOC are relatively small in all experiments and remain within the range of interannual to centennial variations of around 2–3 Sv.
In the Pacific, the deep northward flow that forms the far end of the deep
branch of the conveyor belt circulation also weakens with time during the
Holocene. Between 3000 and 5000 m depth, at latitude 0
As Fig.
In addition to AMOC, the age of water masses can serve as an indicator of deep water formation and the intensity of the global deep water circulation, and help to understand changes in oxygen concentration. We will investigate time series of the water mass age in the deep ocean at the source and end regions of the global conveyor belt circulation, namely the North Atlantic and the North Pacific.
The renewal of water masses in the North Atlantic is indicated by a time
series of the age tracer averaged between 1800 and 2500 m depth and 40 to
10
Also, the BGC-CTL control experiment, however, simulates a sudden decrease in
water mass age similar to the one in BGC-HOL but occurring at a different
time. In the BGC-HOLx10 (brown curve in Fig.
As Fig.
At the far end of the conveyor belt circulation, the deep North Pacific,
changes occur less suddenly than in the North Atlantic, but with a larger
amplitude. Between 2500 and 3500 m depth, 150
This cannot be simulated in the accelerated experiment BGC-HOLx10,
however, which runs for 950 years only. In BGC-HOLx10 deep North
Pacific water mass age decreases to 1400 years at 0 kyr BP. The increase in water mass age in the non-accelerated
experiment BGC-HOL indicates a considerable slowdown of the global conveyor
belt circulation over the Holocene, with significant impact on the marine
biogeochemistry in the Pacific. We will come back to the age of water masses
when investigating the evolution of the EEP OMZ in Sect.
In this section the atmosphere–ocean carbon flux is diagnosed. As
atmospheric
As Fig.
In the early Holocene the atmosphere–ocean carbon flux in BGC-HOL is around
The time-integrated atmosphere–ocean carbon flux in BGC-HOL (blue curve in
Fig.
The zonal mean changes in the atmosphere–ocean carbon flux in BGC-HOL
(Fig.
As Fig.
In BGC-HOL total alkalinity (TA) at the sea surface increases from 2240 to
2250
The increase in TA in BGC-HOL occurs over most latitudes
(Fig.
The global and annual mean pH at the surface follows the temporal variations
in atmospheric
As Fig.
In BGC-HOL, the global mean
The Hovmöller diagram of the zonal mean
As Fig.
As Fig.
Here the focus is on the three major components of the marine ecosystem with
relevance for the carbon cycle, namely the integrated primary production, the
export production, and the calcite (calcium carbonate) export. The primary
production integrated over the euphotic zone (INTPP) is a measure of the
productivity of the marine ecosystem. INTPP in BGC-HOL is around
44 GtC yr
The decrease in global mean INTPP in BGC-HOL originates mainly from latitudes
south of 40
The export production at 100 m depth in BGC-HOL, here computed as the sum of
small and large POC (see Sect.
The zonal mean export production in BGC-HOL decreases mainly at low
latitudes in two bands centred around 20
Time series of EEP OMZ volume for a threshold of
30
The temporal evolution of the calcite export in all experiments is similar to
that of INTPP: In BGC-HOL, calcite export is around 1.08 GtC yr
The zonal mean changes in
Overall the variations of the global marine biological production and export rates remain in the range of about 10 % throughout the Holocene even in the non-accelerated experiment BGC-HOL, with a tendency towards lower values in the mid-Holocene, and variations surprisingly similar in magnitude in the control run.
The largest OMZ in the global ocean resides in the EEP. The EEP here is
defined as the region from 140–74
At the same time as the OMZ volume increases in BGC-HOL, the age of the water
mass within the OMZ increases from around 440 years (9.5–7 kyr BP) to
530 years at 0 kyr BP (Fig.
As Fig.
Time series of the oxygen saturation (
As Fig.
For AOU, there is a more uniform-with-depth tendency to higher values in the
late Holocene, with AOU up to 25
Export production in the EEP is fairly constant over the Holocene at
0.58 GtC a
In contrast to the EEP, for the OMZ in the tropical Atlantic mainly south of
the Equator, the changes over the Holocene are more modest and of opposite
sign. In the region from 5
For the Arabian Sea a steady increase in OMZ volume (here computed for a
70
Comparing the KCM-simulated temporal evolution of global mean SST with
observation-based estimates and other model simulations, there is a notable
difference between models and observations. During the proxy-derived climate
optimum in the mid-Holocene (8 kyr to 5 kyr BP) observation-based global
mean temperature is about 0.4
The largest fraction of the initial post-glacial temperature increase in the
reconstructions of
In support to our model results, and raising the general question of how
representative the mainly land based proxy-derived temperature anomalies can
be transferred to the Holocene SST,
From earlier experiments with KCM with/without a
We note that also in the simulations of
As such, the Holocene simulations of
Possibly there might be a difference in the behaviour of the SST and the
mainly land-based temperature reconstructions. For example,
The model–data mismatch could also imply that at least early Holocene
temperature variations were determined not only by orbital forcing or
greenhouse gases, but also by solar and volcanic forcing, ice sheets, and
internal variability of the system
As proxies might be seasonally biased
The meridional overturning in the Atlantic in KCM-HOL decreases over the
Holocene by about 1.5 Sv/10 %, whereas the deep inflow into the Pacific
decreases by 2.5 Sv/20 % (Sect.
Our original intention in examining the North Atlantic more closely was to
investigate whether the changes in the OMZs could be traced back to the deep
water source regions. It turned out, however, that significant changes
occurred in the North Atlantic that justify further analysis. In
Sect.
An SST time series at 53
This shift is also visible in the concentration of
As also the control simulation shows a sudden shift in deep North Atlantic water mass age, this indicates that small variations in the range of the internal (model) variability are sufficient to trigger shifts in the pattern of North Atlantic deep mixing.
The dominant mechanisms for past and future OMZ variability have yet to be
established
For example,
On glacial to interglacial timescales, on/off changes in the AMOC have been
identified as driving OMZ variations also in the Pacific
Thus, the long-term changes show a decoupling of oxygen saturation and AOU,
the former driven by temperature changes, the latter by a slower circulation
(increase in water mass age). We note that this differs from the compensation
of
In summary, the dominating mechanism for OMZ changes in our transient,
non-accelerated Holocene experiment is a slowdown of the circulation that
develops over thousands of years. This slowdown is not mainly from changes in
AMOC, but is more confined to the deep Pacific. It results in widespread
oxygen consumption and an increase in AOU with an effect on the EEP OMZ
(Fig.
In contrast to studies based on global warming and
A modest deoxygenation of the global ocean has been observed over the last
50 years
Comparing our results with the proxy-derived estimates of OMZ intensity in
the eastern tropical South Pacific
An earlier version of the accelerated experiment BGC-HOLx10 was
compared to sediment core based estimates of Holocene OMZ evolution in the
Arabian Sea. Based on
Concerning the observed changes in atmospheric
Here we can also use our model results to some limited extent to investigate
for our model system if and how the ocean may have contributed to the
observed atmospheric
During the early Holocene the integrated carbon flux is constant
(Fig.
In the mid-Holocene, from 7 to 4 kyr BP, the time-integrated carbon flux is
slowly increasing to a total ocean uptake of 10 GtC. This occurs during a
period of atmospheric
In the late Holocene, after 4 kyr BP, the ocean is outgassing a total of
50 GtC, potentially driven by the simulated increase in SST and damped by
the increasing prescribed atmospheric
The general slowdown of the circulation over the Holocene suggests a
weakening of the physical pump, and a strengthened biological pump as more
dissolved inorganic carbon (DIC) from remineralization of organic matter is
stored in the deep ocean. Finally, as a result of reduced calcite export,
simulated global mean surface alkalinity increases mainly during the
mid-Holocene (6.5 kyr to 5 kyr BP), which would lead to decreasing
To investigate whether the efficiency of the biological pump
In summary, the contribution of oceanic processes to air–sea carbon fluxes
is in line with the prescribed
Finally we discuss the gain from the transient experiments performed here
compared to the earlier time-slice climate model experiments at 9.5, 6, and
0 kyr BP with KCM
One obvious gain from the transient experiments is the more complete time
coverage over the Holocene, potentially allowing better comparison with
proxies and providing more continuous information. The transient experiments
can also be used to determine whether the timing of the time-slice
experiments is appropriate. For the physical fields like global mean SST
(Fig.
Also, the time-slice experiments including a 6 kyr BP simulation do not
capture the extrema of the simulated time series of the transient experiments
BGC-HOL and BGC-HOLx10. For example, the integrated primary
production and export of detritus have their lowest values at 5 kyr to
4 kyr BP (Figs.
So, are non-accelerated experiments generally required, or can the same
knowledge be obtained from experiments with accelerated astronomical forcing?
This has already been investigated for physical models
In this study, a 9500-year simulation of Holocene climate and marine biogeochemistry is analysed together with a 10-fold accelerated simulation and – to our knowledge for the first time – a control run of similar length to the non-accelerated experiment. The simulated climate in terms of global mean SST is characterized by a mid-Holocene cooling, and a late Holocene warming following the temporal evolution of the greenhouse gas forcing and the short-wave radiation at the sea surface. This is in contradiction to a proxy-derived mid-Holocene climate optimum and a late-Holocene cooling. The open question why KCM and other ocean–atmosphere coupled climate models do not simulate the proxy-derived Holocene climate evolution under astronomical and GHG forcing remains a major issue. As long as this issue is not resolved, we have to regard the biogeochemistry simulation as a sensitivity study to orbital and greenhouse gas forcing of the climate system.
Most of the characteristic variables of the marine carbon cycle, like global
atmosphere–ocean
Data are available from the corresponding author by
request. A subset of the data used for plotting and a set of figures with the
completed control run (when available) will be made available at:
Figure
Simulated and observation-based profiles of average
As the threshold for the definition of an OMZ is not very well defined in the
literature, we display the OMZ volume for a range of threshold values for
observations and as simulated in Fig.
Simulated and observation-based
Time series of
Figure
For the atmosphere–ocean carbon flux, the seasonal cycle is almost
2 GtC yr
The authors declare that they have no conflict of interest.
This article is part of the special issue “Progress in quantifying ocean biogeochemistry – in honour of Ernst Maier-Reimer”. It is not associated with a conference.
Joachim Segschneider would like to thank the dearly missed Ernst Maier-Reimer for his countless advice and help, Ernst's perpetual willingness to answer his questions and to discuss scientific issues even beyond office hours – at which time, however, discussions were preferably held not in the office but in more enjoyable surroundings, and strictly had to change subject after the third beer – with a bit of luck to his less known months-long journeys from Germany to India and the Saharan desert by car. This work would not have been possible without him.
The authors acknowledge support by the German Research Foundation through the Collaborative Research Centre Climate-Biogeochemistry Interactions in the Tropical Ocean (SFB754) and the DFG project “Climate impact on marine plankton dynamics during interglacials” (grant DFG SCH 762/3-1) and the Excellence Cluster Future Ocean (grant FO EXC 80/1). We also wish to thank the NEMO/PISCES team for providing their models and general support. Computations were carried out on a NEC-SX-ACE at the computing centre of the Christian-Albrechts-University Kiel, Germany. Finally, the authors wish to acknowledge the use of the Ferret programme for analysis and graphics in this paper. Ferret is a product of NOAA's Pacific Marine Environmental Laboratory. Edited by: Christoph Heinze Reviewed by: two anonymous referees