Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
The impact of mesoscale and submesoscale physical processes on phytoplankton biomass, community composition, and carbon dynamics in the oligotrophic ocean
(USC Thesis Other)
The impact of mesoscale and submesoscale physical processes on phytoplankton biomass, community composition, and carbon dynamics in the oligotrophic ocean
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
THE IMPACT OF MESOSCALE AND SUBMESOSCALE PHYSICAL
PROCESSES ON PHYTOPLANKTON BIOMASS, COMMUNITY
COMPOSITION, AND CARBON DYNAMICS IN THE OLIGOTROPHIC OCEAN
by
Xiao Liu
A dissertation presented to
Faculty of the Graduate School, University of Southern California
Los Angeles, California
In Partial Fulfillment of the Requirements for
the Degree of Doctor of Philosophy in Ocean Sciences
Approved by Advisory Committee:
Dr. Naomi Levine (chair)
Dr. Doug Hammond Dr. John Heidelberg Dr. Burt Jones Dr. Dale Kiefer
August 2017
ii
Dedicated
to
Ms.
崔爱菊
and
Mr.
刘保华,
the
woman
and
man
who
raised
me,
inspired
me,
and
turned
me
into
who
I
am.
iii
Acknowledgements
This
dissertation
cannot
be
possibly
completed
without
the
enormous
amount
of
guidance,
assistance,
and
support
from
my
advisor,
dissertation
committee,
labmates
and
colleagues,
and
family
and
friends.
There
is
no
easy
path
to
a
doctoral
degree
but
I
have
never
felt
helpless
on
this
journey.
Here
I
am
privileged
to
express
my
gratitude
to
the
named
and
many
other
wonderful
people
who
I
interacted
with
in
the
past
almost
six
years.
First
and
foremost,
to
my
advisor,
Dr.
Naomi
Levine.
Thank
you
for
all
your
encouragement
and
help,
and
for
always
trying
to
provide
me
access
to
the
research
projects
that
I
am
interested
in
doing.
But
you
are
also
much
more
than
just
an
advisor
and
mentor
to
me.
Thank
you
for
also
serving
as
my
role
model
as
not
only
a
remarkable
scientist
and
an
inspiring
mentor,
but
a
caring
mother
and
a
joyful
person
–
you
showed
me
that
a
woman
can
do
it
all.
It
has
shed
so
much
light
onto
my
career
path
and
the
journey
of
my
life.
To
my
dissertation
committee,
Drs.
Burt
Jones,
Dale
Kiefer,
Doug
Hammond,
and
John
Heidelberg.
Your
insights
and
feedbacks
have
certainly
broadened
the
scope
of
my
research
and
helped
improve
the
quality
of
this
dissertation.
Thank
you
all
for
always
making
yourselves
available
whenever
I
needed
you.
A
special
‘Thank-‐‑you’
to
you,
Burt,
for
bringing
me
to
USC
and
for
always
trying
your
best
to
support
my
research
and
career
development
no
matter
where
you
are.
To
my
current
and
former
labmates,
Elizabeth
Teel,
Erin
McParland,
Xuening
Wen,
Sara
Rivero-‐‑Calle,
Nathan
Walworth,
Matthew
Ragan,
Bridget
Seegers,
Arvind
Pereira,
Carl
Oberg,
et
al.
I
appreciate
all
the
moments
we
spent
together
discussing
science,
debugging
MATLAB
scripts,
baby-‐‑sitting
gliders,
exchanging
stories,
drinking
coffees
and
beers…
You
are
my
wonderful
friends
and
family.
To
all
the
faculty
and
staff
in
the
Departments
of
Biological
and
Earth
Sciences
at
USC.
Thank
you,
professors,
for
teaching
me
knowledge
and
inspiring
me
in
classes,
and,
administrative
personnel,
for
all
your
assistance
and
support.
You
made
my
life
so
much
easier.
To
Qi
Wang.
Having
computer
geeky
friends
like
you
was
one
of
the
luckiest
things
that
ever
occurred
to
me.
Thank
you
for
guaranteeing
a
lifetime
technical
support
service
after
you
persuaded
me
to
pursue
a
career
in
numerical
modeling.
You
magically
turned
those
obscure
Linux
commands
into
a
beautiful
landscape
for
me.
To
Peng
Zhan,
Yannis
Georgakakis,
Nikolaos
Zarokanellos,
Aya
Hozumi,
and
many
other
friends
at
KAUST
for
all
the
‘hot’
memories
that
we
shared.
I
wish
I
could
have
returned
to
you
at
least
one
more
time.
To
my
friends
Tongtong
Guo,
Zhi
Zhu,
Ling
Zhou,
Yunsheng
Luo,
Qin
iv
Zhu,
Yixin
Luo,
Guang-‐‑Sin
Lu,
and
many
others,
for
all
the
laughter
and
support.
Because
of
you,
the
word
‘homesick’
never
appeared
in
my
dictionary.
To
my
beloved
husband
and
forever
friend,
Yi-‐‑Cheng
Teng.
I
own
you
a
tremendous
amount
of
thanks
for
your
tolerance
of
my
bad
temper
and
(sometimes)
selfishness,
for
you
taking
care
of
our
daughter
so
that
I
could
concentrate
on
finishing
my
dissertation,
and
for
you
always
being
by
my
side
as
a
loyal
listener
and
supporter.
To
my
beautiful
daughter,
Melody
Teng.
You
have
taught
me
so
many
lessons
about
responsibility,
patience,
trust,
and
love.
You
let
me
discover
the
strengths
in
myself.
You
are
truly
the
sunshine
of
my
life.
Last,
but
definitely
not
the
least,
to
my
dearest
parents,
Aiju
Cui
and
Baohua
Liu,
the
woman
and
man
who
raised
me,
inspired
me,
and
turned
me
into
who
I
am.
This
dissertation
is
dedicated
to
the
two
of
you.
Thank
you
for
your
unconditional
love.
v
TABLE OF CONTENTS
CONTENTS PAGE
TITLE .................................................................................................................................................... i
DEDICATION ...................................................................................................................................... ii
ACKNOWLEDGMENTS ................................................................................................................... iii
TABLE OF CONTENTS ...................................................................................................................... v
LIST OF FIGURES ........................................................................................................................... viii
INTRODUCTION ................................................................................................................................. 1
CHAPTER 1 – ENHANCEMENT OF PHYTOPLANKTON CHLOROPHYLL BY SUBMESOSCALE
FRONTAL DYNAMICS IN THE NORTH PACIFIC SUBTROPICAL GYRE ....................................... 9
Abstract .................................................................................................................................. 10
1. Introduction ........................................................................................................................ 11
2. Methods and Data ............................................................................................................... 13
2.1. Heterogeneity Index ............................................................................................ 13
2.2. Satellite data and analyses .................................................................................. 16
3. Results ................................................................................................................................ 16
4. Discussions and Implications ............................................................................................. 19
Acknowledgements ................................................................................................................ 23
References ............................................................................................................................... 24
Figure Captions ...................................................................................................................... 27
Figures .................................................................................................................................... 28
Supporting Information .......................................................................................................... 32
vi
CHAPTER 2 – MODELING THE IMPACT OF FINE-SCALE DISTURBANCES ON
PHYTOPLANKTON COMMUNITY COMPOSITION AND CARBON CYCLING IN THE NORTH
PACIFIC SUBTROPICAL GYRE ........................................................................................................ 53
Abstract .................................................................................................................................. 54
1. Introduction ........................................................................................................................ 55
2. Methods ............................................................................................................................... 58
2.1. Model description ............................................................................................... 58
2.2. Model configuration and simulations ................................................................. 60
2.3 Validation datasets ................................................................................................ 63
3. Results ................................................................................................................................ 64
3.1. SHiP vs. HOT comparisons ................................................................................ 64
3.2. Spatiotemporal heterogeneity ............................................................................. 65
3.3. Ecological relevance of the intensity and duration of disturbances ..................... 67
3.4. Climate change experiments ................................................................................ 68
4. Discussions .......................................................................................................................... 69
5. Conclusions and Implications ............................................................................................. 74
Acknowledgements ................................................................................................................ 75
References ............................................................................................................................... 76
Appendices ............................................................................................................................. 82
Figure Captions ...................................................................................................................... 83
Figures .................................................................................................................................... 85
vii
CHAPTER 3 – SPATIOTEMPORAL VARIABILITY AND ENVIRONMENTAL CONTROLS OF
PHYTOPLANKTON DISTRIBUTIONS IN THE RED SEA ................................................................ 95
Abstract .................................................................................................................................. 96
1. Introduction ........................................................................................................................ 97
2. Data and Methods .............................................................................................................. 101
2.1. Environmental data and eddy identification scheme ........................................ 101
2.2. The bio-optical phytoplankton (bio-opt model) ................................................ 102
2.3 Principal Component Analysis (PCA) ............................................................... 105
3. Results and Discussions ................................................................................................... 105
3.1. Spatial and seasonal variability in phytoplankton chlorophyll ......................... 105
3.2. Impact of cyclonic and anti-cyclonic eddies on chlorophyll ............................ 106
3.3. Environmental controls of chlorophyll and primary production ....................... 109
4. Conclusions ....................................................................................................................... 112
Acknowledgements .............................................................................................................. 113
References ............................................................................................................................. 114
Appendices ........................................................................................................................... 117
Figure Captions .................................................................................................................... 120
Figures .................................................................................................................................. 122
viii
LIST OF FIGURES
FIGURE # PAGE
CHAPTER 1 – ENHANCEMENT OF PHYTOPLANKTON CHLOROPHYLL BY SUBMESOSCALE
FRONTAL DYNAMICS IN THE NORTH PACIFIC SUBTROPICAL GYRE
Figure 1. Feature identification using the Heterogeneity Index .................................................................. 28
Figure 2. Weekly climatologies of SST, Chl, and submesoscale heterogeneity (HI
10
) averaged over the
study region (July 2002 to June 2015) ......................................................................................... 29
Figure 3. Impact of submesoscale heterogeneity (HI
10
) on SST and Chl ................................................... 30
Figure 4. Impact of SST and submesoscale heterogeneity (HI
10
) on Chl ................................................... 31
CHAPTER 2 – MODELING THE IMPACT OF FINE-SCALE DISTURBANCES ON
PHYTOPLANKTON COMMUNITY COMPOSITION AND CARBON CYCLING IN THE NORTH
PACIFIC SUBTROPICAL GYRE
Figure 1. Configuration of SHiP model setup ........................................................................................... 85
Figure 2. Comparisons between model estimates and in situ time-series measurements (HOT) at Station
ALOHA for mixed layer averaged temperature, primary production, and biomass of each of the
three phytoplankton groups .......................................................................................................... 86
Figure 3. Export ratio calculated as particulate carbon export (at 150 m reference depth) divided by depth-
integrated (0–150 m) primary production ................................................................................... 87
Figure 4. Comparisons between model estimated and satellite (MODIS-Aqua) derived sea surface
temperature and chlorophyll concentration in July 2004 ............................................................ 88
Figure 5. The comparison between model estimated, ship-based measurements (HOE-DYLAN), and
profiling float measurements at Station ALOHA for temperature during Year 2012 ................ 89
Figure 6. Impact of intensity (vertical velocity in m day
-1
) and duration (in days) of fine-scale disturbances
on phytoplankton community composition .................................................................................. 90
ix
Figure 7. Climate change experiments with reduced mean fluxes of water and nutrients, and additional
changes in the duration of fine-scale disturbances ...................................................................... 91
Figure 8. Model estimated phytoplankton community composition when the model is run in the
heterogeneous SHiP mode and homogeneous AE mode ............................................................ 92
Figure 9. Model estimated carbon export flux out of the mixed layer when the model is run in the
heterogeneous SHiP mode and homogeneous AE mode ............................................................ 93
Figure 10. Model estimated chlorophyll concentration over the 12-year period ........................................ 94
CHAPTER 3 – SPATIOTEMPORAL VARIABILITY AND ENVIRONMENTAL CONTROLS OF
PHYTOPLANKTON DISTRIBUTIONS IN THE RED SEA
Figure 1. Map of the Red Sea and domain separation into the northern Red Sea (NR, 25-29°N), north
central Red Sea (NCR, 21-25°N), south central Red Sea (SCR, 17-21°N), and the southern Red
Sea (SR, 13-17°N) ..................................................................................................................... 122
Figure 2. A schematic description of the bio-opt model. Modified from Ondercin et al. (1995) ............. 123
Figure 3. Seasonal mean chlorophyll concentration derived from OCI-RG algorithm of Brewin et al. (2015)
and OC3 algorithm averaged over the period of September 1997 to December 2013 ............. 124
Figure 4. Regional and seasonal chlorophyll distributions in the Red Sea, presented as the percentage
chlorophyll contribution of each region (normalized for the area of the region) or season to the
total chlorophyll ........................................................................................................................ 125
Figure 5. 8-day averaged Sea Level Anomaly (SLA) field for November 1-8, 1998, and a cyclonic eddy
and an anti-cyclonic eddy detected from the SLA field ............................................................ 126
Figure 6. Eddy frequency from September 1997 to December 2013 during winter and summer ........... 127
Figure 7. Impact of cyclonic (CE) and anti-cyclonic (AE) eddies on phytoplankton presented as the
percentage change in chlorophyll in the eddy structures relative to the background field during
winter and summer .................................................................................................................... 128
x
Figure 8. A schematic illustration of how a persistent wind exerting on an anticyclonic and a cyclonic eddy
may lead to divergence (upwelling) and convergence (downwelling), respectively, in the eddy
interior ....................................................................................................................................... 129
Figure 9. PCA ordination bi-plots for PP, chlorophyll, and environmental variables in the winter ......... 130
Figure 10. PCA ordination bi-plots for PP, chlorophyll, and environmental variables in the summer .... 131
1
Introduction
The
ocean
plays
an
important
role
in
the
global
carbon
cycle,
sequestering
approximately
half
of
anthropogenic
carbon
dioxide
into
the
deep
ocean
via
both
the
physical
and
biological
ocean
carbon
pump
(Sabine
et
al.,
2004;
Khatiwala
et
al.
2009).
Significant
modifications
in
the
global
radiation
flux,
hydrological
dynamics,
and
nutrient
cycling
as
a
result
of
anthropogenic
activities
(e.g.
emission
of
greenhouse
gases)
have
driven
changes
in
ocean
ecosystems.
These
changes
have
occurred
over
a
range
of
spatial
and
temporal
scales
(Karl
et
al.,
1995;
Gregg
et
al.,
2003),
from
interdecadal
and
basin-‐‑wide
alterations
in
sea
surface
temperature
and
chlorophyll
concentration
(Le
Borgne
et
al.,
2002;
Masotti
et
al.,
2011)
to
changes
in
the
frequency
and
magnitude
of
frontal
cyclone
activities
(Landsea
et
al.,
2010),
which
would
in
turn
alter
the
finer-‐‑scale
biogeochemical
and
ecological
signatures
of
the
ocean.
As
the
climate
warms,
the
strength
of
the
biological
carbon
pump
has
been
predicted
to
diminish
due
to
increased
surface
stratification
that
shifts
phytoplankton
populations
towards
smaller
size
classes
thereby
reducing
the
efficiency
of
gravitational
carbon
export
to
the
deep
ocean
(Li
et
al.,
2009;
Hilligsoe
et
al.,
2011).
Other
studies,
however,
suggest
that
climate
changes
may
increase
the
frequency
of
episodic
processes,
such
as
short-‐‑lived
hurricanes
and
tropical
storms
(Landsea
et
al.,
2010),
which
enhance
mixing
and
inject
nutrients
into
the
surface
ocean.
These
events
may
promote
phytoplankton
growth
and
lead
to
elevated
particulate
export
fluxes.
While
much
has
been
learned
with
regard
to
changes
in
the
ocean
from
traditional,
ship-‐‑based
sampling
efforts,
the
temporal
and
spatial
2
resolutions
in
these
observational
data
are
insufficient
to
resolve
the
fine-‐‑scale
physical
processes
(e.g.
meso-‐‑
and
submesoscale;
1-‐‑100
km)
that
are
extremely
relevant
to
the
dynamics
of
marine
ecosystems.
It
has
not
been
possible
to
investigate
the
dynamics
and
biogeochemical
impact
of
fine-‐‑scale
(<
10
km)
processes
with
global
climate
models
due
to
large
computational
cost
of
running
high-‐‑resolution
model
simulations.
Predictions
of
ecosystem
responses
to
forecasted
changes
in
climate
are
typically
made
using
Earth
System
Models
(e.g.
Orr
et
al.,
2005;
Blackford,
2010)
that
are
run
at
coarse
resolutions
(e.g.
1-‐‑3°).
At
these
resolutions,
climate
models
do
not
resolve
fine-‐‑scale
and
episodic
physical
processes
that
are
ecologically
relevant.
Consequently,
idealized
submesoscale
resolving
models
have
been
used
to
investigate
these
relationships.
These
models
suggest
that
heterogeneity
driven
by
mesocale
eddies
and
submesoscale
fronts
may
be
responsible
for
a
majority
of
the
inter-‐‑annual
variations
in
major
phytoplankton
blooms
(Levy
et
al.,
2014).
Correctly
capturing
physical-‐‑
biological
interactions
at
fine-‐‑scale
may
become
especially
critical
under
future
climate
scenarios
where
non-‐‑linear
biological
responses
to
physical
changes
may
impact
the
efficiency
and
magnitude
of
the
biological
carbon
pump
and
result
in
climate-‐‑feedbacks.
As
such,
computationally
tractable,
submesoscale
resolving
biogeochemical
models
are
needed
in
order
to
fully
disentangle
the
major
physical
drivers
for
biogeochemical
variability
in
the
ocean,
and
to
predict
the
impact
of
forecasted
climate
changes
on
large-‐‑scale
ecosystem
responses
and
global
carbon
cycles.
The
overarching
goal
of
my
research
is
to
understand
the
impact
of
projected
changes
3
in
climate
on
marine
ecosystems
and
global
biogeochemical
cycles.
This
dissertation
investigates
the
role
that
meso-‐‑
and
submesoscale
physical
processes
play
in
mediating
the
large-‐‑scale
climate-‐‑physical-‐‑ecosystem
interactions
in
the
oligotrophic
ocean
using
an
interdisciplinary
approach
that
includes
remote
sensing,
multivariate
data
analysis,
and
numerical
modeling.
Specifically,
there
are
two
parts
of
this
dissertation:
Part
I
addresses
the
impact
of
submesoscale
frontal
dynamics
on
phytoplankton
biomass,
community
structure,
and
carbon
dynamics
in
the
North
Pacific
Subtropical
Gyre
(NPSG);
and
Part
II
aims
to
understand
the
impact
of
mesoscale
processes,
such
as
eddies,
and
a
suite
of
other
environmental
forcings
on
phytoplankton
biomass
and
primary
productivity
in
the
Red
Sea.
Part
I:
Subtropical
gyres
play
a
critical
role
in
regulating
global
ocean
productivity
and
carbon
cycling
(Karl
et
al.,
1996;
Lomas
et
al.,
2010).
While
originally
considered
to
be
the
quiescent
‘deserts’
of
the
oceans,
we
now
know
that
a
substantial
degree
of
patchiness
across
a
range
of
spatial
and
temporal
scales
exists
in
the
oligotrophic
gyres.
The
importance
of
large
and
mesoscale
processes
on
nutrient
and
phytoplankton
dynamics
and
in
the
subtropical
gyres
has
long
been
recognized
(McGillicuddy
et
al.,
1998;
Brown
et
al.,
2008).
However,
in
the
past
decade,
much
attention
has
been
given
to
submesoscale
(1-‐‑10
km)
features
and
their
role
in
promoting
vertical
exchanges
in
the
surface
ocean
particularly
in
nutrient
limited
regions
of
the
ocean.
It
has
been
shown
that
frontal
instabilities
associated
with
these
fine-‐‑scale
features
may
trigger
rapid
upwelling
velocities,
which
in
turn
injects
nutrients
into
the
sunlit
surface
and
can
trigger
short-‐‑lived
phytoplankton
blooms
(Mahadevan
et
al.,
2016;
Levy
et
al.,
2012,
2014).
However,
due
to
the
short
temporal
and
spatial
scale
of
these
features,
they
are
insufficiently
sampled
with
in
situ
ship-‐‑based
4
observations
and
(as
mentioned
above)
not
captured
in
global
model
simulations.
As
a
result,
our
understanding
of
the
dynamics
of
these
fine-‐‑scale
processes
and
their
impact
on
large-‐‑scale
biogeochemical
and
carbon
cycling
is
quite
limited.
Chapters
1
and
2
of
this
dissertation
aim
to
improve
our
understanding
on
this.
In
Chapter
1,
decadal
satellite
imagery
was
combined
with
high
resolution
in
situ
observations
in
order
to
determine
the
impact
of
submesoscale
frontal
dynamics
on
chlorophyll
in
the
NPSG.
Specifically,
a
new
statistical
tool
was
developed
to
identify
frontal
structures
from
sea
surface
temperatures.
I
then
demonstrated
that,
in
these
features,
chlorophyll
concentrations
were
enhanced
by
on
average
38%
(max.
83%)
during
the
late
winter.
Measurements
from
profiling
floats
suggest
that
this
may
be
attributed
to
shoaling
of
nutricline
and
weakened
stratification
that
are
facilitated
by
wintertime
mixing.
In
Chapter
2,
a
new
modeling
framework,
the
Spatially
Heterogeneous
Dynamic
Plankton
(SHiP)
model,
was
applied
to
the
Hawaii
Ocean
Time-‐‑series
site
(Station
ALOHA)
to
study
the
impact
of
submesoscale
processes
on
phytoplankton
community
structure
and
carbon
dynamics
in
the
NPSG.
The
model
was
fully
optimized
and
validated
for
the
site,
and
output
was
compared
against
high
resolution
remotely
sensed
and
in
situ
observations.
The
model
yielded
a
substantially
different
phytoplankton
community
composition
when
run
in
a
temporally
and
spatially
heterogeneous
mode
(which
favors
the
growth
of
large
phytoplankton
and
therefore
facilitates
carbon
export)
relative
to
the
traditional
homogeneous
approach.
The
results
from
these
two
chapters
highlight
the
need
for
an
improved
understanding
of
submesoscale
physical
dynamics
in
the
oligotrophic
ocean,
and
suggest
that
both
large
and
fine-‐‑scale
bio-‐‑physical
interactions
should
be
accounted
for
in
5
the
next-‐‑generation
global
climate
models.
Part
II:
The
semi-‐‑enclosed,
mostly
oligotrophic
Red
Sea
is
home
to
an
extremely
diverse
ecosystem
that
includes
many
endemic
marine
species
(Cantin
et
al.,
2010).
A
unique
characteristic
of
surface
circulation
in
the
Red
Sea
is
the
presence
of
a
chain
of
alternating
mesoscale
cyclonic
and
anti-‐‑cyclonic
eddies
along
its
elongated
basin
(Yao
et
al.,
2014;
Zhan
et
al.,
2014)
forced
by
wind
stress,
buoyancy
fluxes,
and
bathymetric
constraints
(Chen
et
al.,
2014).
These
eddies
have
been
shown
to
interfere
with
the
formation
of
deep
water
circulations
(Zhai
et
al.,
2015)
and
modulate
vertical
nutrient
distributions
(Papadopoulos
et
al.,
2015).
However,
the
ecological
impact
of
eddies
remains
poorly
understood
and
little
is
known
with
regard
to
how
eddies
may
interact
with
other
environmental
forcings
to
drive
phytoplankton
productivity
in
the
Red
Sea.
In
Chapter
3
of
this
dissertation,
a
suite
of
remotely
sensed
observations
was
analyzed
to
identify
the
environmental
variables
that
are
most
relevant
to
phytoplankton
in
the
Red
Sea.
Specific
focus
was
given
to
the
role
of
mesoscale
eddies,
which
were
identified
from
Sea
Level
Anomaly
fields,
in
regulating
chlorophyll
distributions.
Results
show
that
the
impact
of
eddies
on
chlorophyll
varies
both
spatially
and
temporally.
One
interesting
finding
is
the
positive
summertime
chlorophyll
anomaly
associated
with
anti-‐‑cyclonic
eddies,
which
is
in
contrast
to
our
previous
understanding
that
anti-‐‑cyclonic
eddies
often
indicate
downwelling,
depleted
nutrients,
and
reduced
production.
This
analysis
also
suggested
that
dust
deposition
may
play
an
important
role
in
regulating
phytoplankton
chlorophyll
in
the
Red
Sea.
Finally,
a
one-‐‑dimensional,
bio-‐‑optical
phytoplankton
model
in
combination
with
6
results
of
high
resolution
hydrological
simulations
was
used
to
estimate
water
column
distributions
of
phytoplankton
carbon
and
primary
production.
Results
show
a
decoupling
between
satellite
chlorophyll
estimates
and
modeled
primary
production
in
the
southern
regions.
This
suggests
that
careful
attention
should
be
given
when
using
surface
optical
properties
to
interpret
phytoplankton
dynamics
in
the
Red
Sea.
7
References
Blackford,
J.
C.,
2010.
Predicting
the
impacts
of
ocean
acidification:
Challenges
from
an
ecosystem
perspective.
J.
Mar.
Syst.
81,
12-‐‑18.
Chen,
C.,
R.
Li,
L.
Pratt,
R.
Limeburner,
R.
C.
Beardsley
(2014),
Process
modeling
studies
of
physical
mechanisms
of
the
formation
of
an
anticyclonic
eddy
in
the
central
Red
Sea.
J.
Geophys.
Res.,
119,
1445-‐‑1464.
Gregg,
W.
W.,
M.
E.
Conkright,
P.
Ginoux,
J.
E.
O’Reilly,
and
N.
W.
Casey
(2003),
Ocean
primary
production
and
climate:
global
decadal
changes.
Geophys.
Res.
Lett.,
30,
1809.
Hilligsoe,
K.
M.,
K.
Richardson,
J.
Bendtsen,
L.
Sorensen,
T.
G.
Nielsen,
and
M.
M.
Lyngsgaard
(2011),
Linking
phytoplankton
community
size
composition
with
temperature,
plankton
food
web
structure
and
sea-‐‑air
CO2
Flux.
Deep
Sea
Res.
I,
58,
826-‐‑838.
Karl,
D.
M.,
J.
R.
Christian,
J.
E.
Dore,
D.
V.
Hebel,
R.
M.
Letelier,
L.
M.
Tupas,
and
C.
D.
Winn
(1996),
Seasonal
and
interannual
variability
in
primary
production
and
particle
flux
at
Station
ALOHA,
Deep
Sea
Res.
II,
43,
539-‐‑568.
Karl,
T.
R.,
R.
W.
Knight,
and
W.
Plummer
(1995),
Trends
in
High-‐‑Frequency
Climate
Variability
in
the
20th-‐‑Century.
Nature,
377,
217-‐‑220.
Khatiwala,
S.,
F.
Primeau,
and
T.
Hall
(2009),
Reconstruction
of
the
history
of
anthropogenic
CO2
concentrations
in
the
Ocean.
Nature,
462,
346-‐‑349.
Landsea,
C.
W.,
G.
A.
Vecchi,
L.
Bengtsson,
and
T.
R.
Knutson
(2010),
Impact
of
duration
thresholds
on
Atlantic
tropical
cyclone
counts,
J.
Clim.,
23,
2508-‐‑2519.
Le
Borgne,
R.,
R.
T.
Barber,
T.
Delcroix,
H.
Y.
Inoue,
D.
J.
Mackey,
and
M.
Rodier
(2002),
Pacific
warm
pool
and
divergence:
temporal
and
zonal
variations
on
the
equator
and
their
effects
on
the
biological
pump.
Deep
Sea
Res.
II,
49,
2471-‐‑2512.
Levy,
M.,
L.
Resplandy,
and
M.
Lengaigne
(2014),
Oceanic
mesoscale
turbulence
drives
large
biogeochemical
interannual
variability
at
middle
and
high
latitudes,
Geophys.
Res.
Lett.,
41,
2467-‐‑2474.
Levy,
M.,
R.
Ferrari,
P.
J.
S.
Franks,
A.
P.
Martin,
and
P.
Riviere
(2012),
Bringing
physics
to
life
at
the
submesoscale,
Geophys.
Res.
Lett.,
39,
L14602.
Li,
W.
K.
W.,
F.
A.
McLaughlin,
C.
Lovejoy,
and
E.
C.
Carmack
(2009),
Smallest
algae
thrive
as
the
Arctic
Ocean
freshens.
Science,
326,
539-‐‑539.
8
Lomas,
M.
W.,
D.
K.
Steinberg,
T.
Dickey,
C.
A.
Carlson,
N.
B.
Nelson,
R.
H.
Condon,
and
N.
R.
Bates
(2010),
Increased
ocean
carbon
export
in
the
Sargasso
Sea
linked
to
climate
variability
is
countered
by
its
enhanced
mesopelagic
attenuation,
Biogeosciences,
7,
57-‐‑70.
Mahadevan,
A.
(2016),
The
impact
of
submesoscale
physics
on
primary
productivity
of
plankton,
Ann.
Rev.
Mar.
Sci.,
8,
161-‐‑184.
Masotti,
I.,
C.
Moulin,
S.
Alvain,
L.
Bopp,
A.
Tagliabue,
and
D.
Antoine
(2011),
Large-‐‑scale
shifts
in
phytoplankton
groups
in
the
Equatorial
Pacific
during
ENSO
cycles.
Biogeosciences,
8,
539-‐‑550.
McGillicuddy,
D.,
A.
Robinson,
D.
Siegel,
H.
W.
Jannasch,
R.
Johnson,
T.
D.
Dickey,
J.
McNeil,
A.
F.
Michaels,
and
A.
H.
Knap
(1998),
Influence
of
mesoscale
eddies
on
new
production
in
the
Sargasso
Sea.
Nature,
394,
263-‐‑266.
Orr,
J.
C.,
V.
J.
Fabry,
O.
Aumont,
L.
Bopp,
S.
C.
Doney,
R.
A.
Feeley
et
al.
(2005),
Anthropogenic
ocean
acidification
over
the
twenty-‐‑first
century
and
its
impact
on
calcifying
organisms.
Nature,
437,
681-‐‑686.
Sabine,
C.
L.,
R.
A.
Feely,
N.
Gruber,
R.
M.
Key,
K.
Lee,
J.
L.
Bullister,
R.
Wanninkhof
et
al.
(2004),
The
Oceanic
Sink
for
Anthropogenic
CO2.
Science,
305,
367-‐‑371.
Yao,
F.,
I.
Hoteit,
L.
J.
Pratt,
A.
S.
Bower,
P.
Zhai,
A.
Kohl,
and
G.
Gopalakrishnan
(2014),
Seasonal
overturning
circulation
in
the
Red
Sea:
1.
Model
validation
and
summer
circulation.
J.
Geophys.
Res.,
2238-‐‑2262.
Zhai,
P.,
A.
S.
Bower,
W.
M.
Smethie,
and
L.
J.
Pratt
(2015),
Formation
and
spreading
of
Red
Sea
Outflow
Water
in
the
Red
Sea.
J.
Geophys.
Res.
Oceans,
120,
6542-‐‑6563.
Zhan,
P.,
A.
C.
Subramanian,
F.
Yao,
and
I.
Hoteit
(2014),
Eddies
in
the
Red
Sea:
A
statistical
and
dynamical
study.
J.
Geophys.
Res.
Oceans,
119,
3909-‐‑3925.
9
Citation:
Liu,
X.
and
N.
M.
Levine,
2016.
Enhancement
of
phytoplankton
chlorophyll
by
submesoscale
frontal
dynamics
in
the
North
Pacific
Subtropical
Gyre.
Geophys.
Res.
Lett.
43,
1651–1659.
Chapter
1.
Enhancement
of
phytoplankton
chlorophyll
by
submesoscale
frontal
dynamics
in
the
North
Pacific
Subtropical
Gyre
Xiao
Liu
1
and
Naomi
Marcil
Levine
2,*
1
Department
of
Earth
Sciences,
University
of
Southern
California,
Los
Angeles
90089
2
Department
of
Biological
Sciences,
University
of
Southern
California,
Los
Angeles
90089
*:
Corresponding
author;
AHF
M225,
3616
Trousdale
Pkwy,
Los
Angeles,
CA
90089;
n.levine@usc.edu;
+1
(213)
821-‐‑0745
Formatted
for
and
published
in
Geophysical
Research
Letters,
United
States
(Manuscript
received
16
November
2015;
issue
online
16
March
2016)
10
Abstract
Subtropical
gyres
contribute
significantly
to
global
ocean
productivity.
As
the
climate
warms,
the
strength
of
these
gyres
as
a
biological
carbon
pump
is
predicted
to
diminish
due
to
increased
stratification
and
depleted
surface
nutrients.
We
present
results
suggesting
that
the
impact
of
submesoscale
physics
on
phytoplankton
in
the
oligotrophic
ocean
is
substantial
and
may
either
compensate
or
exacerbate
future
changes
in
carbon
cycling.
A
new
statistical
tool
was
developed
to
quantify
surface
patchiness
from
sea
surface
temperatures.
Chlorophyll
concentrations
in
the
North
Pacific
Subtropical
Gyre
were
shown
to
be
enhanced
by
submesoscale
frontal
dynamics
with
an
average
increase
of
38%
(max.
83%)
during
late
winter.
The
magnitude
of
this
enhancement
is
comparable
to
the
observed
decline
in
chlorophyll
due
to
a
warming
of
~1.1°C.
These
results
highlight
the
need
for
an
improved
understanding
of
fine-‐‑scale
physical
variability
in
order
to
predict
the
response
of
marine
ecosystems
to
projected
climate
changes.
Key
points:
1. A
new
statistical
tool
quantifies
spatial
heterogeneity
from
high-‐‑resolution
satellite
images.
2. Submesoscale
dynamics
is
shown
to
enhance
chlorophyll
in
the
North
Pacific
Subtropical
Gyre.
3. The
impact
of
submesoscale
physics
on
phytoplankton
may
modify
the
negative
impact
of
warming.
11
1.
Introduction
The
ocean
and
its
biota
are
undergoing
major
changes
as
a
result
of
natural
and
anthropogenic
forcing.
Over
the
past
decades
much
has
been
learned
with
regard
to
alterations
to
large-‐‑scale
(e.g.
basin-‐‑wide)
circulation
in
the
ocean
[Vecchi
et
al.,
2006],
as
well
as
the
cascading
effects
on
intermediate-‐‑scale
dynamics
such
as
eddies
[Davis
and
Di
Lorenzo,
2015].
The
impact
of
these
physical
variations
on
nutrient
distributions
and
ecosystem
structures
has
been
studied
through
long-‐‑term
time-‐‑series
programs,
field
campaigns,
and
a
variety
of
numerical
modeling
experiments
[e.g.
Corno
et
al.,
2007;
Xiu
and
Chai,
2012].
However,
much
less
is
known
about
the
variations
and
impact
of
another
class
of
ubiquitous
features,
submesoscale
dynamics,
due
to
their
typical
length
(1-‐‑10
km)
and
time
(one
to
several
days)
scales
which
make
them
difficult
to
observe
and
model
[Mahadevan
and
Tandon,
2006].
These
fine-‐‑scale
features
often
arise
through
advective
interactions
with
mesoscale
frontal
jets
and
eddy
peripheries
and
are
associated
with
sharp
density
gradients.
These
gradients
create
enhanced
vertical
velocities,
promoting
effective
exchange
between
the
ocean
interior
and
surface
layers
[Capet
et
al.,
2008;
Klein
and
Lapeyre,
2009;
Levy
et
al.,
2010].
Sensors
mounted
on
autonomous
platforms,
such
as
profiling
floats
and
gliders,
have
captured
enhanced,
intermittent
upwelling
velocities
into
the
euphotic
zone
that
are
hypothesized
to
result
from
submesoscale
frontogenesis
[Johnson
et
al.,
2010;
Niewiadomska
et
al.,
2008].
However,
both
the
net
impact
of
fine-‐‑scale
processes
on
large-‐‑
scale
ocean
biogeochemistry
and
how
these
interactions
might
change
in
the
future
remain
poorly
understood.
12
Various
mechanisms
have
been
proposed
regarding
the
potential
impact
of
submesoscale
physics
on
phytoplankton
dynamics.
In
oligotrophic
regions,
the
upward
branches
of
the
fronts
may
enhance
phytoplankton
growth
and
productivity
by
transporting
nutrients
into
the
euphotic
zone
[Mahadevan
and
Archer,
2000;
Johnson
et
al.,
2010],
while
the
downward
components
may
facilitate
export
production
by
rapidly
subducting
biomass
into
the
subsurface
[Niewiadomska
et
al.,
2008;
Omand
et
al.,
2015].
Using
an
idealized
model,
Levy
et
al.
[2014]
suggested
that
~20%
of
new
production
in
the
oligotrophic
subtropics
could
be
explained
by
submesoscale
dynamics.
Conversely,
in
regimes
where
deep
mixing
frequently
occurs
and
light
is
generally
limiting,
submesoscale
instabilities
may
create
a
re-‐‑stratified
sunlit
layer
that
promotes
productivity
[Mahadevan,
2016].
It
has
also
been
argued
that
the
downwelling
side
of
the
fronts
subducts
much
of
the
phytoplankton
biomass
below
the
euphotic
zone
on
short
enough
time-‐‑scales
that
the
consumption
of
upwelled
nutrients
may
be
incomplete
[Levy
et
al.,
2012].
As
such,
due
to
the
complexity
of
mixed
layer
dynamics
and
light
and
nutrient
availability,
the
net
impact
of
submesoscale
physics
on
phytoplankton
has
been
difficult
to
determine.
In
this
study,
we
focus
on
the
impact
of
fine-‐‑scale
bio-‐‑physical
interactions
in
the
nutrient-‐‑depleted
(oligotrophic)
regions,
such
as
the
subtropical
gyres.
Subtropical
gyres
play
a
critical
role
in
global
ocean
productivity
and
carbon
cycling
[Karl
et
al.,
1996;
Lomas
et
al.,
2010].
As
global
temperatures
continue
to
rise,
the
efficiency
of
carbon
export
within
these
gyres
is
predicted
to
decline
due
to
increased
stratification,
reduced
vertical
nutrient
exchange,
and
shifts
in
phytoplankton
assemblages
towards
smaller
size
classes
[Hilligsoe
et
al.,
2011;
Li
et
al.,
2009].
In
addition,
some
studies
have
13
detected
decadal-‐‑scale
increasing
trends
in
the
frequency
of
oceanic
fronts
and
eddy
kinetic
energy
in
the
oligotrophic
ocean
[Matear
et
al.,
2013;
Hogg
et
al.,
2015].
These
trends
are
hypothesized
to
be
driven
by
climate
and
atmospheric
instabilities.
While
direct
predictions
of
future
changes
in
submesoscale
dynamics
are
lacking,
these
observed
changes
in
large-‐‑
and
mesoscale
processes
may
cause
significant
modifications
to
submesoscale
dynamics.
Over
the
past
two
decades,
technological
advances
in
remote
sensing
have
provided
synoptic
surface
views
of
the
global
ocean
with
improved
temporal
and
spatial
resolutions
[Gaultier
et
al.
2014].
In
this
study,
we
investigated
the
impact
of
submesoscale
physics
on
phytoplankton
distributions
using
high-‐‑resolution
satellite
observations.
Specifically,
we
developed
a
new
metric
(the
Heterogeneity
Index)
that
quantifies
surface
patchiness,
and
used
it
to
identify
signatures
of
fine-‐‑scale,
frontal
structures
in
the
oligotrophic
ocean
from
horizontal
temperature
gradients.
We
then
established
observational
evidence
for
enhanced
chlorophyll
concentrations
associated
with
submesoscale
frontal
dynamics
in
the
North
Pacific
Subtropical
Gyre
(NPSG),
with
an
average
increase
of
up
to
38%
(maximum
of
83%)
during
the
later
winter.
These
results
have
significant
implications
for
understanding
the
impact
of
submesoscale
physics
on
primary
and
export
production
in
the
oligotrophic
ocean.
2.
Methods
and
Data
2.1.
Heterogeneity
Index
Traditional
approaches
for
quantifying
patchiness
in
a
resource
field
have
primarily
focused
on
data
variance
[e.g.
Doney
et
al.,
2003],
which
only
represents
the
average
gradient
14
in
the
field.
Given
the
nonlinearity
in
biological
responses
to
environmental
conditions,
the
high
degree
of
resource
(e.g.
nutrient)
patchiness
created
by
submesoscale
dynamics
is
expected
to
produce
a
greater
impact
than
the
average
gradient
does.
Cayula
and
Cornillon
[1992]
developed
a
method
that
uses
SST
histogram
distributions
to
search
for
bimodality
in
resource
distributions.
This
method
was
adapted
to
identify
sea-‐‑surface
fronts
in
various
regions
such
as
the
California
Current
[Kahru
et
al.,
2012].
Here
we
combine
these
two
approaches
using
measures
of
both
variance
and
bimodality
to
quantify
patchiness
in
SST.
In
addition,
we
add
a
third
term
that
quantifies
the
skewness
of
the
distribution.
This
additional
term
allows
us
to
capture
patchiness
created
by
thin
filaments,
which
often
cause
unimodal,
skewed
SST
distributions.
Our
new
metric
of
spatial
patchiness,
the
Heterogeneity
Index
(HI),
is
defined
as:
=( +
*
+
+)
eq.
(1)
where
is
the
skewness
of
the
distribution,
is
the
standard
deviation,
and
n
is
the
sample
size.
P
describes
the
difference
in
area
between
the
best
5
th
order
polynomial
fit
to
the
data
[()
in
eq.
(2)]
and
a
Gaussian
distribution
with
the
same
sample
mean
()
and
[(,)
in
eq.
(2)]:
=
6 7 89(:,*)
9(:,*)
;<=
(7)
;>?
(7)
eq.
(2)
Coefficients
b,
c,
and
d
(for
the
NPSG:
b
=
1.07,
c
=
1.81,
d
=
1.11)
scale
each
component
between
0
and
1
such
that
equal
weight
is
placed
on
each
component,
and
a
(a
=
0.30)
scales
HI
such
that
HI
=
0
describes
a
homogenous
system,
and
HI
=
1
describes
a
maximally
heterogeneous
system.
Coefficients
a-‐‑d
are
region
specific
and
must
be
retuned
before
HI
15
can
be
applied
to
different
regions
(see
Supporting
Information
S1
and
S2
for
details
regarding
HI
formulation
and
normalization
coefficients
for
other
subtropical
oceans).
HI
is
spatial-‐‑scale
dependent
and
designed
to
identify
physical
processes
occurring
at
the
sub-‐‑domain
scale.
For
example,
elevated
HI
for
a
domain
of
10
km
×
10
km
(HI10)
can
be
caused
by
the
inclusion
of
a
feature
smaller
than
10
km
in
length
(e.g.
a
submesoscale
filamentous
front),
or
a
fraction
of
a
feature
equal
to
or
larger
than
10
km
(e.g.
part
of
a
mesoscale
front
or
the
edge
of
an
eddy).
Figure
1
shows
an
example
of
a
SST
image
in
which
such
frontal
features
result
in
skewed,
high
variance,
and
bimodal
distributions
and,
therefore,
elevated
HI10
values
at
the
fine-‐‑scale.
While
HI
equally
weights
features
with
different
underlying
physical
mechanisms,
it
highlights
sharp
horizontal
density
gradients
occurring
on
the
scale
of
a
few
kilometers
(i.e.
the
submesoscale)
that
are
typically
associated
with
enhanced
vertical
velocities.
As
a
simplification,
hereafter
we
refer
to
all
fine-‐‑scale
frontal
signatures
as
submesoscale
structures
due
to
the
length
scale
of
the
gradients.
For
this
analysis,
we
apply
HI
to
the
oligotrophic
NPSG.
As
density
gradients
are
typically
coincident
with
temperature
gradients
in
this
region,
HI
allows
us
to
identify
submesoscale
structures
in
the
NPSG
from
satellite-‐‑retrieved
SST
fields.
However,
caution
is
needed
when
applying
HI
to
other
oceanographic
regimes
where
this
underlying
assumption
may
need
to
be
revisited.
For
example,
temperature
may
not
be
an
appropriate
indicator
of
water
mass
differences
in
high-‐‑latitude
and
coastal
upwelling
regions.
Detection
of
patchiness
in
these
regions
using
the
HI
metric
may
require
the
use
of
remotely
sensed
altimetry
data
(which
currently
precludes
submesoscale
analyses
due
to
the
spatial
resolution
of
the
data).
16
2.2.
Satellite
data
and
analyses
Level-‐‑2
daily
images
of
MODIS/Aqua
SST
(daytime)
and
chlorophyll-‐‑a
concentration
(Chl)
at
approximately
1
km
resolution
were
retrieved
from
the
NASA
OB.DAAC
for
a
region
in
the
NPSG
(10-‐‑30°N,
160°E-‐‑160°W)
during
a
13-‐‑year
period
(July
2002
-‐‑
June
2015).
The
latest
version
(R20140)
of
the
reprocessed
data
was
used.
A
subset
of
images
were
selected
using
a
filtering
grid
with
a
fixed
window
size
of
100
km
×
100
km
to
ensure
maximal
spatial
coverage
(75%
for
SST
and
70%
for
Chl)
and
optimal
data
quality
(S1).
For
each
of
the
32,222
selected
images
of
100
km
×
100
km,
an
average
HI10
was
calculated
for
each
individual
pixel.
Specifically,
a
grid
with
a
cell
size
of
10
×
10
pixels
was
applied
to
the
SST
data
and
HI10
was
computed
for
each
grid
cell.
The
grid
was
then
shifted
eastward
or
southward
at
increments
of
one
pixel
at
a
time,
and
a
new
HI10
was
calculated
for
each
cell
at
the
new
grid
location.
Pixel-‐‑level
HI10
was
then
estimated
as
the
averaged
HI10
from
all
possible
grid
locations.
To
identify
the
fractional
area
impacted
by
submesoscale
structures,
heterogeneity
maps
of
HI10
(1
km
resolution)
were
examined
at
weekly
intervals.
For
a
single
week,
the
background
field
was
defined
as
those
pixels
with
a
HI10
within
2σ
from
the
mode
of
all
HI10
values
for
the
week,
and
the
region
impacted
by
submesoscale
structures
was
defined
as
pixels
with
a
HI10
at
least
4σ
greater
than
the
mode
(S3).
Several
different
threshold
values
were
tested
and
the
results
were
not
sensitive
to
the
choice
of
4σ.
Weekly
climatologies
of
SST
and
Chl
in
the
impacted
regions
were
then
compared
with
those
in
the
background
field.
17
3.
Results
Seasonal
climatologies
of
SST
and
HI10
over
the
13-‐‑year
period
showed
an
inverse
relationship
between
fine-‐‑scale
heterogeneity
(HI10)
and
SST,
with
winter
dynamics
resulting
in
increased
mixed
layer
depths,
reduced
SST,
and
elevated
HI10
(Fig.
2;
S6).
Overall,
a
positive
relationship
was
observed
between
the
seasonality
of
HI10
and
Chl,
with
elevated
values
in
the
winter
and
spring
and
reduced
values
in
the
summer.
Chl
peaked
in
early
February,
coincident
with
an
increasing
HI10,
and
then
steadily
declined
while
HI10
remained
elevated.
The
fractional
area
impacted
by
submesoscale
structures
(indicated
by
elevated
HI10)
was
greatest
during
the
winter-‐‑spring
period,
and
lowest
in
the
late
summer
and
early
autumn,
with
an
annual
mean
of
5.2%
(Fig.
3a).
This
is
in
agreement
with
a
previous
remote
sensing
study
that
suggests
that
4–10%
of
the
California
Current
System
is
covered
by
fronts
[Woodson
and
Litvin,
2015].
Regions
with
submesoscale
structures
were
also
associated
with
lower
SST
and
elevated
Chl.
The
greatest
change
in
SST
relative
to
the
background
field
was
seen
in
late
February
with
a
weekly
average
difference
of
up
to
1.7°C
(Fig.
3b).
This
is
consistent
with
results
from
current
profilers
that
showed
an
increase
in
the
strength
of
submesoscale
features
during
the
winter
(Jan-‐‑Mar)
due
to
more
frequent
larger-‐‑scale
features
[Callies
et
al.
2015].
Chl
within
submesoscale
structures
showed
the
greatest
enhancement
relative
to
the
surrounding
regions
during
the
wintertime,
with
an
average
increase
of
38%
and
a
maximum
increase
of
83%
(Fig.
3c).
The
average
impact
of
submesoscale
fronts
was
negligible
in
the
summer
and
early
autumn,
during
which
period
the
fractional
area
impacted
by
submesoscale
structures
was
also
at
its
lowest.
We
hypothesize
that
the
decreased
impact
18
during
the
summertime
was
driven
by
a
deepening
of
the
nutricline
coupled
with
increased
stratification
thereby
limiting
the
ability
of
submesoscale
features
to
access
deep
nutrients
(see
Discussion
and
S7).
Ocean
eddies
play
an
important
role
in
facilitating
submesoscale
activities
due
to
baroclinic
instabilities
that
frequently
occur
in
their
vicinity
[Klein
and
Lapeyre,
2009].
A
remote
sensing
based
analysis
of
eddy
location
and
age
[Chelton
et
al.,
2011]
suggests
that
mesoscale
eddies
are
more
frequent
during
the
winter-‐‑spring
period
in
the
NPSG,
and
that
summertime
eddies
are
on
average
older
and
so
theoretically
less
energetic
(S8).
This,
combined
with
the
seasonality
in
HI10,
suggests
a
coupling
between
both
the
frequency
and
intensity
of
mesoscale
and
submesoscale
features
in
the
region.
As
the
average
radius
of
eddies
in
the
NPSG
is
estimated
to
be
~100
km
[Gaube
et
al.,
2015],
HI10
allows
us
to
separate
the
impact
of
eddy-‐‑associated
submesoscale
features
from
that
of
upwelling
in
eddy
interiors.
Thus,
the
enhancement
associated
with
elevated
HI10
is
primarily
due
to
submesoscale
dynamics
and
is
in
addition
to
the
enhancement
that
occurs
within
mesoscale
eddies.
As
the
climate
warms,
changes
in
physical
dynamics
across
many
different
scales
may
alter
nutrient
distributions
in
the
oligotrophic
ocean,
with
the
interactions
between
these
impacts
being
complex
and
difficult
to
predict.
For
example,
temperatures
in
the
upper
ocean
are
anticipated
to
rise,
which
will
enhance
stratification
and
reduce
vertical
nutrient
exchange.
However,
the
frequency
and
amplitude
of
submesoscale
processes
are
also
likely
to
be
modified,
though
the
sign
and
magnitude
of
these
changes
remain
unknown.
To
understand
the
interactions
between
these
processes
and
their
net
impact
on
phytoplankton
19
dynamics
over
a
large
domain,
we
analyzed
the
relationship
between
SST100,
Chl100,
and
@AA
@A
,
which
are
defined
as
the
average
SST,
Chl,
and
HI10
over
a
100
km
×
100
km
region
(Fig.
4).
To
isolate
the
impact
of
submesoscale
dynamics
and
remove
the
strong
relationship
between
SST
and
Chl
in
the
NPSG,
the
correlation
between
@AA
@A
and
Chl100
was
examined
at
each
SST100
level.
We
found
significant
positive
correlations
between
@AA
@A
and
Chl100
for
all
SST100
levels
and
all
seasons,
with
the
exception
of
29.2°C
during
the
summertime
potentially
due
to
limited
data.
In
addition,
these
results
suggest
that
changes
in
SST100
and
@AA
@A
have
opposite
impacts
on
Chl100
of
approximately
the
same
magnitude.
For
example,
a
moderate
change
of
@AA
@A
from
0.242
to
0.266
(indicating
intensified
submesoscale
dynamics
and
enhanced
nutrients
fluxes)
in
the
winter
at
22.14°C
results
in
an
increase
in
Chl100
of
0.015
mg
m
-‐‑3
.
This
change
is
similar
to
the
decline
in
Chl100
due
to
a
warming
of
2.41°C
(indicating
enhanced
stratification
and
reduced
nutrients
fluxes)
with
@AA
@A
remaining
at
0.242.
Similarly,
a
moderate
decline
in
submesoscale
activity
(reduced
@AA
@A
)
combined
with
an
increase
in
SST100
significantly
enhanced
the
negative
impact
of
warming
on
chlorophyll
concentrations.
These
findings
suggest
that
the
impact
of
submesoscale
dynamics
has
the
potential
to
either
compensate
or
exacerbate
nutrient
depletion
caused
by
increased
stratification
of
the
oligotrophic
ocean.
4.
Discussions
and
Implications
Vertical
exchange
of
nutrients
between
the
ocean
interior
and
upper
layers
is
critical
to
phytoplankton
growth
and
productivity.
However,
global
estimates
of
new
production
20
exceed
estimates
of
nutrients
fluxes
from
large-‐‑scale
circulations,
winter
convection,
and
mesoscale
eddies
[McGillicuddy
et
al.,
1998,
2003;
Klein
and
Lapeyre,
2009].
The
impact
of
submesoscale
physics
has
been
proposed
as
one
of
the
missing
physical
mechanisms
behind
this
imbalance
as
these
features
are
associated
with
strong
vertical
velocities
that
are
more
than
an
order
of
magnitude
greater
than
that
associated
with
large-‐‑scale
circulation
and
the
interior
of
eddies
[Thomas
et
al.,
2008].
High-‐‑resolution
surveys
have
found
efficient
vertical
exchange
of
water
properties
in
the
vicinity
of
fronts
and
eddies
where
submesoscale
features
are
prevalent
[Lima
et
al.,
2002;
Omand
et
al.,
2015].
In
the
oligotrophic
ocean
where
the
discrepancy
between
nutrient
requirements
and
replenishment
is
large
[McGillicuddy
et
al.,
1998],
it
is
of
particular
importance
to
understand
the
role
of
submesoscale
physics
in
driving
additional
vertical
nutrient
supply
and,
therefore,
enhanced
productivity.
The
Heterogeneity
Index
(HI)
provides
a
means
of
quantifying
the
impact
of
fine-‐‑scale
frontal
structures
such
as
thin
filaments,
mesoscale
frontal
jets,
and
the
peripheries
of
eddies
on
primary
production
in
this
important
region.
Our
results
demonstrate
that
submesoscale
dynamics
enhanced
the
overall
concentration
of
Chl
in
the
oligotrophic
NPSG
through
most
of
the
year.
These
findings
suggest
both
that
submesoscale
features
increased
nutrient
supply
to
the
surface
ocean
and
that
the
timescales
of
these
fluxes
exceeded
the
doubling
time
of
phytoplankton
cells.
However,
the
impact
of
submesoscale
processes
on
Chl
varied
seasonally
with
diminished
impact
during
the
summer
(Fig.
3c).
This
may
be
due
to
decreases
in
the
effectiveness
of
submesoscale
processes
in
supplying
nutrients
to
the
surface
ocean
caused
by
both
a
deepening
of
the
nutricline
and
a
strengthening
of
the
stratification
in
the
upper
ocean
21
[Mahadevan,
2016].
Specifically,
we
hypothesize
that
enhanced
winds
(maximum
during
March,
Fig.
S9)
and
weakened
stratification
during
the
late
winter
strengthened
the
vertical
motions
associated
with
submesoscale
features
and
facilitated
the
access
of
deep
nutrients
thereby
increasing
the
response
of
phytoplankton
to
these
dynamics.
Conversely,
solar
heating
stratified
the
upper
ocean
during
the
summer
and
nutrients
were
depleted
to
a
greater
depth
resulting
in
a
strong
pycnocline
lying
above
the
nutricline.
We
hypothesize
that
during
summertime
a
significant
fraction
of
submesoscale
structures
could
not
access
the
nutricline
and
thus
had
a
minimal
impact
on
nutrient
transport
and
phytoplankton
growth.
Further
in
situ
observations,
such
as
vertical
measurements
of
density
and
nutrients
made
directly
within
submesoscale
structures,
are
needed
in
order
to
understand
causative
mechanisms
behind
the
differential
impact
of
submesoscale
features
in
the
winter-‐‑spring
relative
to
the
summer.
Using
high
resolution
satellite
data
(1
km,
daily
snapshots),
we
identified
signatures
of
submesoscale
structures
as
heterogeneity
“hotspots”
and
demonstrated
that,
in
the
oligotrophic
subtropical
gyre,
increased
patchiness
in
SST
resulted
in
increased
Chl
concentrations.
However,
in
order
to
understand
the
implications
of
submesoscale
dynamics
on
phytoplankton
productivity
and
carbon
cycling,
we
rely
on
the
assumption
that
remotely
sensed
Chl
is
a
good
proxy
for
phytoplankton
biomass.
Although
we
believe
that
this
assumption
holds
true
as
a
first-‐‑order
approximation
over
large-‐‑scales,
changes
in
environmental
conditions
can
trigger
rapid
physiological
responses
in
phytoplankton,
such
as
modified
intracellular
Chl:C
ratios,
which
may
introduce
some
uncertainty
into
our
results.
Specifically,
phytoplankton
cells
typically
exhibit
significant
increases
in
Chl:C
ratio
22
with
reduced
light
levels
and/or
increased
nutrient
input
[Behrenfeld
et
al.,
2015;
Halsey
et
al.,
2015].
In
the
subtropical
gyres
where
growth
is
primarily
nutrient
limited,
the
input
of
new
nutrients
may
result
in
an
increase
in
cellular
Chl:C
and
therefore
an
increase
in
Chl
concentration
without
necessarily
a
corresponding
increase
in
biomass.
While
increases
in
Chl:C
ratio
are
typically
associated
with
concurrent
increases
in
photosynthesis
and
growth
rates
[Graziano
et
al.,
1996;
Moore
et
al.,
2008;
Li
et
al.,
2015],
such
variations
in
phytoplankton
Chl:C
ratio
may
contribute
significantly
to
the
observed
increase
in
Chl
and
cloud
our
interpretation
of
changes
in
phytoplankton
biomass
and
productivity
associated
with
submesoscale
features.
Additional
work
is
needed
to
better
understand
how
changes
in
nutrient
stoichiometry,
photo-‐‑acclimation,
and
community
composition
impact
variability
in
Chl:C
ratio
[Behrenfeld
et
al.,
2015].
Furthermore,
satellite
records
only
capture
changes
in
the
surface
ocean
and
are
not
fully
indicative
of
water
column
properties.
As
such,
a
necessary
next
step
is
to
merge
satellite
observations
that
resolve
surface
properties
with
in
situ
(e.g.
gliders
and
floats)
profiles
that
diagnose
vertical
dynamics
in
order
to
fully
understand
the
role
of
fine-‐‑scale
processes
in
determining
depth-‐‑integrated
primary
and
export
production.
The
submesoscale
has
been
largely
ignored
by
the
current
generation
of
global
climate
models.
While
these
models
are
powerful
tools
for
exploring
the
impacts
of
large-‐‑scale
climate-‐‑driven
processes
on
marine
biota,
they
are
typically
run
at
coarse
resolutions
(1-‐‑3°)
due
to
computational
constraints
and
thus
only
represent
the
mean
fields
of
a
resource
environment
which,
in
reality,
includes
a
great
deal
of
spatial
and
temporal
heterogeneity
over
much
finer
scales.
Our
findings
provide
observational
evidence
that
fine-‐‑scale
23
processes
may
play
a
significant
role
in
modulating
phytoplankton
growth
and
biomass
distributions
in
the
oligotrophic
ocean,
and
that
the
magnitude
of
the
biological
response
is
comparable
to
that
of
a
warmer,
more
stratified
ocean.
These
results
provide
a
first-‐‑order
estimate
of
fine-‐‑scale
bio-‐‑physical
interactions
that
have
been
previously
under-‐‑determined
by
in
situ
observations.
While
this
study
has
exclusively
focused
on
the
subtropical
gyres,
expanding
this
analysis
to
other
oceanographic
regimes
may
provide
a
means
for
parameterizing
coarse
resolution
global
climate
models
for
the
impact
of
fine-‐‑scale
bio-‐‑
physical
interactions,
ultimately
improving
our
understanding
of
the
response
of
marine
ecosystems
to
future
climate
changes.
Acknowledgments
All
data
for
this
study
are
openly
accessible
from
NASA
OB.DAAC
(http://oceancolor.gsfc.nasa.gov/),
SIO/UCSD
(http://mixedlayer.ucsd.edu),
NOAA/ESRL/PSD
(http://www.esrl.noaa.gov/psd/data/reanalysis/reanalysis.shtml),
and
Chelton
et
al.
(2011,
http://cioss.coas.oregonstate.edu/eddies/).
We
acknowledge
funding
support
from
NSF
(OCE-‐‑RIG
1323319),
NASA
(NNX14AK76H),
and
the
University
of
Southern
California.
We
also
thank
A.
Mahadevan
and
the
two
anonymous
reviewers
for
their
assistance
in
improving
this
manuscript.
24
References
Behrenfeld,
M.
J.,
R.
T.
O’Malley,
E.
S.
Boss,
T.
K.
Westberry,
J.
R.
Graff,
K.
H.
Halsey,
A.
J.
Milligan,
D.
A.
Siegel,
M.
B.
Brown
(2015).
Revaluating
ocean
warming
impacts
on
global
phytoplankton.
Nature
Clim.
Change,
doi:10.1038/nclimate2838.
Callies,
J.,
R.
Ferrari,
J.
M.
Klymak,
and
J.
Gula
(2015),
Seasonality
in
submesoscale
turbulence,
Nat.
Commun.,
6,
6862.
Capet,
X.,
J.
C.
McWilliams,
M.
J.
Molemaker,
and
A.
F.
Shchepetkin
(2008),
Mesoscale
to
submesoscale
transition
in
the
California
current
system.
Part
II:
Frontal
processes,
J.
Phys.
Oceanogr.,
38,
44-‐‑64.
Cayula,
J.
F.
and
P.
Cornillon
(1992),
Edge
detection
algorithm
for
SST
images.
J.
Atmos.
Oceanic
Technol.,
9,
67–80.
Chelton,
D.
B.,
M.
G.
Schlax,
and
R.
M
Samelson
(2011),
Global
observations
of
nonlinear
mesoscale
eddies.
Prog.
Oceanogr.,
91,
167-‐‑216.
Corno,
G.,
D.
M.
Karl,
M.
J.
Church,
R.
M.
Letelier,
R.
Lukas,
R.
R.
Bidigare,
and
M.
R.
Abbott
(2007),
Impact
of
climate
forcing
on
ecosystem
processes
in
the
North
Pacific
Subtropical
Gyre,
J.
Geophys.
Res.
Oceans,
112,
14.
Davis,
A.,
and
E.
Di
Lorenzo
(2015),
Interannual
forcing
mechanisms
of
California
Current
transports
II:
Mesoscale
eddies.,
Deep
Sea
Res.
I,
112,
31-‐‑41.
Doney,
S.
C.,
D.
M.
Glover,
S.
J.
McCue,
and
M.
Fuentes
(2003),
Mesoscale
variability
of
Sea-‐‑
viewing
Wide
Field-‐‑of-‐‑view
Sensor
(SeaWiFS)
satellite
ocean
color:
Global
patterns
and
spatial
scales,
J.
Geophys.
Res.,
108,
3024.
Gaube,
P.,
D.
B.
Chelton,
R.
M.
Samelson,
M.
G.
Schlax,
and
L.
W.
O'Neill
(2015),
Satellite
observations
of
mesoscale
eddy-‐‑Induced
Ekman
pumping,
J.
Phys.
Oceanogr.,
45,
104-‐‑
132.
Gaultier,
L.,
B.
Djath,
J.
Verron,
J.
M.
Brankart,
P.
Brasseur,
and
A.
Melet
(2014),
Inversion
of
submesoscale
patterns
from
a
high-‐‑resolution
Solomon
Sea
model:
Feasibility
assessment,
J.
Geophys.
Res.
Oceans,
119,
4520-‐‑4541.
Graziano,
L.
M.,
R.
J.
Geider,
W.
K.
W.
Li,
and
M.
Olaizola.
Nitrogen
limitation
of
North
Atlantic
phytoplankton:
analysis
of
physiological
condition
in
nutrient
enrichment
experiments.
Aquat.
Microb.
Ecol.,
11,
53-‐‑64.
Halsey,
K.H.,
and
B.
M.
Jones
(2015).
Phytoplankton
strategies
for
photosynthesis
energy
allocation.
Ann.
Rev.
Mar.
Sci.,
7,
265-‐‑297.
Hilligsoe,
K.
M.,
K.
Richardson,
J.
Bendtsen,
L.
L.
Sorensen,
T.
G.
Nielsen,
and
M.
M.
Lyngsgaard
(2011),
Linking
phytoplankton
community
size
composition
with
temperature,
plankton
food
web
structure
and
sea-‐‑air
CO2
flux,
Deep
Sea
Res.
I,
58,
826-‐‑838.
25
Hogg,
A.
McC.,
M.
P.
Meredith,
D.
P.
Chambers,
E.
P.
Abrahamsen,
C.
W.
Hughes,
and
A.
K.
Morrison
(2015),
Recent
trends
in
the
Southern
Ocean
eddy
field,
J.
Geophys.
Res.
Oceans,
120,
257–267.
Johnson,
K.
S.,
S.
C.
Riser,
and
D.
M.
Karl
(2010),
Nitrate
supply
from
deep
to
near-‐‑surface
waters
of
the
North
Pacific
subtropical
gyre,
Nature,
465,
1062-‐‑1065.
Kahru,
M.,
E.
Di
Lorenzo,
M.
Manzano-‐‑Sarabia
and
B.
G.
Mitchell
(2012),
Spatial
and
temporal
statistics
of
sea
surface
temperature
and
chlorophyll
fronts
in
the
California
Current,
J.
Plankton
Res.,
34,
749-‐‑760.
Karl,
D.
M.,
J.
R.
Christian,
J.
E.
Dore,
D.
V.
Hebel,
R.
M.
Letelier,
L.
M.
Tupas,
and
C.
D.
Winn
(1996),
Seasonal
and
interannual
variability
in
primary
production
and
particle
flux
at
Station
ALOHA,
Deep
Sea
Res.
II,
43,
539-‐‑568.
Klein,
P.,
and
G.
Lapeyre
(2009),
The
oceanic
vertical
pump
induced
by
mesoscale
and
submesoscale
turbulence,
Ann.
Rev.
Mar.
Sci.,
1,
351-‐‑375.
Levy,
M.,
L.
Resplandy,
and
M.
Lengaigne
(2014),
Oceanic
mesoscale
turbulence
drives
large
biogeochemical
interannual
variability
at
middle
and
high
latitudes,
Geophys.
Res.
Lett.,
41,
2467-‐‑2474.
Levy,
M.,
R.
Ferrari,
P.
J.
S.
Franks,
A.
P.
Martin,
and
P.
Riviere
(2012),
Bringing
physics
to
life
at
the
submesoscale,
Geophys.
Res.
Lett.,
39,
13.
Levy,
M.,
P.
Klein,
A.
M.
Treguier,
D.
Iovino,
G.
Madec,
S.
Masson,
and
K.
Takahashi
(2010),
Modifications
of
gyre
circulation
by
sub-‐‑mesoscale
physics,
Ocean
Model.,
34,
1-‐‑15.
Li,
Q.,
L.
Legendre,
and
N.
Jiao
(2015).
Phytoplankton
responses
to
nitrogen
and
iron
limitation
in
the
tropical
and
subtropical
Pacific
Ocean.
J.
Plankton
Res.,
37,
306-‐‑319.
Li,
W.
K.
W.,
F.
A.
McLaughlin,
C.
Lovejoy,
and
E.
C.
Carmack
(2009),
Smallest
algae
thrive
as
the
Arctic
Ocean
freshens,
Science,
326,
539-‐‑539.
Lima,
I.
D.,
D.
B.
Olson,
and
S.
C.
Doney
(2002),
Biological
response
to
frontal
dynamics
and
mesoscale
variability
in
oligotrophic
environments:
Biological
production
and
community
structure,
J.
Geophys.
Res.
Oceans,
107,
21.
Lomas,
M.
W.,
D.
K.
Steinberg,
T.
Dickey,
C.
A.
Carlson,
N.
B.
Nelson,
R.
H.
Condon,
and
N.
R.
Bates
(2010),
Increased
ocean
carbon
export
in
the
Sargasso
Sea
linked
to
climate
variability
is
countered
by
its
enhanced
mesopelagic
attenuation,
Biogeosciences,
7,
57-‐‑
70.
Mahadevan,
A.,
and
D.
Archer
(2000),
Modeling
the
impact
of
fronts
and
mesoscale
circulation
on
the
nutrient
supply
and
biochemistry
of
the
upper
ocean.
J.
Geophys.
Res.,
105,
1209–25
Mahadevan,
A.,
and
A.
Tandon
(2006),
An
analysis
of
mechanisms
for
submesoscale
vertical
motion
at
ocean
fronts,
Ocean
Model.,
14,
241-‐‑256.
Mahadevan,
A.
(2016),
The
impact
of
submesoscale
physics
on
primary
productivity
of
plankton,
Ann.
Rev.
Mar.
Sci.,
8,
161-‐‑184.
26
Matear,
R.
J.,
M.
A.
Chamberlain,
C.
Sun,
and
M.
Feng
(2013),
Climate
change
projection
of
the
Tasman
Sea
from
an
Eddy-‐‑resolving
Ocean
Model,
J.
Geophys.
Res.
Oceans,
118,
2961-‐‑
2976.
McClain,
C.
R.
(2009),
A
decade
of
satellite
ocean
color
observations,
Ann.
Rev.
Mar.
Sci.,
1,
19-‐‑
42.
McGillicuddy,
D.
J.,
A.
R.
Robinson,
D.
A.
Siegel,
H.
W.
Jannasch,
R.
Johnson,
T.
D.
Dickey,
J.
McNeil,
A.F.
Michaels,
and
A.
H.
Knap
(1998),
Influence
of
mesoscale
eddies
on
new
production
in
the
Sargasso
Sea.
Nature,
394,
263-‐‑265.
McGillicuddy,
D.
J.,
L.
A.
Anderson,
S.
C.
Doney,
and
M.
E.
Maltrud
(2003),
Eddy-‐‑driven
sources
and
sinks
of
nutrients
in
the
upper
ocean:
result
from
a
0.1
°
resolution
model
of
the
north
atlantic.
Glob.
Geochem.
Cycles,
17,
1035.
Moore
C.
M.,
M.
M.
Mills,
R.
Langlois,
A.
Milne,
E.
P.
Achterberg,
J.
La
Roche,
and
R.
J.
Geider
(2008).
Relative
influence
of
nitrogen
and
phosphorous
availability
on
phytoplankton
physiology
and
productivity
in
the
oligotrophic
sub-‐‑tropical
North
Atlantic
Ocean.
Limnol.
Oceanogr.,
53,
291-‐‑305.
NASA
Ocean
Biology
(OB.DAAC).
Moderate
Resolution
Imaging
Spectroradiometer
(MODIS)
Ocean
Color
Data,
2014.0
Reprocessing.
NASA
OB.DAAC,
Greenbelt,
MD,
USA.
http://oceancolor.gsfc.nasa.gov,
accessed
on
2015/08/03.
Niewiadomska,
K.,
H.
Claustre,
L.
Prieur,
and
F.
d'Ortenzio
(2008),
Submesoscale
physical-‐‑
biogeochemical
coupling
across
the
Ligurian
Current
(northwestern
Mediterranean)
using
a
bio-‐‑optical
glider,
Limnol.
Oceanogr.,
53,
2210-‐‑2225.
Omand,
M.
M.,
E.
A.
D'Asaro,
C.
M.
Lee,
M.
J.
Perry,
N.
Briggs,
I.
Cetinic.,
A.
Mahadevan
(2015),
Eddy-‐‑driven
subduction
exports
particulated
organic
carbon
from
the
spring
bloom,
Science,
348,
222-‐‑225.
Thomas
L.N.,
A.
Tandon,
and
A.
Mahadevan
(2008),
Submesoscale
processes
and
dynamics.
In
Ocean
Modeling
in
an
Eddying
Regime,
ed.
MW
Hecht,
H
Hasumi,
pp.
17–38.
Geophys.
Monogr.
Vol.
177.
Washington,
DC:
Am.
Geophys.
Union
Vecchi,
G.
A.,
B.
J.
Soden,
A.
T.
Wittenberg,
I.
M.
Held,
A.
Leetmaa,
and
M.
J.
Harrison
(2006),
Weakening
of
tropical
Pacific
atmospheric
circulation
due
to
anthropogenic
forcing,
Nature,
441,
73-‐‑76.
Werdell,
P.
J.,
and
S.
W.
Bailey
(2005),
An
improved
in-‐‑situ
bio-‐‑optical
data
set
for
ocean
color
algorithm
development
and
satellite
data
product
validation,
Remote
Sens.
Environ.,
98,
122-‐‑140.
Woodson,
C.
B.,
and
S.
Y.
Litvin
(2015),
Ocean
fronts
drive
marine
fishery
production
and
biogeochemical
cycling,
Proc.
Natl.
Acad.
Sci.
USA,
112,
1710-‐‑1715.
Xiu,
P.,
and
F.
Chai
(2012),
Spatial
and
temporal
variability
in
phytoplankton
carbon,
chlorophyll,
and
nitrogen
in
the
North
Pacific,
J.
Geophys.
Res.
Oceans,
117,
17.
27
Figure
Captions
Figure
1.
Feature
identification
using
the
Heterogeneity
Index.
An
example
of
MODIS-‐‑Aqua
SST
image
from
04/04/2003
is
shown.
Sub-‐‑regions
(10
km
×
10
km)
associated
with
submesoscale
structures
are
identified
by
high
HI10
values
while
the
background
field
is
characterized
by
low
HI10.
Figure
2.
Weekly
climatologies
of
SST,
Chl,
and
submesoscale
heterogeneity
(HI10)
averaged
over
the
study
region
(July
2002
to
June
2015).
Error
bars
represent
±0.25σ
from
the
mean.
Figure
3.
Impact
of
submesoscale
heterogeneity
(HI10)
on
SST
and
Chl.
Panel
a)
shows
the
fractional
area
impacted
by
submesoscale
features
(high
HI10).
Panels
b)
and
c)
show
the
difference
in
SST
and
Chl
between
the
background
field
and
the
feature-‐‑impacted
regions.
In
all
panels,
the
central
mark
of
each
box
plot
is
the
median,
edges
of
the
box
are
the
25
th
and
75
th
percentiles,
and
whiskers
extend
to
the
most
extreme
data
points
excluding
outliers
which
are
denoted
by
red
+.
The
solid
and
dashed
lines
are
generated
using
a
3-‐‑point
moving
average
filter.
Note
the
changes
in
y-‐‑axis
scales
for
panels
a)
and
c).
Figure
4.
Impact
of
SST
and
submesoscale
heterogeneity
(HI10)
on
Chl.
SST100,
Chl100,
and
@AA
@A
are
defined
as
the
average
SST,
Chl,
and
HI10
over
100
km
×
100
km
regions.
Results
are
presented
by
season
with
bins
colored
by
Chl100.
Chl100
increases
with
decreasing
SST100
(horizontal
axis)
and
increasing
@AA
@A
(vertical
axis).
The
significance
of
the
positive
relationship
between
@AA
@A
and
Chl100
for
each
SST100
bin
is
shown
on
top
of
each
column
with
p
<
0.01
denoted
by
two
stars
(**)
and
p
<
0.05
denoted
by
one
star
(*).
White
bins
indicate
conditions
where
less
than
15
images
(100
km
×
100
km
with
good
spatial
coverage)
were
available.
The
arrows
demonstrate
the
comparable,
and
compensating,
change
in
Chl100
(0.015
mg
m
-‐‑3
)
that
would
result
from
a
moderate
increase
in
@AA
@A
from
0.242
to
0.266
(solid
arrow)
versus
a
warming
of
the
same
waters
by
2.41°C
(dashed
arrow).
21 21.5 22
Frequency
0
5
10
15
20
25
30
20 20.5 21 21.5
Frequency
0
5
10
15
20
25
30
SST (
o
C)
20 20.5 21 21.5
Frequency
0
5
10
15
20
25
30
21.5 22 22.5 23
Frequency
0
5
10
15
20
25
30
HI10 = 0.13
HI10 = 0.34 HI10 = 0.73 HI10 = 0.64
‘feature-impacted’ regions (high HI10)
‘background’ field (low HI10) MODIS-Aqua SST
SST (
o
C)
SST (
o
C)
SST (
o
C)
Figure 1. Feature identication using the Heterogeneity Index. An example of
MODIS-Aqua SST image from 04/04/2003 is shown. Sub-regions (10 km × 10 km)
associated with submesoscale structures are identied by high HI
10
values while the
background eld is characterized by low HI
10.
28
Month
J F M A M J J A S O N D
Heterogeneity Index (HI
10
)
0.21
0.22
0.23
0.24
0.25
0.26
0.27
0.28
Month
J F M A M J J A S O N D
SST (
o
C)
24
25
26
27
28
Chl (μg L
-1
)
0.05
0.06
0.07
0.08
Figure 2. Weekly climatologies of SST, Chl, and submesoscale heterogeneity (HI
10
)
averaged over the study region (July 2002 to June 2015). Error bars represent ± 0.25
σ from the mean.
29
J F M A M J J A S O N D
Fraction of Impacted Area (%)
0
2
4
6
8
10
16
J F M A M J J A S O N D
Δ SST(
o
C)
-3
-2
-1
0
1
Month
J F M A M J J A S O N D
Δ Chlorophyll (%)
-20
0
20
40
80
b
a
c
Figure 3. Impact of submesoscale heterogeneity (HI
10
) on SST and Chl. Panel a)
shows the fractional area impacted by submesoscale features (high HI
10
). Panels b)
and c) show the dierence in SST and Chl between the background eld and the
feature-impacted regions. In all panels, the central mark of each box plot is the
median, edges of the box are the 25
th
and 75
th
percentiles, and whiskers extend to
the most extreme data points excluding outliers which are denoted by red +. The
solid and dashed lines are generated using a 3-point moving average lter. Note
the changes in y-axis scales for panels a) and c).
30
SST
100
(
o
C)
23.58 25.31 27.04 28.77
0.221
0.233
0.244
0.256
*
*
*
*
*
*
*
*
*
*
*
*
*
*
Fall
Chl
100
(mg m
- 3
)
0.041
0.050
0.061
0.074
0.091
0.111
0.135
SST
100
(
o
C)
23.54 25.42 27.31 29.19
0.212
0.230
0.247
0.265
*
*
* *
*
*
*
*
*
*
*
Summer
19.74 22.14 24.55 26.96
0.218
0.242
0.266
0.290
*
*
*
*
*
*
*
*
*
*
*
*
*
*
Spring
Chl
100
(mg m
- 3
)
0.041
0.050
0.061
0.074
0.091
0.111
0.135
20.38 22.57 24.77 26.96
0.219
0.236
0.254
0.271
*
*
*
*
*
*
*
*
*
*
*
*
*
*
Winter
Heterogeneity (HI
100
)
10
Heterogeneity (HI
100
)
10 Figure 4. Impact of SST and submesoscale heterogeneity (HI
10
) on Chl. SST
100
, Chl
100
, and
are dened as the average SST, Chl, and HI
10
over 100 km × 100 km regions.
Results are presented by season with bins colored by Chl
100
. Chl
100
increases with
decreasing SST
100
(horizontal axis) and increasing (vertical axis). The signicance
of the positive relationship between and Chl100 for each SST
100
bin is shown on
top of each column with p < 0.01 denoted by two stars (**) and p < 0.05 denoted by one
star (*). White bins indicate conditions where less than 15 images (100 km × 100 km with
good spatial coverage) were available. The arrows demonstrate the comparable, and
compensating, change in Chl
100
(0.015 mg m-3) that would result from a moderate
increase in from 0.242 to 0.266 (solid arrow) versus a warming of the same waters
by 2.41°C (dashed arrow).
31
32
Geophysical
Research
Letters
Supporting
Information
for
Enhancement
of
phytoplankton
chlorophyll
by
submesoscale
frontal
dynamics
in
the
North
Pacific
Subtropical
Gyre
Xiao
Liu
1
and
Naomi
Marcil
Levine
2
,*
1
Department
of
Earth
Sciences,
University
of
Southern
California,
Los
Angeles,
USA
2
Department
of
Biological
Sciences,
University
of
Southern
California,
Los
Angeles,
USA
Contents
of
this
file
1.
Text
S1
to
S8
2.
Figures
S1
to
S11
Introduction
Section
S1
includes
a
map
of
the
study
region
in
the
North
Pacific
Subtropical
Gyre
(Figure
S1)
and
describes
the
satellite
data
used
for
our
analyses
(Figure
S2)
and
the
procedure
for
data
quality
control.
Section
S2
provides
detailed
information
on
the
formulation
of
the
Heterogeneity
Index
(HI).
Since
the
normalization
coefficients
in
the
equation
are
region-‐‑
specific,
we
provide
a
look-‐‑up
table
(Table
S1)
with
the
normalization
coefficients
for
the
major
subtropical
ocean
gyres,
and
Figure
3S
shows
maps
of
the
applicable
regions
in
these
gyres.
We
provided
a
few
examples
of
satellite
images
and
their
corresponding
HI
values
in
Figure
S4.
Section
S3
and
Figures
S5-‐‑6
demonstrate
in
further
detail
the
statistical
method
employed
to
identify
regions
impacted
by
submesoscale
physics,
and
how
SST
and
Chl
differ
in
these
regions
relative
to
the
surrounding
homogenous
fields.
Section
S4
and
Figure
S7
focus
on
the
skewness
component
of
the
HI
equation
and
discuss
the
different
impacts
of
a
negatively
and
a
positively
skewed
SST
distribution
on
Chl.
Section
S5
and
Figure
S8
demonstrate
that
the
reduced
33
number
of
summertime
satellite
images
due
to
increased
cloud
coverage
does
not
introduce
bias
into
our
results.
Sections
S6-‐‑8
and
Figures
S9-‐‑11
show
the
seasonal
dynamics
of
surface
winds,
mixed
layer
depth,
nutricline
depth,
and
vertical
density
gradient,
as
well
as
the
frequency
and
age
of
mesoscale
eddies
in
the
study
region.
These
sections
also
include
a
discussion
of
how
these
properties
might
relate
to
the
submesoscale
physics
and
the
seasonality
of
chlorophyll
response.
Text
S1.
Satellite
data
processing
and
quality
control
MODIS/Aqua
Level-‐‑2
swaths
of
Sea
Surface
Temperature
(SST)
and
chlorophyll
concentration
(Chl)
were
retrieved
from
NASA
OB.DAAC
and
processed
for
the
region
(10-‐‑30°N,
160°E-‐‑
160°W;
Figure
S1)
using
EXELIS
IDL.
The
default
MODIS/Aqua
Chl
algorithm
OC3
was
used,
and
the
standard
pixel
size
for
the
mapped
products
was
~1.04
km
after
processing.
A
filtering
grid
with
a
cell
size
of
96
×
96
pixels
(or
100
km
×
100
km)
was
applied,
and
images
were
extracted
from
the
grid
cells
where
the
spatial
coverage
was
greater
or
equal
to
75%
for
SST
and
70%
for
Chl.
Extracted
data
were
then
flagged
‘selected’
on
the
original
swaths.
This
procedure
was
repeated
a
total
of
64
times
for
each
swath
by
shifting
the
grid
eastward
or
southward
at
increments
of
12
pixels
or
1/8
of
a
grid
cell
each
time.
Only
the
extracted
data
with
high
coverage
(a
total
of
32,222
images;
Fig.
S2)
were
used
for
our
analyses.
The
purpose
of
this
filtering
was
to
minimize
the
artificial
spatial
patchiness
caused
by
missing
pixels
(e.g.
aerosol-‐‑
and
cloud-‐‑contaminated).
The
average
coverage
for
the
extracted
data
was
86%
for
SST
and
79%
for
Chl.
The
Heterogeneity
Index
(HI)
was
then
computed
for
the
full
domain
and
for
subdomains
of
a
variety
of
spatial
scales,
such
as
50
km
×
50
km.
Here
we
focus
on
submesoscale
dynamics
and
so
only
present
results
of
HI10
(10
km
×
10
km).
Text
S2.
Heterogeneity
Index
A
new
metric
of
spatial
patchiness
(the
Heterogeneity
Index,
HI)
was
developed,
which
combines
measures
of
data
variance
and
normalcy
of
a
distribution.
HI
was
designed
and
validated
for
the
oligotrophic
ocean
where,
as
a
first-‐‑order
assumption,
SST
can
be
used
as
a
proxy
for
density.
HI
has
not
been
applied
to
other
regions
and
significant
testing
would
need
to
be
done
before
it
can
be
used
in
other
oceanographic
regimes.
Specifically,
HI
calculated
34
using
remotely
sensed
Sea
Surface
Height
data
maybe
more
applicable
in
higher
latitude
regions.
HI
is
defined
as:
eq.
(1)
where
the
first
term
quantifies
the
skewness
of
the
distribution,
the
second
term
quantifies
the
data
variance,
and
the
third
terms
describes
how
this
distribution
deviates
from
a
Gaussian
distribution.
We
calculated
HI
by
first
generating
a
probability
distribution
was
generated
using
1
km
SST
extracted
from
an
individual
image
(see
examples
in
Fig.
S4).
The
width
of
each
bin
was
set
to
0.1
(°C),
and
the
number
of
bins
was
determined
according
to
the
range
of
SST.
We
then
computed
the
following:
1)
The
absolute
value
of
the
skewness
of
the
distribution,
,
which
is
a
measure
of
the
asymmetry
of
the
data
around
the
sample
mean
.
The
skewness
relates
to
the
third
moment
of
the
distribution,
and
is
defined
as:
eq.
(2)
2)
The
standard
deviation
()
of
the
sample
(e.g.
SST)
distribution
normalized
to
the
sample
size
,
which
represents
the
average
gradient
in
SST.
3)
The
polygon
area
(P),
which
is
defined
as
the
difference
between
a
Gaussian
distribution
with
the
same
sample
mean
()
and
standard
deviation
()
[(,)
in
equation
(3)]
and
the
area
of
the
best
5
th
order
polynomial
fit
to
the
data
[()
in
equation
(3)]
integrated
over
the
range
of
x.
=
- . /0(1,2)
0(1,2)
345
(.)
367
(.)
eq.
(3)
This
component
quantifies
how
far
the
distribution
is
from
normal
with
a
particular
emphasis
on
bimodal
and
multimodal
distributions.
The
5
th
order
polynomial
was
created
using
the
MATLAB
built-‐‑in
command
polyfit
function.
In
order
to
equally
weight
each
component
of
HI,
the
three
components
were
calculated
independently
for
each
pixel
and
then
scaled
between
0
and
1:
HI =a(bγ +c
σ
n
+dP)
γ =
(x
i
−µ)
3
σ
3
1
n
i=1
n
∑
n
35
eq.
(4)
where
and
are
the
normalized
and
original
values
of
the
j
th
component
of
HI
(j
=
1-‐‑3)
calculated
for
the
i
th
pixel
(i
=
1
to
N,
where
N
is
the
total
number
of
pixels
from
all
images,
~3
×
10
8
),
respectively,
and
and
are
the
maximum
and
minimum
values
for
the
j
th
component
of
HI,
respectively.
A
normalization
coefficient
was
then
estimated
for
each
component
of
HI
as
the
slope
of
the
best
fit
linear
regression
(forced
through
the
origin)
between
the
normalized
and
original
values
(for
the
NPSG:
b
=
1.07,
c
=
1.81,
d
=
1.11).
Finally,
the
normalization
coefficients
(b-‐‑d)
were
applied
to
the
HI
components,
and
the
sum
of
the
three
values
for
each
HI
was
normalized
again
(a
=
0.30)
so
that
the
final
HI
values
range
between
0
and
1.
HI
=
0
describes
a
homogenous
system
and
HI
=
1
describes
a
maximally
heterogeneous
system.
Table
S1
provides
sets
of
normalization
coefficients
(a-‐‑d)
for
the
five
major
subtropical
ocean
gyres
and
and
Fig.
S3
shows
maps
of
these
gyres.
HI
is
spatial
scale
dependent
and
can
be
used
to
quantify
surface
patchiness
at
large-‐‑,
meso-‐‑,
and
fine
scales.
Fig.
1
(main
text)
shows
examples
of
fine-‐‑scale
frontal
features
identified
by
elevated
HI
at
the
submesoscale
and
Fig.
S4
shows
additional
examples
of
different
spatial
patterns
in
SST
fields.
These
figures
highlight
the
importance
of
the
three
components
of
HI
in
capturing
the
full
range
of
ocean
patchiness.
Text
S3.
Separation
of
regions
impacted
by
submesoscale
structures
from
the
background
field.
To
identify
regions
impacted
by
submesoscale
dynamics,
we
define
‘background’
and
‘impacted’
regions
using
the
distribution
of
HI10
over
the
entire
domain.
Specifically,
for
a
given
time
interval
(i.e.
weekly
in
our
analysis),
a
frequency
distribution
was
generated
using
all
HI10
values.
The
number
of
bins
was
set
to
50,
and
the
width
of
each
bin
was
determined
according
to
the
range
of
HI10
for
the
given
week.
The
distributions
were
always
positively
skewed
with
the
mode
centered
at
low
HI10
values
and
a
long
tail
with
elevated
values.
A
typical
distribution
is
shown
in
Fig.
S5.
This
is
expected
as
fine-‐‑scale
features
are
known
to
cover
only
a
small
fractional
area
of
the
open
ocean,
despite
being
ubiquitous
and
accounting
for
much
of
the
vertical
kinetic
energy.
We
defined
the
background
field
as
the
region
associated
with
low
HI10
norm(HI
i,j
)=
HI
i,j
−HI
min,j
HI
max,j
−HI
min,j
norm(HI
i,j
) HI
i,j
HI
max,j
HI
min,j
36
values.
We
assumed
that
these
values
were
normally
distributed
around
the
sample
mode
such
that
the
background
field
was
defined
as
regions
with
a
HI10
within
2σ
from
the
mode.
σ
was
calculated
using
the
best
Gaussian
fit
to
the
data
within
κ
from
the
mode
(fitdist
in
MATLAB),
where
is
the
absolute
difference
between
the
mode
and
the
minimum
value
of
HI10.
The
area
impacted
by
submesoscale
structures
was
then
defined
as
those
regions
with
a
HI10
at
least
4σ
greater
than
the
mode.
Several different threshold values were tested; while the threshold value
impacts absolute value of the fractional area associated with submesoscale structures, it does not
impact the seasonality of HI or the chlorophyll response (i.e. greater in the late winter and
negligible during the summertime).
Therefore,
the
main
conclusions
of
this
study
are
not
impacted
by
the
choice
of
the
threshold
value.
Text
S4.
Differential
impact
of
negatively
and
positively
skewed
SST
distributions
on
Chl.
HI
uses
the
absolute
value
of
the
skewness.
This
formulation
was
chosen
to
allow
for
the
calculation
of
pixel-‐‑level
HI
that
requires
averaging
all
computed
values
for
each
HI
component
at
each
pixel.
However,
the
sign
of
the
skewness
term
potentially
contains
additional
information.
We
conducted
an
additional
analysis
to
investigate
the
differential
impact
of
negatively
and
positively
skewed
SST
distributions
on
Chl.
We
hypothesize
that
these
results
are
due
to
difference
in
the
spatial
structures
of
submesoscale
features.
Specifically,
we
flagged
HI10
values
based
on
the
sign
of
the
skewness
component.
Each
pixel
was
then
identified
as
high
HI10
with
negative
skewness
(more
than
75%
skewness
values
are
negative),
high
HI10
with
positive
skewness
(more
than
75%
skewness
values
are
positive),
or
high
HI10
with
neutral
skewness.
We
then
re-‐‑ran
our
analysis
on
these
three
sub-‐‑datasets
to
differentiate
between
the
impact
of
negatively
and
positively
skewed
SST
distributions
on
chlorophyll.
We
found
a
very
small
(statistically
insignificant)
increase
in
the
impact
of
negatively
skewed
distributions
relative
to
positively
skewed
distributions
(Fig.
S7).
Both
positively
and
negatively
skewed
distributions
resulted
in
the
same
seasonal
cycle
as
the
full
dataset.
This
is
consistent
with
our
findings
from
visual
examination
of
paired
satellite
images
of
SST
and
Chl.
In
the
open
ocean,
submesoscale
structures
associated
with
high
HI
values
include
1)
submesoscale
filaments,
2)
parts
of
meso-‐‑
and
large-‐‑scale
frontal
features,
and
3)
edges
of
mesoscale
eddies
(provided
these
edges
and
fronts
are
associated
with
sharp
κ
37
density/temperature
gradients
less
than
10
km
in
width).
In
case
1,
submesoscale
processes
that
result
in
upwelling
of
cooler,
potentially
nutrient-‐‑rich,
waters
to
the
surface
will
lead
to
a
negatively
skewed
distribution
of
SST.
In
this
case,
we
would
expect
to
observe
enhanced
chlorophyll
in
response
to
negatively
skewed
distributions
of
SST.
We
would
not
expect
to
see
chlorophyll
responses
to
warm
filaments,
which
would
result
in
positively
skewed
distributions.
However,
in
cases
2
and
3,
the
opposite
relationship
is
expected:
positively
(negatively)
skewed
distributions
are
expected
on
the
cooler
(warmer),
theoretically
nutrient-‐‑
rich
(poor),
sides
of
the
fronts.
The
history
of
larger-‐‑scale
features
and
temporal
and
spatial
variations
in
vertical
physical
dynamics
will
also
impact
the
magnitude
of
the
signature.
In
addition,
any
horizontal
transport
could
act
to
re-‐‑distribute
some
upwelled
nutrients
from
the
cold
to
the
warm
side
of
the
front.
In
fact,
visual
examination
of
paired
satellite
images
of
SST
and
Chl
has
identified
many
examples
of
submesoscale
fronts
(case
2
or
3)
that
result
in
elevated
chlorophyll
concentrations
along
both
sides
of
the
front.
While
our
analysis
confirms
that
the
negatively
skewed
SST
distributions
indeed
have
a
slightly
greater
impact
(as
expected
in
case
1),
this
difference
was
not
statistically
significant
as
is
expected
due
to
the
inclusion
of
cases
2
and
3.
Text
S5.
Data
limitation
during
the
summertime
in
the
NPSG.
In
the
NPSG,
cloud
coverage
significantly
limits
the
overall
quality
of
ocean
color
satellite
images
during
the
summertime,
resulting
in
a
large
discrepancy
in
total
number
of
observations
between
summertime
and
wintertime
(Fig.
S2).
We
conducted
an
additional
analysis
to
investigate
the
impact
of
this
discrepancy
on
our
findings,
and
demonstrated
that
the
reduced
number
of
images
during
the
summertime
does
not
bias
our
results.
Specifically,
we
randomly
subsampled
our
data
to
create
artificial
datasets
with
an
equal
number
of
images
for
each
week.
Specifically,
we
stochastically
selected
55
images
from
each
week,
where
n
=
55
is
the
minimum
number
of
images
available
during
summer
weeks.
This
subsampling
was
repeated
5
times
to
generate
5
independent
subsets.
We
then
re-‐‑ran
our
analyses
on
these
5
subsets.
The
results
from
this
exercise
(Fig.
S8)
are
consistent
with
the
findings
using
the
complete
dataset
(Fig.
3c
in
the
main
text)
indicating
that
the
reduced
number
of
images
during
the
summertime
does
not
bias
our
results.
38
Text
S6.
Seasonal
dynamics
of
Argo
mixed
layer
depth
and
NCEP
reanalysis
winds
in
the
NPSG.
The
monthly
climatology
of
mixed
layer
depth
for
the
study
region
(10-‐‑30°N,
160°E-‐‑160°W;
Fig.
S1)
was
obtained
directly
from
the
Talley
Lab
at
Scripps
Institution
of
Oceanography/UCSD
(http://mixedlayer.ucsd.edu).
This
product
was
determined
using
Argo
profiles
and
a
hybrid
method
[Holte
et
al.,
2010].
The
monthly
climatology
of
wind
speed
for
the
region
was
derived
from
NCEP/NCAR
Reanalysis
monthly
means
on
a
2.5
degree
grid
(Kalnay
et
al.,
1996;
http://www.esrl.noaa.gov/psd/data/reanalysis/;
Fig.
S9).
Text
S7.
Seasonal
dynamics
of
vertical
density
gradients
and
depth
of
nutricline
in
the
NPSG.
Vertical
profiles
of
density
anomalies
and
nitrate
concentrations
for
the
study
region
(10-‐‑30°N,
160°E-‐‑160°W)
were
determined
using
MBARI
Apex/ISUS
profiling
float
data
from
2009
to
2015
(accessed
on
2015/08/03
from
http://www.mbari.org/chemsensor/floatviz.htm).
Negative
values
of
nitrate
concentration
were
first
set
to
zero,
and
the
depth
of
nitracline
(nutricline)
was
determined
as
the
shallowest
depth
below
which
the
nitrate
concentration
was
consistently
greater
than
0.5
μM.
The
vertical
density
gradient
between
a
stable
surface
layer
(10
m)
and
the
nutricline
depth,
Δσ,
was
then
calculated
for
each
profile.
This
term
is
used
as
a
proxy
for
accessibility
of
deep
nutrients,
with
a
greater
Δσ
suggesting
that
more
energy
was
required
to
generate
sufficient
vertical
velocities
to
access
the
nutricline.
The
monthly
climatologies
of
nutricline
depth
and
Δσ
are
shown
in
Fig.
S10.
Text
S8.
Seasonal
dynamics
of
mesoscale
eddies
in
the
NPSG.
Mesoscale
ocean
eddies
play
an
important
role
in
facilitating
the
formation
of
submesoscale
instabilities
[Klein
and
Lapeyre,
2009].
An
understanding
of
the
dynamics
of
mesoscale
eddies
may,
therefore,
provide
crucial
information
on
the
spatial
and
temporal
variations
in
submesoscale
activities.
Using
the
analysis
of
Chelton
et
al.
[2011]
that
identified
and
tracked
mesoscale
eddies
from
Sea
Surface
Height
anomalies,
we
quantified
the
abundance
and
age
of
eddies
within
our
study
region.
We
found
that
during
the
early
spring
(i.e.
March)
mesoscale
39
eddy
abundance
was
at
its
peak
and
that
these
eddies
were
in
general
younger.
Conversely,
late
summertime
(August
to
September)
in
the
NPSG
typically
had
fewer
and
older
(less
energetic)
eddies
(Fig.
S11).
This
is
in
agreement
with
the
observed
seasonal
variability
of
submesoscale
heterogeneity
as
estimated
by
HI,
suggesting
a
coupling
between
both
the
frequency
and
intensity
of
mesoscale
and
submesoscale
features
in
the
region.
40
Table
S1.
Normalization
coefficients
(a-‐‑d)
in
eq(1)
for
the
major
subtropical
ocean
gyres:
the
North
Pacific
Subtropical
Gyre
(NPSG,
this
study),
the
North
Atlantic
Subtropical
Gyre
(NASG),
the
South
Pacific
Subtropical
Gyre
(SPSG),
the
South
Atlantic
Subtropical
Gyre
(SASG),
and
the
Indian
Ocean
(IO).
a
b
c
d
NPSG
10-‐‑30°N,
160°E-‐‑160°W
0.3036
1.0650
1.8137
1.1116
NASG
15-‐‑30°N,
30-‐‑60°W
0.2803
0.8728
1.7976
1.3178
SPSG
18-‐‑38°N,
120-‐‑160°W
0.2955
1.1142
2.1597
1.1902
SASG
5-‐‑35°S,
0-‐‑35°W
0.2836
0.8386
2.1533
1.1810
IO
12-‐‑34°S,
65-‐‑105°W
0.3038
1.0983
1.9949
1.1622
41
Figure
S1.
Map
of
the
study
region
in
the
North
Pacific
Subtropical
Gyre
(10-‐‑30°N,
160°E-‐‑
160°W).
Color
shows
the
winter
climatology
of
MODIS-‐‑Aqua
chlorophyll
a
concentration
during
the
13-‐‑year
period
of
July
2002
through
June
2015.
42
Figure
S2.
Number
of
extracted,
high-‐‑coverage
images
(100
km
×
100
km)
used
in
our
analyses.
Total
images
for
each
week
during
the
13
year
period
(2002-‐‑2015)
are
shown.
The
decrease
in
good
quality
images
during
the
summertime
is
attributed
to
seasonality
in
cloud
coverage
in
this
region.
Month
J F M A M J J A S O N D
Total number of images (100km)
0
200
400
600
800
1000
1200
1400
1600
1800
2000
43
Figure
S3.
Maps
of
the
analyzed
regions
in
the
a)
North
Atlantic
Subtropical
Gyre,
b)
South
Pacific
Subtropical
Gyre,
c)
South
Atlantic
Subtropical
Gyre,
and
d)
Indian
Ocean.
Color
shows
the
winter
climatology
of
MODIS/Aqua
chlorophyll
a
concentration
during
the
13-‐‑year
period
of
July
2002
through
June
2015.
44
Figure
S4.
Satellite
images
(200
km
×
200
km)
of
SST
associated
with
low
HI
(a),
intermediate
HI
(b,
c,
d),
and
high
HI
(e).
Image
a)
shows
a
quiescent,
normally
distributed,
nearly
homogenous
field
with
a
small
temperature
gradient.
Image
b)
displays
a
negatively
skewed
distribution
due
to
a
distinct
patch
of
cooler
water
that
elevates
HI,
despite
a
low
standard
deviation
of
temperature.
Image
c)
has
a
normally
distributed
temperature
field
yet
is
characterized
by
elevated
HI
as
a
result
of
a
higher
standard
deviation.
Image
d)
captures
two
water
masses
with
a
sharp
frontal
gradient
that
results
in
a
bimodal
distribution
and
therefore
higher
HI.
Image
e)
is
associated
with
the
highest
HI,
which
is
a
result
of
not
only
the
largest
standard
deviation
but
also
a
highly
skewed
distribution
and
the
presence
of
two
water
masses.
45
Figure
S5.
Selection
of
impacted
versus
background
regions.
An
example
of
a
typical
distribution
of
HI10
over
the
entire
region
for
a
single
week
is
shown.
The
background
field
(±2σ
from
the
mode)
is
highlighted
in
blue
and
regions
impacted
by
submesoscale
structures
(>=
4σ
from
the
mode)
are
highlighted
in
red.
Submesoscale Heterogeneity (HI
10
)
0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55
Frequency
0
100
200
300
400
500
600
700
800
900
1000
Background Impacted regions
46
Figure
S6.
Impact
of
submesoscale
physics
(elevated
HI
10
)
on
temperature
(SST)
and
chlorophyll
(Chl).
Panels
a),
b)
and
c)
show
HI
10
,
SST,
and
Chl
in
the
impacted
area
(solid
lines)
and
in
the
background
field
(dashed
lines).
Panel
d)
shows
the
fraction
of
Chl
located
within
the
submesoscale
structures,
calculated
as
Chl
i
/Chl
T
,
where
Chl
i
is
the
mean
Chl
within
these
structures
multiplied
by
their
fractional
area,
and
Chl
T
is
the
overall
mean
Chl
in
the
region.
In
all
panels
the
central
mark
of
each
box
plot
is
the
median,
the
upper
and
lower
bounds
of
the
box
are
the
25
th
and
75
th
percentiles,
and
the
whiskers
extend
to
the
most
extreme
data
points
excluding
outliers
which
are
denoted
by
red
+.
The
solid
and
dashed
lines
are
generated
using
a
3-‐point
moving
average
filter.
Note
the
change
in
y-‐axis
scale
for
panel
d.
47
Figure
S7.
Impact
of
SST
skewness
on
chlorophyll
response.
The
percentage
chlorophyll
increase
within
the
impacted,
high
HI10
regions
relative
to
the
background
field
is
shown
for
pixels
with
a)
negatively
and
b)
positively
skewed
distributions.
While
negatively
skewed
pixels
show
a
slightly
elevated
response,
the
increase
is
not
significant
and
the
seasonal
pattern
is
consistent
with
that
seen
in
the
full
dataset
(Fig.
3c
in
the
main
text).
48
Figure
S8.
Impact
of
summertime
data
limitation
on
chlorophyll
response.
The
percentage
chlorophyll
increase
within
the
impacted,
high
HI10
regions
relative
to
the
background
field
is
shown
using
5
independent
subsets
of
data.
Each
subset
was
generated
by
stochastically
selecting
55
images
from
each
week,
where
n
=
55
is
the
minimum
number
of
images
available
during
summer
weeks.
49
Figure
S9.
Monthly
climatologies
of
Argo
mixed
layer
depth
and
NCEP
reanalysis
winds
for
the
region
during
the
period
January
2000
through
May
2015.
Month
J F M A M J J A S O N D
MLD (m)
-80
-70
-60
-50
-40
Wind Velocity (m/s)
6
6.5
7
7.5
8
8.5
50
Figure
S10.
Monthly
climatologies
of
a)
the
nutricline
depth
and
b)
the
vertical
density
gradient
(Δσ)
between
a
stable
surface
layer
(10
m)
and
the
nutricline
for
the
study
region
during
the
period
December
2009
through
June
2015.
Error
bars
represent
±1σ
from
the
mean.
51
Figure
S11.
Monthly
climatologies
of
a)
total
number
of
mesoscale
eddies,
b)
number
of
newly
generated
eddies,
and
c)
age
of
eddies
for
the
study
region
during
the
period
October
1992
through
April
2012.
Error
bars
represent
±0.5σ
from
the
mean.
52
References:
Chelton,
D.,
M.
Schlax,
R.
Samelson
(2011),
Global
observations
of
nonlinear
mesoscale
eddies,
Prog.
Oceanogr.,
91,
167-‐‑216,
http://cioss.coas.oregonstate.edu/eddies,
accessed
on
2015/05/06.
Holte,
J.,
J.
Gilson,
T.
Talley,
D.
Roemmich
(2010),
Argo
Mixed
Layers,
Scripps
Institution
of
Oceanography/UCSD,
http://mixedlayer.ucsd.edu,
accessed
on
2015/03/15.
Kalnay,
E.,
M.
Kanamitsu,
R.
Kistler
et
al.
(1996),
The
NCEP/NCAR
40-‐‑year
reanalysis
project,
Bull.
Amer.
Meteor.
Soc.,
77,
437-‐‑470,
http://www.esrl.noaa.gov/psd/data/reanalysis/,
accessed
on
2015/03/15.
Klein,
P.,
and
G.
Lapeyre
(2009),
The
Oceanic
Vertical
Pump
Induced
by
Mesoscale
and
Submesoscale
Turbulence,
Ann.
Rev.
Mar.
Sci.,
1,
351-‐‑375.
53
Chapter
2.
Modeling
the
impact
of
fine-‐‑scale
disturbances
on
phytoplankton
community
composition
and
carbon
cycling
in
the
North
Pacific
Subtropical
Gyre
Xiao
Liu
1
and
Naomi
Marcil
Levine
2,*
1
Department
of
Earth
Sciences,
University
of
Southern
California,
Los
Angeles
90089
2
Department
of
Biological
Sciences,
University
of
Southern
California,
Los
Angeles
90089
*:
Corresponding
author;
AHF
M225,
3616
Trousdale
Pkwy,
Los
Angeles,
CA
90089;
n.levine@usc.edu;
+1
(213)
821-‐‑0745
Formatted
for
submission
to
Global
Change
Biology
54
Abstract
In
the
subtropical
gyres,
spatial
and
temporal
patchiness
exists
across
a
wide
range
of
scales.
As
ocean
surface
temperatures
continue
to
rise,
global
climate
models
suggest
that
the
strength
of
these
gyres
as
a
biological
carbon
pump
may
diminish
due
to
increased
stratification
and
depleted
nutrients
over
large
scales.
However,
such
predictions
often
ignore
climate-‐‑physical-‐‑ecosystem
interactions
on
much
finer
scales
due
to
computational
constraints
and
a
lack
of
observational
evidence.
In
order
to
assess
the
ecosystem
responses
to
fine-‐‑scale
patchiness
in
the
physical
and
biogeochemical
environment,
we
introduced
a
new
and
computationally
tractable
modeling
approach,
the
Spatially
Heterogeneous
Dynamic
Plankton
(SHiP)
model,
which
allows
for
subgrid-‐‑scale
heterogeneities
in
the
resource
environment
through
a
probabilistic
representation
of
fine-‐‑scale
disturbances.
We
applied
the
SHiP
model
to
the
Hawaii
Ocean
Time-‐‑series
(HOT)
site
(Station
ALOHA)
in
the
North
Pacific
Subtropical
Gyre,
and
compared
the
model
output
against
high-‐‑resolution
observations
from
satellite,
automatous
sampling
platforms,
and
ship-‐‑based
field
campaigns.
We
show
that
the
model
successfully
captured
both
the
mean
dynamics
and
the
full
range
of
variability
observed
at
Station
ALOHA.
The
model
also
yielded
a
substantially
different
phytoplankton
community
composition
and
carbon
cycling
when
run
in
a
temporally
and
spatially
heterogeneous
mode
relative
to
the
traditional
homogeneous
approach.
In
general,
subgrid-‐‑scale
heterogeneities
in
the
nutrient
field
favors
the
growth
of
large
phytoplankton
and
enhances
dissolved
organic
carbon
production
under
oligotrophic
conditions.
Our
findings
also
suggest
that
both
the
temporal
scale
(i.e.
duration)
and
the
intensity
(i.e.
magnitude)
of
fine-‐‑scale
disturbances
are
critical
for
determining
the
response
of
oligotrophic
ecosystems
to
fine-‐‑scale
processes.
This
indicates
that
future
changes
in
both
large
and
fine-‐‑scale
dynamics
may
significantly
impact
marine
ecosystem
structures
and
global
carbon
cycling,
and
both
should
be
accounted
for
in
the
next-‐‑generation
global
climate
models.
55
1.
Introduction
Phytoplankton
serve
as
the
base
of
the
marine
food
web
and
play
a
critical
role
in
determining
the
ocean’s
capacity
for
sequestering
atmospheric
inorganic
carbon
(Falkowski
2012).
General
patterns
in
marine
phytoplankton
biomass
and
community
size
structure
are
shown
to
directly
relate
to
temperature
and
nutrient
availability,
resulting
in
a
clear
contrast
between
the
oligotrophic,
picophytoplankton
dominated
(sub)tropical
oceans
and
the
more
productive,
microphytoplankton
dominated
(sub)polar
regions
(Chisholm,
1992;
Maranon
et
al.,
2001;
Cabre
et
al.,
2016).
While
these
patterns
hold
true
at
the
global
scale,
the
surface
ocean
also
presents
physical
and
biogeochemical
patchiness
at
finer
spatial
(e.g.
meso-‐‑
and
submeso-‐‑)
and
temporal
scales
that
may
act
to
modulate
the
local
landscape
of
marine
phytoplankton
(McGillicuddy
et
al.,
2007;
Levy
et
al.,
2015;
Mahadevan,
2016).
However,
how
this
fine-‐‑scale
(particularly
the
submesoscale)
heterogeneity
may
mediate
phytoplankton
community
composition,
higher
trophic
level
dynamics,
and
marine
ecosystem
functioning,
and
how
such
interactions
may
propagate
into
larger-‐‑scale
biogeochemical
and
carbon
cycling
remains
poorly
understood.
The
submesoscale
(1-‐‑10
km)
processes,
such
as
sharp
density
fronts,
are
ubiquitous
features
in
the
surface
ocean
(Capet
et
al.,
2008;
Mahadevan
et
al.,
2010).
Technological
advances
have
allowed
for
improved
detection
of
these
processes
revealing
their
potential
impact
on
resource
distributions,
phytoplankton
biomass,
and
export
production
(Johnson
et
al.,
2010;
Omand
et
al.,
2015;
Liu
and
Levine,
2016).
It
has
been
suggested
that
baroclinic
instabilities
associated
with
sharp
density
gradients
can
trigger
rapid
vertical
upwelling
and
downwelling,
with
velocities
much
greater
than
those
56
associated
with
meso-‐‑
and
large-‐‑scale
processes
(Mahadevan
and
Tandon,
2006;
McWilliams,
2016).
These
‘disturbance
events’
are
particularly
important
in
the
oligotrophic
ocean,
such
as
the
subtropical
gyres,
as
the
upwelling
branches
alongside
the
fronts
may
transport
nutrients
into
the
sunlit,
oligotrophic
surface
layers
and
promote
the
growth
of
phytoplankton
(Levy
et
al.,
2012;
Mahadevan
2016),
particularly
the
large,
faster-‐‑growing
functional
groups
(Levy
et
al.,
2015).
Previous
studies
have
shown
evidence
that
submesoscale
processes
may
have
a
significant
impact
on
marine
phytoplankton,
particularly
in
the
subtropical
gyres
that
account
for
~
60%
of
total
primary
production
in
the
ocean
(Karl
et
al.,
1995;
Lomas
et
al.,
2010).
For
example,
Liu
and
Levine
(2012)
used
satellite
images
to
detect
submesoscale
fronts
and
identified
a
chlorophyll
enhancement
of
up
to
38%
associated
with
submesoscale
features
in
the
North
Pacific
Subtropical
Gyre.
Levy
et
al.
(2015)
applied
a
submesoscale-‐‑permitting,
off-‐‑line
ecosystem
model
to
the
North
Atlantic
and
suggested
that
phytoplankton
diversity
is
much
greater
at
fronts
than
in
the
background
fields.
Finally,
in
situ
experiments
have
shown
that
the
injections
of
deep
sea-‐‑water
into
surface
waters,
such
as
those
predicted
to
occur
during
submesoscale
disturbance
events,
will
simulate
both
enhanced
productivity
and
a
drastic
shift
in
community
composition
(Alexander
et
al.,
2015a).
As
the
atmospheric
and
ocean
surface
temperatures
continue
to
rise
(Vecchi
and
Soden,
2007),
global
climate
models
suggest
that
the
strength
of
the
ocean
as
a
biological
carbon
pump
may
diminish
due
to
increased
stratification
and
depleted
nutrients
over
large
scales
(e.g.
Wang
et
al.,
2014;
Fu
et
al.,
2016).
While
these
models
are
powerful
tools
for
predicting
large
scale
patterns,
they
are
unable
to
resolve
submesoscale
dynamics
due
57
to
computational
constraints.
As
a
result,
each
grid
cell
in
these
models
only
represents
the
‘mean’
field
of
a
resource
environment
that,
in
reality,
includes
a
great
deal
of
spatial
and
temporal
heterogeneities
over
much
finer
scales.
Given
the
nonlinearities
in
bio-‐‑
physical
interactions,
the
ecosystem
impact
of
a
homogenous
resource
environment
does
not
equal
and
may
not
be
appropriately
used
to
parameterize
that
of
a
homogenous
resource
environment
(Levy
and
Martin,
2013).
As
such,
the
dynamics
of
submesoscale
processes
and
their
impact
on
marine
ecosystem
structures
cannot
be
addressed
by
the
current
generation
of
global
climate
models
and,
therefore,
how
these
fine-‐‑scale
interactions
might
change
in
the
future
remains
unclear.
In
this
study,
we
investigate
the
impact
of
submesoscale
disturbances
on
phytoplankton
community
composition
and
carbon
cycling
in
the
oligotrophic
North
Pacific
Subtropical
Gyre
(NPSG)
using
a
new
and
computationally
tractable
modeling
approach,
the
Spatially
Heterogeneous
Dynamic
Plankton
(SHiP)
model.
SHiP
represents
the
ecosystem
response
to
subgrid-‐‑scale
heterogeneities
in
the
resource
environment
through
the
probabilistic
representation
of
intermittent
disturbances
(Levine,
submitted).
Here,
we
focus
on
understanding
the
sensitivity
of
ecosystem
dynamics
to
changes
in
the
intensity
and
duration
of
‘fine-‐‑scale
disturbances’
and
the
resulting
impact
on
larger-‐‑scale
dynamics
(e.g.
carbon
sequestration).
We
define
‘fine-‐‑scale
disturbances’
as
all
physical
dynamics
(e.g.
submesoscale
fronts)
that
result
in
elevated
upwelling
velocities
on
short
temporal
(days)
and
fine-‐‑spatial
(<10
km)
scales.
We
applied
the
model
to
the
Hawaii
Ocean
Time-‐‑series
(HOT)
site
(Station
ALOHA)
in
the
NPSG.
The
model
successfully
captured
both
the
mean
dynamics
and
the
full
range
of
variability
observed
at
Station
ALOHA.
The
SHiP
framework
provides
the
unique
ability
to
explicitly
58
test
the
sensitivity
of
this
ecosystem
to
changes
in
fine-‐‑scale
physics.
We
show
that
both
the
temporal
scale
(i.e.
duration)
and
the
intensity
of
fine-‐‑scale
disturbances
significantly
impacts
the
response
of
phytoplankton
communities.
Furthermore,
future
changes
to
the
nature
of
‘fine-‐‑scale’
disturbances
could
substantially
alter
ecosystem
structures
and
carbon
cycling
in
the
gyres.
As
such,
an
improved
understanding
of
how
fine-‐‑scale
physical
processes
might
change
is
necessary
to
adequately
predict
future
ecosystem
and
carbon
dynamics.
2.
Methods
2.1
Model
description
The
Spatially
Heterogeneous
Dynamic
Plankton
(SHiP)
model
used
in
this
study
has
been
previously
described
in
Levine
(submitted)
and
is
only
briefly
summarized
here.
Here,
SHiP
was
run
as
a
mixed
layer
model
where
the
depth
of
the
vertically
well
mixed
surface
layer
varied
seasonally.
Subgrid-‐‑scale
heterogeneities
in
resources
(e.g.
nutrients)
were
created
through
a
probabilistic
representation
of
the
frequency
and
intensity
of
fine-‐‑scale
physical
disturbances.
Specifically,
at
any
given
time,
each
subgrid-‐‑
scale
horizontal
point
in
the
model
has
a
probability
(χ)
of
being
disturbed
by
a
disturbance
event
that
enhances
vertical
velocities,
where
χ
varies
temporally
as
a
function
of
density
gradients
(ρ).
At
each
disturbance
time
step
(e.g.
beginning
of
each
3-‐‑
day
period
when
the
disturbance
time
step
=
3
days),
this
disturbance
event
created
two
new
‘daughter’
environments
from
each
existing
‘parent’
where
the
area
of
each
‘daughter’
environment
was
given
as:
59
"#$%&'()
, =
.
/
(,)
4#)(5'
,−1 (1)
where
s(a,
t)
represents
the
fractional
area
of
an
existing
environment
with
a
disturbance
history
of
a
and
at
time
t.
The
‘daughter’
environments
were
paired,
with
one
experiencing
enhanced
upwelling
and
one
enhanced
downwelling
velocities,
while
the
remaining
fraction
of
the
‘parent’
environment
maintained
a
smaller
background
velocity.
The
enhanced
fluxes
injected
nutrient-‐‑enriched
water
from
the
deep
boundary
layer
into
the
‘daughter’
environments,
thereby
modifying
the
physical
and
biogeochemical
properties.
After
each
disturbance
period
(e.g.
3
days),
the
‘daughter’
patches
were
de-‐‑
coupled
and
the
vertical
velocities
returned
to
the
background
levels.
As
the
model
simulation
progressed
and
disturbance
dynamics
continued
to
create
new
environments,
individual
environments
with
similar
properties
(here
defined
as
2.5%
similarity
between
all
state
variables)
were
merged
together.
This
allowed
the
model
to
run
with
approximately
100
individual
environments.
The
ecosystem
framework
was
adapted
from
Doney
et
al.
(1996)
and
Moore
et
al.
(2004,
2013),
and
tracked
dissolved
nutrients
(nitrate
and
ammonium),
three
phytoplankton
functional
groups
(large
phytoplankton
such
as
diatoms,
small
phytoplankton
such
as
Prochlorococcus,
and
diazotrophs
such
as
Trichodesmium),
zooplankton
grazers,
particulate
organic
carbon
(POC),
and
dissolved
organic
carbon
(DOC).
The
evolution
of
phytoplankton
biomass
over
time
was
calculated
as:
89
:,;
8'
=−
8
8<
>
@,>
+
8
8B
Κ
89
:,;
8B
+
EFG
(@)
@,>
J
@,>
K
@,>
L
@,>
−
@,>
−
@,>
(2)
where
w
is
the
advective
exchange
with
the
bottom
boundary
condition,
K
is
the
diffusive
transport
between
‘parent’
and
‘daughter’
environments,
µmax
is
the
maximum
growth
60
rate
which
is
constrained
by
the
most
limiting
element
among
temperature,
light,
and
nutrients
(Geider
et
al.,
1998).
Phytoplankton
biomass
were
subject
to
losses
due
to
natural
mortality,
and
zooplankton
grazing.
POC
was
exported
out
of
the
surface
ocean
through
aggregation
and
sinking.
2.2
Model
configuration
and
simulations
The
SHiP
model
was
used
to
simulate
the
Hawaiian
Ocean
Time-‐‑series
(HOT)
site
(Station
ALOHA,
22°
45'N,
158°
00'W)
in
the
North
Pacific
Subtropical
Gyre
from
2003-‐‑
2014.
The
horizontal
dimension
of
the
model
grid
cell
was
defined
as
a
5°
×
5°
domain
centered
at
Station
ALOHA.
Observed
monthly
mean
mixed
layer
over
the
model
domain
(Talley
Lab
at
Scripps
Institution
of
Oceanography,
http://mixedlayer.ucsd.edu)
was
used
to
set
the
model
vertical
dimensions.
This
product
was
calculated
from
ARGO
float
profiles
using
a
hybrid
method
based
on
traditional
threshold
and
gradient
methods
(Holte
et
al.,
2010).
The
model
was
forced
with
monthly
photosynthetically
active
radiation
(PAR,
NASA
OB.DAAC
MODIS-‐‑Aqua
L3),
and
air
temperature
and
wind
velocity
(NCEP/NCAR
reanalysis).
The
bottom
boundary
condition
was
defined
using
the
average
of
observed
temperature
and
nitrate
at
station
ALOHA
between
the
bottom
of
the
mixed
layer
and
the
first
depth
at
which
density
was
2-‐‑units
(kg
m
-‐‑3
)
greater
than
the
average
density
of
the
mixed
layer
(data
retrieved
from
HOT
Data
Organization
and
Graphical
System,
HOT-‐‑DOGS;
http://hahana.soest.hawaii.edu/hot/hot-‐‑dogs/).
All
monthly
drivers
were
first
interpolated
into
pseudo-‐‑daily
values.
Hourly
PAR
data
was
then
generated
following
Stull
(1988).
61
SHiP
dynamics
disturbance
dynamics
were
driven
using
the
probability
of
disturbance
(χ)
calculated
from
MODIS-‐‑Aqua
L3
Sea
Surface
Temperature
following
Liu
and
Levine
(2016).
The
disturbance
vertical
velocity
was
set
at
10
m
day
-‐‑1
.
The
background
velocity
was
calculated
as
(0.1
+
0.02
×
E)
m
day
-‐‑1
,
where
E
is
the
number
of
cyclonic
eddies
passing
through
the
region
during
the
month.
E
was
estimated
using
eddy
track
data
calculated
from
Sea
Surface
Height
fields
(Chelton
et
al.,
2011,
http://cioss.coas.oregonstate.edu/eddies).
This
results
in
a
dynamic
background
velocity
in
the
range
of
0.1
and
0.24
m
day
-‐‑1
,
which
aims
to
reflect
the
temporal
large
and
mesoscale
dynamics
in
the
region.
A
schematic
of
SHiP
model
configuration
is
shown
in
Figure
1.
The
model
was
spun-‐‑up
for
two
years
using
repeating
annual
forcings
of
2003
to
an
approximate
steady-‐‑state,
then
run
for
a
total
of
12
years
between
2003
and
2014
with
time
steps
of
60
minutes
while
employing
a
4
th
order
Runge-‐‑Kutta
scheme
in
MATLAB
R2016a
(Shampine
and
Reichelt,
1997).
Key
physical
and
ecosystem
parameters
(n
=
10,
Appendix
A)
were
optimized
using
the
Nelder-‐‑Mead
technique
to
minimize
the
model-‐‑
observational
differences
in
climatological
mixed
layer
averaged
temperature,
nitrate
concentration,
and
total
primary
production
for
a
6-‐‑year
period
between
2003
and
2008.
This
6-‐‑year
period
was
only
used
for
model
optimization,
and
all
results
were
based
on
output
from
2009-‐‑2014.
An
ensemble
of
sensitivity
modeling
experiments
were
performed
to
investigate
the
ecosystem
responses
to
fine-‐‑scale
disturbances
with
variant
characteristics,
namely
their
intensity
and
duration.
In
these
experiments,
the
intensity
of
disturbances
(i.e.
upwelling
velocity)
was
adjusted
between
2
and
50
m
day
-‐‑1
,
and
the
duration
was
varied
62
between
1
and
5
days,
while
all
other
conditions
were
held
fixed.
These
numbers
were
chosen
to
reflect
the
range
of
variability
for
submesoscale
features
found
in
the
literature
(e.g.
Mahadevan
and
Tandon,
2006;
Thomas
et
al.,
2008;
Levy
et
al.,
2012).
To
allow
for
comparison
across
the
simulations,
the
mean
and
total
upwelling
fluxes
into
the
model
grid
cell
were
kept
the
same
for
all
experiments:
such
that,
when
the
intensity
or
duration
of
disturbances
was
increased
(decreased),
the
frequency
of
disturbance
(i.e.
the
fractional
area
of
each
subgrid-‐‑scale
environment
being
disturbed
during
each
disturbance
time
step)
was
decreased
(increased)
proportionally.
To
investigate
the
impact
of
warming-‐‑induced
stratification
on
ecosystem
dynamics,
model
simulations
were
conducted
with
reduced
background
and
disturbance
vertical
velocities
at
0.75
and
0.5
times
present-‐‑day
values,
while
other
conditions
being
held
fixed.
To
test
the
additional
impact
of
modifications
in
the
characteristics
of
disturbances
under
the
warming
conditions,
a
second
set
of
sensitivity
experiments
were
conducted
varying
the
intensity
and
duration
of
disturbance
events
under
reduced
vertical
velocities.
Finally,
the
results
from
the
SHiP
physical
framework
(SHiP
mode)
were
compared
against
the
ecosystem
model
run
in
a
traditional,
Averaged
Environment
(AE)
mode
in
which
the
entire
grid
cell
responded
homogenously
to
the
surface
and
bottom
boundary
forcing.
The
vertical
velocity
used
in
the
AE
mode
was
equal
to
the
grid-‐‑cell
mean
velocity
in
the
SHiP
mode.
As
such,
the
total
upwelling
fluxes
into
the
model
grid
cell
were
consistent
between
the
SHiP
and
AE
mode
simulations.
63
2.3
Validation
datasets
The
SHiP
model
output
(2009-‐‑2014)
was
compared
against
both
remotely
sensed
and
in
situ
observations.
Measurements
of
temperature,
nitrate,
and
net
primary
and
export
production
at
station
ALOHA
were
obtained
from
HOT-‐‑DOGS.
MODIS-‐‑Aqua
L2
Sea
Surface
Temperature
and
chlorophyll
concentration
were
retrieved
from
NASA
OB.DAAC
and
processed
to
1-‐‑km
products
using
SeaDAS
packages.
Quality
controlled
CTD
data
from
Apex
profiling
floats
were
obtained
from
MBARI
FloatViz
6.0
(www.mbari.org/science/upper-‐‑ocean-‐‑systems/chemical-‐‑sensor-‐‑group/floatviz/).
Finally,
modeled
short-‐‑term
temporal
variability
was
compared
against
temperature
profiles
from
an
intensive
field
campaign
(HOE-‐‑DYLAN,
http://hahana.soest.hawaii.edu/,
e.g.
Fitzsimmons
et
al.,
2015)
at
Station
ALOHA
during
the
summer
of
2012.
3.
Results
3.1
SHiP
vs.
HOT
comparisons
The
SHiP
model
estimates
of
temperature
and
total
net
primary
production
(NPP)
were
compared
against
HOT
measurements
at
Station
ALOHA
for
a
6-‐‑year
period
between
2009
and
2014
(Figure
2).
This
dataset
is
independent
from
the
one
used
to
optimize
the
model
parameters
(6-‐‑year
period
between
2003
and
2008).
To
enable
direct
comparisons,
HOT-‐‑DOGS
measurements
were
averaged
over
the
mixed
layer
and
daily
model
estimates
were
averaged
over
the
entire
grid
cell,
before
monthly
and
climatological
means
were
generated.
The
modeled
means
and
seasonal
cycles
of
temperature
and
NPP
were
in
good
agreement
with
observations.
The
modeled
monthly
climatology
for
mixed
layer
temperature
was
highly
correlated
with
the
observed
values
64
(r
=
0.96,
p
<
0.01).
The
magnitude
of
the
seasonal
cycle
was
slightly
underestimated
in
the
model
likely
due
to
a
lack
of
horizontal
advection
which
as
not
represented
in
the
SHiP
mixed-‐‑layer
model
framework.
A
comparison
between
modeled
and
observed
NPP
also
yielded
a
significant
correlation
(r
=
0.67,
p
<
0.05).
Moderate
model-‐‑observation
mismatches
occurred
in
October
most
likely
due
to
an
underestimation
of
diazotroph
biomass
in
the
model
(see
Discussions).
Small
phytoplankton
dominated
the
modeled
phytoplankton
community
(Figures
2e,
f)
accounting
for
an
average
of
73.1%
of
total
particulate
organic
carbon
over
the
6-‐‑year
period
between
2009
and
2014.
This
is
in
consistent
with
White
et
al.
(2015),
which
used
a
HPLC-‐‑based
approximation
to
suggest
that
ca.
68%
of
total
pigments
at
Station
ALOHA
were
related
to
small-‐‑sized
cyanobacteria.
Both
of
these
estimates
excluded
diazotrophs
as
they
are
not
sufficiently
extractable
in
the
methanol
solution
used
in
HPLC.
Wintertime
conditions,
such
as
low
temperatures
and
nutrient
enrichment
from
deep
convection,
favored
the
growth
of
large
phytoplankton
in
the
model,
while
diazotroph
biomass
peaked
in
the
summer
when
nitrate
was
depleted
(<
0.05
μM).
Modeled
diazotroph
blooms
were
mainly
restricted
to
July
and
August
and
less
variant
relative
to
the
values
reported
at
Station
ALOHA
(bloom
season
between
June
and
October,
White
et
al.,
2007).
We
attribute
this
primarily
to
the
iron
dynamics
which
are
not
represented
in
SHiP
but
are
a
critical
control
on
diazotrophs
in
the
NPSG.
The
export
ratio
of
particulate
organic
carbon
(POC),
defined
as
the
export
flux
of
POC
at
the
150
m
reference
depth
divided
by
depth-‐‑integrated
NPP
over
the
top
150
m
was
compared
against
previously
reported
values
from
Station
ALOHA
(Karl
and
Church,
2014).
As
the
model
grid
cell
only
represents
the
surface
mixed
layer
with
a
depth
ranging
65
from
~
45
to
100
m,
depth-‐‑integrated
NPP
over
the
top
150
m
was
estimated
using
the
fractional
contribution
of
mixed-‐‑layer-‐‑integrated
NPP
to
the
top
150
m
integrated
NPP
from
on
HOT
measurements.
The
model
estimated
export
ratio
falls
in
a
range
between
0.013
and
0.12
(Figure
3),
which
is
comparable
to
previously
reported
values
for
the
region
of
0.09
(Buesseler
et
al.,
2007)
and
0.01–0.20
(Karl
and
Church,
2014).
3.2
Spatiotemporal
heterogeneity
While
station
ALOHA
allows
us
to
compare
modeled
grid-‐‑cell
mean
dynamics
against
in
situ
observations
(Figures
2,
3),
this
data-‐‑set
only
samples
temporal
variability
at
a
single
fixed
location.
In
order
to
evaluate
the
predicted
spatial
variability
across
the
model’s
heterogeneous
grid
cell,
we
compared
SHiP
model
output
against
high
resolution
satellite
observations.
Figure
4
suggests
that
the
model
run
with
only
~
85
subgrid-‐‑scale
environments
was
able
to
predict
the
full
range
of
spatial
and
temporal
variability
in
temperature
and
chlorophyll
captured
by
MODIS-‐‑Aqua
images
(~
4,600
valid
daily
pixels)
for
this
period
(July
2004).
When
the
model
was
run
in
the
traditional,
Averaged
Environment
(AE)
mode
with
a
single
homogenous
environment,
it
was
only
able
to
capture
10–20%
of
the
observed
variability
(Figures
4c,
f).
This
suggests
that
the
SHiP
framework
is
able
both
to
represent
the
full
range
of
subgrid
cell
resource
environments
and
the
corresponding
biological
response.
The
APEX
profiling
floats
are
capable
of
providing
high
resolution
vertical
measurements
of
the
surface
ocean
(e.g.
0–1000
m)
at
temporal
intervals
between
3–10
days
(Schmid
et
al.,
2007).
As
these
floats
are
designed
to
follow
currents
in
a
Lagrangian
manner
(Perry
and
Rudnick,
2003),
each
of
them
records
the
‘history’
of
a
water
parcel
66
and
therefore
the
temporal
variability
its
experiences.
Since
2007,
more
than
10
floats
equipped
with
Sea-‐‑Bird
CTD
sensors
(some
also
with
optical,
pH,
nitrate,
or
oxygen
sensors)
have
been
deployed
near
Station
ALOHA.
With
multiple
floats
sampling
the
same
region
simultaneously,
both
temporal
and
spatial
(horizontal
and
vertical)
variability
in
the
surface
ocean
can
be
sampled
at
high
resolution.
Mixed
layer
averaged
temperatures
calculated
from
each
float
profile
were
compared
against
the
temperatures
predicted
in
all
subgrid
environments
in
the
model
for
each
monthly
period
between
2009
and
2014.
Figure
5
shows
results
for
Year
2012,
during
which
period
both
float
data
and
ship-‐‑based
HOE-‐‑DYLAN
measurements
were
available.
For
temperature,
model
estimates
are
on
average
0.6°C
higher
than
the
observed
values
during
this
period.
This
is
likely
due
to
dynamics
(e.g.
cooling
induced
by
horizontal
convections)
that
are
currently
missing
in
the
model.
To
enable
a
direct
comparison
between
the
ranges
of
observed
and
modeled
variability,
the
model
estimates
were
corrected
for
the
mean
of
both
sets
of
observations
(i.e.
subtracted
by
0.6°C).
In
general,
profiling
floats
captured
a
greater
range
of
variability
than
ship-‐‑based
measurements
likely
due
to
broader
spatial
and
temporal
coverages
(Figure
5,
July
to
September).
In
June
of
2012,
for
example,
a
total
of
3
floats
provided
12
vertical
profiles
for
the
5°
×
5°
region,
which
showed
a
temperature
range
of
1.6°C;
while
the
variability
at
a
single
site
captured
during
HOE-‐‑DYLAN
cruises
was
relatively
small
with
a
range
of
0.7°C.
The
modeled
temperature
showed
a
greater
variability
relative
to
both
float
and
ship-‐‑based
observations
with
a
monthly
range
of
~
3°C.
This
difference
was
particularly
large
when
fewer
measurements
were
available.
For
example,
a
temperature
range
of
0.4°C
was
captured
in
March
of
2012
when
only
5
float
profiles
were
available
for
the
region.
This
supports
the
hypothesis
that
in
situ
observations,
not
67
surprisingly,
may
have
significantly
under-‐‑sampled
the
heterogeneity
that
exists
in
the
environment.
3.3
Ecological
relevance
of
the
intensity
and
duration
of
disturbances
The
modeled
phytoplankton
community
composition
showed
substantial
variability
when
the
model
was
run
with
disturbances
of
varying
intensity
and
duration
but
the
same
grid
cell
mean
flux.
The
average
contribution
of
large
phytoplankton
to
total
biomass
ranged
between
12–23%
over
the
12
year
simulations
(Figure
6a).
This
suggests
that
the
characteristics
of
disturbance
dynamics
can
have
a
significant
impact
on
ecosystem
structure
despite
the
same
mean
flux
into
the
model’s
grid
cell.
We
observed
a
‘sweet
spot’
for
large
phytoplankton
at
intermediate
frequencies
and
intensities:
if
the
disturbances
were
too
mild
and
too
short
(high
frequency)
large
phytoplankton
were
more
likely
to
be
out-‐‑completed
by
small
phytoplankton,
while
if
disturbances
were
too
strong
and
too
persistent
(low
frequency)
grazing
pressure
on
the
large
phytoplankton
increased
again
giving
the
small
phytoplankton
an
advantage.
The
combination
of
intensity
and
frequency
which
yielded
the
maximal
advantage
for
large
phytoplankton
is
considered
to
be
a
nonlinear
function
of
the
maximum
growth
rate
relative
to
that
of
the
small
phytoplankton
and
the
grazing
rates.
3.4
Climate
change
experiments
Increased
stratification
(reduced
mean
flux
of
nitrate
into
the
model)
resulted
in
a
fairly
linear
reduction
in
phytoplankton
biomass
and
corresponding
decrease
in
percentage
of
large
phytoplankton
from
22%
(present-‐‑day)
to
14%
at
a
50%
reduction
in
68
mean
flux
(Figure
7).
These
reductions
(up
to
37%
change
relative
to
the
present-‐‑day)
are
consistent
with
previous
observational
and
modeling
estimates
(Hilligsoe
et
al.,
2011;
Wang
et
al.,
2014;
Fu
et
al.,
2016)
suggesting
that
SHiP
is
capturing
the
same
large-‐‑scale
dynamics
represented
by
more
conventional
models.
However,
in
addition
to
large-‐‑scale
changes
in
the
resource
environment,
the
characteristics
of
disturbances
may
also
be
modified
under
future
climate
conditions.
Our
model
suggests
that
changes
in
the
intensity
and
duration
of
fine-‐‑scale
disturbances
(without
changing
mean
and
total
fluxes)
have
the
potential
to
substantially
impact
the
ecological
response
to
warming
by
either
exacerbating
or
partially
compensating
the
impact
of
nutrient
depletion.
For
example,
shortening
the
duration
of
disturbances
from
3
days
to
1
day
results
an
additional
decline
of
2.5%
(a
11%
change
relative
to
the
present-‐‑day)
in
large
phytoplankton
contribution.
This
additional
impact
due
to
changes
in
the
characteristics
of
disturbances
cannot
be
reflected
in
the
traditional,
coarser
grid
climate
models
without
explicitly
representing
fine-‐‑scale
dynamics
(Bachman
and
Taylor
2014;
Brannigan,
2016).
When
run
in
the
AE
mode
for
the
present-‐‑day,
the
total
particulate
carbon
pool
was
similar
(p
=
0.63)
to
that
estimated
by
the
SHiP
mode,
but
the
ecosystem
was
significantly
different
with
large
phytoplankton
contributing
only
12%
of
the
POC
pool
in
the
AE
model
as
compared
to
22%
in
the
SHiP
mode.
The
AE
mode
run
also
predicts
a
different
ecosystem
response
to
reduced
nutrient
flux,
where
a
50%
reduction
results
in
a
negligible
1.5%
decrease
in
percentage
of
large
phytoplankton
as
compared
to
the
37%
reduction
in
the
SHiP
mode.
69
4.
Discussions
The
SHiP
model
framework
captures
subgrid-‐‑scale
heterogeneities
in
resource
environment
that
results
from
fine-‐‑scale
physical
dynamics
that
are
unresolved
in
the
model
and
simulates
the
full
ecosystem
responses
to
these
disturbances
in
a
computationally
tractable
approach.
This
framework
allows
us
to
investigate
the
importance
of
environmental
fluctuations
in
driving
ecosystem
dynamics
and
carbon
cycling
and
understand
biases
that
might
be
introduced
when
these
unresolved
dynamics
are
represented
as
a
correction
to
the
mean
flux
(Average
Environment,
AE)
as
opposed
to
spatially
and
temporally
variable
(SHiP
mode).
The
spatiotemporally
heterogeneous
(SHiP)
mode
yielded
a
substantially
different
phytoplankton
community
composition
relative
to
the
traditional
homogeneous
(AE)
mode
(Figure
7).
Under
oligotrophic
conditions
(i.e.
the
average
nutrient
concentration
is
low
within
the
large-‐‑scale
field),
the
existence
of
spatial
heterogeneity
in
nutrient
distributions
generated
in
the
SHiP
mode
increased
the
growth
of
large
phytoplankton
in
subgrid
cell
environments
with
elevated
nutrient
concentrations.
As
these
environments
were
temporally
dynamic,
this
promoted
the
coexistence
of
different
phytoplankton
groups
and
the
diversity
in
phytoplankton
communities
(Levine,
submitted).
Figure
8
shows
the
percentage
contribution
of
each
phytoplankton
group
to
total
phytoplankton
biomass
relative
to
the
grid
cell
average
NO3
flux.
When
NO3
flux
was
low
(e.g.
<
10
mmol
m
-‐‑2
day
-‐‑1
),
the
SHiP
and
AE
modes
produced
similar
phytoplankton
compositions,
with
small
phytoplankton,
diazotrophs,
and
large
phytoplankton
contributing
to
ca.
60%,
30%,
and
10%
of
the
total
biomass,
respectively.
At
elevated
NO3
fluxes
(e.g.
during
wintertime
conditions),
large
phytoplankton
gained
a
competitive
advantage
in
the
SHiP
mode
contributing
up
to
45%
of
total
biomass,
similar
70
to
that
of
small
phytoplankton.
Diazotrophs
accounted
for
only
7%
of
total
biomass
when
NO3
flux
was
the
highest
(i.e.
>
35
mmol
m
-‐‑2
day
-‐‑1
).
In
contrast,
in
the
AE
mode,
small
phytoplankton
always
dominated
(~
80%)
regardless
of
NO3
flux.
This
difference
is
attributed
to
the
varying
dynamics
between
the
two
model
modes.
In
the
AE
simulations,
nitrate
concentrations
were
always
below
the
threshold
where
small
phytoplankton,
which
have
higher
affinity
for
nitrate,
out-‐‑competed
large
phytoplankton.
In
the
SHiP
simulations,
as
NO3
flux
increases,
the
heterogeneous
nutrient
field
resulted
in
subgrid
cell
environments
where
nitrate
concentrations
were
temporarily
above
this
threshold,
allowing
large
phytoplankton
to
gain
an
advantage.
Consequently,
the
SHiP
mode
runs
yielded
an
enhanced
export
of
particulate
carbon
relative
to
the
AE
mode
due
to
a
greater
contribution
of
large
and
faster-‐‑sinking
phytoplankton
(Figure
9),
with
an
average
increase
of
8.4%
over
the
12-‐‑year
period
and
a
maximum
increase
of
34%
(Figure
9b).
This
impact
on
carbon
export,
however,
is
less
important
compared
to
that
through
the
the
dissolved
organic
carbon
(DOC)
pool.
Previous
studies
have
suggested
that
DOC
pool
may
also
play
a
major
role
in
carbon
export
in
the
subtropical
oceans
(Ducklow
et
al.,
1995;
Hansell
et
al.,
2012).
In
this
study,
the
SHiP
mode
runs
maintained
a
substantially
larger
DOC
pool
relative
to
the
AE
mode
(116
vs
87
mg
C
m
-‐‑3
,
p
<
0.01).
In
the
model,
a
portion
of
this
DOC
was
sequestered
into
the
deep
layers
through
vertical
mixing
with
the
deep
boundary.
The
rate
of
carbon
export
through
DOC
sequestration
was
estimated
at
40
and
28
mg
C
m
-‐‑2
day
-‐‑1
in
the
SHiP
and
AE
mode,
respectively.
Given
that
these
rates
are
on
the
same
order
of
magnitude
with
the
particulate
carbon
export
(Figure
3),
the
total
carbon
exported
out
of
the
surface
ocean
was
appropriately
20%
greater
in
the
SHiP
mode
runs
relative
to
the
AE
mode.
Our
71
model
results
agree
with
previous
studies
(Ducklow
et
al.,
1995;
Hansell
et
al.,
2012),
suggesting
that
the
role
of
DOC
plays
in
sequestering
inorganic
carbon
is
important
in
the
NPSG
and
might
be
significantly
underestimated
in
the
traditional
homogeneous
models.
Although
both
SHiP
and
AE
mode
runs
successfully
captured
the
mean
seasonal
cycles
of
phytoplankton
dynamics
at
the
site,
the
SHiP
mode
was
also
able
to
produce
the
observed
variability
(Figures
4,
5)
including
intermittent
‘bloom’
events
that
are
captured
in
the
HOTS
dataset
(Figure
10).
For
example,
in
January
2012
a
bloom
was
observed
at
Station
ALOHA
with
a
significantly
elevated
chlorophyll
concentration
of
~
0.22
mg
m
-‐‑3
.
While
the
SHiP
and
AE
mode
runs
yield
a
similar
average
chlorophyll
concentration
of
~
0.12
mg
m
-‐‑3
,
the
chlorophyll
within
the
individual
subgrid
cell
environments
in
the
SHiP
mode
reached
0.20
mg
m
-‐‑3
(95%
percentile)
and
0.27
mg
m
-‐‑3
(99%
percentile),
which
is
more
comparable
to
the
observed
level
during
this
bloom
event.
This
highlights
a
challenge
with
time-‐‑series
programs.
Even
though
these
sampling
efforts
aim
to
record
the
mean
dynamics
and
long-‐‑term
trends
at
a
given
a
site,
the
measurements
also
capture
significant
variability
driven
by
poorly
sampled
episodic
events
such
as
fine-‐‑scale
disturbances.
Therefore,
it
is
expected
that
the
full
range
of
the
observed
variability
could
only
be
predicted
when
both
large
and
fine-‐‑scale
dynamics
were
accounted
for
in
a
model.
While
the
SHiP
model
presents
a
novel
framework
for
representing
subgrid
cell
bio-‐‑physical
interactions,
the
current
version
includes
a
highly
simplified
ecosystem
module
with
only
three
phytoplankton
groups
and
two
limiting
nutrients
(i.e.
nitrate
and
ammonium).
This
ecosystem
module
provides
a
first
order
understanding
of
ecosystem
dynamics
in
the
NPSC
by
representing
new
and
regenerated
production
and
nitrogen
limitation,
believed
to
be
the
primary
limiting
nutrient
at
Station
ALOHA
(Alexander
et
al.,
72
2015b).
However,
it
has
been
shown
that
phytoplankton
dynamics
at
Station
ALOHA
can
be
controlled
by
a
combination
of
physical
and
biogeochemical
factors
including
both
macronutrients,
namely
nitrogen,
silica,
and
micronutrients
such
as
iron
(Karl
et
al.,
2001;
White
et
al.,
2007;
Church
et
al.,
2009;
Alexander
et
al.,
2015a,
b).
Iron
limitations
are
particularly
important
in
constraining
the
growth
of
the
nitrogen-‐‑fixing
diazotrophs
such
as
Trichodesmium
spp.
(Sanudo-‐‑Wilhelmy
et
al.,
2001;
Moore
et
al.,
2006;
Sohm
et
al.,
2011).
Consistent
with
observations,
modeled
diazotroph
biomass
increased
in
the
summer
once
nitrogen
became
limiting
for
large
and
small
phytoplankton.
However,
modeled
diazotroph
biomass
declined
in
September
due
to
grazing
pressure
while
observations
show
that
diazotroph
blooms
could
be
found
both
in
the
summer
and
fall.
For
example,
intensive
diazotroph
blooms
were
reported
in
October
of
2003
(White
et
al.,
2007)
and
September-‐‑October
of
2005
and
2006
(Church
et
al.,
2009).
It
is
hypothesized
that
these
blooms
are
driven
by
nutrient
depletion
and
warmer
surface
waters
as
well
as
mesoscale
physical
forcing
such
as
anti-‐‑cyclonic
eddies
(Church
et
al.,
2009).
A
lack
of
limiting
nutrients
(e.g.
iron)
in
our
simplified
ecosystem
module
and
a
lack
of
realistic
representation
of
mesoscale
processes
(e.g.
eddies)
in
the
mixed
layer
model
framework
introduce
a
bias
in
the
model
and
limits
our
ability
to
predicted
the
timing
and
magnitude
of
diazotroph
blooms.
We
hypothesize
that
a
more
complete
ecosystem
component
will
likely
improve
the
model’s
estimates
for
diazotroph
dynamics.
In
addition,
mixtrophic
phytoplankton
such
as
dinoflagellates
are
shown
to
contribute
to
ca.
35%
of
the
total
biomass
at
Station
ALOHA
(Alexander
et
al.,
2015a).
These
phytoplankton
groups
are
not
explicitly
represented
in
the
model
due
to
a
lack
of
understanding
of
their
mobility
and
more
complex
mixtrophic
behaviors.
73
While
global
climate
models
produce
predictions
of
long-‐‑term
trends
in
warming
(IPCC,
2014),
little
is
known
with
regard
to
how
fine-‐‑scale
(<10
km)
dynamics
may
change
in
the
next
100
years.
This
lack
of
knowledge
is
due
to
insufficient
direct
observations
of
these
processes
and
an
incomplete
understanding
of
how
they
may
interact
with
and
be
impacted
by
global
climate
patterns.
However,
since
these
fine-‐‑scale
features
are
driven
by
larger-‐‑scale
dynamics,
it
is
reasonable
to
anticipate
that
large-‐‑scale
trends
in
the
climate
system
will
also
impact
the
dynamics
of
fine-‐‑scale
disturbances,
such
as
their
intensity
and
duration.
Previous
studies
have
suggested
that
atmospheric
submesoscale
processes,
such
as
tropical
storms,
have
increased
in
frequency
and
shortened
in
duration
over
the
past
~60
years
largely
due
to
a
warming
climate
(Landsea
et
al.,
2010;
Villarini
et
al.,
2011).
To
test
the
potential
impact
of
shorter,
more
frequent
submesoscale
features
might
have
on
phytoplankton
dynamics,
we
performed
a
climate
change
sensitivity
experiment
based
on
the
assumption
that
the
submesoscale
processes
in
the
ocean
may
follow
the
same
trends
as
atmospheric
dynamics.
Specifically,
we
both
reduced
the
large-‐‑
scale
mean
flux
and
reduced
the
mean
duration
of
disturbances
from
3
days
to
2
days
and
then
to
1
day
(frequency
was
increased
accordingly
to
maintain
a
same
flux
among
these
experiments).
Model
output
suggests
that
a
shortening
of
feature
duration
and
an
increase
in
frequency
may
result
in
an
additional
decline
in
large
phytoplankton
biomass
with
a
magnitude
similar
to
that
of
a
25%
decline
in
mean
flux
(Figure
7).
While
this
is
only
an
example
of
how
changes
in
fine-‐‑scale
dynamics
may
result
in
an
additional
ecological
impact,
our
results
suggest
that
the
dynamics
of
fine-‐‑scale
disturbances
may
have
the
potential
to
either
exacerbate
or
compensate
the
impact
of
nutrient
depletion
due
to
warming
and
increased
stratification.
Additional
work
is
needed
in
order
to
74
constrain
how
long-‐‑term
warming
trends
might
impact
the
frequency
and
intensity
of
submesoscale
features
in
the
oceans
might
change.
5.
Conclusions
and
Implications
We
used
a
new
and
computationally
tractable
modeling
approach
to
assess
phytoplankton
community
responses
to
enhanced
nutrient
supplies
created
by
fine-‐‑scale
disturbances.
Our
results
showed
that,
in
the
oligotrophic
subtropical
gyres,
spatiotemporal
heterogeneity
in
the
resource
environment
(i.e.
nutrient
field)
favored
the
growth
of
larger-‐‑sized
phytoplankton
and
allowed
for
a
more
diverse
phytoplankton
community
and
enhanced
carbon
sequestration.
This
suggests
that
coarser
grid
traditional
models
might
misrepresent
ecosystem
structure
and
underestimate
carbon
sequestration
by
not
accounting
for
non-‐‑linear
biological
responses
to
fine-‐‑scale
features.
We
also
suggest
that
future
changes
in
both
large
and
fine-‐‑scale
dynamics
may
significantly
impact
phytoplankton
and
ecosystem
structures.
Although
we
validated
and
analyzed
the
model
at
a
single
study
site
(Station
ALOHA
in
the
NPSG),
the
model
framework
and
the
conclusions
from
this
case
study
might
be
extended
to
the
rest
of
the
oligotrophic
regime,
which
covers
a
majority
of
the
global
ocean.
Considerations
of
bio-‐‑
physical
interactions
at
fine-‐‑scales
in
the
next
generation
climate
models
will
advance
our
understanding
of
global
biogeochemical
and
carbon
budgets
and
increase
our
ability
to
predict
shifts
in
these
critical
processes
under
both
current
and
future
climate
conditions.
75
Acknowledgements
This
study
was
supported
by
the
National
Science
Foundation
(OCE-‐‑RIG
1323319
awarded
to
N.
Levine),
NASA
Earth
and
Space
Science
Fellowship
(NNX14AK76H
awarded
to
X.
Liu),
and
University
of
Southern
California.
We
acknowledge
the
invaluable
contributions
of
P.
Calil,
A.
White,
and
Y.
Teng
who
assisted
with
the
validation
and
analysis
of
our
model.
76
References
Alexander,
H.,
M.
Rouco,
S.
T.
Haley,
S.
T.
Wilson,
D.
M.
Karl,
and
S.
T.
Dyhrman
(2015a),
Functional
group-‐‑specific
traits
drive
phytoplankton
dynamics
in
the
oligotrophic
ocean.
Proc.
Natl.
Acad.
Sci.,
112,
E5972-‐‑E5979.
Alexander,
H.,
B.
D.
Jenkins,
T.
A.
Rynearson,
and
S.
T.
Dyhrman
(2015b),
Metatranscriptome
analyses
indicate
resource
partitioning
between
diatoms
in
the
field.
Proc.
Natl.
Acad.
Sci.,
112,
E2182-‐‑E2190.
Bachman,
S.
D.,
and
J.
R.
Taylor
(2014),
Modelling
of
partially-‐‑resolved
oceanic
symmetric
instability.
Ocean
modelling,
82,
15-‐‑27.
Bopp,
L.,
P.
Monfray,
O.
Aumont,
J.-‐‑L.
Dufresne,
H.
Le
Treut,
G.
Madec,
L.
Terray,
and
J.
C.
Orr
(2001),
Potential
impact
of
climate
change
on
marine
export
production,
Global
Biogeochem.
Cycles,
15(1),
81–99,
doi:10.1029/1999GB001256.
Brannigan,
L.
(2016),
Intense
submesoscale
upwelling
in
anticyclonic
eddies,
Geophys.
Res.
Lett.,
43,
3360–3369,
doi:10.1002/2016GL067926.
Buesseler,
K.
O.,
C.
H.
Lamborg,
P.
W.
Boyd,
P.
J.
Lam,
T.
W.
Trull,
R.
R.
Bidigare,
et
al.
(2007),
Revisiting
carbon
flux
through
the
ocean's
twilight
zone.
Science,
316,
567-‐‑570.
Capet,
X.,
J.
C.
McWilliams,
M.
J.
Molemaker,
and
A.
F.
Shchepetkin
(2008),
Mesoscale
to
submesoscale
transition
in
the
California
current
system.
Part
II:
Frontal
processes,
J.
Phys.
Oceanogr.,
38,
44-‐‑64.
Cabre,
A.,
D.
Shields
D,
I.
Marinov,
and
T.
S.
Kostadinov
(2016),
Phenology
of
size-‐‑
partitioned
phytoplankton
carbon-‐‑biomass
from
ccean
color
remote
sensing
and
CMIP5
models.
Front.
Mar.
Sci.
3,
39.
Chelton,
D.,
M.
Schlax,
R.
Samelson
(2011),
Global
observations
of
nonlinear
mesoscale
eddies,
Prog.
Oceanogr.,
91,
167-‐‑216,
http://cioss.coas.oregonstate.edu/eddies,
accessed
on
2015/05/06.
Chisholm,
S.
W.
(1992),
Phytoplankton
size.
In:
Falkowski
P.
G.
and
A.
D.
Woodhead
(eds)
Primary
productivity
and
biogeochemical
cycles
in
the
sea.
Plenum
Press,
New
York,
213-‐‑217.
Church,
M.
J.,
C.
Mahaffey,
R.
M.
Letelier,
R.
Lukas,
J.
P.
Zehr,
and
D.
M.
Karl
(2009),
Physical
forcing
of
nitrogen
fixation
and
diazotroph
community
structure
in
the
North
Pacific
subtropical
gyre,
Global
Biogeochem.
Cycles,
23,
GB2020.
77
Doney,
S.
C.,
I.
Lima,
J.
K.
Moore,
K.
Lindsay,
M.
J.
Behrenfeld,
T.
K.
Westberry,
N.
Mahowald,
D.
M.
Glover,
T.
Takahashi
(2009),
Skill
metrics
for
confronting
global
upper
ocean
ecosystem-‐‑biogeochemistry
models
against
field
and
remote
sensing
data.
J.
Mar.
Syst.,
76,
95-‐‑112.
Ducklow,
H.
W.,
C.
A.
Carson,
N.
R.
Bates,
A.
H.
Knap,
A.
F.
Michaels,
et
al.
(1995),
Dissolved
Organic
Carbon
as
a
Component
of
the
Biological
Pump
in
the
North
Atlantic
Ocean.
Phil.
Trans.
R.
Soc.
B.,
348,
161-‐‑167.
Falkowski,
P.
(2012),
Ocean
Science:
The
power
of
plankton.
Nature,
483,
S17-‐‑S20.
Fitzsimmons,
J.
N.,
C.
T.
Hayes,
S.
N.
Al-‐‑Subiai,
R.
Zhang,
P.
L.
Morton,
R.
E.
Weisen,
F.
Ascani,
and
E.
A.
Boyle
(2015),
Daily
to
decadal
variability
of
size-‐‑fractionated
iron
and
iron-‐‑binding
ligands
at
the
Hawaii
Ocean
Time-‐‑series
Station
ALOHA.
Geochimica
et
Cosmochimica
Acta,
171,
303-‐‑324.
Fu,
W.,
J.
T.
Randerson,
and
J.
K.
Moore
(2016),
Climate
change
impacts
on
net
primary
production
(NPP)
and
export
production
(EP)
regulated
by
increasing
stratification
and
phytoplankton
community
structure
in
the
CMIP5
models.
Biogeosciences,
13,
5151-‐‑5170.
Geider,
R.
J.,
H.
L.
MacIntyre,
T.
M.
Kana
(1998),
A
dynamic
regulatory
model
of
phytoplanktonic
acclimation
to
light,
nutrients,
and
temperature.
Limnology
and
Oceanography,
43,
679-‐‑694.
Hansell,
D.
A.,
C.
A.
Carlson,
and
R.
Schlitzer
(2012),
Net
removal
of
major
marine
dissolved
organic
carbon
fractions
in
the
subsurface
ocean.
Global
Biogeochemical
Cycles,
26,
GB1016,
doi:10.1029/2011GB004069.
Hilligsoe,
K.
M.,
K.
Richardson,
J.
Bendtsen,
L.
L.
Sorensen,
T.
G.
Nielsen,
and
M.
M.
Lyngsgaard
(2011),
Linking
phytoplankton
community
size
composition
with
temperature,
plankton
food
web
structure
and
sea-‐‑air
CO2
flux,
Deep
Sea
Res.
I,
58,
826-‐‑838.
Holte,
J.,
J.
Gilson,
T.
Talley,
D.
Roemmich
(2010),
Argo
Mixed
Layers,
Scripps
Institution
of
Oceanography/UCSD,
http://mixedlayer.ucsd.edu,
accessed
on
2015/03/15.
Johnson,
K.
S.,
S.
C.
Riser,
and
D.
M.
Karl
(2010),
Nitrate
supply
from
deep
to
near-‐‑
surface
waters
of
the
North
Pacific
subtropical
gyre,
Nature,
465,
1062-‐‑1065.
Jensen,
J.
L.
W.
V.
(1906),
Sur
les
fonctions
convexes
et
les
inégalités
entre
les
valeurs
moyennes.
Acta
Mathematica,
30,
175–93.
78
Karl,
D.
M.,
J.
R.
Christian,
J.
E.
Dore,
D.
V.
Hebel,
R.
M.
Letelier,
L.
M.
Tupas,
and
C.
D.
Winn
(1996),
Seasonal
and
interannual
variability
in
primary
production
and
particle
flux
at
Station
ALOHA,
Deep
Sea
Res.
II,
43,
539-‐‑568.
Karl,
D.
M.,
and
M.
J.
Church
(2014),
Microbial
oceanography
and
the
Hawaii
Ocean
Time-‐‑series
programme.
Nature
Reviews
Microbiology
12,
699-‐‑713.
Karl,
D.
M.,
R.
R.
Bidigare,
R.
M.
Letelier
(2001),
Long-‐‑term
changes
in
plankton
community
structure
and
productivity
in
the
North
Pacific
Subtropical
Gyre:
The
domain
shift
hypothesis.
Deep
Sea
Res.
II,
48,
1449-‐‑1470.
Landsea,
C.
W.,
G.
A.
Vecchi,
L.
Bengtsson,
and
T.
R.
Knutson
(2010),
Impact
of
duration
thresholds
on
Atlantic
tropical
cyclone
counts,
J.
Clim.,
23,
2508–2519,
doi:10.1175/2009JCLI3034.1.
Lomas,
M.
W.,
D.
K.
Steinberg,
T.
Dickey,
C.
A.
Carlson,
N.
B.
Nelson,
R.
H.
Condon,
and
N.
R.
Bates
(2010),
Increased
ocean
carbon
export
in
the
Sargasso
Sea
linked
to
climate
variability
is
countered
by
its
enhanced
mesopelagic
attenuation,
Biogeosciences,
7,
57-‐‑70.
Levine,
N.
(submitted),
Life
in
a
patchy
ocean:
the
impact
of
fine-‐‑scale
spatiotemporal
heterogeneity
on
ecosystem
dynamics.
Ecological
Modelling.
Levy,
M.,
R.
Ferrari,
P.
J.
S.
Franks,
A.
P.
Martin,
and
P.
Riviere
(2012),
Bringing
physics
to
life
at
the
submesoscale,
Geophys.
Res.
Lett.,
39,
L14602.
Levy
M.,
O.
Jahn,
S.
Dutkiewicz,
M.
J.
Follows,
F.
d’Ovidio
(2015),
The
dynamical
landscape
of
marine
phytoplankton
diversity.
J.
R.
Soc.
Interface,
12,
20150481.
Levy,
M.,
L.
Resplandy,
and
M.
Lengaigne
(2014),
Oceanic
mesoscale
turbulence
drives
large
biogeochemical
interannual
variability
at
middle
and
high
latitudes,
Geophys.
Res.
Lett.,
41,
2467-‐‑2474.
Levy,
M.,
A.
P.
Martin
(2013),
The
influence
of
mesoscale
and
submesoscale
heterogeneity
on
ocean
biogeochemical
reactions.
Global
Biogeochemical
Cycles,
27,
1139-‐‑1150.
Liu,
X.,
and
N.
M.
Levine
(2016),
Enhancement
of
phytoplankton
chlorophyll
by
submesoscale
frontal
dynamics
in
the
North
Pacific
Subtropical
Gyre.
Geophys.
Res.
Lett.,
43,
1651–1659.
Marafion,
E.,
P.
M.
Holligan,
R.
Barciela,
N.
Gonzalez,
B.
Mourifio,
M.
J.
Pazo,
and
M.
Varela
(2001),
Patterns
of
phytoplankton
size
structure
and
productivity
in
contrasting
open-‐‑ocean
environments,
Mar.
Ecol.
Prog.
Ser.,
216,
43-‐‑56.
79
McGillicuddy,
D.
J.,
L.
A.
Anderson,
N.
R.
Bates,
T.
Bibby,
K.
O.
Buesseler,
C.
A.
Carlson,
C.
S.
Davis,
C.
Ewart,
P.
G.
Palkowski,
S.
A.
Goldthwait,
D.
A.
Hansell,
W.
J.
Jenkins,
R.
Johnson,
V.
K.
Kosnyrev,
J.
R.
Ledwell,
Q.
P.
Li,
D.
A.
Siegel,
D.
K.
Steinberg
(2007).
Eddy/Wind
interactions
stimulate
extraordinary
mid-‐‑ocean
plankton
blooms.
Science,
316,
1021-‐‑1026.
Mahadevan,
A.
(2016),
The
impact
of
submesoscale
physics
on
primary
productivity
of
plankton,
Ann.
Rev.
Mar.
Sci.,
8,
161-‐‑184.
Mahadevan,
A.,
A.
Tandon,
and
R.
Ferrari
(2010),
Rapid
changes
in
mixed
layer
stratification
driven
by
submesoscale
instabilities
and
winds.
J.
Geophys.
Res.,
115,
C03017.
Mahadevan,
A.,
and
A.
Tandon
(2006),
An
analysis
of
mechanisms
for
submesoscale
vertical
motion
at
ocean
fronts,
Ocean
Model.,
14,
241-‐‑256.
McWilliams
J.
C.
(2016),
Submesoscale
currents
in
the
ocean.
Proc.
R.
Soc.
A,
472,
20160117.
Moore,
J.
K.,
S.
C.
Doney,
K.
Lindsay
(2004),
Upper
ocean
ecosystem
dynamics
and
iron
cycling
798
in
a
global
three-‐‑dimensional
model.
Global
Biogeochemical
Cycles
18,
GB4028.
Moore,
J.
K.,
K.
Lindsay,
S.
C.
Doney,
M.
C.
Long,
K.
Misumi
(2013),
Marine
ecosystem
dynamics
and
biogeochemical
cycling
in
the
Community
Earth
System
Model
[CESM1(BGC)]:
Comparison
of
the
1990s
with
the
2090s
under
the
RCP4.5
and
RCP8.5
scenarios.
J.
Clim.,
26,
9291-‐‑9312.
Omand,
M.
M.,
E.
A.
D'Asaro,
C.
M.
Lee,
M.
J.
Perry,
N.
Briggs,
I.
Cetinic.,
A.
Mahadevan
(2015),
Eddy-‐‑driven
subduction
exports
particulated
organic
carbon
from
the
spring
bloom,
Science,
348,
222-‐‑225.
Perry,
M.
J.
and
D.
L.
Rudnick.
2003.
Observing
the
oceans
with
autonomous
and
Lagrangian
platforms
and
sensors:
the
role
of
ALPS
in
sustained
ocean
observing
systems.
Oceanography,
16,
31-‐‑36.
Pickett,
E.
J.,
D.
L.
Thomson,
T.
A.
Li,
S.
Xing
(2015),
Jensen’s
Inequality
and
the
impact
of
short-‐‑term
environmental
variability
on
long-‐‑term
population
growth
rates.
Riebesell,
U.,
A.
Kortzinger,
and
A.
Oschlies
(2009),
Sensitivities
of
marine
carbon
fluxes
to
ocean
change.
Proc.
Natl.
Acad.
Sci.
U.
S.
A,
106(49),
20,602–20,609.
Sheridan
C.
C.,
and
M.
R.
Landry
(2004),
A
9-‐‑year
increasing
trend
in
mesozooplankton
biomass
at
the
Hawaii
Ocean
Time-‐‑series
Station
ALOHA,
ICES
J.
Mar.
Sci.
61,
457-‐‑463.
80
Saba
V.
S.,
M.
A.
M.
Friedrichs,
M.E.
Carr,
D.
Antoine,
R.
A.
Armstrong,
et
al.
(2010)
Challenges
of
modelling
depth-‐‑integrated
marine
primary
productivity
over
multiple
decades:
a
case
study
at
BATS
and
HOT.
Global
Biogeochemical
Cycles,
24,
GB3020.
Stull,
B.
R.
(1988),
An
Introduction
to
Boundary
Layer
Meteorology,
Kluwer
Academic
Publishers,
Dordrecht,
the
Netherlands.
Shampine,
L.
F.,
and
M.
W.
Reichelt
(1997),
The
MATLAB
ODE
Suite,
SIAM
Journal
on
Scientific
Computing,
18,
1-‐‑22.
Sohm
J.
A.,
E.
A.
Webb
E.
A.,
and
D.
G.
Capone
(2011),
Emerging
Patterns
of
Marine
Nitrogen
Fixation.
Nature
Reviews
Microbiology,
9,
499-‐‑508.
Schmid,
C.,
R.
L.
Molinari,
R.
Sabina,
Y.
H.
Daneshzadeh,
X.
Xia,
E.
Forteza,
and
H.
Yang
(2007),
The
real-‐‑time
data
management
system
for
Argo
profiling
float
observations,
J.
Atmos.
Oceanic
Technol.,
24,
1608–1628.
Sarmiento,
J.
L.,
et
al.
(2004),
Response
of
ocean
ecosystems
to
climate
warming.
Global
Biogeochem.
Cycles,
18(3),
GB3003,
doi:10.1029/
2003GB002134.
Sanudo-‐‑Wilhelmy,
S.
A.,
A.
Kustka,
C.
Gobler,
M.
Yang,
D.
Hutchins,
J.
Burns,
K.
Lwiza,
D.
Capone,
J.
Raven
and
E.
Carpenter
(2001),
Phosphorus
limitation
of
nitrogen
fixation
by
Trichodesmium
in
the
central
Atlantic
Ocean.
Nature,
411,
66-‐‑69.
Thomas
L.N.,
A.
Tandon,
and
A.
Mahadevan
(2008),
Submesoscale
processes
and
dynamics.
In
Ocean
Modeling
in
an
Eddying
Regime,
ed.
MW
Hecht,
H
Hasumi,
pp.
17–38.
Geophys.
Monogr.
Vol.
177.
Washington,
DC:
Am.
Geophys.
Union
Vecchi,
G.
A.,
and
B.
J.
Soden
(2007),
Global
warming
and
the
weakening
ofthe
tropical
circulation.
J.
Clim.,
20,
4316-‐‑4340.
Vecchi,
G.
A.,
and
B.
J.
Soden
(2007),
Increased
tropical
Atlantic
wind
shear
in
model
projections
of
global
warming,
Geophys.
Res.
Lett.,
34,
L08702.
Villarini,
G.,
G.
A.
Vecchi,
T.
R.
Knutson,
and
J.
A.
Smith
(2011),
Is
the
recorded
increase
in
short-‐‑duration
North
Atlantic
tropical
storms
spurious?
J.
Geophys.
Res.,
116,
D10114,
doi:10.1029/2010JD015493.
Wang,
S.
J.,
L.
Cao,
and
N.
Li
(2014),
Responses
of
the
ocean
carbon
cycle
to
climate
change:
Results
from
an
earth
system
climate
model
simulation,
Advances
in
Climate
Change
Research,
5,
123-‐‑130.
81
White,
A.
E.,
A.
L.
Whitmire,
B.
Barone,
R.
M.
Letelier,
D.
M.
Karl,
and
M.
J.
Church
(2015),
Phenology
of
particle
size
distributions
in
the
North
Pacific
gyre.
Journal
of
Geophysical
Research-‐‑Oceans,
120,
7381-‐‑7399.
doi:
10.1002/2015JC010897.
White,
A.
E.,
Y.
H.
Spitz
and
R.
M.
Letelier
(2007),
What
factors
are
driving
summer
phytoplankton
blooms
in
the
North
Pacific
Subtropical
Gyre?
Journal
of
Geophysical
Research-‐‑Oceans,
112,
C12006,
doi:10.1029/2007JC004129.
82
Appendix
A:
Key
physical
and
ecosystem
parameters
used
in
the
SHiP
model.
Parameters
optimized
for
Station
ALOHA
are
denoted
with
asterisks
(n
=
10).
Same
parameter
values
from
Levine
(submitted)
are
used
when
not
listed
here.
Physical
parameters
Values
E-‐‑scaling
factor
used
to
and
estimate
background
vertical
flux
from
E
(number
of
cyclonic
eddies)
0.02*
χ-‐‑scaling
factor
used
to
estimate
disturbance
rate
from
χ
0.6*
PAR-‐‑scaling
factor
used
to
estimate
incoming
radiation
from
PAR
5.45*
Biological
parameters
Values
small
phytoplankton
maximum
growth
rate
(day
-‐‑1
)
1.62*
large
phytoplankton
maximum
growth
rate
(day
-‐‑1
)
2.60*
diazotroph
maximum
growth
rate
(day
-‐‑1
)
0.78*
small
phytoplankton
nitrate
uptake
half-‐‑saturation
coeff.
(mmol
m
-‐‑3
)
0.1*
large
phytoplankton
nitrate
uptake
half-‐‑saturation
coeff.
(mmol
m
-‐‑3
)
0.5*
diazotroph
nitrate
uptake
half-‐‑saturation
coeff.
(mmol
m
-‐‑3
)
2.0
small
phytoplankton
ammonium
uptake
half-‐‑saturation
coeff.
(mmol
m
-‐‑3
)
0.015
large
phytoplankton
ammonium
uptake
half-‐‑saturation
coeff.
(mmol
m
-‐‑3
)
0.04
diazotroph
ammonium
uptake
half-‐‑saturation
coeff.
(mmol
m
-‐‑3
)
0.25
zooplankton
maximum
growth
rate
when
grazing
small
phytoplankton
(day
-‐‑1
)
2.0*
zooplankton
maximum
growth
rate
when
grazing
large
phytoplankton
(day
-‐‑1
)
2.2*
zooplankton
maximum
growth
rate
when
grazing
diazotrophs
(day
-‐‑1
)
1.0
small
phytoplankton
maximum
aggregation
rate
(day
-‐‑1
)
0.3
large
phytoplankton
maximum
aggregation
rate
(day
-‐‑1
)
0.9
fraction
of
grazed
matter
added
to
zooplankton
biomass
when
grazing
small
and
large
phytoplankton
0.25
fraction
of
grazed
matter
added
to
zooplankton
biomass
when
grazing
diazotrophs
0.3
small
phytoplankton
optimal
temperature
for
growth
(°C)
22.5
large
phytoplankton
optimal
temperature
for
growth
(°C)
25
diazotroph
phytoplankton
optimal
temperature
for
growth
(°C)
27.5
fraction
of
zooplankton
biomass
loss
to
detrital
matter
when
grazing
small
phytoplankton
0.15
fraction
of
zooplankton
biomass
loss
to
detrital
matter
when
grazing
large
phytoplankton
0.20
diazotroph
non-‐‑grazing
mortality
(day
-‐‑1
)
0.15
83
Figure
Captions
Figure
1.
Configuration
of
SHiP
model
setup.
The
model
was
run
as
mixed
layer
model
for
a
5°
×
5°
domain
centered
at
Station
ALOHA.
This
schematic
shows
realistic
changes
in
mixed
layer
depth
(dz
for
the
model
colored
in
red)
and
bottom
boundary
layer
(colored
in
light
blue)
at
the
site
for
the
first
three
years
of
the
model
run.
Vertical
lines
indicate
the
frequency
(i.e.
at
monthly
intervals)
of
in
situ
sampling
at
the
site.
Figure
2.
Comparisons
between
model
estimates
and
in
situ
time-‐‑series
measurements
(HOT)
at
Station
ALOHA
for
mixed
layer
averaged
temperature
(a-‐‑b),
primary
production
(c-‐‑d),
and
biomass
of
each
of
the
three
phytoplankton
groups
(e-‐‑f).
Both
time-‐‑series
(a,
c,
e)
and
climatological
results
(b,
d,
f)
are
shown.
In
b
and
d,
the
central
mark
of
each
boxplot
is
the
median,
edges
of
the
box
are
the
25
th
and
75
th
percentiles.
Note
that
in
situ
measurements
are
not
sufficient
to
provide
an
accurate
approximation
for
the
three
phytoplankton
groups
so
are
excluded
in
the
comparisons
(e-‐‑f).
Figure
3.
Export
ratio
calculated
as
particulate
carbon
export
(at
150
m
reference
depth)
divided
by
depth-‐‑integrated
(0–150
m)
primary
production
with
contours
of
0.01
and
0.20.
Monthly
estimates
from
SHiP
model
(green
squares;
2009-‐‑2014)
were
compared
against
observations
at
Station
ALOHA
(blue
filled
circles;
1998-‐‑2011).
Observational
data
were
retrieved
by
digitizing
Figure
4d
in
Karl
and
Church
(2014).
Figure
4.
Comparisons
between
model
estimated
and
satellite
(MODIS-‐‑Aqua)
derived
sea
surface
temperature
(a-‐‑c)
and
chlorophyll
concentration
(d-‐‑f)
in
July
2004.
Histogram
distributions
are
generated
using
all
available
1-‐‑km,
daily
data
(MODIS-‐‑Aqua,
a,
d)
and
model
estimates
from
all
its
subgrid-‐‑scale
environments
(SHiP,
b,
e)
for
the
5°
×
5°
region
during
this
monthly
period.
Results
from
a
model
run
in
a
traditional,
average
environment
mode
(AE,
only
one
environment,
c,
f)
are
also
shown
as
a
comparison.
Figure
5.
The
comparison
between
model
estimated,
ship-‐‑based
measurements
(HOE-‐‑
DYLAN),
and
profiling
float
measurements
at
Station
ALOHA
for
temperature
during
Year
2012.
Mixed
layer
averaged
temperature
was
calculated
from
each
float
(green
stars)
and
HOE-‐‑DYLAN
(red
segments)
profile,
and
compared
against
the
temperature
predicted
in
all
subgrid-‐‑scale
environments
in
the
model
(bean
shape
shading).
Note
that
the
model
estimated
temperatures
are
corrected
for
the
differences
in
the
means
(subtracted
by
0.6°C).
84
Figure
6.
Impact
of
intensity
(vertical
velocity
in
m
day
-‐‑1
)
and
duration
(in
days)
of
fine-‐‑
scale
disturbances
on
phytoplankton
community
composition.
Results
are
presented
as
and
colored
by
the
percentage
contribution
of
large
phytoplankton
biomass
averaged
over
the
12-‐‑year
period.
Note
that
the
frequency
of
disturbances
is
adjusted
based
on
the
intensity
and
duration
to
maintain
the
same
mean
fluxes
of
water
and
nutrients
among
these
experiments.
Figure
7.
Climate
change
experiments
with
reduced
mean
fluxes
of
water
and
nutrients,
and
additional
changes
in
the
duration
of
fine-‐‑scale
disturbances.
Results
are
presented
as
and
colored
by
the
percentage
contribution
of
large
phytoplankton
biomass
within
the
mixed
layer
averaged
over
the
12-‐‑year
period.
Bars
and
circles
represent
results
from
the
experiments
with
reduction
in
mean
flux
only,
while
squares
represent
results
from
the
experiments
with
reduction
in
both
mean
flux
and
duration
of
disturbances.
Figure
8.
Model
estimated
phytoplankton
community
composition
when
the
model
is
run
in
the
heterogeneous
SHiP
mode
(a)
and
homogeneous
AE
mode
(b).
Results
are
presented
as
the
averaged
percentage
contribution
of
each
phytoplankton
group
in
biomass
at
variant
levels
of
nitrate
flux.
Figure
9.
(a)
Model
estimated
carbon
export
flux
out
of
the
mixed
layer
when
the
model
is
run
in
the
heterogeneous
SHiP
mode
and
homogeneous
AE
mode.
(b)
Percentage
change
in
carbon
export
relative
to
the
AE
mode,
with
the
dashed
line
representing
the
average
over
the
12-‐‑year
period.
Figure
10.
Model
estimated
chlorophyll
concentration
over
the
12-‐‑year
period.
The
thick
line
represents
the
averages,
while
thin
lines
represent
the
5
th
and
95
th
percentiles
of
model
estimates
from
all
its
subgrid-‐‑scale
environments
during
each
monthly
period.
Figure 1. Conguration of SHiP model setup. The model was run as mixed layer model for
a 5° × 5° domain centered at Station ALOHA. This schematic shows realistic changes in
mixed layer depth (dz for the model colored in red) and bottom boundary layer (colored in
light blue) at the site for the rst three years of the model run. Vertical lines indicate the
frequency (i.e. at monthly intervals) of in situ sampling at the site.
! "# $ % & ' ( # ) * %+ , -. /
! ""# $ % & ' ()* + , "- $ . / $ 0 1
2
345 6
7 8 ! $ 3 45 6
9 5:&-#;" +, <$ = & +<,(> - ?9@/ $ . / $ A , ()-$ $ 3 45 6
.":#&+*B$ " C &B ' 5, &(
D"*+$ 0 D " * +$ 0 EF D"*+$ 0EG
85
2009 2010 2011 2012 2013 2014
Year
22
23
24
25
26
27
28
Temperature (
o
C)
SHiP
HOT-DOGS
2009 2010 2011 2012 2013 2014
Year
3
4
5
6
7
8
9
10
Prim.Prod.( mgC m
-3
day
-1
)
SHiP
HOT-DOGS
a
c
J F M A M J J A S O N D
23
23.5
24
24.5
25
25.5
26
26.5
Temperature (
o
C)
SHiP
HOT-DOGS
J F M A M J J A S O N D
4
5
6
7
8
9
10
Prim.Prod.( mgC m
-3
day
-1
)
SHiP
HOT-DOGS
Month
b
d
2009 2010 2011 2012 2013 2014
Month
0
5
10
15
POC (mgC m
-3
)
Small
Large
Diazo
J F M A M J J A S O N D
Month
0
5
10
15
POC (mgC m
-3
)
Small
Large
Diazo
e
f
Figure 2. Comparisons between model estimates and in situ time-series measurements
(HOT) at Station ALOHA for mixed layer averaged temperature (a-b), primary production
(c-d), and biomass of each of the three phytoplankton groups (e-f). Both time-series (a, c, e)
and climatological results (b, d, f) are shown. In b and d, the central mark of each boxplot is
the median, edges of the box are the 25th and 75th percentiles. Note that in situ
measurements are not sucient to provide an accurate approximation for the three
phytoplankton groups so are excluded in the comparisons (e-f).
86
0 200 400 600 800 1000 1200
Prim. Prod. mgC m
-2
day
-1
0
20
40
60
80
PC-flux mgC m
-2
day
-1
SHiP (2009-14)
Karl & Church
Figure 3. Export ratio calculated as particulate carbon export (at 150 m reference depth)
divided by depth-integrated (0–150 m) primary production with contours of 0.01 and 0.20.
Monthly estimates from SHiP model (green squares; 2009-2014) were compared against
observations at Station ALOHA (blue lled circles; 1998-2011). Observational data were
retrieved by digitizing Figure 4d in Karl and Church (2014).
e = 0.01
e = 0.20
87
26 28
SST (
o
C)
0
0.05
0.1
Probablity
MODIS
24 26 28
SST (
o
C)
0
0.05
0.1
Probablity
SHiP
24 26
SST (
o
C)
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
Probablity
Traditional
0 0.1 0.2
Chlorophyll (mg m-3)
0
0.05
0.1
0.15
Probablity
MODIS
0.1 0.2
Chlorophyll (mg m-3)
0
0.05
0.1
0.15
Probablity
SHiP
0.1 0.2
Chlorophyll (mg m-3)
0
0.1
0.2
0.3
Probablity
Traditional
a
b c
d e f
Figure 4. Comparisons between model estimated and satellite (MODIS-Aqua) derived sea
surface temperature (a-c) and chlorophyll concentration (d-f) in July 2004. Histogram
distributions are generated using all available 1-km, daily data (MODIS-Aqua, a, d) and model
estimates from all its subgridscale environments (SHiP , b, e) for the 5° × 5° region during this
monthly period. Results from a model run in a traditional, average environment mode (AE,
only one environment, c, f) are also shown as a comparison.
88
22
23
24
25
26
J F M A M J J A S O N D
Temperature (
o
C)
HOE-DYLAN
Argo floats
SHiP environ.
Figure 5. The comparison between model estimated, ship-based measurements (HOE-DYLAN),
and proling oat measurements at Station ALOHA for temperature during Year 2012. Mixed
layer averaged temperature was calculated from each oat (green stars) and HOE-DYLAN (red
segments) prole, and compared against the temperature predicted in all subgridscale
environments in the model (bean shape shading). Note that the model estimated temperatures
are corrected for the dierences in the means (subtracted by 0.6°C).
89
2 6 10 20 30 50
Intensity (velocity in m day
-1
)
5
4
3
2
1
Duration (day)
12
14
16
18
20
22
Large phytoplankton C (%)
Figure 6. Impact of intensity (vertical velocity in m day
-1
) and duration (in days) of
ne-scale disturbances on phytoplankton community composition. Results are
presented as and colored by the percentage contribution of large phytoplankton
biomass averaged over the 12-year period. Note that the frequency of disturbances
is adjusted based on the intensity and duration to maintain the same mean uxes
of water and nutrients among these experiments.
90
5
10
15
20
25
Large Phyto C (%)
Present flux
75% flux
50% flux
SHiP mode AE mode
Figure 7. Climate change experiments with reduced mean uxes of water and nutrients,
and additional changes in the duration of ne-scale disturbances. Results are presented as
and colored by the percentage contribution of large phytoplankton biomass within the
mixed layer averaged over the 12-year period. Bars and circles represent results from the
experiments with reduction in mean ux only, while squares represent results from the
experiments with reduction in both mean ux and duration of disturbances.
91
5 10 15 20 25 30 >35
NO
3
flux (mmol m
-2
day
-1
)
0
50
100
POC (%)
b
Small
Large
Diazo
5 10 15 20 25 30 >35
NO
3
flux (mmol m
-2
day
-1
)
0
50
100
POC (%)
a
Small
Large
Diazo
Figure 8. Model estimated phytoplankton community composition when the model is
running in the heterogeneous SHiP mode (a) and homogeneous AE mode (b). Results are
presented as the averaged percentage contribution of each phytoplankton group in
biomass at variant levels of nitrate ux.
92
Figure 9. (a) Model estimated carbon export ux out of the mixed layer when the model
is run in the heterogeneous SHiP mode and homogeneous AE mode. (b) Percentage
change in carbon export relative to the AE mode, with the dashed line representing the
average over the 12-year period.
2004 2006 2008 2010 2012 2014
Year
50
100
150
200
Export C (mgC m
-2
day
-1
)
AE
SHiP
2004 2006 2008 2010 2012 2014
Year
0
10
20
30
Δ Export C (%)
a
b
Figure 9. (a) Model estimated carbon export flux out of the mixed layer when the model
is run in the heterogeneous SHiP mode and homogeneous AE mode. (b) Percentage
change in carbon export relative to the AE mode, with the dashed line representing the
average over the 12-year period.
93
2004 2006 2008 2010 2012 2014
Year
0
0.1
0.2
0.3
Chlorophyll (mg m
-3
)
SHiP
HOT-DOGS
Figure 10. Model estimated chlorophyll concentration over the 12-year period. The
thick line represents the averages, while thin lines represent the 5
th
and 95
th
percentiles
of model estimates from all its subgridscale environments during each monthly period.
94
95
Chapter
3.
Spatiotemporal
variability
and
environmental
controls
of
phytoplankton
distributions
in
the
Red
Sea
Xiao
Liu
1,*
,
Burton
H.
Jones
2
,
Dale
A.
Kiefer
3
,
Peng
Zhan
4
,
Ibrahim
Hoteit
4
,
Nikolaos
Zarokanellos
2
,
Aditya
R.
Kartadikaria
4
,
and
Naomi
M.
Levine
3
1
Department
of
Earth
Sciences,
University
of
Southern
California,
Los
Angeles,
USA,
90089
2
Red
Sea
Research
Center,
Division
of
Biological
and
Environmental
Sciences
and
Engineering,
King
Abdullah
University
of
Science
and
Technology,
Thuwal,
Saudi
Arabia,
23955
3
Department
of
Biological
Sciences,
University
of
Southern
California,
Los
Angeles,
USA,
90089
4
Division
of
Physical
Sciences
and
Engineering,
King
Abdullah
University
of
Science
and
Technology,
Thuwal,
Saudi
Arabia,
23955
*
:
Corresponding
author;
AHF
M226,
3616
Trousdale
Pkwy,
Los
Angeles,
CA,
USA
90089;
liu284@usc.edu;
+1
(213)
821-‐‑3146
Formatted
for
submission
to
Journal
of
Geophysical
Research
-‐‑
Oceans
96
Abstract
The
Red
Sea
is
home
to
an
extremely
diverse
marine
population
and
also
one
of
the
most
underexplored
marine
ecosystems
on
Earth.
Due
to
the
lack
of
direct
observations,
the
spatiotemporal
dynamics
of
the
complex
Red
Sea
ecosystem
remains
poorly
understood,
and
even
less
is
known
about
the
environmental
controls
over
these
dynamics.
In
this
study,
we
used
a
one-‐‑dimensional,
bio-‐‑optical
phytoplankton
model
in
combination
with
results
of
high-‐‑resolution
hydrological
simulations
to
estimate
depth-‐‑integrated
phytoplankton
growth
and
primary
production
in
the
Red
Sea.
We
also
analyzed
a
suite
of
remotely
sensed
observations
to
identify
the
environmental
variables
that
are
most
relevant
for
controlling
chlorophyll
and
primary
production
in
the
region,
with
a
specific
focus
on
the
role
of
mesoscale
eddies.
Our
results
demonstrate
that
cyclonic
eddies
were
associated
with
elevated
chlorophyll,
with
the
exception
of
in
the
north
central
basin
where
summer
(winter)
chlorophyll
was
significantly
(weakly)
enhanced
by
persistent
anti-‐‑cyclonic
eddies.
Overall,
we
hypothesize
that
nutrient
replenishment
driven
by
dust
deposition
(proxied
by
satellite
aerosol
optical
thickness)
and
the
prevailing
winds
played
an
important
role
in
regulating
primary
production
in
the
Red
Sea.
These
environmental
controls,
however,
varied
both
spatially
and
seasonally.
Further
in
situ
evidence
is
required
in
order
to
assess
the
underlying
mechanisms
hypothesized
here
and
to
fully
understand
the
environmental
regulations
of
the
Red
Sea
ecosystem.
97
1.
Introduction
Largely
due
to
geopolitical
concerns,
the
Red
Sea
is
one
of
the
most
underexplored
Large
Marine
Ecosystems
(LME)
on
Earth.
However,
this
semi-‐‑enclosed
basin,
which
is
home
to
an
immense
diversity
of
marine
biota,
is
hypothesized
to
be
particularly
vulnerable
to
environmental
fluctuations
(Cantin
et
al.,
2010;
Raitsos
et
al.,
2011)
and
might
serve
as
a
case
study
for
how
larger
ocean
basins
might
respond
to
climate.
Climate-‐‑
related
dynamics
have
resulted
in
substantial
modifications
to
the
Red
Sea
ecosystem
over
the
past
several
decades.
For
example,
Raitsos
et
al.
(2011)
documented
an
abrupt
warming
of
0.7°C
in
the
mid-‐‑90s
for
the
region.
Such
a
drastic
change
in
the
environment
is
hypothesized
to
threaten
the
vulnerable
species
and
reduce
biodiversity
in
the
Red
Sea
ecosystem.
However,
systematic
hydrographic
surveys
have
only
been
conducted
over
the
past
two
decades
(e.g.
summer
cruise
in
2001
described
in
Sofianos
and
Johns,
2007).
Due
to
the
scarcity
of
observational
data,
our
understanding
of
the
complex
dynamics
of
the
Red
Sea
ecosystem
and
how
it
may
respond
to
abrupt
and
long-‐‑term
changes
in
the
climate
has
been
severely
hindered
(Triantafyllou
et
al.,
2014).
Overall,
the
Red
Sea
is
considered
an
oligotrophic
region
subject
to
negligible
riverine
inputs
and
precipitation.
It
has
been
hypothesized
that
the
major
flux
of
new
nutrients
into
the
basin
comes
from
the
intrusion
of
nutrient-‐‑rich
waters
from
the
open
ocean
(i.e.
the
Gulf
of
Aden;
Sofianos
and
Johns,
2007).
As
a
result,
on
average,
the
chlorophyll
concentrations
are
much
higher
in
the
south
relative
to
the
north.
However,
this
latitudinal
gradient
in
chlorophyll
is
also
mediated
by
distinct
seasonal
overturning
circulations
in
the
Red
Sea
that
results
in
substantial
spatiotemporal
variability
in
98
nutrient
availability
(Yao
et
al.,
2014a,
b).
In
the
northern
basin,
wintertime
convection,
triggered
by
enhanced
cooling
and
strengthened
winds,
plays
a
critical
role
in
replenishing
nutrients
and
promoting
the
growth
of
phytoplankton
(Raitsos
et
al.,
2013;
Wafar,
2016).
In
contrast,
summertime
chlorophyll
in
the
north
is
significantly
reduced
due
to
intensified
stratification
and
depleted
surface
nutrients
(Brewin
et
al.,
2015).
In
the
south,
a
two-‐‑layer
circulation
is
established
during
the
winter
driven
by
southwesterly
winds,
which
result
in
a
surface
flow
of
fresher,
nutrient-‐‑rich,
Gulf
of
Aden
waters
into
the
Red
Sea
and
a
corresponding
outflow
of
more
saline
Red
Sea
water
at
depth.
The
circulation
switches
in
the
summertime
into
a
three-‐‑layer
system
driven
by
the
prevailing
northeasterly
winds,
with
a
surface
and
a
deep
outflow
from
the
Red
Sea,
and
a
subsurface
intrusion
of
the
nutrient-‐‑rich
Gulf
of
Aden
Intermediate
Water
(GAIW).
GAIW
is
believed
to
serve
as
an
important
nutrient
source
for
the
southern
Red
Sea
during
the
summer
thereby
sustaining
a
high
level
of
summertime
chlorophyll
in
the
region
(Raitsos
et
al.,
2013).
A
unique
feature
of
the
Red
Sea
surface
circulation
is
the
presence
of
a
chain
of
alternating
cyclonic
and
anti-‐‑cyclonic
mesoscale
structures
(10-‐‑100
km;
hereafter
referred
as
eddies)
along
its
elongated
basin,
many
of
which
are
shown
to
be
persistent
features
that
occur
on
time
scales
of
months
to
over
a
year
(Zhan
et
al.,
2014).
Papadopoulos
et
al.
(2015)
suggested
that
cyclonic
eddies
may
play
an
important
role
in
enhancing
Red
Sea
phytoplankton
biomass
in
the
northern
Red
Sea
by
feeding
the
surface
layers
with
upwelled
nutrients.
It
has
also
been
shown
that
eddies
may
interfere
with
the
99
formation
of
deep
water
outflows
in
the
southern
Red
Sea
(Zhai
et
al.,
2015),
although
the
implications
to
ecosystem
dynamics
remain
poorly
understood.
In
addition
to
eddies,
a
number
of
other
environmental
factors
are
thought
to
play
important
roles
in
regulating
nutrient
and
phytoplankton
dynamics
in
the
Red
Sea.
For
example,
Raitsos
et
al.
(2015)
found
that
chlorophyll
concentrations
in
the
southern
Red
Sea
were
correlated
with
the
strength
of
wintertime
Arabian
monsoon,
which
they
hypothesized
was
due
to
increased
horizontal
advection
of
nutrient-‐‑rich
waters
from
the
Gulf
of
Aden.
In
contrast,
Wafar
et
al.
(2016)
suggested
that
monsoonal
processes
may
only
contribute
17–40%
of
the
total
nutrient
supply
to
the
Red
Sea,
and
that
other
sources
of
new
nutrients,
such
as
atmospheric
deposition
and
upwelling
of
regenerated
nutrients,
must
be
significant
in
order
to
sustain
the
measured
levels
of
primary
production.
Due
to
the
proximity
of
a
number
of
active
deserts,
the
Red
Sea
experiences
frequent
dust
storms
that
have
been
captured
by
satellite
images
and
model
simulations
(Kalenderski
and
Stenchikov,
2016;
Anisimov
et
al.,
2017).
Although
few
studies
have
been
conducted
to
investigate
the
direct
impact
of
dust
storms
on
the
Red
Sea
ecosystems,
such
storms
have
been
found
to
stimulate
phytoplankton
blooms
in
other
oligotrophic
regimes,
such
as
the
Mediterranean
Sea,
by
depositing
macronutrients
and/or
micronutrients
into
the
surface
waters
(e.g.
Gallisai
et
al.,
2014).
Due
to
the
lack
of
long-‐‑term
in
situ
observations,
most
previous
work
on
Red
Sea
phytoplankton
dynamics
has
relied
on
satellite
products
such
as
ocean
color
chlorophyll
proxies
(e.g.
Acker
et
al.,
2008;
Raitsos
et
al.,
2011,
2013).
However,
the
global
ocean
color
algorithms
used
to
estimate
chlorophyll
concentrations
from
satellite
data
(e.g.
MODIS-‐‑
100
Aqua
OC3
and
SeaWiFS
OC4)
are
subject
to
larger
uncertainties
in
optically
complex
water
bodies
such
as
the
coastal
and
southern
Red
Sea,
the
optical
properties
of
which
are
complicated
by
the
shallow
bathymetry,
coral
reefs,
and
the
presence
of
enriched
suspended
particles
and
colored
dissolved
materials
(Acker
et
al.,
2008;
Brewin
et
al.,
2015).
In
addition,
chlorophyll
records
may
also
reflect
physiological
adjustments
in,
for
example,
intracellular
chlorophyll
to
carbon
ratio,
due
to
nutrient
limitation
and
photo-‐‑
acclimation
(Behrenfeld
et
al.,
2015),
the
impact
of
which
can
be
particularly
profound
in
the
Red
Sea
where
planktonic
organisms
periodically
experience
intensified
stratification,
sustained
heating,
and
super-‐‑saturated
irradiation.
Finally,
satellite
observations
only
capture
the
pigment
dynamics
within
the
surface
optical
layers,
and
may
not
always
be
indicative
of
growth
and
depth-‐‑integrated
phytoplankton
properties.
As
a
result,
significant
uncertainty
remains
with
regard
to
the
long-‐‑term
dynamics
of
phytoplankton
biomass
and
primary
productivity
in
the
Red
Sea.
This
study
investigates
the
environmental
controls
on
the
spatiotemporal
variability
of
phytoplankton
dynamics
in
the
Red
Sea
with
a
specific
focus
on
the
role
of
mesoscale
eddies.
Specifically,
eddies
were
identified
from
sixteen
years
of
satellite-‐‑based
estimates
of
sea-‐‑surface
height,
and
the
concurrent
satellite
measurements
of
chlorophyll
were
analyzed.
For
chlorophyll
retrieval
we
used
a
new
merged
ocean
color
product
(ESA
CCI)
and
a
new
algorithm
specific
to
the
Red
Sea
(Brewin
et
al.,
2015)
to
obtain
more
accurate
estimates.
We
then
used
a
one-‐‑dimensional,
bio-‐‑optical
phytoplankton
model
in
combination
with
results
of
high-‐‑resolution
hydrological
simulation
to
estimate
vertical
distributions
of
phytoplankton
biomass,
growth
rate,
and
primary
production.
Finally,
to
101
decipher
the
environmental
controls
on
both
chlorophyll
and
primary
production
in
the
Red
Sea,
we
performed
a
series
of
multivariate
analyses
on
a
suite
of
remotely
sensed
observations.
Our
results
demonstrate
the
important
roles
of
eddies,
dust
deposition,
and
wind-‐‑driven
nutrient
replenishment
in
regulating
the
fertility
of
the
Red
Sea.
2.
Data
and
Methods
2.1.
Environmental
data
and
eddy
identification
scheme
Sixteen
years
(September
1997
–
December
2013)
of
remote
sensing
data
for
the
following
environmental
variables
was
analyzed:
air
and
sea
surface
temperature,
sea
surface
height,
radiation,
aerosol
optical
depth,
chlorophyll,
and
winds.
Monthly
average
9-‐‑km
night-‐‑time
Sea
Surface
Temperature
(SST)
was
retrieved
from
both
the
Advanced
Very
High
Resolution
Radiometer
(AVHRR-‐‑Pathfinder)
provided
by
NOAA
NODC
and
the
Moderate
Resolution
Imaging
Spectroradiometer
(MODIS-‐‑Aqua)
provided
by
NASA
OB.DAAC.
Monthly
average
9-‐‑km
photosynthetically
active
radiation
(PAR)
and
aerosol
optical
thickness
(AOT,
a
proxy
for
dust
deposition)
were
retrieved
from
Sea-‐‑Viewing
Wide
Field-‐‑of-‐‑View
Sensor
(SeaWiFS)
and
MODIS-‐‑Aqua
both
provided
by
NASA
OB.DAAC.
Chlorophyll
concentration
(Chl)
was
estimated
from
spectral
Remote
Sensing
Reflectance
(Rrs
(λ))
provided
by
ESA
Ocean
Colour
Climate
Change
Initiative
using
an
improved
regional
algorithm
calibrated
and
validated
for
the
Red
Sea
(OCI-‐‑RG
algorithm,
Brewin
et
al.,
2015).
Monthly
NCEP/NCAR
reanalysis
air
temperature,
and
wind
vectors
and
velocity
(u,
v,
w)
mapped
on
a
2.5°
grid
from
NOAA/OAR/ESRL/PSD
were
used.
8-‐‑day
102
Sea
Level
Anomaly
(SLA)
fields
were
provided
by
AVISO
Altimetry
project.
All
data
were
mapped
onto
a
1/4°
grid
using
linear
interpolation
to
enable
direct
comparisons
and
multivariate
analysis.
In
order
to
determine
how
phytoplankton
dynamics
and
the
environmental
controls
over
these
dynamics
have
varied
spatially
and
seasonally,
we
divided
our
data
first
into
four
regions
–
the
northern
Red
Sea
(NRS),
the
north
central
Red
Sea
(NCR),
the
south
central
Red
Sea
(SCR),
and
the
southern
Red
Sea
(SRS;
Figure
1)
–
and
then
by
season,
winter
(December
–
February),
spring
(March
–
May),
summer
(June
–
August),
and
fall
(September
–
November).
Anticyclonic
(AE)
and
cyclonic
(CE)
eddies
were
detected
from
the
SLA
fields
following
Zhan
et
al.
(2014).
Specifically,
the
geostrophic
velocity
fields
were
first
calculated
from
SLA
gradients,
and
then
an
improved
winging-‐‑angle
method
was
used
to
identify
closed
streamlines,
indicating
complete
eddy
structures,
from
the
geostrophic
velocity
fields
(Ari
Sadarjoen
and
Post,
2000;
Zhan
et
al.,
2014).
A
total
of
748
8-‐‑day
SLA
fields
were
analyzed
and
2711
(1,387
CEs
and
1,324
AEs)
eddies
were
identified.
The
concurrent
chlorophyll
concentrations
inside
the
AE
and
CE
structures
were
compared
against
the
background
field
at
each
grid
cell
to
examine
the
respective
impact
of
AEs
and
CEs
on
chlorophyll.
2.2.
The
bio-‐‑optical
phytoplankton
(bio-‐‑opt)
model
Depth-‐‑integrated
primary
production
was
estimated
using
a
steady
state,
one-‐‑
dimensional,
bio-‐‑optical
phytoplankton
model
(hereafter
referred
as
the
bio-‐‑opt
model;
103
Kiefer
and
Mitchell,
1983;
Pak
et
al.,
1988)
for
the
Red
Sea.
The
model
predicts
vertical
distributions
of
chlorophyll
concentration,
phytoplankton
carbon
(phytoC),
phytoplankton
growth
rate,
and
primary
production
(PP)
in
the
upper
(i.e.
0–200
m)
water
column.
This
simple
model
has
been
shown
to
reproduce
in
situ
observations
at
multiple
sites
in
the
North
Atlantic
(e.g.
Stramska
and
Dickey,
1994;
Ondercin
et
al.,
1995).
The
physical
fields
(e.g.
temperature
and
density
profiles)
from
a
high-‐‑resolution
simulation
of
the
MIT
general
circulation
model
(MITgcm;
Marshall
et
al.,
1997)
for
the
Red
Sea
(Yao
et
al.
2014a,
b;
Zhai
et
al.,
2015)
were
used
to
drive
the
bio-‐‑opt
model.
The
MITgcm
has
been
shown
to
capture
both
the
general
overturning
circulation
patterns
and
the
mesoscale
structures,
including
both
cyclonic
and
anticyclonic
eddies,
in
the
Red
Sea.
The
model
was
run
for
a
total
of
62
years
(January
1952
to
December
2013)
forced
with
high-‐‑frequency
realistic
forcing
following
a
10-‐‑year
spin-‐‑up
to
a
steady
state.
A
detailed
description
of
the
model
configurations
can
be
found
in
Yao
et
al.
(2014a)
and
is
only
briefly
summarized
here.
The
model
domain
included
the
Red
Sea
and
the
Gulfs
of
Suez
Aqaba,
and
Aden,
with
an
open
boundary
located
in
the
Gulf
of
Aden.
The
horizontal
resolution
of
the
model
was
~1.8
km,
and
the
vertical
grid
was
defined
on
25
levels
with
layer
depths
varying
from
10
to
442
m.
The
K-‐‑Profile
Parameterization
(KPP)
scheme
of
Large
et
al.
(1994)
was
used
for
vertical
mixing
dynamics.
The
bio-‐‑opt
model
estimates
vertical
distributions
of
PP
using
surface
values
of
PAR
and
chlorophyll,
vertical
profiles
of
temperature
and
nitrate
concentration
(NO3),
and
the
mixed
layer
depth
(MLD).
The
vertical
temperature
and
density
profiles
were
obtained
from
MITgcm
simulations.
MLD
was
calculated
as
the
first
depth
where
the
density
was
104
0.03
kg
m
-‐‑3
greater
than
that
of
the
surface
(here
we
used
10
m
due
to
fluctuations
in
surface
density).
The
vertical
profiles
of
NO3
were
retrieved
from
NOAA
World
Ocean
Atlas
database
(2013.v2)
due
to
limited
in
situ
NO3
observations.
The
bio-‐‑opt
model
estimated
phytoplankton
growth
rate
based
on
the
factor
that
was
most
limiting
to
growth
among
temperature,
PAR,
and
NO3.
Equations
and
a
list
of
parameters
used
in
the
model
are
given
in
Appendices
A
and
B.
The
water
column
was
first
divided
into
three
idealized
layers:
1)
the
surface
mixed
layer,
defined
as
waters
between
surface
and
MLD
where
vertical
turbulent
transport
is
rapid
and
significantly
faster
than
rate
of
photo-‐‑adaptation,
2)
the
euphotic
zone,
defined
as
stratified
waters
between
MLD
and
the
light
compensation
depth
(DC),
where
there
is
net
growth
of
phytoplankton
and
increasing
chlorophyll
due
to
photo-‐‑adaptation,
and
3)
the
aphotic
zone,
defined
as
waters
below
DC
where
there
is
net
respirational
loss
of
phytoplankton
(Figure
2).
Chlorophyll
within
the
mixed
layer
was
assumed
to
be
constant
and
equal
to
the
surface
value;
phytoC
was
estimated
from
chlorophyll
and
phytoplankton
intracellular
chlorophyll
to
phytoC
ratio
(Chl:C),
which
was
a
function
of
phytoplankton
growth
rate
and
incident
PAR
at
a
given
depth.
Within
the
stratified
euphotic
zone,
phytoC
was
assumed
to
remain
constant,
and
chlorophyll
was
estimated
from
the
Chl:C.
Chlorophyll
increased
with
depth
and
reached
a
maximum
at
DC.
Within
the
aphotic
zone,
incident
PAR
decreased
roughly
exponentially,
and
phytoC
decreased
accordingly.
The
model
was
run
in
MATLAB
(R2014b)
for
each
grid
cell
(size
=
1/4°)
to
estimate
the
distributions
of
phytoplankton
properties
in
the
upper
200
m
at
1-‐‑m
vertical
resolution
105
for
each
month
from
September
1997
to
December
2013.
Monthly
depth-‐‑integrated
PP
was
then
calculated
for
each
grid
cell.
2.3.
Principal
Component
Analysis
(PCA)
A
principal
component
analysis
(PCA)
was
performed
using
the
following
data
to
assess
the
environmental
controls
on
phytoplankton
dynamics
in
the
Red
Sea:
chlorophyll
(CHL),
air
temperature
(airT),
sea
surface
temperature
(SST),
mixed
layer
depth
(MLD),
aerosol
optical
thickness
(AOT,
a
proxy
for
dust
deposition),
longitudinal
wind
vector
(u),
latitudinal
wind
vector
(v),
and
wind
velocity
(τ).
Note
that
the
signs
for
v
data
were
reversed
in
all
analyses
except
the
one
for
the
SCR
and
SR
winter
and
u
data
reversed
in
the
one
for
SCR
and
SR
winter,
so
that
wind
vectors
in
PCA
were
parallel
in
direction
to
the
prevailing
winds
for
each
region
and
season
(i.e.
mostly
northeasterlies
except
southwesterlies
during
the
SCR
and
SR
winter;
Raitsos
et
al.,
2015).
For
each
subset
of
data
separated
by
season
and
region,
a
PCA
was
used
to
decipher
the
correlations
among
multiple
environmental
variables
and
identify
their
respective
relationships
to
both
phytoplankton
chlorophyll
and
PP.
3.
Results
and
discussions
3.1.
Spatial
and
seasonal
variability
in
phytoplankton
chlorophyll
The
16-‐‑year
averaged
chlorophyll
concentration
showed
a
southward
increasing
trend
during
all
four
seasons
(Figure
3),
which
was
consistent
with
Brewin
et
al.
(2015).
This
pattern
was
seen
in
the
chlorophyll
estimates
from
both
the
OCI-‐‑RG
(Figure
3a)
and
106
MODIS-‐‑Aqua
OC3
algorithms
(Figure
3b),
with
the
northern
regions
contributing
only
7-‐‑
8.2%
(NR)
and
11.3-‐‑11.8%
(NCR)
of
the
total
phytoplankton
biomass
in
the
Red
Sea
(Figure
4a).
OC3
algorithm,
however,
yielded
a
mean
chlorophyll
of
0.39
mg
m
-‐‑3
over
the
entire
Red
Sea,
which
was
nearly
twice
the
mean
estimated
by
the
Red
Sea
specific
OCI-‐‑
RG
algorithm
(0.21
mg
m
-‐‑3
).
The
difference
between
the
two
estimates
was
particularly
large
during
the
winter
and
fall
in
the
SR
with
the
OC3
estimates
on
average
250%
greater
than
the
OCI-‐‑RG
estimates.
In
addition,
the
OCI-‐‑RG
algorithm
estimated
that
chlorophyll
was
fairly
evenly
distributed
across
seasons
with
summertime
biomass
accounting
for
26%
of
the
total
annual
biomass
and
winter
accounting
for
31.5%.
In
contrast,
the
OC3
algorithm
estimated
that
the
majority
of
the
chlorophyll
occurred
during
the
winter
months
(40.5%)
with
very
little
biomass
during
the
summertime
(12.3%;
Figure
4b).
3.2.
Impact
of
cyclonic
and
anti-‐‑cyclonic
eddies
on
chlorophyll
Both
cyclonic
(n
=
1,387)
and
anti-‐‑cyclonic
(n
=
1,324)
eddies
were
successfully
identified
from
the
8-‐‑day
averaged
SLA
fields
(e.g.
Figure
5).
As
expected,
the
CE
(AE)
corresponded
with
grid
cells
with
negative
(positive)
SLA,
indicating
upwelling
(downwelling),
which
was
consistent
with
our
common
understanding
of
the
vertical
velocity
fields
in
the
eddy
interiors.
Both
CEs
and
AEs
were
abundant
in
the
central
basins
with
approximately
13.3%
of
the
grid
cells
impacted
by
an
eddy
in
the
winter
and
12.6%
in
the
summer
(Figure
6).
Fewer
eddies
were
observed
in
the
southern
region
(SR).
This
region
(south
of
17°N)
was
therefore
excluded
from
this
analysis
due
to
a
limited
number
of
detected
eddies.
107
The
impact
of
AEs
and
CEs
on
chlorophyll
varied
significantly
both
seasonally
and
spatially,
particularly
in
the
central
basin
(Figure
7).
In
the
northernmost
region,
positive
chlorophyll
anomalies
were
detected
in
the
CEs
in
both
seasons,
while
AEs
in
this
region
were
associated
with
negative
chlorophyll
anomalies
(Figure
7a).
The
enhancement
by
CEs
was
most
significant
in
the
winter
(16.5%)
and
the
reduction
by
AEs
was
most
significant
in
the
summer
(-‐‑13.9%).
This
northernmost
region
is
believed
to
be
extremely
oligotrophic
with
depleted
surface
nutrients
year
round
such
that
phytoplankton
growth
is
likely
driven
by
nutrient
injections
from
deeper
layers
(Raitsos
et
al.,
2013).
Our
results
are
consistent
with
this
hypothesis,
with
CEs
(AEs),
which
correspond
with
upwelled
(downwelled)
nutrients,
resulting
in
enhanced
(reduced)
chlorophyll
(so
called
‘eddy-‐‑
pumping
mechanism’).
Very
different
dynamics
were
observed
in
the
northern
central
basin.
In
this
region,
AEs
were
associated
with
positive
chlorophyll
anomalies
in
both
seasons
with
the
strongest
response
(29.5%)
observed
during
the
summer.
CEs
were
associated
with
weak
negative
chlorophyll
anomalies
(-‐‑5.8%
for
NCR)
in
the
winter
and
positive
anomalies
(8.2%)
in
the
summer.
While
these
dynamics
are
not
consistent
with
the
eddy-‐‑pumping
mechanism,
similar
relationships
have
been
observed
in
nutrient
limited
regions
such
as
the
Sargasso
Sea
and
the
South
Indian
Ocean
(McGillicuddy
et
al.
2007;
Gaube
et
al.,
2014;
Mcgillicuddy
et
al.,
2016
and
references
therein).
We
hypothesize
that
two
mechanisms
could
be
responsible
for
the
observed
dynamics
in
the
central
Red
Sea:
1)
Eddy-‐‑wind
interaction.
Persistent
surface
winds,
regardless
of
directions,
can
impact
eddy
dynamics
causing
a
greater
stress
on
the
edge
of
the
eddy
where
the
direction
108
of
the
winds
opposes
that
of
the
surface
current,
and
a
weaker
stress
on
the
edge
of
the
eddy
where
the
directions
of
the
two
are
the
same
(Mcgillicuddy
et
al.,
2016;
Figure
8).
Such
a
pair
of
imbalanced
stresses
on
eddy
edges
gives
rise
to
upwelling
(positive
chlorophyll
anomalies)
in
the
AEs
and
downwelling
(negative
anomalies)
in
the
CEs
(Dewar
and
Flierl
1987,
Martin
and
Richards,
2001).
Over
the
Red
Sea,
the
prevailing
winds
were
in
stronger
in
the
NCR
(τ
=
4.4
m
s
-‐‑1
)
and
SCR
(4.3
m
s
-‐‑1
)
than
in
the
NR
(3.7
m
s
-‐‑1
;
p<0.05),
with
the
strongest
northeasterlies
found
in
the
NCR
(τ
=
4.6
m
s
-‐‑1
)
in
the
summer.
This
interaction
between
elevated
wind
stress
and
persistent
eddies
may
explain
the
chlorophyll
enhancement
in
AEs
particularly
during
the
summertime
and
the
reduction
in
CEs
in
the
wintertime.
2)
Lateral
eddy
stirring.
In
the
summer,
upwelling
along
the
eastern
coast
driven
by
the
northwesterly
winds
has
been
observed
both
in
MITgcm
simulations
and
in
situ
observations
(Yao
et
al.,
2014a).
This
suggests
that
elevated
nutrient
concentrations
associated
with
the
intruding
GAIW
may
be
upwelled
along
the
eastern
coast
and
stimulate
phytoplankton
growth.
Elevated
chlorophyll
along
the
eastern
coast
relative
to
the
western
coast
and
open
waters
were
observed
in
previous
remote
sensing
studies
(e.g.
Raitsos
et
al.,
2013
in
its
Figure
2)
and
in
this
study
(Figure
3a,
e.g.
summer).
The
stirring
of
this
coastal
upwelling
region
by
both
AEs
and
CEs
in
the
central
basins
may
transport
the
upwelled
nutrients
and
elevated
chlorophyll
of
coastal
origins
laterally
offshore
to
the
central
basin.
However,
in
contrast
to
CEs
in
which
upwelled
waters
disperse
the
stirred
nutrients
and
chlorophyll
outward,
the
AEs
may
act
to
entrain
the
trapped
coastal
nutrients
and
chlorophyll
into
their
interiors
thereby
creating
positive
109
chlorophyll
anomalies.
This
is
supported
by
an
analysis
on
the
lifespan
of
eddies,
which
was
found
to
be
much
longer
in
the
central
basins
(~8
weeks)
than
in
the
NR
(~4
weeks;
Zhan
et
al.,
2014).
The
continuous
entrainment
of
coastal
nutrients
and
chlorophyll
by
persistent
anticyclonic
eddies
may
explain
the
strong
(weak)
positive
summertime
(wintertime)
chlorophyll
anomalies
associated
with
the
AEs
in
the
NCR.
It
is
unclear,
however,
why
CEs
and
AEs
in
the
SCR
and
CEs
in
NCR
in
the
summertime
did
not
follow
the
same
patterns
but
instead
showed
responses
similar
to
the
eddy-‐‑pumping
effect
(i.e.
positive
chlorophyll
anomalies
in
the
CEs
and
negative
anomalies
in
the
AEs).
Most
likely,
the
net
impact
of
eddies
is
the
combined
consequence
of
multiple
processes
(i.e.
eddy-‐‑pumping,
eddy-‐‑wind
interactions,
lateral
eddy
stirring)
discussed
here.
Additional
mechanisms,
such
as
the
formation
of
mode
water
eddies
which
cannot
be
distinguished
from
anticyclonic
eddies
from
SLA
based
eddy-‐‑centric
analysis,
are
also
possible.
Further
observational
evidence
is
needed
in
order
to
assess
these
hypothesized
mechanisms.
3.3.
Environmental
controls
of
phytoplankton
chlorophyll
and
primary
production
Principal
component
analyses
(PCA;
Figures
9,
10)
suggested
seasonally
and
spatially
differential
environmental
controls
of
chlorophyll
and
primary
production
(PP)
in
the
Red
Sea.
A
slight
decoupling
between
chlorophyll
and
PP,
as
indicated
by
the
slight
angles
between
the
two
vectors
on
the
bi-‐‑plots,
was
observed
for
all
regions
and
all
seasons.
However,
chlorophyll
and
PP
were
overall
highly
correlated
and
so
we
focus
on
patterns
of
PP
in
PCAs.
110
In
general
PP
displayed
a
greater
variability
in
the
winter
than
summer,
suggested
by
its
weights
on
the
first
two
axes
of
PCA
(PC1
and
PC2).
For
example,
in
the
winter,
PP
places
a
substantial
weight
on
PC2
in
both
NR
and
NCR,
and
on
both
PCs
in
SCR,
while
this
weight
is
much
reduced
in
the
summer
except
in
the
NCR
where
we
hypothesize
that
anticyclonic
eddies
may
act
to
promote
nutrient
delivery
and
enhance
phytoplankton
production.
The
variability
in
PP
is
found
to
be
smallest
in
the
NR
summer,
with
negligible
weights
on
both
PCs.
This
is
consistent
with
our
finding
that
chlorophyll
was
lowest
in
this
region
and
season
(Figure
3a).
The
PCA
highlighted
nutrient
deposition
from
dust
as
an
important
mechanism
associated
with
enhanced
PP
in
the
Red
Sea.
Specifically,
in
the
winter,
AOT
(a
proxy
for
dust
deposition)
was
correlated
with
PP
in
both
the
northern
(NR)
and
north
central
(NCR)
Red
Sea
(Figures
10a,
b).
This
relationship
was
also
observed
in
the
summer
in
the
NCR
(Figure
10b).
Although
direct
studies
of
the
impact
of
dust
transport
into
the
Red
Sea
on
phytoplankton
growth
is
lacking,
dust
storms
have
been
shown
to
promote
phytoplankton
growth
in
other
oligotrophic
oceans
by
depositing
macro-‐‑
and
micronutrients
into
the
surface
layers
(Gallisai
et
al.,
2014).
We
therefore
hypothesize
that
the
positive
relationship
observed
between
AOT
and
PP
may
be
due
to
the
impact
of
nutrient
deposition
driven
by
dust
storms.
The
relationship
between
AOT
and
PP
was
strongest
when
the
correlation
between
PP
and
wind
velocity
was
the
weakest
(perpendicular
on
the
bi-‐‑plot).
For
example,
in
the
winter
for
the
NR
and
NRC,
PC1
was
highly
correlated
with
wind
direction
(both
u
and
v)
while
PC2
was
highly
correlated
with
111
PP,
Chl,
and
AOT.
This
also
suggests
that
AOT
was
not
always
driven
by
the
prevailing
northeasterlies
in
these
regions.
PCAs
suggested
that
wind-‐‑driven
nutrient
replenishment
plays
the
major
control
on
PP
in
the
SCR
during
both
summer
and
winter
(Figures
9c,
10c).
In
the
winter,
PP
co-‐‑
varies
with
w
and
u,
indicating
the
impact
of
monsoonal
(south)westerlies.
In
contrast
to
the
north,
other
environmental
variables,
such
as
MLD
and
AOT,
explain
much
less
variance
in
the
data
and
show
no
correlation
with
PP.
This
is
consistent
with
previous
studies
which
have
shown
that
the
northward
surface
inflows
from
the
nutrient-‐‑rich
Gulf
of
Aden
driven
by
the
monsoonal
winds
serve
as
a
major
nutrient
source
to
the
southern
regions
(e.g.
Raitsos
et
al.,
2014).
In
the
summer,
wind
direction
reverses
over
this
region,
leading
to
a
surface
southward
outflow
sitting
on
top
of
a
layer
of
northward
intruding
inflow
from
the
Gulf
of
Aden
(Yao
et
al.,
2014a).
It
is
hypothesized
that
stronger
northeasterlies
may
expedite
the
surface
outflows,
which
must
be
balanced
by
faster
water
(and
nutrient)
intrusions
at
the
intermediate
depths.
This
hypothesis
is
supported
by
PCA
(Figure
10c),
which
shows
that
PP
is
positively
correlated
with
u
and
v
(northeasterlies)
and
w
(the
strength
of
winds).
The
combination
of
the
PCA
and
eddy
analysis
suggests
that
drivers
of
variability
in
PP
in
the
Red
Sea
are
spatially
and
temporally
variable
but
that
the
region
can
be
summarized
into
three
main
regimes:
1)
the
northern
Red
Sea
which
is
chronically
nutrient-‐‑limited
and
responds
strongly
to
dust
deposition
and
eddy-‐‑pumping,
2)
the
southern
Red
Sea
where
the
primary
driver
of
PP
dynamics
is
the
entrainment
of
nutrient-‐‑
rich
Gulf
of
Aden
waters
by
northwesterly
winds,
and
3)
the
central
basins
where
we
112
hypothesize
that
persistent
eddies
may
both
interact
with
strong
surface
winds
and
act
to
entrain
coastally
upwelled
waters.
In
addition,
in
the
north
central
region
we
see
enhanced
PP
with
increased
dust
deposition,
similar
to
the
response
in
the
northern
region.
4.
Conclusions
This
study
is
one
of
the
first
to
use
long-‐‑term
and
large
datasets
to
examine
the
spatiotemporal
variability
and
environmental
controls
of
phytoplankton
dynamics
in
the
Red
Sea.
We
analyzed
a
suite
of
remotely
sensed
observations
and
results
from
a
bio-‐‑
optical
phytoplankton
model
to
identify
the
environmental
variables
that
are
most
relevant
to
chlorophyll
and
primary
production
in
the
region,
with
a
specific
focus
on
the
role
of
mesoscale
eddies.
Our
results
demonstrate
that
the
impact
of
eddies
on
chlorophyll
varied
significantly
both
spatially
and
seasonally.
Possible
underlying
mechanisms,
including
eddy-‐‑pumping,
wind-‐‑eddy
interactions,
and
lateral
eddy
stirring
are
discussed.
In
addition
to
eddies,
we
hypothesize
that
nutrient
replenishment
driven
by
dust
deposition
(proxied
by
satellite
aerosol
optical
thickness)
and
the
prevailing
winds
played
an
important
role
in
regulating
primary
production
in
the
Red
Sea.
Further
in
situ
evidence
is
required
in
order
to
assess
the
underlying
mechanisms
hypothesized
here
and
to
fully
understand
the
environmental
regulations
of
the
Red
Sea
ecosystem.
113
Acknowledgements
This
study
was
supported
by
a
NASA
Earth
and
Space
Science
Fellowship
(NNX14AK76H
awarded
to
X.
Liu),
The
University
of
Southern
California,
and
the
King
Abdullah
University
of
Science
and
Technology
(KAUST).
We
thank
F.
Yao,
E.
Teel,
and
A.
Hozumi
for
constructive
comments.
114
References
Acker,
J.,
G.
Leptoukh,
S.
Shen,
T.
Zhu,
S.
Kempler
(2008),
Remotely-‐‑sensed
chlorophyll
a
observations
of
the
northern
red
sea
indicate
seasonal
variability
and
influence
of
coastal
reefs.
J.
Mar.
Syst.
69,
191-‐‑204.
Anisimov,
A.,
W.
Tao,
G.
Stenchikov,
S.
Kalenderski,
P.
Jish
Prakash,
Z.
L.
Yang,
and
M.
Shi
(2017),
Quantifying
local-‐‑scale
dust
emission
from
the
Arabian
Red
Sea
coastal
plain.
Atmos.
Chem.
Phys.,
17,
993-‐‑1015.
Ari
Sadarjoen,
I.,
and
F.
H.
Post
(2000),
Detection,
quantification,
and
tracking
of
vortices
using
streamline
geometry.
Comput.
Graph.,
24,
333–341.
Behrenfeld, M. J., R. T. O’Malley, E. S. Boss, T. K. Westberry, J. R. Graff, K. H. Halsey, A. J.
Milligan, D. A. Siegel, M. B. Brown (2015). Revaluating ocean warming impacts on
global phytoplankton. Nature Clim. Change, doi:10.1038/nclimate2838.
Brewin,
R.
J.
W.,
D.
E.
Raitsos,
G.
Dall’Olmo,
N.
Zarokanellos,
T.
Jackson,
M.
F.
Racault,
E.
Boss,
S.
Sathyendranath,
B.
H.
Jones,
and
I.
Hoteit
(2015),
Regional
ocean-‐‑colour
chlorophyll
algorithms
for
the
Red
Sea.
Remote
Sensing
of
Environment,
165,
64-‐‑
85.
Cantin,
N.
E.,
A.
L.
Cohen,
K.
B.
Karnauskas,
A.
M.
Tarrant,
and
D.
C.
McCorkle
(2010),
Ocean
warming
slows
coral
growth
in
the
central
Red
Sea.
Science,
329,
322-‐‑325.
Dewar,
W.
K.,
G.
R.
Flierl
(1987),
Some
effects
of
wind
on
rings.
J.
Phys.
Oceanogr.
17,
1653–
67
Dreano,
D.,
D.
E.
Raitsos,
J.
Gittings,
G.
Krokos,
and
I.
Hoteit
(2016),
The
Gulf
of
Aden
Intermediate
Water
Intrusion
Regulates
the
Southern
Red
Sea
Summer
Phytoplankton
Blooms.
PLoS
One,
12,
E0168440.
Gallisai,
R.,
F.
Peters,
G.
Volpe,
S.
Basart,
and
J.
M.
Baldasano
(2014),
Saharan
dust
deposition
may
affect
phytoplankton
growth
in
the
Mediterranean
Sea
at
ecological
time
scales,
PLoS
One,
9,
E110762.
Gaube
P.,
D.
J.
McGillicuddy,
D.
B.
Chelton,
M.
J.
Behrenfeld,
and
P.
G.
Strutton
(2014),
Regional
variations
in
the
influence
of
mesoscale
eddies
on
near-‐‑surface
chlorophyll.
J.
Geophys.
Res.
Oceans
119,
8195–220.
115
Kalenderski,
S.,
and
G.
Stenchikov
(2016),
High-‐‑resolution
regional
modeling
of
summertime
transport
and
impact
of
African
dust
over
the
Red
Sea
and
Arabian
Peninsula.
J.
Geophys.
Res.
Atmos.,
121.
Kiefer,
D.
A.,
and
B.
Mitchell.
(1983),
A
simple,
steady-‐‑state
description
of
phytoplankton
growth
based
on
absorption
cross-‐‑section
and
quantum
efficiency.
Limnol.
Oceanogr.
28,
770-‐‑776.
Large,
W.,
J.
McWilliams,
and
S.
Doney
(1994),
Oceanic
vertical
mixing
-‐‑
a
review
and
a
model
with
a
nonlocal
boundary-‐‑layer
parameterization.
Rev.
Geophys.
32,
363-‐‑
403.
Marshall,
J.,
A.
Adcroft,
C.
Hill,
L.
Perelman,
C.
Heisey
(1997),
A
finite-‐‑volume,
incompressible
Navier
Stokes
model
for
studies
of
the
ocean
on
parallel
computers.
J
Geophys
Res-‐‑Oceans,
102,
5753-‐‑5766.
Martin
A.
P.,
K.
J.
Richards
(2001),
Mechanisms
for
vertical
nutrient
transport
within
a
North
Atlantic
mesoscaleeddy.
Deep-‐‑Sea
Res.
II
48,
757-‐‑773.
McGillicuddy
D.
J.,
L.
A.
Anderson,
N.
R.
Bates,
T.
Bibby
T,
K.
O.
Buesseler,
et
al.
(2007),
Eddy/wind
interactions
stimulate
extraordinary
mid-‐‑ocean
plankton
blooms.
Science
316,
1021–26.
McGillicuddy
D.
J.
(2016),
Mechanisms
of
physical-‐‑biological-‐‑biogeochemical
interaction
at
the
oceanic
mesoscale.
Annu.
Rev.
Mar.
Sci.
2016,
125-‐‑159.
Ondercin,
D.,
C.
Atkinson,
D.
A.
Kiefer
(1995),
The
distribution
of
bioluminescence
and
chlorophyll
during
the
late
summer
in
the
North
Atlantic
-‐‑
maps
and
a
predictive
model.
J.
Geophys.
Res.
100,
6575-‐‑6590.
Pak,
H.,
D.
A.
Kiefer,
J.
Kitchen
(1988),
Meridional
variations
in
the
concentration
of
chlorophyll
and
microparticles
in
the
north
pacific-‐‑ocean.
Deep
Sea
Res.
I,
35,
1151-‐‑1171.
Papadopoulos,
V.
P.,
P.
Zhan,
S.
S.
Sofianos,
D.
E.
Raitsos,
M.
Qurban,
Y.
Abualnaja,
A.
Bower,
H.
Kontoyiannis,
A.
Pavlidou,
T.
T.
M.
Asharaf,
N.
Zarokanellos,
and
I.
Hoteit
(2015),
Factors
governing
the
deep
ventilation
of
the
Red
Sea.
J
Geophys
Res-‐‑Oceans,
120,
7493-‐‑7505.
Raitsos
D.
E.,
I.
Hoteit,
P.
K.
Prihartato,
T.
Chronis,
G.
Triantafyllou,
and
Y.
Abualnaja
(2011),
Abrupt
warming
of
the
Red
Sea.
Geophysical
Research
Letters,
38,
L14601
116
Raitsos,
D.
E.,
Y.
Pradhan,
R.
J.
W.
Brewin,
G.
Stenchikov,
and
I.
Hoteit
(2013),
Remote
sensing
the
phytoplankton
seasonal
succession
of
the
Red
Sea.
PLoS
One,
8,
64909.
Raitsos,
D.
E.,
X.
Yi,
T.
Platt,
M.
F.
Racault,
R.
J.
W.
Brewin,
Y.
Pradhan,
V.
P.
Papadopoulos,
S.
Sathyendranath,
and
I.
Hoteit
(2015).
Monsoon
oscillations
regulate
fertility
of
the
Red
Sea.
Geophysical
Research
Letters,
42,
doi:10.1002/2014GL062882.
Sofianos,
S.
S.,
and
W.
E.
Johns
(2007),
Observations
of
the
summer
Red
Sea
circulation.
J.
Geophys.
Res.,
112,
C06025,
doi:10.1029/2006JC003886.
Stramska,
M.,
and
T.
D.
Dickey
(1994),
Modeling
phytoplankton
dynamics
in
the
northeast
Atlantic
during
the
initiation
of
the
spring
bloom.
J.
Geophys.
Res.,
99,
C5,
10241.
Triantafyllou,
G.,
F.
Yao,
G.
Petihakis,
K.
Tsiaras,
D.
E.
Raitsos,
I.
Hoteit
(2014),
A
3D
biophysical
model
to
explore
the
ecosystem
functioning
of
the
Red
Sea.
Journal
of
Geophysical
Research,
119,
1791–1811.
Wafar,
M.,
M.
A.
Qurban,
M.
Ashraf,
K.
P.
Manikandan,
A.
V.
Flandez,
and
A.
C.
Balala
(2016),
Patterns
of
distribution
of
inorganic
nutrients
in
Red
Sea
and
their
implications
to
primary
production.
Journal
of
Marine
Systems,
156,
86-‐‑98.
Yao,
F.,
I.
Hoteit,
L.
J.
Pratt,
A.
S.
Bower,
P.
Zhai,
A.
Kohl,
and
G.
Gopalakrishnan
(2014a),
Seasonal
overturning
circulation
in
the
Red
Sea.
Part-‐‑I:
Model
validation
and
summer
circulation.
J.
Geophys.
Res.
DOI:
10.1002/2013JC009004.
Yao,
F.,
I.
Hoteit,
L.
J.
Pratt,
A.
S.
Bower,
A.
Kohl,
G.
Gopalakrishnan,
and
D.
Rivas
(2014b),
Seasonal
overturning
circulation
in
the
Red
Sea.
Part-‐‑II:
Winter
circulation.
J.
Geophys.
Res.
DOI:
10.1002/2013JC009331.
Zhai,
P.,
A.
S.
Bower,
W.
M.
Smethie,
and
L.
J.
Pratt
(2015a),
Formation
and
spreading
of
Red
Sea
Outflow
Water
in
the
Red
Sea.
J.
Geophys.
Res.
Oceans,
120,
6542-‐‑6563.
Zhai,
P.,
L.
J.
Pratt,
and
A.
S.
Bower
(2015b),
On
the
crossover
of
boundary
currents
in
an
idealized
model
of
the
Red
Sea.
J.
Phys.
Oceanogr.,
45,
1410-‐‑1425.
Zhan,
P.,
A.
C.
Subramanian,
F.
Yao,
A.
R.
Kartadikaria,
D.
Guo,
and
I.
Hoteit
(2016),
The
eddy
kinetic
energy
budget
in
the
Red
Sea.
J.
Geophys.
Res.
Oceans,
121,
4732-‐‑
4747.
Zhan,
P.,
A.
C.
Subramanian,
F.
Yao,
and
I.
Hoteit
(2014),
Eddies
in
the
Red
Sea:
A
statistical
and
dynamical
study.
J.
Geophys.
Res.
Oceans,
119,
3909-‐‑3925.
117
Appendix
A.
Equations
used
in
the
bio-‐‑optical
phytoplankton
model
following
Ondercin
et
al.
(1995).
1,2,3
denotes
the
equations
that
are
applied
to
1)
the
surface
mixed
layer
between
surface
and
MLD,
2)
the
euphotic
zone
between
MLD
and
Dc,
and
3)
the
aphotic
zone
below
Dc.
1
1,2,3
=
%
+
'
+0.76∗
-
∗ℎ()
2
1,2,3
=
7
8
97
:
97
;<=
∗>?@
A(B)
3
1,2,3
=
D
− − −
D
∗
HIJ∗- BH'B ∗'B
4
1,2,3
D
=
7
8
97
:
97
;<=
∗>?@
7
8
97
:
97
;<=
∗>?@
9K.L∗-(B)
5
1,2,3
= ∗
B
K
6
1,2,3
N
=
N
∗
HO B ∗B
7
1,2,3
Q
=
R
S
I9T
UVW∗(X W UY
X
)
8
1,2,3
Z
=
R
S
∗Z(K)
O
[
9Z(K)
9
1,2,3
ZQ
=min
(
Q
,
Z
)
10
1,2,3
`
=
O
a
S
∗R
[X
R
S
11
1
ℎ =ℎ(0)
12
1
c
d
B
=
I
?
R
S
∗c
d
B
O
e
d
9c
d
B
?
K
∗
13
1,2,3
=min
(
c
d
B
,
ZQ
)
14
1
>?@(B)
>(B)
=
I
?
R B ∗(c
d
B 9O
a
)
7
;<=
∗`
S
∗c
d
B ∗O
a
?
K
∗
15
1,2,3
c
d
B
=
R
S
∗c
d
B
O
e
d
9c
d
B
16
2,3
>?@(B)
>(B)
=
R B ∗(c
d
B 9O
a
)
7
;<=
∗`
S
∗c
d
B ∗O
a
118
17
2,3
ℎ = ∗
>?@(B)
>(B)
18
2,3
-
=
I
O B
c
d
K ∗R
S
Hg
g∗O
e
d
19
2
=
>?@(K)
>?@(B)/>(B)
20
3
= 0 ∗
IK∗-(K)(B
;
HB)
119
Appendix
B.
Parameters
(constants)
used
in
the
bio-‐‑optical
Phytoplankton
model.
Parameters
(constants)
Definitions
Values
%
attenuation
coeff.
of
pure
water
0.022
'
attenuation
coeff.
of
detrital
matter
0.18
-
attenuation
coeff.
of
phytoplankton
carbon
0.97
%
absorption
coeff.
of
pure
water
'
absorption
coeff.
of
detrital
matter
0.10
-?@
absorption
coeff.
of
chlorophyll
0.02
j
maximum
growth
rate
2.0
Z
half-‐‑saturation
coeff.
for
nitrogen-‐‑limiting
growth
0.02
Q
half-‐‑saturation
coeff.
for
temperature-‐‑limiting
growth
8.0
j
maximum
photosynthetic
quantum
yield
0.1
`
S
irradiance
at
which
quantum
yield
is
half
its
maximal
value
10
c
d
irradiance
at
which
photosynthetic
rate
is
half
maximal
1.0
120
Figure
Captions
Figure
1.
Map
of
the
Red
Sea
and
domain
separation
into
the
northern
Red
Sea
(NR,
25-‐‑
29°N),
north
central
Red
Sea
(NCR,
21-‐‑25°N),
south
central
Red
Sea
(SCR,
17-‐‑21°N),
and
the
southern
Red
Sea
(SR,
13-‐‑17°N).
The
blue
(winter)
and
red
(summer)
arrows
show
the
prevailing
wind
directions
over
each
region.
Figure
2.
A
schematic
description
of
the
bio-‐‑opt
model.
Modified
from
Ondercin
et
al.
(1995).
MLD
referes
the
mixed
layer
depth
and
Dc
refers
to
the
light
compensation
depth.
These
two
depths
seperate
the
water
column
into
three
idealized
layers:
1)
the
surface
mixed
layer
between
surface
and
MLD,
2)
the
euphotic
zone
between
MLD
and
Dc,
and
3)
the
aphotic
zone
below
Dc.
Figure
3.
Seasonal
mean
chlorophyll
concentration
derived
from
a)
OCI-‐‑RG
algorithm
of
Brewin
et
al.
(2015)
and
b)
OC3
algorithm
averaged
over
the
period
of
September
1997
to
December
2013.
This
figure
is
generated
following
Fig.
11
in
Brewin
et
al.
(2015)
with
data
from
extended
years.
Figure
4.
a)
Regional
and
b)
seasonal
chlorophyll
distributions
in
the
Red
Sea,
presented
as
the
percentage
chlorophyll
contribution
of
each
region
(normalized
for
the
area
of
the
region)
or
season
to
the
total
chlorophyll.
For
comparison,
chlorophyll
estimated
using
both
the
OC3
algorithm
(MODIS-‐‑Aqua
product)
and
the
OCI-‐‑RG
algorithm
(Brewin
et
al.,
2015;
OC-‐‑CCI
merged
product).
Figure
5.
a)
8-‐‑day
averaged
Sea
Level
Anomaly
(SLA)
field
for
November
1-‐‑8,
1998
and
b)
a
cyclonic
eddy,
shown
in
blue,
and
an
anti-‐‑cyclonic
eddy,
shown
in
red,
detected
from
the
SLA
field.
Figure
6.
Eddy
frequency
from
September
1997
to
December
2013
during
a)
winter
and
b)
summer.
The
frequency
is
calculated
at
each
grid
cell
as
the
percenage
period
occupied
by
eddy
structures
relative
to
the
total
period.
Figure
7.
Impact
of
cyclonic
(CE)
and
anti-‐‑cyclonic
(AE)
eddies
on
phytoplankton
presented
as
the
percentage
change
in
chlorophyll
in
the
eddy
structures
relative
to
the
background
field
during
a)
winter
and
b)
summer.
The
percentage
change
was
calculated
at
each
grid
cell
and
then
averaged
over
each
region
(4°
zonal
band).
Results
for
SR
(south
of
17°N)
is
not
shown
due
to
a
limited
number
of
detected
eddies.
121
Figure
8.
A
schematic
illustration
of
how
a
persistent
wind
exerting
on
a)
an
anticyclonic
and
b)
a
cyclonic
eddy
may
lead
to
divergence
(upwelling)
and
convergence
(downwelling),
respectively,
in
the
eddy
interior.
The
wind
causes
a
greater
stress
on
the
edge
of
the
eddy
where
the
direction
of
the
winds
opposes
that
of
the
surface
current,
and
a
weaker
stress
on
the
edge
of
the
eddy
where
the
directions
of
the
two
are
the
same.
This
results
in
a
difference
in
the
magnitude
of
the
associated
Ekman
transport,
causing
divergence/convergence
in
the
eddy
interior.
Modified
from
Figure
6
in
Mcgillicuddy
et
al.
(2016).
Figure
9.
PCA
ordination
bi-‐‑plots
for
PP,
chlorophyll,
and
environmental
variables
for
a)
the
northern
Red
Sea
(NR),
b)
the
north
central
Red
Sea
(NCR),
and
c)
the
south
central
Red
Sea
(SCR)
in
the
winter
(CHL
=
chlorophyll;
AOT
=
aerosol
optical
thickness
as
a
proxy
for
dust
deposition;
MLD
=
mixed
layer
depth;
u
=
longitudinal
wind
vector;
v
=
latitudinal
wind
vector;
τ
=
wind
velocity).
Directions
of
u
and
v
are
determined
by
the
prevailing
winds
over
each
region
and
denoted
in
the
brackets;
e.g.
in
panel
(a)
directions
for
u
and
v
winds
are
reversed,
and
u(e)
indicates
easterly
winds
and
v(s)
indicates
southerly
winds.
For
clarity
purposes,
environmental
variables
that
are
not
correlated
with
PP
and
Chl
or
not
specifically
discussed
(e.g.
temperatures)
are
omitted
on
the
plots.
Figure
10.
PCA
ordination
bi-‐‑plots
for
PP,
chlorophyll,
and
environmental
variables
for
a)
the
northern
Red
Sea
(NR),
b)
the
north
central
Red
Sea
(NCR),
and
c)
the
south
central
Red
Sea
(SCR)
in
the
summer
(CHL
=
chlorophyll;
AOT
=
aerosol
optical
thickness
as
a
proxy
for
dust
deposition;
MLD
=
mixed
layer
depth;
u
=
longitudinal
wind
vector;
v
=
latitudinal
wind
vector;
τ
=
wind
velocity).
Directions
of
u
and
v
are
determined
by
the
prevailing
winds
over
each
region
and
denoted
in
the
brackets.
For
clarity
purposes,
environmental
variables
that
are
not
correlated
with
PP
and
Chl
or
not
specifically
discussed
(e.g.
temperatures)
are
omitted
on
the
plots.
NR
Figure 1. Map of the Red Sea and domain separation into the northern Red Sea
(NR, 25-29°N), north central Red Sea (NCR, 21-25°N), south central Red Sea (SCR,
17-21°N), and the southern Red Sea (SR, 13-17°N). The blue (winter) and red
(summer) arrows show the prevailing wind directions over each region.
NCR
SCR
SR
122
D
C
MLD
Chlorophyll
Phytoplankton Carbon
Figure 2. A schematic description of the bio-opt model. Modied from Ondercin et al.
(1995). MLD referes the mixed layer depth and Dc refers to the light compensation
depth. These two depths seperate the water column into three idealized layers: 1) the
surface mixed layer between surface and MLD, 2) the euphotic zone between MLD and
Dc, and 3) the aphotic zone below Dc.
123
24
o
N
21
o
N
18
o
N
15
o
N
44
o
E 42
o
E 40
o
E 38
o
E 36
o
E 34
o
E 32
o
E
27
o
N
0.03
0.1
0.4
1.5
24
o
N
21
o
N
18
o
N
15
o
N
44
o
E 42
o
E 40
o
E 38
o
E 36
o
E 34
o
E 32
o
E
27
o
N
0.03
0.1
0.4
1.5
24
o
N
21
o
N
18
o
N
15
o
N
44
o
E 42
o
E 40
o
E 38
o
E 36
o
E 34
o
E 32
o
E
27
o
N
0.03
0.1
0.4
1.5
24
o
N
21
o
N
18
o
N
15
o
N
44
o
E 42
o
E 40
o
E 38
o
E 36
o
E 34
o
E 32
o
E
27
o
N
0.03
0.1
0.4
1.5
b
24
o
N
21
o
N
18
o
N
15
o
N
44
o
E 42
o
E 40
o
E 38
o
E 36
o
E 34
o
E 32
o
E
27
o
N
0.03
0.1
0.4
1.5
-3
24
o
N
21
o
N
18
o
N
15
o
N
44
o
E 42
o
E 40
o
E 38
o
E 36
o
E 34
o
E 32
o
E
27
o
N
0.03
0.1
0.4
1.5
-3
24
o
N
21
o
N
18
o
N
15
o
N
44
o
E 42
o
E 40
o
E 38
o
E 36
o
E 34
o
E 32
o
E
27
o
N
0.03
0.1
0.4
1.5
-
Winter Spring Summer
24
o
N
21
o
N
18
o
N
15
o
N
44
o
E 42
o
E 40
o
E 38
o
E 36
o
E 34
o
E 32
o
E
27
o
N
0.03
0.1
0.4
1.5
mg m
- -3
Fall
Chlorophyll
-3
mg m
Chlorophyll
Winter Spring Summer
Fall
a
Figure 3. Seasonal mean chlorophyll concentration derived from a) OCI-RG algorithm of
Brewin et al. (2015) and b) OC3 algorithm averaged over the period of September 1997
to December 2013. This gure is generated following Fig. 11 in Brewin et al. (2015) with
data from extended years.
124
Chl (OC3) Chl (OCI-RG)
0
20
40
60
80
100
Percentage contribution (%)
27.8%
12.3%
19.4%
40.5%
23.4%
26%
19.2%
31.5%
Winter
Spring
Summer
Fall
Chl (OC3) Chl (OCI-RG)
0
20
40
60
80
100
Percentage contribution (%)
56.6%
25.1%
11.3%
7%
53.4%
26.6%
11.8%
8.2%
NR
NCR
SCR
SR
b
Figure 4. a) Regional and b) seasonal chlorophyll distributions in the Red Sea, presented as the
percentage chlorophyll contribution of each region (normalized for the area of the region) or
season to the total chlorophyll. For comparison, chlorophyll estimated using both the OC3
algorithm (MODIS-Aqua product) and the OCI-RG algorithm (Brewin et al., 2015; OC-CCI
merged product).
a
125
SLA (cm)
a b
Figure 5. a) 8-day averaged Sea Level Anomaly (SLA) eld for November 1-8, 1998 and b) a
cyclonic eddy, shown in blue, and an anti-cyclonic eddy, shown in red, detected from the SLA
eld.
126
a
b
Eddy Frequency (%)
CE AE
Figure 6. Eddy frequency from September 1997 to December 2013 during a) winter and b)
summer. The frequency is calculated at each grid cell as the percenage period occupied by
eddy structures relative to the total period.
CE AE
127
a
b
Δ Chlorophyll (%)
CE AE
Figure 7. Impact of cyclonic (CE) and anti-cyclonic (AE) eddies on phytoplankton presented
as the percentage change in chlorophyll in the eddy structures relative to the background
eld during a) winter and b) summer. The percentage change was calculated at each grid
cell and then averaged over each region (4° zonal band). Results for SR (south of 17°N) is not
shown due to a limited number of detected eddies.
AE CE
128
Figure 8. A schematic illustration of how a persistent wind exerting on a) an anticyclonic
and b) a cyclonic eddy may lead to divergence (upwelling) and convergence
(downwelling), respectively, in the eddy interior. The wind causes a greater stress on the
edge of the eddy where the direction of the winds opposes that of the surface current, and
a weaker stress on the edge of the eddy where the directions of the two are the same. This
results in a dierence in the magnitude of the associated Ekman transport, causing
divergence/convergence in the eddy interior. Modied from Figure 6 in Mcgillicuddy et al.
(2016).
Divergence
Convergence
Ekman Transport Wind Eddy currents
Anticyclonic Cyclonic
a b
129
-0.5 0 0.5
Principal Component 1 (29.3%)
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
Principal Component 2 (17.8%)
CHL
MLD
AOT
τ
u(e)
v(s)
PP
-0.6 -0.4 -0.2 0 0.2 0.4 0.6
Principal Component 1 (26.8%)
-0.5
0
0.5
Principal Component 2 (20.8%)
CHL
AOT
τ
u(w)
v(n)
MLD
PP
-0.6 -0.4 -0.2 0 0.2 0.4 0.6
Principal Component 1 (40.7%)
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
Principal Component 2 (18.4%)
CHL
AOT
τ
u(w)
v(n)
PP
Figure 9. PCA ordination bi-plots for PP , chlorophyll, and environmental variables for a) the northern Red Sea (NR),
b) the north central Red Sea (NCR), and c) the south central Red Sea (SCR) in the winter (CHL = chlorophyll; AOT =
aerosol optical thickness as a proxy for dust deposition; MLD = mixed layer depth; u = longitudinal wind vector;
v = latitudinal wind vector; τ = wind velocity). Directions of u and v are determined by the prevailing winds over
each region and denoted in the brackets; e.g. in panel (a) directions for u and v winds are reversed, and u(e)
indicates easterly winds and v(s) indicates southerly winds. For clarity purposes, environmental variables that are
not correlated with PP and Chl or not specically discussed (e.g. temperatures) are omitted on the plots.
Primary production (mmol C m
-2
day
-1
)
a
b
c
130
-0.6 -0.4 -0.2 0 0.2 0.4 0.6
Principal Component 1 (40.6%)
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
Principal Component 2 (18.4%)
CHL
MLD
AOT
τ
u(w)
v(n)
PP
-0.6 -0.4 -0.2 0 0.2 0.4 0.6
Principal Component 1 (37.7%)
-0.5
0
0.5
Principal Component 2 (17.8%)
CHL
MLD
AOT
u(w)
v(n)
airT
PP
-0.6 -0.4 -0.2 0 0.2 0.4 0.6
Principal Component 1 (46.8%)
-0.5
0
0.5
Principal Component 2 (17.2%)
CHL
MLD
AOT
τ
u(w)
v(n)
PP
Figure 10. PCA ordination bi-plots for PP , chlorophyll, and environmental variables for a) the northern Red Sea
(NR), b) the north central Red Sea (NCR), and c) the south central Red Sea (SCR) in the summer (CHL =
chlorophyll; AOT = aerosol optical thickness as a proxy for dust deposition; MLD = mixed layer depth; u =
longitudinal wind vector; v = latitudinal wind vector; τ = wind velocity). Directions of u and v are determined by
the prevailing winds over each region and denoted in the brackets. For clarity purposes, environmental
variables that are not correlated with PP and Chl or not specically discussed (e.g. temperatures) are omitted on
the plots.
Primary production (mmol C m
-2
day
-1
)
b
c
a
131
Abstract (if available)
Abstract
The ocean plays an important role in the global carbon cycle, sequestering approximately half of anthropogenic carbon dioxide into the deep ocean via both the physical and biological ocean carbon pump (Sabine et al., 2004
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Multi-dataset analysis of bacterial heterotrophic variability at the San Pedro Ocean Time-series (SPOT): an investigation into the necessity and feasibility of incorporating a dynamic bacterial c...
PDF
Spatial and temporal investigations of protistan grazing impact on microbial communities in marine ecosystems
PDF
The impact of Lagrangian environmental variability on the growth of phytoplankton
PDF
The dynamic regulation of DMSP production in marine phytoplankton
PDF
Thermal diversity within marine phytoplankton communities
PDF
The impact of the concentration and distribution of dissolved and particulate B-vitamins and their congeners on marine microbial ecology
PDF
Marine biogenic halocarbons: potential for heterotrophic bacterial production and seasonality at San Pedro Ocean Time-series
PDF
Carbonate dissolution at the seafloor: fluxes and drivers from a novel in situ porewater sampler
PDF
Future impacts of warming and other global change variables on phytoplankton communities of coastal Antarctica and California
PDF
How open ocean calcifiers broke the link between large igneous provinces and mass extinctions
PDF
New insights into glacial-interglacial carbon cycle: multi-proxy and numerical modeling
PDF
Identifying functional metabolic guilds: a computational approach to classifying heterotrophic diversity in the marine system
PDF
A multi-omics investigation into breeding shellfish for ocean acidification resilience in the California current system
PDF
The connection of the phosphorus cycle to diazotrophs and nitrogen fixation
PDF
Investigating the global ocean biogeochemical cycling of alkalinity, barium, and copper using data-constrained inverse models
PDF
Harmful algal blooms in the urbanized coastal ocean: an application of remote sensing for understanding, characterization and prediction
PDF
Photophysiological parameters for CO2 and N2 fixation in trichodesmium spp. in natural populations and culture nutrient limitation experiments
PDF
Proteorhodopsin quantification in marine bacteria by LCMS measurement of the retinal chromophore
PDF
The distributions and geochemistry of iodine and copper in the Pacific Ocean
PDF
Carbon dioxide capture using silica supported organoamine adsorbents
Asset Metadata
Creator
Liu, Xiao
(author)
Core Title
The impact of mesoscale and submesoscale physical processes on phytoplankton biomass, community composition, and carbon dynamics in the oligotrophic ocean
School
College of Letters, Arts and Sciences
Degree
Doctor of Philosophy
Degree Program
Ocean Sciences
Publication Date
06/02/2017
Defense Date
04/05/2017
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
carbon cycling,mesoscale,OAI-PMH Harvest,oligotrophic ocean,phytoplankton,submesoscale
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Levine, Naomi (
committee chair
), Hammond, Doug (
committee member
), Heidelberg, John (
committee member
), Jones, Burt (
committee member
), Kiefer, Dale (
committee member
)
Creator Email
liuxiao37k@gmail.com,xiaoliu@princeton.edu
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c40-378067
Unique identifier
UC11257873
Identifier
etd-LiuXiao-5357.pdf (filename),usctheses-c40-378067 (legacy record id)
Legacy Identifier
etd-LiuXiao-5357.pdf
Dmrecord
378067
Document Type
Dissertation
Rights
Liu, Xiao
Type
texts
Source
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Access Conditions
The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...
Repository Name
University of Southern California Digital Library
Repository Location
USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA
Tags
carbon cycling
mesoscale
oligotrophic ocean
phytoplankton
submesoscale