Introduction
The Arabian Sea biological productivity is influenced by the strong seasonal
activity of the atmospheric circulation
. During the
boreal summer, the south-west monsoon consists of strong winds blowing from
the south-west to the north-east of the Indian Ocean. These winds result from
the rapid heating of the land mass relative to the ocean, which creates a
pressure gradient between the southern Indian Ocean high-pressure cell and
the low-pressure cell over the Tibetan Plateau. During this season, heavy
precipitation occurs over India and south-east Asia. In the Arabian Sea, the
alongshore winds off the coast of Somalia focus into a low-level jet, called
the Somali Jet, and generate a strong coastal upwelling . In addition, the wind's tendency to turn on itself in the
horizontal plane, quantified by the wind stress curl, also drives upward and
downward water transport . Between the axis of the jet
and the western coast, the wind stress is cyclonic and the wind stress curl
is positive; it drives a divergent flow that causes upward Ekman pumping
. On the other
side of the jet axis, however, the wind stress is anticyclonic. The wind
stress curl is therefore negative and drives a convergent flow and downward
Ekman pumping. Coastal upwelling and upward Ekman pumping are responsible for
increased productivity in the western coastal Arabian Sea thanks to a higher
supply of nutrients to the surface layer (;
). The upwelled nutrients are advected from the
coast to the north and the interior of the sea. Thus, productivity in the
central and northern Arabian Sea also increases during the south-west monsoon
. Wind stress and mixing of the
upper layers, as well as Ekman pumping generated by the positive wind stress
curl, also contribute to the supply of nutrients to the surface layers and
increase productivity in those regions . On the mesoscale, filaments contribute to the lateral advection of
nutrients from the coast to the central Arabian Sea .
Monsoon intensity can be characterised in different ways, depending on the
observational scale and on the studied processes. Precipitation is a major
indicator of monsoonal changes. For example, the rainfall-based index,
defined as the seasonally averaged precipitation over all the Indian
subcontinent from July to September, is used to monitor the strength of the
monsoon over India . However, indicate
that, in the context of anthropogenic climate change, an increase in rainfall
is not necessarily associated with an increase in the associated circulation
due to changes in atmospheric stability. This result questions the
reliability of such an indicator for the monsoon intensity. A second
indicator of the monsoon strength is based on the sea level pressure (SLP)
that is a large-scale fingerprint of the monsoon. The monsoon strength can be
determined by the SLP anomaly gradient between a northern region over the
Tibetan Plateau, where the Tibetan low develops during the monsoon months,
and a southern region over the southern Indian Ocean, where the Mascarene
High develops. The large-scale changes in SLP impact the local dynamics over
the Arabian Sea . Monsoon intensity can also be related to
the strength of the winds over the Arabian Sea and the associated upwelling.
The general paradigm is that a stronger summer monsoon generates stronger
upwelling that enhances productivity. Based on this paradigm, past monsoon
intensities have been reconstructed using proxies of productivity from marine
sediment cores .
Monsoon reconstructions and modelling studies have shown that insolation
variations are the major drivers of fluctuations in the summer monsoon
intensity: the monsoon is stronger when the Northern Hemisphere summer
insolation is higher (e.g. during the Holocene). Changes in the astronomical
parameters, such as the precession that is defined as the longitude of the
perihelion, or the obliquity that is defined as the angle between the Equator
and the orbital plane, modify the seasonal cycle of insolation. Along with
astronomical parameters, changes in ice sheet height between glacial and
interglacial climates also have an impact on the monsoon intensity
(; ;
).
There has been some concern about the fact that marine proxies for
productivity may be influenced by processes other than monsoon intensity,
such as changes in ice volume, especially in the Northern Hemisphere; aeolian
transport of nutrients; or the Atlantic Meridional Overturning Circulation
. Moreover, most studies linking
monsoon and productivity in the past have focused on the monsoon intensity
but the monsoon pattern, e.g wind orientation, can also change in time.
have shown that summer monsoon mean position shifted
southward during glacial periods. The monsoon pattern affects the position
and the orientation of the low-level jet over the Arabian Sea, which modifies
the upwelling of nutrients in the Arabian Sea .
Furthermore, showed opposite evolutions of the upwelling
behaviour in the western coastal Arabian Sea and the south-western tip of
India during the Holocene, which they related to a southward shift of the
monsoonal winds.
Here, we investigate the relationship between the summer monsoon intensity
and the Arabian Sea biological productivity. How do changes in the summer
monsoon pattern along with changes in its intensity impact productivity in
the Arabian Sea? Could the variations in the summer monsoon
pattern explain higher
productivity rates in some glacial climates? To answer these questions, we
test the effects of a range of astronomical parameters and different ice
sheet states on the Arabian Sea productivity.
Boundary conditions for the eight simulations studied in this work.
Precession is defined as the longitude of the perihelion, relative to the
moving vernal equinox, minus 180∘. Ice sheets are represented in
Fig. . Pmip3 ice sheet stands for the PMIP3 ice sheet
reconstruction , ICE6g-16k stands for the ICE6G
reconstruction at 16 kyr BP .
Interglacial climates
Glacial climates
Simulation name
CTRL
MH
EH
LGM
MIS3
MIS4F
MIS4M
MIS4D
Time (kyr BP)
0
6
9.5
21
46
60
66
72
Ice sheets
present
present
present
pmip3
Ice6g-16k
Ice6g-16k
Ice6g-16k
Ice6g-16k
Sea level difference vs. CTRL (m)
0
0
0
-120
-70
-70
-70
-70
Eccentricity
0.016715
0.018682
0.0193553
0.018994
0.0138427
0.018469
0.021311
0.024345
Obliquity (∘)
22.391
24.105
24.2306
22.949
24.3548
23.2329
22.493
22.391
Precession (ω-180∘)
102.7
0.87
303.032
114.42
101.337
266.65
174.82
80.09
CO2 (ppm)
284
280
284
185
205
200
195
230
N2O (ppm)
275
270
275
200
260
230
217
230
CH4 (ppb)
791
650
791
350
500
426
450
450
In Sect. 2, we describe the model we use and the experiments we performed, we
evaluate the model results for the pre-industrial period and detail the analyses we
performed. In Sect. 3, we explain the changes in productivity in the early
Holocene and then look at several glacial and interglacial climates to link
productivity changes to local dynamics and boundary conditions. In Sect. 4,
we discuss our results in the light of the summer monsoon paradigm and we
perform a simple model–data comparison and discuss the effects of seasonality
on productivity. Finally, we summarise our results and give some perspectives.
Model, experiments, evaluation and diagnostics
The model
This study uses an Earth system model (ESM) that explicitly represents the
global climate, oceanic circulation and marine productivity. We use the
IPSL-CM5A-LR model developed at the Institut Pierre Simon Laplace (IPSL)
. This ESM is composed of the LMDZ5A atmospheric general
circulation model coupled to the ORCHIDEE land-surface
model and the NEMO v3.2 ocean model ,
which includes the OPA9 ocean general circulation model, LIM-2 the sea-ice
component and the PISCES biogeochemical model
. These components are coupled once a day using the
OASIS coupler .
We use the low-resolution (LR) version of the model with a regular
atmospheric grid of 96 × 96 points horizontally, 39 vertical
levels and an irregular horizontal oceanic grid (ORCA2.0) with
182 × 149 points corresponding to a nominal resolution of
2∘, enhanced near the Equator and over the Arctic and subpolar North
Atlantic. The ocean vertical grid comprises 31 levels with intervals from
10 m for the first 150 m and up to 500 m for the bottom of the ocean.
(a) The three ice sheet covers used as boundary conditions.
The 0k ice sheet is used in the CTRL, MH and EH simulations. The pmip3 ice
sheet is used in the LGM simulation. The ICE6g-16k ice
sheet is used in the MIS3 and all the MIS4 (F, M and D)
simulations and (b) the different values of the other forcing
parameters for all the simulations.
The PISCES model simulates marine biogeochemistry and lower trophic levels.
PISCES includes two phytoplankton types (nanophytoplankton and diatoms), two
zooplankton size classes (micro- and mesozooplankton) and two detritus
compartments distinguished by their vertical sinking speed (small and large
organic matter particles), a semi-labile dissolved organic carbon pool, and
five nutrients (Fe, NO3-, NH4+, Si, and PO43-)
. In PISCES, phytoplankton growth is a function of
temperature, light, mixed-layer depth and nutrient concentrations.
Experiments
Here, we exploit eight simulations of IPSL-CM5A-LR forced by different boundary
conditions (astronomical parameters, greenhouse gas concentrations and ice
sheet cover) to account for different climates throughout the last
glacial–interglacial cycle, as detailed in Table and
Fig. .
The reference (CTRL) simulation is a pre-industrial climate with no external
forcing such as volcanoes or anthropogenic activities ,
forced by pre-industrial CMIP5 forcings and the present-day ice sheet (0k in Fig. ). A mid-Holocene (MH) simulation,
6 kyr BP , part of Paleoclimate Modeling
Intercomparison Project phase 3 (PMIP3) and an
early Holocene (EH) simulation, 9.5 kyr BP, are used to study productivity
changes in different interglacial climates. The EH simulation trace gas
concentrations are the same as for the CTRL simulation, whereas CH4 and
N2O concentrations are slightly lower for the MH simulation compared
with
the CTRL (Table ). MH and EH simulations mainly differ in their
astronomical parameters, especially in the precession value
(Table , Fig. ). Both Holocene simulations are
forced by the present-day ice sheet cover (Fig. ). Five glacial
simulations have also been performed, including the last glacial maximum (LGM,
21 kyr BP), the marine isotopic stage 3 (MIS3, 46 kyr BP) and three marine
isotopic stage 4 states: MIS4F (60 kyr BP), MIS4M (66 kyr BP) and MIS4D
(72 kyr BP). The LGM, which has also been performed for PMIP3
, has the largest ice sheet (Fig. )
, which modifies the land–sea distribution and topography
since the sea-level is reduced by about 120 m. The LGM run has the lowest
greenhouse gas concentrations of this set of eight simulations
(Table ). Two of the three MIS4 simulations (MIS4F and MIS4D) are
described in . The MIS4 ice sheets have been prescribed by
using the 16 kyr BP ice sheet, which is the period for which we have an ice
sheet reconstruction for the same sea level as during MIS4, i.e. 70 m lower
than today (Ice6g-16k in Fig. ) . This is the
most realistic scenario we could simulate given the available reconstruction at the time of
running the MIS4 experiments . However, our MIS4 runs are
different from the ones described in since we added the
nutrient inputs from dust, rivers and sediments that are essential to marine
productivity. Large changes in precession occur between the three MIS4
simulations (Table , Fig. ). The MIS3 simulation
uses the same ice sheet reconstruction as MIS4 and it has the lowest
eccentricity and highest obliquity of all eight simulations (Table ,
Fig. ).
In PISCES, three source terms contribute to the input of nutrients in the
ocean: atmospheric dust deposition, river input and sediment mobilisation.
The change in sea level in glacial climate simulations modifies the land–sea
mask; thus, in the LGM, MIS3 and all MIS4 (F, M and D) simulations, the source
terms were adjusted so that the ocean receives the same quantity of
associated nutrient supply as in the CTRL. In these simulations, no attempt was
made to account for the dustier glacial states . All
productivity changes are therefore due to other factors.
Our analyses are performed on 100 years of monthly outputs from the last
stable part of each simulation.
(a) Coastal (orange shading) and northern (black dotted
contour) Arabian Sea areas, position of core MD04-2873 (yellow circle) and
(b) seasonal cycles of current-day productivity in the coastal (bold
orange line) and northern (black dashed line) Arabian Sea. Productivity data
come from the SeaWiFS satellite data for the period 1998–2005
.
Modern evaluation of the summer mean
This study focuses on primary productivity in the Indian Ocean for the last
glacial–interglacial cycle as simulated by the IPSL-CM5A-LR coupled model.
Figure shows the seasonal cycle of observed present-day
productivity (data from SeaWiFS 1998–2014; ) for two areas
in the Arabian Sea: a coastal area in the western Arabian Sea
(Fig. a, orange area) and a northern region
(Fig. a, black box). Current-day productivity has two periods of
bloom: one in summer and one in winter. In the coastal Arabian Sea, the
summer season is the most productive period of the year (Fig. b)
and contributes the most to the bulk sediment composition. In the northern
Arabian Sea, both seasons are equally productive in the data
(Fig. b). In boreal winter, the mechanisms behind productivity
peaks are different compared with the summer maximum; the winds reverse and blow from the north-east to the
south-west. The presence of strong southward winds generates a convective
overturning that induces vertical mixing and brings nutrients to the surface.
During this period, productivity is high in the north-western Arabian Sea
(dashed line in Fig. b). Following these observations, we focus
our analyses on the boreal summer season, defined as
June–July–August–September (JJAS) to account for the whole summer monsoon,
and we will especially analyse the coastal Arabian Sea (orange area in
Fig. a).
A comparative global evaluation of the marine bio-geochemical component of
the ESM has been published in . Even if the model poorly
represents the deep-ocean circulation, especially in the Southern Ocean, it
has a quite good representation of annual wind patterns, wind stress,
mixed-layer depth and geostrophic circulation. The model is able to represent
the global ocean biological fields such as macronutrients, with correlations
higher than 0.9, and surface chlorophyll concentration, with a correlation
coefficient of 0.42 . We focus here on the representation
of the physical processes and productivity distributions in the Indian Ocean,
especially in the Arabian Sea. We use satellite products from remote sensing
from NASA's Sea-viewing Field-of-view Sensor (SeaWiFS) during the period
1998–2005 processed with the algorithm to
obtain monthly productivity , multiple-satellite
products from NOAA for the 1995–2005 climatological cycle for wind intensity and wind
stress and the ERA-Interim reanalysis (1979–2014) for sea
surface temperature (SST) . We compute the observed and
modelled wind stress curl intensity from the wind stress data and model
output, respectively. We compare the observations to the pre-industrial
(CTRL) simulation outputs.
Figure a shows that simulated boreal summer productivity
integrated over the whole water column is underestimated relative to the
reconstructed boreal summer productivity, especially in the regions of
upwelling, along the coast of the Arabian Peninsula and Somalia. The spatial
Pearson's correlation coefficient, R, between the observed and simulated
productivity is 0.44. Underestimation of productivity is first caused by an
underestimated wind intensity (Fig. b), which affects the extent
and intensity of the coastal upwelling and the supply of nutrients to the
surface layer. The boreal summer wind patterns, which are characteristic of
the boreal summer monsoon system, are better represented than productivity,
with a correlation of 0.86. studied the representation of the Asian summer
monsoon in the CMIP5 models, which comprise the IPSL model. They showed that
the monsoon was better represented in the CMIP5 models compared with the
CMIP3 models, especially the monsoonal winds. We can however note that the
alongshore winds in the western Arabian Sea have a more northerly orientation
in the CTRL simulation than in the observations, which can affect the
dynamical processes in the region (Fig. b).
Modelled and observed seasonal (JJAS) patterns of
(a) productivity (molC m-2 yr-1), (b) surface
wind intensity (m s-1), (c) surface wind stress intensity
(10-3 N m-2), (d) surface wind stress curl intensity
(10-7 N m-3) and (e) sea surface temperature (SST)
(∘C). We used SeaWiFS data in 1998–2005 for productivity
, multiple-satellite products from NOAA
1995–2005 for the climatology of surface wind, wind stress
and wind stress curl intensity and the ERA-Interim reanalysis 1979–2014 for
SST .
In the Arabian Sea, summer productivity is affected by the winds through
different mechanisms driven by the wind stress and the
wind stress curl . The strong winds along the Arabian
coast, called the Somali Jet, generate a positive wind stress, which
increases Ekman transport off the coast. The water that leaves the coastal
area is replaced by subsurface water: this is the coastal upwelling. Similar
to the wind intensity, the CTRL simulation wind stress intensity is
underestimated compared with the reanalyses: the maximum wind stress
intensity is lower and it does not extend as far north in the Arabian Sea as
the reconstructed wind stress (Fig. c). The wind stress
orientation is also more zonal in the simulation than in the observations,
which causes the simulated wind stress to be higher close to the Oman coast
relative to the observations (Fig. b, c). Figure d
represents the wind stress curl, computed from the wind stress, in the
simulation and in the observations. The simulated distribution resembles the
reconstructed one: on the left-hand side of the strong low-level wind jet,
between the coast and the maximum wind intensity, the curl in the wind stress
is positive and on the other side of the jet the wind stress curl is
negative. The differences seen in the jet position and width are transmitted
to the wind stress and wind stress curl intensity and distribution, with an
overall less positive curl close to the coast and less negative curl offshore
in the simulation. In Fig. e, we can see that the modelled SST
anomalies, relative to the global averaged SST in boreal summer, are
underestimated in the Arabian Sea, especially close to the Oman coast,
suggesting a less intense upwelling activity compared with the observations.
This is coherent with the underestimated wind stress and wind stress curl
intensities, which control the upward Ekman pumping intensity
(Fig. c, d).
Discrepancies between our pre-industrial simulation and the observations may
be due to the coarse model resolution. In , a
higher-resolution version of the model was used to study the effects of
mesoscale dynamics on productivity. They showed that the model is able to
reproduce the observed mesoscale dynamics, such as the Great Whirl and
filaments that transport nutrients from the coast to the open sea. They
highlighted the major role of the eddy-driven transports in the establishment
of biological blooms in the Arabian Sea and the model's ability to represent
the different physical processes at stake behind productivity blooms in
summer and in winter in the region. Nevertheless, even though both the winds
and productivity are underestimated in CTRL by the lower-resolution version of the model, the physical
mechanisms playing a role in the marine productivity are represented, which
therefore makes the simulations suitable for our study.
Figure illustrates the combined effects of wind stress and
wind stress curl on productivity in the coastal Arabian Sea (orange area in
Fig. a) on the inter-annual timescale. It shows that if the
summer (JJAS) wind stress and wind stress curl intensities are higher than
their summer average, productivity is higher than average in the coastal
Arabian Sea (upper-right quadrant). It also highlights that the higher the
wind stress and the wind stress curl anomalies, the higher the productivity
change, and conversely. This is coherent with the fact that both the wind
stress and the wind stress curl act to bring nutrient-rich waters to the
surface and fuel productivity in boreal summer
. Figure also shows that a
high wind stress curl (wind stress) can compensate
a reduced wind stress (wind stress curl) intensity and lead to
higher-than-average productivity (lower-right and upper-left quadrants in
Fig. , respectively).
Anomalies of total primary productivity (TPP,
molC m-2 yr-1, represented by the colour scale) integrated over
the whole water column as a function of wind stress anomalies
(10-3 N m-2, y axis) and wind stress curl anomalies
(10-7 N m-3, x axis) in the western coastal Arabian Sea (see
Fig. a) for the CTRL simulation. Anomalies are computed as the
difference between each summer average (JJAS) and the seasonal summer mean of
the corresponding variable in the CTRL simulation. The colour of the circles
represents the value of the change in TPP compared with the CTRL. The colour scale,
x axis range and y axis range are the same as those in Figs. and
.
(a, b) Seasonal cycle of the insolation at the top of the
atmosphere (W m-2) in (a) the CTRL simulation and
(b) the difference between EH and CTRL. The Northern Hemisphere (0–90∘ N) mean
summer (JJAS) insolation, NHmean(JJAS), in CTRL and the difference between
the EH and CTRL simulations,
ΔNHmean(JJAS), is given under panels (a, b),
respectively. (c, d) Upper-tropospheric temperature (TT) averaged
between 200 and 500 hPa in (c) the CTRL simulation and (d) the difference between EH and CTRL.
ΔTT value, under panel (c), is the TT gradient between a
northern region (60–120∘ E, 10–45∘ N) and a southern
region (60–120∘ E, 25∘ S–10∘ N) (black boxes on
the maps). Δ(ΔTT), under panel (d), is the difference of
TT gradients between EH and CTRL. (e, f) Boreal summer (JJAS) sea
level pressure (SLP) anomaly (from the annual mean) lower than -5 hPa for
(e) the CTRL simulation and
(f) the difference between the EH and CTRL. The SLPa-5 barycentre (i.e. barycentre of the SLP
anomalies lower than -5 hPa over the region
20∘ W–150∘ E, 30∘S–60∘ N) is represented
by a black star for CTRL
in (e, f) and by a red star for EH in (f).
Diagnostics
In this section we briefly describe the variables and the methods we use
throughout the paper. We are interested in the links between the large-scale
Indian summer monsoon system and the Arabian Sea primary productivity. To
characterise the boreal summer monsoon intensity, we use the meridional
gradient of upper-tropospheric temperature (TT, averaged from 200 to
500 hPa) between a northern region covering India, south-east Asia, the
Tibetan Plateau (60–120∘ E, 10–45∘ N) and a southern
region over the tropical Indian Ocean (60–120∘ E,
25∘ S–10∘ N) (;
). This gradient, ΔTT, is associated with the
land–sea contrast in temperature . ΔTT is
averaged over the boreal summer period (JJAS) and the higher its value, the
stronger the Indian summer monsoon.
Changes in the monsoon intensity and pattern affect the SLP field. We compute the SLP from the model outputs (i.e. model air
temperature and pressure, within and between the atmospheric grid levels, and
orography) using the extrapolation described in . We define
SLP anomalies as the SLP minus the global annual average of SLP. In order
to characterise the monsoon pattern, we compute the barycentre of the region
defined by an SLP anomaly lower than -5 hPa over the region covering the
African, east Asian and Indian monsoon regions of influence
(20∘ W–150∘ E, 30∘ S–60∘ N) and we call
SLPa-5 the region delimited by the -5 hPa contour in SLP anomalies
(Fig. c–f). This SLPa-5 barycentre is representative of the
balance between the different monsoons as well as of the Indian summer
monsoon wind position and direction over the Arabian Sea. A modification of
the monsoon pattern, which can have impacts on productivity through
atmospheric forcing onto the ocean circulation, can then be related to
movements of the SLPa-5 barycentre. We only focus on the Tibetan low since
the Mascarene High, the region of high SLP in the southern Indian Ocean,
barycentre remains quite similar in the different simulations.
showed that the wind stress intensity generates coastal
upwelling and that the positive wind stress curl is responsible for upward
Ekman pumping offshore. We focus our work on these two wind variables in the
coastal western Arabian Sea, a region of positive curl, for the CTRL climate,
between the axis of the Somali Jet and the coast (Fig. a, orange area).
In the following sections of the paper, total primary productivity (TPP) is
defined as the sum of nanophytoplankton and diatoms total net primary
productivity integrated over the whole water column. We also analyse nitrate
concentrations in the first 30 m of the water column, nitrate being the
major limiting nutrient in the region, and its supply to the surface layers
being mainly driven by atmospheric changes via coastal upwelling and upward
Ekman pumping.
Boreal summer mean differences between EH and CTRL for (a) total primary productivity
(TPP, molC m-2 yr-1) integrated over the whole water column,
(b) NO3- concentration
(mmol m-2) integrated in the first 30 m
of the water column, (c) wind stress intensity (N m-2) and
direction (arrows), (d) mixed-layer depth (MLD, m),
(e) wind stress curl intensity (10-7 N m-3), and
(f) Ekman pumping (m yr-1, positive upward). The black contour
in each panel represents the coastal area on which we averaged the variables
throughout the text. Coastal areas for atmospheric and oceanic variables
differ slightly because of the different model grids.
We use the CTRL simulation as a reference. All changes are then defined
relative to this pre-industrial simulation.
Simulated palaeo-productivity and monsoon changes
In this section, we investigate the changes in summer productivity in past
climate simulations with respect to the CTRL simulation, starting with the
early Holocene and then generalising to all the climates.
The early Holocene case
The EH experiences a stronger Indian summer monsoon than the
pre-industrial period (Fig. ). Therefore, we would expect higher
productivity in the Arabian Sea. However, the EH simulation shows lower
levels of productivity than in the CTRL (Fig. ). We explain this
counter-intuitive result by a change in the monsoon pattern instead of a
change in its intensity (Fig. ).
Seasonal (JJAS) anomalies of total primary productivity (TPP,
molC m-2 yr-1), integrated over the whole water column, as a
function of wind stress anomalies (10-3 N m-2) and wind stress
curl anomalies (10-7 N m-3) in the coastal western Arabian Sea
(see Fig. a). Anomalies are computed as the difference between
each yearly summer average in the EH simulation and the seasonal summer mean
of the CTRL simulation. The colour of the circles represents the value of the
change in TPP between EH and CTRL. The colour scale, x axis range and y axis
range are the same as those in Figs. and .
The early Holocene, which we choose to represent with a snapshot at
9.5 kyr BP, is an interglacial period that mainly differs from the
pre-industrial period because of the imposed obliquity (24.2306∘ vs.
22.391∘) and precession (303.032∘ vs. 102.7∘)
(Table ). These changes in astronomical parameters cause the
boreal summer insolation in the Northern Hemisphere to be higher than in the
pre-industrial climate . In our simulation of EH, the
boreal summer (JJAS) Northern Hemisphere (0–90∘ N, NH) mean
insolation is 20 W m-2 higher than in the CTRL (Fig. a,
b). This change in insolation modifies the upper-tropospheric temperature
gradient, ΔTT, represented in Fig. c, d: in
EH, ΔTT is 1 K higher
than in CTRL, supporting a
stronger monsoon intensity consistent with previous studies
.
The large-scale spatial pattern of the summer monsoon is also different in
EH compared with
CTRL. Maps (e) and (f) in
Fig. show the SLP anomalies and the location of the SLPa-5
barycentre. The EH depression extends further to the north-west and onto
Africa and the Arabian peninsula than in CTRL, and the EH minimum SLP anomaly moves to the north-west
compared with the CTRL one (Fig. e, f). The SLPa-5 barycentre,
which is representative of the balance between the different monsoons and of
the Somali Jet position and direction, moves to the north-west in
EH relative to
CTRL (Fig. f). This
suggests a modification of the monsoon structure with potential impacts on
productivity through atmospheric forcing onto the ocean.
Observation-based and model-based
studies have shown that a stronger monsoon, with
increased wind strength over the Arabian Sea, leads to higher productivity
through intensified supply of nutrients in the photic zone. The EH Indian
summer monsoon is enhanced compared with the CTRL simulation. One would therefore expect the EH Arabian Sea productivity to be higher in EH than in
CTRL. However, our model shows
that the EH productivity is reduced in the Arabian Sea (Fig. a).
In this region, productivity is mainly nutrient-limited
and the levels of surface NO3-
concentrations are lower for EH
than for CTRL
(Fig. b). Most of the nutrients are moved to the surface layer
via Ekman dynamics: coastal upwelling, upward Ekman pumping, mixing of the
upper layers close to the coast and offshore, and mixing and advection
processes further offshore
.
The shift of the SLPa-5 barycentre (Fig. f) leads to a poleward
and westward shift of the monsoon jet (Fig. c). This causes
weaker alongshore winds in the Somalia upwelling and stronger alongshore
winds in the Oman upwelling with visible effects on the mixed-layer depth
(Fig. c, d). However, this also brings the Ekman downwelling to
the right of the jet axis closer to both the Oman and Somalia coasts
(Fig. e). Both factors (alongshore stress and offshore curl)
contribute to a Somalia upwelling reduction, while the wind stress curl
change seems to overwhelm the increased alongshore winds in the Oman region,
leading to overall upwelling reduction and fewer nutrients even though the
monsoon is more intense (Fig. b, f).
In Fig. , we represented the 100 summers of the EH simulation
minus the seasonal summer mean of the CTRL simulation for the variables TPP,
wind stress and wind stress curl in the coastal Arabian Sea (see
Fig. a). In the coastal area, even though wind stress is always
higher in EH (y axis),
productivity is always lower in EH than in CTRL
(Fig. a). This highlights the role of the wind stress curl that
is always less positive in EH
than in CTRL (x axis): a less
positive wind stress curl is responsible for lower productivity levels in the
western coastal Arabian Sea. The wind stress intensity also affects the
intensity of the changes in TPP: the stronger the wind stress, the smaller
the reduction in TPP. Increased wind stress can oppose the negative effect of
reduced wind stress curl but not overcome it as it is restricted very close
to the coast (Fig. ).
In summary, in the EH simulation the summer monsoon intensity is stronger
than in the CTRL simulation but the productivity in the Arabian Sea is lower.
This is caused by a shift in the Somali Jet position, which reduces coastal
upwelling and upward Ekman pumping. This shift in the maximum wind intensity
position closer to the coast can be inferred from the north-western movement
of the SLPa-5 barycentre that translates into a modification of the monsoon
pattern (Fig. ).
Seasonal (JJAS) productivity (molC m-2 yr-1, colour scale)
and Ekman pumping (black contour every 20 m yr-1) changes compared
with
the CTRL simulation for the (a) MH, (b) LGM, (c) MIS3,
(d) MIS4F, (e) MIS4M and (f) MIS4D simulations in
the Indian Ocean. The yellow circle indicates the position of core MD04-2873
used later in this paper.
Generalisation
In the previous section, we saw that a stronger monsoon in the
EH simulation does not imply more productivity
in the Arabian Sea and that it is important to consider the spatial movements
of the monsoonal winds. We now examine the links between productivity,
monsoon intensity and boundary conditions in the remaining set of six glacial
and interglacial simulations.
Box plots of summer (JJAS) (a) ΔTT (K)
values, coastal Arabian Sea (b) productivity
(molC m-2 yr-1) integrated over the whole water column,
(c) nitrate concentration (molN m-2) in the first 30 m,
(d) wind stress intensity (10-3 N m-2) and
(e) wind stress curl intensity (10-7 N m-3) for all
eight
simulations. Dashed grey line indicates the CTRL value for each variable. The
simple black line in (e) panel indicates the zero value. The
box plots highlight the median value (bold line), the first and the third
quartiles (lower and higher limits of the box), and the 95 % confidence
interval of the median (upper and lower horizontal dashes). The dots are
extreme values that happened during the 100 years of simulation.
Figure shows the changes in productivity in the Arabian Sea
in all the remaining climates compared with the CTRL climate. Similar to the EH
results, the MH, LGM, MIS4M and MIS4D coastal productivities are reduced
(Fig. a–b, d–f). MH and MIS4F coastal productivities are reduced
on average but present a dipole-like pattern in the western coastal Arabian
Sea, with higher productivity in the north and reduced productivity in the
south compared with the CTRL (Fig. a, d). Coastal productivity in the
western Arabian Sea is enhanced in the MIS3 simulation (Fig. c).
Along with these TPP changes, Ekman pumping changes are represented in
Fig. in black contour. In the simulations where coastal
productivity is higher than in the CTRL, upward Ekman pumping also increases
close to the coast (Fig. a, c, d) and inversely
(Fig. a, b, e, f). Ekman pumping patterns suggest that the wind
orientation has changed throughout the different time periods, differently affecting
the supply of nutrients and then productivity.
The tropospheric temperature gradient (ΔTT) for each simulation, in
Fig. a, shows that the Indian summer monsoon intensity is
stronger in the MH, EH, MIS3 and MIS4F (i.e. higher ΔTT values) and
less intense compared with the CTRL in LGM, MIS4M and MIS4D. The changes in
productivity for the western coastal Arabian Sea are also summarised in
Fig. b. By only looking at these two variables, ΔTT and
TPP, we cannot draw a conclusion on a direct link between monsoon intensity
and productivity because stronger monsoons compared with
CTRL, as characterised by
ΔTT, do not necessarily imply higher productivity
(Fig. a, b), in particular for the MH and
EH simulations.
Productivity and local dynamics
Productivity is nutrient-limited in the region and coastal productivity
changes are similar to the changes in the nitrate content of the upper 30 m
of the ocean (Fig. b, c). When the upper layer receives more
nutrients from the subsurface, there is either a stronger upwelling or a
higher macronutrient (NO3- and PO43-) concentration under the
mixed layer associated with enhanced entrainment. The stronger monsoon
intensity, characterised by a higher ΔTT value, is associated with
higher values of coastal wind stress (Fig. a, d). However,
changes in wind stress curl are independent of the monsoon intensity since it
is lower than CTRL in MH,
EH and MIS4F and more positive in all the other glacial simulations
(Fig. e). Wind stress curl intensity is also more positive in
all the glacial climates compared with the Holocene. In
MH, EH and MIS4F, even though
wind stress intensity is stronger and should generate more coastal upwelling;
the highly reduced wind stress curl overcomes this positive effect on
productivity and induces lower levels of macronutrients, which in turn limit
productivity (Fig. b–e). Conversely, in LGM, MIS4M and MIS4D,
even though the wind stress curl is more positive than in the CTRL, the lower
wind stress intensity seems to prevail and productivity is reduced because of
lower concentrations of nutrients (Fig. b–e). In MIS3, both the
wind stress and the wind stress curl are more positive, more nutrients are
brought to the surface and productivity increases (Fig. b–e).
Figure summarises the links between productivity, wind stress
and wind stress curl intensities in the coastal Arabian Sea. It shows that
changes in wind stress and wind stress curl are drivers of productivity
changes. Both variables modulate the change in productivity, with higher
values associated with higher productivity.
Coastal Arabian Sea seasonal (JJAS) productivity changes
(molC m-2 yr-1) related to wind stress intensity
(10-3 N m-2, y axis) and wind stress curl intensity
(10-7 N m-3, x axis) changes compared with the CTRL simulation.
The colour scale, x axis range and y axis range are the same as those in
Figs. and .
(a) Position of the SLPa-5 barycentre for the eight
simulations. Seasonal (JJAS) coastal (b) productivity
(molC m-2 yr-1), (c) wind stress intensity
(10-3 N m-2, y axis) and (d) wind stress curl
intensity (10-7 N m-3, x axis) as a function of the SLP
barycentre longitude and latitude. Errors bars give standard deviation of the
100 summers. The dotted black box in panel (a) represents the region
on which we zoomed for panels (b–d).
Relation to the large-scale forcing and boundary conditions
In order to understand how changes in the monsoon pattern are linked to the
imposed boundary conditions and influence productivity and local monsoonal
changes, we use the SLPa-5 barycentre. The position of the barycentre of each
simulation is plotted on a map in Fig. a. We also added in
Fig. , the mean boreal summer values (colour scales) of
productivity, wind stress and wind stress curl as a functions of the longitude
(x axis) and the latitude (y axis) of the SLPa-5 barycentre.
Productivity shows an increasing trend with the longitude of the SLPa-5
barycentre between 70 and 94∘ E (Fig. b). It reaches a
maximum of 30 molCm-2yr-1 around 94∘ E. For higher
values of longitude, which correspond to simulations where the monsoon
intensity is reduced, productivity decreases (Fig. b). The
higher values of productivity occur in the simulations for which the SLPa-5
barycentre's longitude and latitude have medium values (MH, CTRL, MIS4F and
MIS3) (Fig. b). The trends in productivity can be explained by
the variations in wind stress (Fig. c) and wind stress curl
(Fig. d) with the SLPa-5 barycentre's position.
Wind stress exhibits an overall decrease with the longitude and latitude of
the SLPa-5 barycentre as it moves south-eastward (Fig. c). If we
omit the CTRL simulation, wind stress is quite constant for the first five
simulations (i.e. lower value of longitude) and then it decreases strongly
for MIS4M, MIS4D and LGM. The latitude of the barycentre exerts a strong
control on the coastal wind stress amplitude (Fig. c). Wind
stress curl shows an increasing trend with longitude for the five simulations
with lower values of longitude and then the wind stress curl becomes quite
constant for MIS4M, MIS4D and LGM (Fig. d). The wind stress
curl has also a tendency to decrease with the SLPa-5 barycentre's latitude
(Fig. d).
The increase in productivity with longitude is mostly due to an increase in
wind stress curl intensity, and the reduction in productivity with higher
longitude is caused by a strong reduction in wind stress while the wind
stress curl remains constant (Fig. b–d). These plots also show
that all the glacial simulations have higher values of SLPa-5 barycentre
longitude compared with the CTRL. This demonstrates a major role of the ice sheet
cover in the longitudinal position of the SLPa-5 barycentre. The
simulations with a stronger monsoon intensity than the CTRL have the highest
value of SLPa-5 barycentre latitude, which suggests an influence of the
astronomical parameters on the latitudinal position of the SLPa-5 barycentre.
This latitudinal movement of the SLPa-5 barycentre with the monsoon strength
also seems to be dependent on the glacial or interglacial state of the
simulation. Indeed, in glacial simulations the SLPa-5 barycentre is north of
the CTRL barycentre even if the monsoons are less intense (LGM, MIS4D and
MIS4M); however, the other glacial simulations with higher summer monsoon
intensity (MIS3 and MIS4F) have their barycentre located north of these
glacial simulations (Fig. ). Similarly, in interglacial
climates, the Holocene simulations have a barycentre north of the CTRL
barycentre
(Fig. ).
In Fig. , we plotted the values of the climatic precession
(defined as e×sin(ω-180∘), with e the eccentricity
and ω the precession), which modulates the Northern Hemisphere
insolation, and obliquity, which controls the temperature contrast between
the high and low latitudes, relative to the SLPa-5 barycentre position, in
order to analyse the relationship between the astronomical parameters and the
SLPa-5 barycentre's position. Climatic precession influences the SLPa-5
barycentre position in both longitude and latitude: when the climatic
precession is high, the barycentre tends to move to the south-east
(Fig. a). Obliquity modulates the latitudinal changes in the
SLPa-5 barycentre: high obliquities are associated with a SLPa-5 barycentre
farther north (Fig. b). MIS3 has a climatic precession value
similar to the CTRL one and a much
higher obliquity than the CTRL one.
Thus, the changes in MIS3 winds and productivity related to insolation are
mostly obliquity-driven (Fig. ). Inversely, MIS4F has a similar
obliquity to the CTRL one and a smaller
climatic precession, which implies that the changes in monsoon intensity in
MIS4F are related to precession (Fig. ). The Holocene
simulations are influenced by both obliquity and precession, while the LGM,
MIS4M and MIS4D simulations seem to reflect a stronger link with the
obliquity signal than with the climatic precession (Fig. ).
Mean (a) climatic precession (e⋅sin(ω-180∘) where e is the eccentricity and
ω-180∘ is the precession) and (b) obliquity values as
a function of the longitude and latitude of the SLPa-5 barycentre for the
eight
climate simulations. Errors bars give the standard deviation of the SLPa-5
position over the 100 summers of each simulation.
Discussion
The summer monsoon paradigm
In the simulations, the general paradigm stating that a stronger summer
monsoon intensity induces a stronger upwelling and therefore increases marine
productivity, is not always verified. Our results show that the
characterisation of the summer monsoon intensity is probably insufficient to
assess past productivity changes and vice versa (Fig. ).
Our results for the summer productivity are consistent with the reconstructed
productivity of . In their study, they analyse two marine
sediment cores in the Arabian Sea: one in the south-east
(5∘04′ N–73∘52′ E) and one in the upwelling region
close to the Oman coast (13∘42′ N–53∘15′ E). They show
that palaeo-productivity in the south-eastern core was higher in glacial
stages than in interglacial stages, which they interpret as the fingerprint
of a stronger winter monsoon. In the other core, the productivity signal is
more complex and they could find some glacial stages (e.g. stage 2) with high
productivity, some interglacial stages with low productivity (e.g. stage 1)
and stage 3 with high productivity. Similarly, in the simulations, in the
central Arabian Sea, glacial productivity is higher than interglacial
productivity (except for MIS4F) (Fig. ). In the western coastal
Arabian Sea, the simulated MIS3 productivity is higher than the CTRL, while the
other climates' productivity is lower than CTRL. Hints on the sources of
discrepancies between our results and the results of and their
interpretation of the productivity changes are given later in this section.
We explain the simulations' summer productivity changes by analysing the
variations in the wind forcing (Fig. ). Given the productivity
changes in the different simulations, the summer monsoon intensity alone is
not able to explain the
changes in productivity and therefore we also investigate the changes in the
monsoon pattern (Fig. ).
The large-scale definition of the monsoon intensity, via ΔTT, is
mainly driven by astronomical changes. The simulations with a strong summer
monsoon either have a high obliquity, which enhances the temperature contrast
between low and high latitudes in summer (e.g. MIS3), or a small climatic
precession that intensifies the summer insolation (MIS4F), or both (MH and
EH) (Fig. ) . Simulations with a weak summer
monsoon all have a small obliquity forcing (Fig. b). The local
wind stress that affects productivity is tightly coupled to the monsoon
intensity (Fig. a, b) and is associated with a SLPa-5 barycentre
movement to the north (Fig. c). This simulated latitudinal
movement of the SLPa-5 barycentre, according to the monsoon intensity, is
consistent with the study of . In this study, the
authors show that with a stronger monsoon the Somali Jet moves further
north, which would indeed translate into a northward movement of the SLP
barycentre (and inversely). In , the authors investigate
the variations in the precipitation records of stalagmites located in Oman
and Yemen. They link the changes in precipitation to the position and
structure of the Inter-Tropical Convergence Zone (ITCZ), which affects the tropical climate. They explain that
during the early Holocene, a northward movement of the mean latitudinal
position of the summer ITCZ is responsible for the decrease in precipitation.
Throughout the Holocene, they show that the ITCZ shifted southward
concomitantly with a decrease in the monsoon precipitation, induced by the
reduction in Northern Hemisphere summer insolation . The
changes in the ITCZ position highlighted by are
consistent with our results, especially with the changes in latitude in the position of
the SLPa-5 barycentre given a similar glacial or interglacial
background state (Fig. ).
Marine productivity is not only influenced by the wind stress but also by the
wind stress curl and the latter is also strongly
influenced in our simulations by the glacial or interglacial state of the
climate (Fig. ). The glacial–interglacial distribution of the
wind stress curl is associated with a longitudinal movement of the SLPa-5
barycentre: the SLPa-5 barycentre moves to the east in glacial climates
and to the west in interglacial climates (Fig. d).
, who analysed the effects of different LGM boundary
conditions on the atmospheric circulation, indeed found that ice sheet
topography is responsible for changes in many features of the SLP field, e.g.
position of lows and highs and their variability. also
evoke an eastward shift in the low-level jet position in summer as a possible
mechanism to explain some productivity changes in the eastern Arabian Sea.
Furthermore, we showed that climatic precession can also act to move the
SLPa-5 barycentre eastward (Fig. a) and therefore affect wind
stress curl and productivity.
We find a valid physical explanation for our simulations' productivity
changes through the effects of the simulations' boundary conditions
(astronomical parameters and ice sheets) on the monsoon intensity and
pattern. However, even if the simulated summer productivity compares quite
well with the data in in the south-eastern Arabian Sea
(except for MIS4F), there are more discrepancies with their second core in
the western coastal Arabian Sea (core in the southern part of the our coastal
area), and also when compared with the reconstructions in
(core in the northern part of the coastal area)
(Fig. ). These differences can arise from several sources, the
first one being linked to the area in which we computed our averages
(Fig. a). Indeed, in the core in the
upwelling region is taken in the southern part of our coastal area
(Fig. a) and the core close to the Oman coast in
is located in the northern part of our coastal area
(Fig. a). Productivity reconstructed by is
high in the early Holocene and gradually decreases throughout the Holocene,
whereas, in our simulations, the Holocene productivity is low compared with
the pre-industrial productivity. A closer look at the productivity changes in
the MH simulation in Fig. a can reconcile our simulations and
this reconstructed productivity. In the simulated MH climate, the northern
part of the coastal area exhibits a positive productivity change, whereas the
southern part of the coastal area is characterised by a negative productivity
change compared with the CTRL (Fig. a). Therefore, our results
are coherent with the results of for the mid-Holocene
since their core is located in the northern part of our coastal area where we
simulate higher MH productivity than in CTRL. However, we do not observe this dipole-like pattern in the EH
productivity (Fig. a), and consequently, the EH simulation does
not agree with this reconstruction. In the EH simulation, we do not have the
remnant Laurentide Ice Sheet that is supposed to be present at this time
period . The simulation with this model version was not
available at the time of our analyses. The addition of a remnant ice-sheet
over Europe and North America in the EH has been shown in
, to induce a southward shift of the ITCZ and a
strengthening of the Indian monsoon. A southward shift of the jets would
modify the large-scale pattern of the SLP and therefore the wind stress and
wind stress curl effects on the Arabian Sea. Based on our findings, the
addition of this residual ice sheet would move the SLPa-5 barycentre to the
south-east, which could increase the wind stress curl and therefore
productivity (Fig. ).
Another source of mismatch between our simulations and data resides in the
fact that we looked at productivity and not at the export production that
will eventually reach the bottom of the ocean. We also focused on the boreal
summer season while it is often advanced that the winter monsoon is
responsible for higher productivity in glacial climates compared with
interglacial climates, e.g. for the south-eastern core productivity in
. It is likely that the boreal winter productivity may also
affect the overall recorded signal in the sediments. In the next section, we
especially discuss the effect of seasonality on productivity.
Box plots of (a) yearly reconstructed productivity from
core MD04-2873 in the north-western Arabian Sea and (b) annual
total primary productivity (TPP), (c) summer total primary
productivity, (d) annual export production (EPC) at 100 m and
(e) summer export production at 100 m in the northern Arabian Sea
(60–68∘ E, 20–68∘ N) for the Holocene and glacial time
periods. The box plots highlight the median value (bold line), the first and
the third quartiles (lower and higher limits of the box), and the 95 %
confidence interval of the median (upper and lower horizontal dashes). The dots
are extreme values that happened during the 100 years of simulation
Seasonality
Here, we investigate new palaeo-productivity reconstructions for the Arabian
Sea and we compare them to our simulations. The simulations largely agree
with the reconstructions, with glacial productivities that are higher than Holocene
productivities in the north-western Arabian Sea, even during boreal summer.
We use data from the sediment core MD04-2873 located at
23∘32′ N–63∘50′ E in the northern Arabian Sea on the
Murray Ridge (Figs. a and ,
yellow circle). This core is well dated by 14C dates from 50 kyr to
present and has a marked Toba Ash layer (74 kyr BP; )
giving a significant robust stratigraphic marker. The coccolithophores are
well preserved and abundant at this location. Their assemblages are used to
reconstruct palaeo-productivity by using a transfer function that has been
designed for the Indian Ocean, including the Arabian Sea
. Samples have been prepared by settling onto cover slips
every 10 cm for stratigraphic intervals covering
2000 years above and below each time period simulated by the model. On
average, six samples were studied by time intervals. Coccolithophore
analysis was automatically generated by a software, SYRACO, that has been
trained to recognise coccolithophores .
Figure a shows the resulting palaeo-productivity for seven of
the eight time periods we previously analysed. This reconstruction indicates
that glacial productivity is higher than Holocene productivity in this core.
At the core location, the effect of the boreal winter monsoon on productivity
is known to be strong . Consequently, the stronger winter
monsoons during glacial time periods are often used to explain how glacial
productivity can be higher than interglacial productivity (e.g.
).
In Fig. , we also plotted the box plots of simulated annual and
summer (JJAS) productivity and export production at 100 m for the same time
periods in the northern Arabian Sea (60–68∘ E, 20–28∘ N).
Our simulated annual and boreal summer productivity shows smaller differences
between glacial and interglacial climates than the reconstructions
(Fig. a–c). The export production shows a clearer separation
between the glacial and interglacial climates in both the summer and annual
plots (Fig. d, e). The differences between productivity and
export production (Fig. b–e) highlight that water column
processes modify the recorded signal, which adds difficulties when comparing
the model results with data.
In all simulated climates, the mean boreal summer productivity is lower than
the mean annual productivity (Figs. and ). This
indicates that in this region in the simulations the mean boreal winter
productivity is higher than the mean boreal summer productivity. Indeed, in
the simulations in the region of the core boreal winter productivity
accounts for more than 40 % of the annual productivity and boreal summer
productivity accounts for more than 20 % (not shown). The hypothesis,
stating that glacial productivity is higher than interglacial productivity
because of a stronger winter monsoon, could explain the variations in this
core (e.g. ). However, the observed present-day seasonal
cycle of productivity in the northern Arabian Sea shows equal contributions
of the winter and summer seasons to the annual productivity
(Fig. b), suggesting that the simulations may underestimate the
boreal summer contribution to productivity compared with the boreal winter
contribution.
Interestingly, the annual and boreal summer productivity plots look alike,
with higher glacial than interglacial productivity in the model
(Fig. ). Since the summer monsoon is able to affect the
north-western Arabian Sea, as seen in Fig. b, it contributes to
the recorded signal in the sediment (e.g. ). The boreal
summer monsoon effect on the recorded signal is then non-negligible and we
see that, even during the boreal summer season, the simulations show higher
glacial than interglacial coastal productivity (Figs. and
), as well as in the central Arabian Sea (Fig. ).
Consequently, the boreal winter productivity is not the sole contributor to
the higher glacial productivity signal compared with the interglacial
productivity in the model, even in the northern Arabian Sea.
Summary and perspectives
We use the coupled IPSL-CM5A-LR model to study the Arabian Sea
palaeo-productivity in eight different climates of the past. We focus on the
processes behind the boreal summer productivity changes in the western
coastal Arabian Sea. We show that a stronger Indian summer monsoon, which is
mostly driven by higher NH insolation, does not necessarily enhance the
Arabian Sea productivity.
We show that glacial climates can be more productive in boreal summer in
the Arabian Sea compared with the pre-industrial climate (Figs. and
). Furthermore, the glacial climates are more productive than the
early Holocene, which was supposed to be the most productive period in the
region (Figs. and ). We found that the paradigm
between monsoon intensity and productivity is valid for MIS3 in the western
coastal sea: a stronger monsoon leads to more productivity. The paradigm is
also valid in the coastal Arabian Sea for the LGM, MIS4M and MIS4D
simulations: a reduced monsoon intensity leads to a reduction in
productivity. However, this is not the case for the MH, EH and MIS4F simulations for
which a stronger summer monsoon is associated with reduced productivity.
Our analyses highlight the importance of considering the monsoon pattern,
especially the position of the maximum wind intensity over the Arabian Sea.
The mechanisms behind productivity changes are summarised in
Fig. . We examine the monsoon pattern through the SLP
barycentre position of the depression covering the summer monsoon regions
(SLPa-5 barycentre). The SLPa-5 barycentre is moved to the east in glacial
climates and far north in climates where the monsoon is enhanced
(Fig. ), which highlights the influence of the ice sheet cover
and of the astronomical parameters .
The monsoon pattern affects the wind stress and wind stress curl efficiency
to bring more or less nutrients to the surface layers. A change in the
pattern can reduce or increase the area in which the winds are effective.
This study also highlights the combined effects of wind stress and processes related to wind
stress curl on productivity. Neither wind intensity nor
wind stress curl alone can explain productivity changes.
Identified seasonal (JJAS) processes behind productivity changes in
glacial and interglacial climates. Bold lines highlight the major pathways.
We need to keep in mind that the model's coarse resolution does not allow for
a very precise representation of the region dynamics. This may have altered
the relative weight of the processes related to wind stress and wind stress
curl and can explain why the astronomical signal is weak in the productivity
changes. Arabian Sea productivity in the CTRL simulation shows quite large
differences with observations, especially for the high-coastal-productivity
extension in the western Arabian Sea (Fig. a). This can result
from the model's coarse resolution, which prevents the representation of
mesoscale processes such as eddies. These fine-scale processes are shown to
be of importance for the coupling between biology and physics
. Moreover, these mesoscale processes contribute
strongly to the export of nutrients offshore and thus can explain why our
high-productivity area is more restricted to the coast than in the
observations. From our set of simulations we cannot assess the effect of the
underestimation of productivity on our results. Our simulations may
underestimate productivity levels and variations but since most of the main
physical processes are represented, we can draw conclusions on the link
between these processes and productivity and their changes through time. A
way to overcome these limitations and to quantify their effects would be to
work with several other models and to analyse the coupling between biology
and physics in those different models.
We demonstrated that both changes in wind stress and wind stress curl can
affect productivity on the timescales of thousands of years
(Fig. ). The same effects of changes in productivity from wind stress and wind stress curl
can be found on the inter-annual timescale
(Fig. ). The relationship between changes in stress or curl and
productivity is similar to the one we found for the glacial–interglacial
climate changes (Figs. and ). It could be
interesting to further investigate these relationships by looking at high-resolution models and reanalyses.
This study allows us to draw attention to certain points that may affect the
reconstruction of past climate and productivity as well as the comparison
between model and data. In addition, in regard to projected changes in the
monsoon intensity and structure, these results can add some constraints on
future productivity changes in the region. In chapter 14 of the 2013 IPCC
report , it has been shown, through the use of climate
projections, that the future Indian summer monsoon is expected to strengthen
in regard to precipitation but become less intense in regard to the monsoon
flow. Moreover, have shown that the projected low-level
jet over the Arabian Sea will shift northward because of global warming. A
northward shift of the low-level jet is consistent with an increased monsoon
intensity in our simulations. Then, if a stronger summer monsoon calls for
increased productivity, a northward shift of the Somali Jet can either lead
to reduced productivity, as in the Holocene and MIS4F simulations, or to an
increased productivity as in the MIS3 simulation, depending on the degree of
the shift and on the change in wind stress curl.