00104682 1

00104682 1
Tracking phytoplankton from space in a
changing Southern Ocean
Dissertation zur Erlangung des akademischen Grades
Doktor der Naturwissenschaften (Dr. rer. nat.)
vorgelegt dem Fachbereich 1 Physik der Universität Bremen
Erstgutachter: Prof. Dr. Astrid Bracher
Zweitgutachter: Prof. Dr. Reiner Schlitzer
Datum der Abgabe: 26.05.2015
Mariana Altenburg Soppa
geb. in Blumenau
2015
1st Reviewer:
Prof. Dr. Astrid Bracher,
Institute of Environmental Physics, Universität Bremen,
Bremen and Alfred-Wegener-Institute for Polar and Marine
Research, Bremerhaven
2nd Reviewer:
Prof. Dr. Reiner Schlitzer,
Alfred-Wegener-Institute for Polar and Marine Research,
Bremerhaven
Anlage zur Dissertation
Eidesstattliche Erklärung
(Erklärung gemäß § 5 Abs. 2 Nr.3 und § 7 Abs. 1 Nr. 2)
Hiermit versichere ich, dass ich
1) die vorliegende Arbeit ohne unerlaubte, fremde Hilfe angefertigt habe
2) keine anderen, als die von mir im Text angegebenen Quellen und Hilfsmittel benutzt habe
3) die den benutzen Werken wörtlich oder inhaltlich entnommenen Stellen als solche
kenntlich gemacht habe.
Bremen, 26.05.2015
___________________
Mariana Altenburg Soppa
Valeu a pena? Tudo vale a pena se a alma não é pequena.
Was it worth it? Everything is worthy If the soul is not small.
Fernando Pessoa, Portuguese Sea, 1934
Acknowledgments
This work would not have been possible without the support of many important people and
institutions. I would like to especially thank:
My supervisor Astrid Bracher for her guidance, assistance and encouragement throughout my
PhD. She has supported me during all the phases of my PhD.
My colleagues of Phytooptics group (Helmholtz-University Young Investigators Group
PHYTOOPTICS): Tilman Dinter, Sonja Wiegmann, Wee Cheah, Aleksandra Wolanin, Rafael
Gonçalves Araujo, Bettina Taylor and Alexandra Cherkasheva. I am thankful for the friendly
atmosphere and great scientific and non-scientific moments we spent together.
Marc Taylor, Christoph Volker, Werner Wosnok, Volker Strass, Ilka Peeken, Robert Johnson,
Mathias van Caspel and Lilian Krug for their help, constructive comments and discussions. Those
were very helpful.
Takafumi Hirata for the orientation in the second paper and for providing the financial support,
together with the Global Observation Mission - Climate Project by Japan Aerospace Exploration
Agency, for my scientific stay at Hokkaido University.
Prof. Reiner Schlitzer for his willingness to review and comment my thesis.
The graduate school POLMAR for organizing nice courses and for providing travel grants to attend conferences and summer schools.
The Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES, Brazil) and AlfredWegener-Institut Helmholtz-Zentrum für Polar- u. Meeresforschung (AWI) for the financial
support of my PhD. I was mainly founded by the CAPES by the research grant BEX 3483/09-6.
Additional support was given by the Phytoscope Project and the Climate Sciences Division at
AWI.
AWI for the nice offices and equipment to work.
Prof. J. Burrows for hosting me as a guest scientist at the Institute of Environmental Physics in the
University of Bremen in the last year of my thesis.
All principal investigators and contributors for collecting, analyzing and sharing the in situ data
used in this thesis available in SeaBASS, MAREDAT, Lter database, Bonus Good Hope project
(LOV, Josephine Ras, Amelie Tale, Herve Claustre), KEOPS cruise (Herve Claustre) and from all
other individual cruises.
The captains, crew, principal investigators and scientists onboard the RV Sonne SO218 and RV
Polarstern cruise ANTXXVIII-3 for their support during these two cruises I took part of.
The Ocean Biology Processing Group of NASA and the NASA Goddard Space Flight Center's
Ocean Data Processing System (ODPS) for the production and distribution of the SeaWiFS and
MODIS satellite data.
The Ocean Colour Climate Change Initiative sponsored by ESA for producing and providing the
merged ocean colour data.
Deutscher Akademischer Austauschdienst (DAAD) for the six months of German classes before
starting my PhD which made my life in Germany much easier.
Tilman, Lilian and Brenner for reading my thesis.
Simone, Tanja, Eva, Doris, Dorothea, Petra, Ina for helping me to stay positive.
I am most grateful to:
My parents and grandparents for their trust in me and for their values passed on to me that made
me believe that this long way is worth.
Brenner for his unending love, patience and encouragement every day of the way. I have been
fortunate to have him.
2
Table of Contents
Abstract .............................................................................................................................................6
List of Publications ............................................................................................................................8
List of Abbreviations .........................................................................................................................9
List of Figures .................................................................................................................................12
List of Tables ...................................................................................................................................16
1
Introduction ..........................................................................................................................18
1.1
Motivation .............................................................................................................................18
1.2
Outline of the thesis ...............................................................................................................21
2
2.1
Scientific background ..........................................................................................................24
Ocean colour retrievals – an overview ..................................................................................24
2.1.1
Chlorophyll-a concentration .........................................................................................26
2.1.2
Phytoplankton absorption .............................................................................................28
2.1.3
Euphotic Depth .............................................................................................................30
2.1.4
Phytoplankton functional types ....................................................................................31
2.1.4.1
Abundance based approach ..........................................................................................32
2.1.4.2
Spectral based approach ...............................................................................................34
2.1.4.3
Potentials and limitations of the methods .....................................................................36
2.2
The Southern Ocean dynamics and phytoplankton blooms ..................................................37
2.2.1
Southern Ocean circulation ..........................................................................................37
2.2.2
Phytoplankton blooms ..................................................................................................40
2.2.2.1
Hypothesis on phytoplankton bloom initiation ........................................................40
2.2.2.1.1
Sverdrup’s Critical Depth Hypothesis ..................................................................41
2.2.2.1.2
Critical Turbulence Hypothesis ............................................................................41
2.2.2.1.3
Disturbance-Recovery-Hypothesis .......................................................................41
2.3
Climate oscillations and the influence in the Southern Ocean ..............................................42
2.3.1
El Niño Southern Oscillation - ENSO..........................................................................42
2.3.2
Southern Annular Mode - SAM ...................................................................................44
3
Study 1: Satellite derived euphotic depth in the Southern Ocean: implication for primary
production modeling........................................................................................................................46
3.1
Motivation .............................................................................................................................46
3.2
Material and Methods ............................................................................................................46
3.2.1
In situ data ....................................................................................................................46
3.2.2
Satellite data ..................................................................................................................47
3.2.3
Zeu derived from ocean colour ......................................................................................48
3
3.2.4
Primary production model ............................................................................................48
3.2.5
Validation and statistical analysis .................................................................................49
3.3
Results ...................................................................................................................................49
3.3.1
Comparison of satellite and in situ Zeu .........................................................................49
3.3.2
Spatial distribution of Zeu-Chla and Zeu-IOP ................................................................51
3.3.3
Primary Production ......................................................................................................53
3.3.3.1
Validation of SeaWiFS and MODIS derived aph ......................................................53
3.3.3.2
Spatial distribution of NPP-Zeu-Chla and NPP-Zeu-IOP ...........................................54
3.4
Discussion..............................................................................................................................55
3.4.1
Validation of Zeu and Chla ............................................................................................55
3.4.2
Zeu spatial distribution ..................................................................................................57
3.4.3
Validation of aph ............................................................................................................58
3.4.4
Primary Production ......................................................................................................58
3.5
Conclusions ...........................................................................................................................59
4
Study 2: Global retrieval of diatoms abundance based on phytoplankton pigments and
satellite data .....................................................................................................................................61
4.1
The role of diatoms in the Southern Ocean ...........................................................................61
4.2 Motivation .................................................................................................................................61
4.3
Data and Methods ..................................................................................................................63
4.3.1
In Situ Measurements of Phytoplankton Pigments.......................................................63
4.3.2
Satellite Data ................................................................................................................63
4.3.3
An Improved Abundance Based Approach ..................................................................63
4.3.3.1
4.3.4
4.4
A Regional Model for the SO ...................................................................................65
Statistical Analysis of Trends .......................................................................................66
Results and Discussion ..........................................................................................................66
4.4.1
The ABAZpd ..................................................................................................................66
4.4.2
Satellite Retrieval of Diatoms using ABAZpd ...............................................................72
4.4
Conclusions ...........................................................................................................................73
5
Study 3: Mean patterns and interannual variability of diatom phenology in the Southern
Ocean ……………………………………………………………………………………………..76
5.1
Motivation .............................................................................................................................76
5.2
Data and Methods ..................................................................................................................76
5.2.1
Satellite data .................................................................................................................76
5.2.2 Polar fronts position .............................................................................................................77
5.2.3
Maximum Sea Ice Extent .............................................................................................77
5.2.4
Climate indices .............................................................................................................77
4
5.2.5
Phenological indices .....................................................................................................78
5.2.6
Statistical analysis ........................................................................................................79
5.3
Results and discussion ...........................................................................................................80
5.3.1
Mean spatial distribution of the phenological indices ..................................................80
5.3.2
Interannual variability ..................................................................................................83
5.3.2.1
Trends .......................................................................................................................84
5.3.2.2
Relationships with ENSO and SAM ........................................................................86
5.3.2.2.1
Correlation maps .......................................................................................................86
5.3.2.2.2
Composite maps of anomalies ...................................................................................88
5.4
6
Concluding remarks...............................................................................................................90
Synthesis and major outcomes .............................................................................................93
Appendix .........................................................................................................................................97
References .....................................................................................................................................104
5
Abstract
Changes in the Southern Ocean (SO) have global consequences. The SO region is responsible for
about half of the global annual uptake of anthropogenic carbon dioxide (CO2) from the atmosphere. As part of the atmospheric CO2 uptake is driven by phytoplankton primary production, a
significant impact on the feedback of phytoplankton is expected under climate change. Indeed,
changes in the atmospheric and ocean temperature, wind patterns and sea ice concentration have
already been documented in the SO region. However, our understanding on how phytoplankton
respond to ongoing and future environmental changes strongly depends on consistent large scale
and long term observations.
As a remote region, substantial time and costs are required to obtain a comprehensive dataset
for the SO. The use of satellite remote sensing is a cost-effective alternative and has led to important insights into the current knowledge of phytoplankton dynamics in this region. However,
this technique does not come without limitations. Ocean colour remote sensing at high latitudes
has to deal with different issues as for example high cloudiness and the limited number of in situ
observations for development and calibration/validation of algorithms. Consequently, there is a
strong need to assess the performance of ocean colour derived-products in the SO.
Ocean colour remote sensing can be used to estimate net primary production (NPP), abundance
of phytoplankton functional types (PFT), as well as their spatial and temporal dynamic. Although
accurate information on NPP is fundamental, large differences have been observed among models
hitherto applied in the SO. Apart from that, different PFTs play specific roles in the oceanic
biogeochemical cycle and this information is of key importance on quantifying oceanic NPP.
Diatoms, for instance, are the main primary producers in the region. Furthermore, additional
insights into their variability due to environmental changes can be gained by studying the
phenology of diatom blooms. The underlying aim of this thesis is to shed light into the above
mentioned topics with a focus on the SO.
Four main objectives have been pursued: 1) to evaluate the satellite retrievals of euphotic depth
(Zeu) and how they influence NPP satellite retrievals; 2) to evaluate and improve the satellite
retrievals of diatom abundance; 3) to investigate the mean patterns and interannual variability of
diatom bloom phenology and 4) to examine the potential of ocean colour products to access
environmental changes in the SO.
The first study analyses satellite retrievals of Zeu, which is the lower limit of the euphotic zone
and where most of the primary production takes place. Although the Zeu is a key parameter in
modelling oceanic NPP from satellite data, assessments of the uncertainties of satellite Zeu
products are scarce. This study investigated existing approaches and sensors to evaluate how
different Zeu products might affect the estimation of NPP in the SO. Zeu was derived from MODIS
and SeaWiFS products of (i) surface chlorophyll-a (Zeu-Chla) and (ii) inherent optical properties
(Zeu-IOP). After comparison with in situ measurements, both approaches have shown robust
results of Zeu retrievals, but spatial differences were of up to 30% over specific regions.
Differences between the sensors were less evident. It was also shown that differences larger than
30% are expected in NPP, depending on the method used to estimate Zeu.
In the second study, focus is given to the major marine primary producer - the diatoms - and to
6
the importance of the SO in developing a global algorithm for the retrieval of diatom abundance
using the Abundance Based Approach (ABA). A large global in situ dataset of phytoplankton
pigments was compiled, particularly with more samples collected in the SO. The ABA was
revised to account for the penetration depth (Zpd) and to improve the relationship between diatoms
and total chlorophyll-a (TChla). The results showed a distinct relationship between diatoms and
TChla in the SO and a new global model (ABAZpd) was suggested to improve the estimation of
diatoms abundance, which improved the uncertainties by 28% in the SO compared with the
original ABA model. In addition, a regional model for the SO was developed which further
improved the retrieval of diatoms by 17% compared with the global ABA Zpd model. The main
finding of this study is that diatom may be more abundant in the SO than previously thought.
In the third study, the new regional model was used to examine the mean pattern and the
interannual variability of the diatom bloom phenology from 1997 to 2012. Ten phenological
indices were used to describe the timing, duration and magnitude of the diatom blooms. The
results show that the mean spatial patterns are generally associated to the position of the Southern
Antarctic Circumpolar Current Front and of the maximum sea ice extent. Furthermore, in several
areas of the SO the interannual variability of the anomalies of the phenological indices is found to
be correlated with the large scale climate oscillations El Niño Southern Oscillation (ENSO) and
Southern Annular Mode (SAM). The composite maps of the anomalies revealed distinct spatial
patterns and opposite events of ENSO and SAM have similar effects in the diatom phenology. For
example, in the Ross Sea region, later start of the bloom and lower biomass were observed
associated with El Niño and negative SAM events; likely influenced by an increase in sea ice
concentration during these events. These results confirm that climate variability and diatom
blooms in the SO are closely linked through environmental changes and these processes can be
accessed using ocean colour remote sensing.
7
List of Publications
This thesis is comprised of the following manuscripts:
Soppa, M. A., Dinter, T., Taylor, B., Bracher, A. (2013). Satellite derived euphotic depth in the
Southern Ocean: implication for primary production modeling. Remote Sensing of Environment,
137, 198-211. doi:10.1016/j.rse.2013.06.017.
Soppa, M. A., Hirata, T., Silva, B., Dinter, T., Peeken, I., Wiegmann S., Bracher, A. (2014). Global retrieval of diatoms abundance based on phytoplankton pigments and satellite data. Remote
Sensing. Remote Sensing, 6 (10), 10089-10106. doi:10.3390/rs61010089.
Soppa, M. A., Völker, C., Bracher, A. Mean and interannual variability of diatom blooms phenology in the Southern Ocean. To be submitted.
Other peer-reviewed manuscripts not included in the thesis:
Sadeghi, A., Dinter, T., Vountas, M., Taylor, B., Peeken, I., Soppa, M. A. and Bracher, A. (2012).
Improvements to the PhytoDOAS method for identification of coccolithophores using hyperspectral satellite data. Ocean Sciences, 8 (6), 1055-1070.
Sadeghi, A., Dinter, T., Vountas, M., Taylor, B., Soppa, M. A. and Bracher, A. (2012). Remote
sensing of coccolithophore blooms in selected oceanic regions using the PhytoDOAS method
applied to hyper-spectral satellite data. Biogeosciences, 9 (6), pp. 2127-2143.
Brandt, A., Vanreusel, A., Bracher, A., Hoppe, C., Lins, L., Meyer-Löbbecke, A., Soppa, M. A.,
Würzberg, L. (2014). Are boundary conditions in surface productivity at the Southern Polar Front
are reflected in benthic activity? Deep-Sea Research Part II-Topical Studies in Oceanography,
108, 51-59.
Hoppe, C. J. M., Ossebaar, S., Soppa, M. A., Cheah, W., Klaas, C., Rost, B., Wolf-Gladrow, D.,
Hoppema, M., Bracher, A., Strass, V., de Baar, H. J. W., Trimborn, S. Controls of primary production in two different phytoplankton blooms in the Antarctic Circumpolar Current. Submitted to
Deep-Sea Research Part II-Topical Studies in Oceanography, May 2015.
8
List of Abbreviations
a
aph
aw
bb
bbw
AAIW
AAO
ABA
ABA*
ABA**
ABAZpd
ABPM
ACC
Allo
AOP
AZ
BD
BDD
BED
BGD
BSD
Butfuco
CA
CAV
CCI
CDOM
Chla
Chlb
Chlc
CI
CM
CMD
Dia
DOAS
DP
DPA
E
Ed
ENSO
ESA
total absorption coefficient
absorption coefficient of phytoplankton
absorption coefficient of pure seawater
total backscattering coefficient
backscattering coefficient of pure seawater
Antarctic Intermediate Water
Antarctic Oscillation index
Absorption Based Approach
Original model of Hirata et al. (2011) parameterized with a new dataset
Original model and fitting parameters of Hirata et al. (2011)
Modified ABA model
Absorption Based Primary Production Model
Antarctic Circumpolar Current
Alloxanthin
Apparent Optical Properties
Antarctic Zone
Bloom Duration
Bloom Decline Duration
Bloom End Date
Bloom Growth Duration
Bloom Start Date
19’-butanoyloxyfucoxanthin
Dia-Chla Amplitude
Dia-Chla averaged over BGD
Climate Change Initiative
Colored dissolved organic matter
Monovinyl chlorophyll-a plus chlorophyllid-a, allomers and epimers
Monovinyl chlorophyll-b
Monovinyl chlorophyll-c
Dia-Chla integrated over BGD
Dia-Chla Maximum
Date of Dia-Chla Maximum
Diatom
Differential optical absorption spectroscopy method
Concentration of the diagnostic pigments
Diagnostic Pigment Analysis
Average absolute percentage of error
Downwelling irradiance
El Niño Southern Oscillation
European Space Agency
9
f-PFT
Fuco
Hexfuco
HNLC
HPLC
IDW
IOP
ITF
Kd
L
LCDW
MAE
MEI
MERIS
MOC
MODIS
N
NADW
NASA
nLw
NPP
NSIDC
OC
PAR
PC
PDW
Perid
PF
PFT
PFZ
PIG
PPC
PSC
QAA
Ra
Rrs
RMSE
SACCF
SAF
SAM
SAMW
SAZ
SeaWiFS
fraction of TChla attributed to a specific PFT
Fucoxanthin
19’-hexanoyloxyfucoxanthin
High nutrient low chlorophyll
High Performance Liquid Chromatography
Indian Deep Water
Inherent Optical Properties
Indonesian Throughflow
Attenuation coefficient
Radiance
Lower Circumpolar Deep Water
Mean Absolute Error
Multivariate El Niño Southern Oscillation index
Medium Resolution Imaging Spectrometer
Meridional overturning circulation
Moderate Resolution Imaging Spectroradiometer
Number of data points/samples
North Atlantic Deep Water
National Aeronautics and Space Administration
Normalized water leaving radiance
Net primary production
National Snow and Ice Data Center
Ocean Colour
Photosynthetic Available Radiation
Photosynthetic carotenoids
Pacific Deep Water
Peridinin
Polar Front
Phytoplankton Functional Type
Polar Front Zone
Quality controlled HPLC pigment dataset
Photoprotective carotenoids
Phytoplankton Size Classes
Quasi Analytical Algorithm
Radiance anomaly
Remote sensing reflectance
Root-mean-square error
Southern Antarctic Circumpolar Front
Subantarctic Front
Southern Annular Mode
Subantarctic Mode Water
Subantarctic Zone
Sea-viewing Wide Field-of-view Sensor
10
SCIAMACHY Scanning Imaging Absorption Spectrometer for Atmospheric Chartography
SNR
Signal-to-noise ratio
SO
Southern Ocean
SOM
Self-Organizing Maps
SST
Sea surface temperature
STF
Subtropical Front
TACC
Sum of all accessory pigments
TChla
Total Chlorophyll-a (Chla plus divinyl chlorophyll-a)
TChlb
Monovinyl chlorophyll-b plus divinyl chlorophyll-b
UCDW
Upper Circumpolar Deep Water
UV
Ultraviolet
VIS
Visible
VRS
Vibrational Raman Scattering
Zea
Zeaxanthin
Zeu
Euphotic depth
Zpd
Penetration depth
11
List of Figures
Figure 1.1. On the top: illustration of the major phytoplankton functional types living in the
global ocean. On the bottom: simplified scheme of biological carbon pump. Adapted from
http://earthobservatory.nasa.gov/. ................................................................................................... 19
Figure 2.1. A schematic showing the components of the total upwelling radiance at the sensor:
water leaving radiance (Lw), atmospheric radiance (La) and surface-reflected radiance (Lr). Based
on Martin (2004). ........................................................................................................................... 24
Figure 2.2. (a) Specific pigment absorption spectra for major phytoplankton pigments:
chlorophyll-a (Chla), chlorophyll-b (Chlb), chlorophyll-c (Chlc), photosynthetic carotenoids (PC)
and photoprotective carotenoids (PPC). Adapted from IOCCG (2014); originally from Bidigare et
al. (1990). (b) Spectral changes in the Rrs with respect to changes in Chla. Numbers above the
lines indicates the Chla concentration in mg m-3. Adapted from Dierssen (2010). ........................ 27
Figure 2.3. (a) Relationship between Chla and Diatom (% Chla). To get % Chla, the f-Diatom is
multiplied by 100. The orange line represents the model and fitting parameters of Hirata et al.
(2011) for Diatoms, as presented in Table 3, red line represents a running mean of the in situ data.
Modified from Hirata et al. (2011). (b) Mean % of Chla of Diatoms over 1998-2010 for January.
Modified from IOCCG (2014). ...................................................................................................... 34
Figure 2.4. Climatology of the dominance of multiple PFTs for January (1997-2010) applied to
SeaWiFS. White areas represent missing data or unidentified phytoplankton type. Modified from
Ben Mustapha et al. (2014). ........................................................................................................... 35
Figure 2.5. Monthly Chla for specific phytoplankton groups – October 2009. The coloured circles
are the Chla of the respective groups derived from HPLC pigment concentration and CHEMTAX
analysis of in situ samples taken during TransBrom Sonne cruise (9–23 October 2009) (IOCCG
2014)............................................................................................................................................... 36
Figure 2.6. (a) Two-dimensional view of the Southern Ocean part of the meridional overturning
circulation (MOC). NADW, North Atlantic Deep Water; Indian Deep Water – IDW; Pacific Deep
Water – PDW; Indonesian Throughflow – ITF; Upper Circumpolar Deep Water - UCDW; Lower
Circumpolar Deep Water - LCDW; Subantarctic Mode Water (SAMW), Antarctic Intermediate
Water (AAIW). Adapted from Talley (2013). (b) Southern Ocean bathymetry (m) overlaid with
the mean position of the maximum sea ice extent (1997-2012, solid white line, Fetterer et al.
2002), Subtropical Front (Orsi et al. 1995) and the Antarctic Circumpolar Current fronts (Orsi et
al. 1995, Salle et al. 2008). From north to south: Subtropical Front (STF, dashed line),
Subantarctic Front (SAF, solid line), Polar Front (PF, dashed line), and Southern Antarctic
Circumpolar Front (SACCF, solid line). SAZ, Subantarctic Zone; PFZ, Polar Front Zone; AZ,
Antarctic Zone. ............................................................................................................................... 39
Figure 2.7. On the left: ocean-atmosphere processes that occur during (a) normal years, (b) El
Niño event and (c) La Niña event. Modified from Robinson (2010). On the right: composite of
anomalies of Sea Surface Temperature from November to March in (d) El Niño (1965, 1972,
1982, 1987, 1991, 1993, 1994, 1997, 2002) and (e) La Niña (1950, 1955, 1956, 1964, 1971, 1974,
1988, 1998, 1999) events. The maps were produced from the data display pages of the
NOAA/ESRL
Physical
Sciences
Division,
Boulder
Colorado,
available
at
12
http://www.esrl.noaa.gov/psd/. ....................................................................................................... 43
Figure 2.8. Regression of anomaly patterns of (a) atmospheric pressure at 700 mb and (b) wind
stress (dyne cm−2) onto the SAM index. Modified from Lovenduski (2007). ............................... 44
Figure 3.1. On the left, location of the in situ measurements in light grey and the matched ones
with satellite in black: (a) Zeu (1288), (b) Chla (1032) and (c)ܽ‫(݄݌‬465). On the right, the
respective relative frequency distribution of the matched in situ measurements. .......................... 47
Figure 3.2. Scatterplots of satellite Zeu against in situ Zeu. (a) and (c) Zeu is derived from Chla
approach (Zeu-Chla), (b) and (d) Zeu is derived from the IOP approach (Zeu-IOP). The solid line
represents the regression and the dotted line represents 1:1 line as reference. .............................. 50
Figure 3.3. (a) Scatterplots of satellite and in situ Chla. The dotted line represents the 1:1 line as
reference. (b) Relative differences between satellite Chla and in situ Chla. The dotted line
represents the zero line. .................................................................................................................. 51
Figure 3.4. Spatial distribution of Zeu in the Southern Ocean (climatology of February). The white
pixels correspond to areas with no data.......................................................................................... 52
Figure 3.5. Spatial distribution of the relative percentage of difference between SeaWiFS and
MODIS. The white pixels correspond to areas with no data. ......................................................... 53
Figure 3.6. Spatial distribution of net primary production (in the figure caption called PP) in the
Southern Ocean (climatology of February). NPP-Zeu-Chla (left), NPP-Zeu-IOP (right) and relative
percentage of difference between NPP-Zeu-Chla and NPP-Zeu-IOP (center). The white pixels
correspond to areas with no data. ................................................................................................... 55
Figure 4.1. Distribution of the quality controlled in situ measurements. The SO, region south of
50°S, is the portion of the global ocean presented in blue. ............................................................ 62
Figure 4.2. A flow chart of the processing steps conducted to retrieve diatom abundance using
ABAZpd............................................................................................................................................ 65
Figure 4.3. Relationship between TChlaZpd and f-DiatomZpd: (a) Global dataset (N = 2806), (b)
global dataset excluding SO data (N = 1737) and (c) SO data (N = 1069). The datasets were
smoothed with a 5 point running mean to improve the signal-to-noise ratio (Hirata et al. 2011) The
green and blue lines represent the new model (ABAZpd) and the model of Hirata et al. (2011)
(ABA*) parameterized with the DPZpd dataset. The red line represents the original model and
fitting parameters of Hirata et al. (2011) (ABA**). The fitting parameters are presented in Table
4.2. The MAE values refer to the errors in terms of f-DiatomZpd. Note that we could not fit the
global models to the SO dataset exclusively. The cyan and green lines in (c) represent the regional
model for the SO and the ABAZpd plotted with the global fitting parameters as reference. ........... 68
Figure 4.4. Monthly mean TChlaZpd (mg m-3) of diatoms for February 2003 using the ABAZpd
model parameterized with: (a) Global dataset (average = 0.060 mg m-3) and (b) global dataset
excluding SO data (average = 0.041 mg m-3). White areas correspond to waters with depths
shallower than 200 m or without satellite information. .................................................................. 70
Figure 4.5. On the left: relationship between TChlaZpd and DiatomZpd in the SO with the fit
function plotted in blue (log10 transformed data). On the right: validation calculated with both
log10 transformed data (e.g. log10(y+0.00003)). The red line represents the 1:1 line. .................... 71
Figure 4.6. Climatology of TChlaZpd of diatoms (mg m-3) for the months of January to December
13
based on the period 2003-2013 retrieved using the ABAZpd model. White areas correspond to
waters with depths shallower than 200 m or without satellite information. ................................... 73
Figure 5.1. Time series (adimensional) of annual Multivariate ENSO Index (MEI, solid line) and
Antarctic Oscillation index (SAM, dashed line). ........................................................................... 78
Figure 5.2. Schematic of the indices used to describe the diatom phenology. .............................. 79
Figure 5.3. Spatial distribution of the mean diatom phenology in 1997 – 2012: (left) bloom start
date – BSD, (center) date of Dia-Chla maximum – CMD, (right) bloom end date - BED. Grey
areas represent missing data. Black solid lines show the mean position of the Polar Front (Sallee
et al. 2008) over 1997-2012. Dashed lines show the Southern Antarctic Circumpolar Front (Orsi et
al. 1995). Purple line displays the mean position of the maximum sea ice extent over 1997-2012
(Fetterer et al. 2002). ...................................................................................................................... 81
Figure 5.4. Same as Figure 5.3, but for bloom growth duration (BGD), bloom decline duration
(BDD) and total duration (BD) of the blooms. Units are in week. ................................................ 82
Figure 5.5. Same as Figure 5.4, but for Dia-Chla maximum (CM), Dia-Chla amplitude (CA), DiaChla average (CAV) and Dia-Chla integrated over the growth duration (CI). Units are in mg m -3.
........................................................................................................................................................ 83
Figure 5.6. Schematic representation of the latitudinal variability (longitudinal average) of the
phenological indices: bloom start date (BSD), date of Dia-Chla maximum (CMD), bloom end date
(BED), bloom growth duration (BGD), bloom decline duration (BDD), bloom duration (BD), DiaChla maximum (CM), Dia-Chla amplitude (CA), Dia-Chla averaged BGD (CAV), Dia-Chla
integrated over BGD (CI). .............................................................................................................. 83
Figure 5.7. Spatial distribution of the relative standard deviation (RSD) of the bloom start date
(BSD) and Dia-Chla amplitude (CA) for 15 years of data (1997-2012). Grey areas represent
missing data. Black continuous lines represent the mean Polar Front position (Sallee et al. 2008)
and dashed black lines the standard deviation of the position over 1997-2012. Purple continuous
lines indicate mean of the maximum sea ice extent (Fetterer et al. 2002) and dashed purple lines
the standard deviation of the position for the years of 1997 to 2012. White line indicates the mean
position of the Southern Antarctic Circumpolar Front (Orsi et al. 1995). ..................................... 84
Figure 5.8. Trends of the standardized anomalies of date of Dia-Chla maximum (CMD) and DiaChla maximum (CM). Reddish colour indicates a positive trend and bluish indicates a negative
trend. Only statistically significant trends (p < 0.05) are shown. The stars highlight the regions
between Malvinas and South Georgia Islands (green) and south of 60°S between 120°E to 150°E
(black) and 60°E to 120°E (grey). .................................................................................................. 85
Figure 5.9. Correlation coefficients of the standardized anomalies of date of Dia-Chla maximum
(CMD) and Dia-Chla maximum (CM) vs. ENSO (MEI) and SAM (AAO) indices. Only
statistically significant trends (p < 0.05) are shown. Black and purple lines indicate the mean
position of the Polar Front and maximum sea ice extent over 1997-2012, respectively. ............... 87
Figure 5.10. Composites of bloom start date (BSD) standardized anomalies during El Niño
(N=6), La Niña (N=8), positive SAM (N=7) and negative SAM (N=4) years. Grey areas represent
missing data. Black lines show the mean position of the Polar Front (Sallee et al. 2008) over
1997-2012. Purple line displays the mean position of the maximum sea ice extent (Fetterer et al.
2002) over 1997-2012. The white boxes depict the Weddell Sea region (dashed) and the sector
14
between 120°W and 180°W. .......................................................................................................... 88
Figure 5.11. Same as Figure 5.10 but for Dia-Chla maximum. ..................................................... 89
Figure A1. Scatterplots of satellite Zeu-Chla and Zeu-IOP against in situ Zeu south of 60°S. The
solid line represents the regression and the dotted line represents the 1:1 line as reference. ......... 97
Figure A2. Scatterplot of the validation for the global DPZpd dataset (N= 1182): (a) new model
(ABAZpd), (b) model of Hirata et al. (2011) parameterized with the DPZpd dataset (ABA*) and (c)
original model and fitting parameters of Hirata et al. (2011) (ABA**). The samples located in the
SO are presented in grey (N = 460), together with the statistics of the validation. The red line
represents the 1:1 line. The statistics were calculated with log10 transformed data (e.g.
log10(y+0.00003)). .......................................................................................................................... 98
Figure A3. Climatology of TChlaZpd of diatoms (mg m-3) using the regional algorithm for the SO
based on 2003-2013 period. The austral winter months of May, June, July and August are not
presented due to too few number observations available in these months. White areas correspond
to waters with depths shallower than 200 m or without satellite information. ............................... 99
Figure A4. Spearman correlation coefficients between the time series of phenological indices (15yr, 1997 – 2012): bloom start date (BSD), date of Dia-Chla maximum (CMD), bloom end date
(BED), bloom growth duration (BGD), bloom decline duration (BDD), bloom duration (BD), DiaChla maximum (CM), Dia-Chla amplitude (CA), Dia-Chla averaged BGD (CAV), Dia-Chla
integrated over BGD (CI). Only statistically significant trends (p < 0.05) are shown. White areas
correspond to non-significant correlations or missing data. Black and purple lines indicate the
mean position of the Polar Front and the mean position of the maximum sea ice extent over 19972012, respectively. ........................................................................................................................ 100
Figure A5. Partial correlation coefficients of the standardized anomalies of date of Dia-Chla
maximum (CMD) and Dia-Chla maximum (CM) vs. ENSO (MEI) and SAM (AAO) indices. Only
statistically significant trends (p < 0.05) are shown. Black and purple lines indicate the mean
position of the Polar Front and the mean position of the maximum sea ice extent over 1997-2012,
respectively. .................................................................................................................................. 101
Figure A6. Composites of bloom duration standardized anomalies during El Niño (N=6), La Niña
(N=8), positive SAM (N=7) and negative SAM (N=4) years. Grey areas represent missing data.
Black lines show the mean position of the Polar Front (Sallee et al. 2008) over 1997-2012. Purple
line displays the mean position of the maximum sea ice extent (Fetterer et al. 2002) over 19972012. The white boxes depict the Weddell Sea region (dashed) and the sector between 120°W and
180°W........................................................................................................................................... 102
Figure A7. Composites of bloom start date (BSD) and Dia-Chla maximum (CM) standardized
anomalies during amplified years. Left plot: El Niño and negative SAM (N=3). Right plot: La
Niña and positive SAM (N=6). Grey areas represent missing data. Black and purple lines indicate
the mean position of the Polar Front and the mean position of the maximum sea ice extent over
1997-2012, respectively. The white boxes depict the Weddell Sea region (dashed) and the sector
between 120°W and 180°W. ........................................................................................................ 103
15
List of Tables
Table 2.1. Phytoplankton functional types and their respective diagnostic pigments (Hirata et al.
2011; IOCCG 2014; Vidussi et al. 2001)........................................................................................ 32
Table 2.2. Equations used to calculate f-PFT/PSC (adapted from Hirata et al. 2011). DPw is the
sum of the weighted concentration of the DP and ai are the partial coefficients as in Uitz et al.
(2006). ............................................................................................................................................ 33
Table 2.3. Models and parameters used to estimate the f-PFT/PSC (adapted from Hirata et al.
2011). .............................................................................................................................................. 33
Table 2.4. Potential and limitations of the algorithms described in section 2.3. Based on IOCCG
(2014). ............................................................................................................................................ 37
Table 3.1. Statistical results of the comparison between QAA-satellite derived aph and in situ aph.
........................................................................................................................................................ 54
Table 4.1. The partial regression coefficients and standard deviation (in brackets) where available.
The number of samples is indicated by N. The empty fields indicate that the coefficient is not
statistically significant…………………………………………….................................................67
Table 4.2. Models of f-DiatomZpd as a function of TChlaZpd and their respective fitting parameters
used to plot the lines in Figure 4.3a and 4.3b. Note that we could not fit the global models to the
SO data exclusively. The fitting parameters of the original ABA model of Hirata et al. (2011)
(ABA**) do not change and therefore they are presented only once in the
table……………………………………………………………………...………………………...69
Table 4.3. Statistical results of the fits for the global dataset and global excluding SO data using
the fitting parameters of Table 4.2. Note that we could not fit the global models to the SO data
exclusively. The fitting statistics for the SO dataset refer to the regional SO model (Figure 4.5).
The MAE is given in f-DiatomZpd for the global models and for the regional model in mg m -3
(log10-transformed data). ................................................................................................................ 69
Table 4.4. Statistical results of the validation in terms of diatoms abundance. Note that we could
not fit the global models to the SO data exclusively. The results for the SO dataset correspond to
the global models using the global fitting parameters and the regional model. The MAE is given in
mg m-3. The statistics were calculated with log10-transformed data (e.g., log10(y+0.00003)). ....... 70
Table 5.1. Phenological indices…………………………………………………………………...79
16
Chapter 1
Introduction
17
1
Introduction
1.1 Motivation
Phytoplankton are microscopic unicellular algae and bacteria (cyanobacteria) living in freshwater
and marine ecosystems. Perhaps the most important aspect of phytoplankton is the fact that they
are autotroph photosynthetic active organisms and primary producers of organic compounds.
During photosynthesis they absorb light to break water molecules and assimilate carbon dioxide
(CO2) by transforming the dissolved inorganic carbon into organic carbon. Subtracting respiratory
and non-respiratory carbon release gives the net primary production (NPP), which is the main
food supply of the aquatic ecosystem and a key component of the global carbon cycle.
What makes marine phytoplankton unique is that although their biomass represents less than
1% of the global photosynthetic biomass, they are responsible for about half of the total NPP of
the earth (Falkowski et al. 1998; Falkowski et al. 2004; Field et al. 1998). Estimates of marine
NPP vary from 39 (Rousseaux and Gregg 2014) to 50 PgC yr-1 (Carr et al. 2006), similar to the
magnitude of terrestrial NPP (~ 56 PgC yr–1) (Ito 2011). However, the turnover time of marine
phytoplankton is short, about one week, ultimately higher than terrestrial plants (Falkowski et al.
1998), which makes marine NPP more efficient than terrestrial NPP. Microalgae, seagrasses and
other marine primary producers are responsible for other ~ 5 PgC yr–1 (Duarte and Cebrian 1996).
The current number of marine phytoplankton species in the global oceans is not clear. In the
early 1990’s the estimate was of ca. 4000 species (Sournia et al. 1991), but nowadays, with the
advances in technology and sampling effort over the last decades, this number is likely to be
underestimated. Generally, phytoplankton can be categorized into major groups according to their
size or biogeochemical function (Nair et al. 2008). There are three main phytoplankton size
classes (PSC): microplankton (> 20 μm), nanoplankton (2-20 μm) and picoplankton (< 2 μm), and
four main phytoplankton functional types (PFT): calcifiers (e.g. coccolithophores), silicifiers (e.g.
diatoms, chrysophytes), nitrogen fixers (e.g. cyanobacteria) and dimethyl sulphide producers (e.g.
dinoflagellates, haptophytes) (Figure 1.1 upper panel). From a biogeochemical perspective, the
classification of phytoplankton based on their functionality is preferred to size-based since the
same PSC may contain phytoplankton with common biogeochemical function (Nair et al. 2008).
Different PFTs occupy different niches and functionality in the biogeochemical cycle of the
ocean. For example, diatoms are the major silicifiers in the ocean with frustules made of dissolved
silicic acid, influencing the cycle of silica in the oceans. Coccolithophores build their external
plates (coccolith) from calcium carbonate, altering the seawater alkalinity by the release of CO2
and, in turn, influencing the marine carbon cycle. In the marine carbon cycle, the size of
phytoplankton is associated with sinking rates. Overall, larger and denser cells have higher
sinking rates and hence increased export of carbon to the deeper ocean whereas compared to
smaller cells. Smaller cells tend to sink more slowly, remaining for longer period in the mixed
layer and increasing their chance to be grazed (IOCCG 2014).
The fixed organic carbon in the upper ocean is transformed by several pathways in the water
column. These include zooplankton grazing and excretion, particle aggregation and bacteria
18
decomposition (Siegel et al. 2014) (Figure 1b). These processes are part of the ocean’s biological
carbon pump. Although most of the organic particles are recycled in the upper ocean (Henson et
al. 2011), the sinking of organic particles from the euphotic zone exports approximately 6 - 11
PgC yr-1 to the deeper ocean (Siegel et al. 2014; Yao and Schlitzer 2013).
Figure 1.1. On the top: illustration of the major phytoplankton functional types living in the
global ocean. On the bottom: simplified scheme of biological carbon pump. Adapted from
http://earthobservatory.nasa.gov/.
In the recent years, the interest on quantitative estimation of the abundance and dynamics of
different phytoplankton types has increased under the eminent climate change. To what extent the
PFTs and primary production will respond to changes in the climate is still unknown. One
potential effect is the acidification of the oceans. Ocean acidity increases positively with increase
of anthropogenic CO2 emissions, threating calcifying organisms. In addition, there is mounting
evidence that the increase in the sea surface temperature will result in a more stratified ocean and
changes in phytoplankton composition, favoring smaller phytoplankton due to reduced nutrient
supply (Bopp et al. 2005; Finkel et al. 2010).
Among the Earth’s oceans, the Southern Ocean (SO) is responsible for about half of the annual
19
global anthropogenic atmospheric CO2 uptake, around 1 PgC (Takahashi et al. 2012). Within this
region, the major sink occurs between 30°S and 50°S due to biological and physical mechanisms;
biological utilization of CO2 in summer and the cooling of surface waters in winter (Takahashi et
al. 2012). In the marginal ice zone, phytoplankton blooms, elevate phytoplankton biomass
(Behrenfeld and Boss 2014), develop with the retreat of sea ice in springtime and summer, which
turns surface waters into a strong sink of CO2 (Takahashi et al. 2009; Takahashi et al. 2012).
Because of the complex spatial and temporal dynamics of phytoplankton, investigations
usually rely on satellite remote sensing observations. For several years, ocean colour remote
sensing was regarded as the study of the chlorophyll-a concentration (Chla), the main
photosynthetic pigment present in phytoplankton and a proxy of their biomass (Martin 2004).
During the last four decades since the launch of the first ocean colour sensor, the Coastal Zone
Colour Scanner (1978-1986), many sensors have been launched leading to the development of
several satellite products (e.g. NPP, PFTs) and significant improvement of our understanding on
phytoplankton. However, uncertainties still remain.
Current NPP models differ in terms of complexity and despite the efforts to accurately retrieve
NPP from ocean colour, large differences among estimates have been observed for the SO
(Campbell et al. 2002; Carr et al. 2006; Saba et al. 2011). These differences can result from
uncertainties in the input variables of the models or because the models do not properly represent
the “reality” (Milutinovic and Bertino 2011). For example, Saba et al. (2011) investigated how the
skill of twenty one ocean color models are affected by the uncertainties in the input variables (e.g.
Chla, sea surface temperature) and found that the skills of the models are improved if the errors in
the input variables are considered.
Of major interest in studying the oceanic biogeochemical cycle is to know which PFTs are
present, their abundance and distribution. Within this context, one PFT of special interest includes
the diatoms, which are the major contributors to the oceanic primary production (Rousseaux and
Gregg 2014), carbon export and cycling of silica (Smetacek 1999), and together with
dinoflagellates, the most diverse PFT (Armbrust 2009; Leblanc et al. 2012). Diatoms are also one
of the largest PFTs in terms of size, ranging from micrometers to a few millimeters (Armbrust
2009). Given the biogeochemical and ecological importance of diatoms, several methods based on
satellite remote sensing data have been developed to retrieve their global abundance or dominance
(Alvain et al. 2005; Bracher et al. 2009; Hirata et al. 2011). Since most of the methods are built on
empirical relationships and rely on in situ data for model development and validation, refinement
is needed when additional data are available to improve the retrieval of diatoms for both global
and under-sampled oceans.
Furthermore, much of the primary production occurs during phytoplankton bloom events. The
term “phenology” is used to describe periodic events and their relation to environmental
conditions (Schwartz 2013). In the case of phytoplankton it refers to blooms. The number of
studies on phytoplankton phenology has increased in the recent years since changes in the
phytoplankton phenology (e.g. start, duration) can have a large effect on the marine ecosystem
(Edwards and Richardson 2004) and are indicators of environmental changes (Racault et al.
2012). Moreover, studies have suggested that the phytoplankton phenology is influenced by
climate oscillations (e.g. Southern Annual Mode – SAM, El Niño Southern Oscillation - ENSO).
Arrigo and van Dijken (2004) observed a later development of the bloom in the Ross Sea caused
20
by increased sea ice cover during the El Niño of 1997–1998. Alvain et al. (2013) observed an
increase in the dominance of diatoms in the SO during a positive SAM phase, possibly induced by
stronger winds which lead to increased mixing and nutrient supply. In spite of these findings, none
of the studies investigated in detail the mean patterns and the interannual variability of the
different characteristics and stages of blooms (e.g. start date, end date, duration of the growth and
decline phases, maximum biomass), their relationship with climate oscillations, neither focused on
the abundance of diatoms. In the SO, diatom contribution to primary production was estimated
from the NASA Ocean Biogeochemical Model to represent 89% of the total NPP (Rousseaux and
Gregg 2014). In addition to that, iron fertilization studies have demonstrated the potential of
diatoms as major contributors for the biological drawdown of atmospheric CO2 and for the export
of carbon from the surface to the deep ocean (Smetacek 1999; Smetacek et al. 2012).
Considering the knowledge gaps addressed in the previous paragraphs, this work aims to
complement and extend existing studies on ocean colour in the SO. Four main objectives have
been pursued:
1) to evaluate the uncertainties of satellite retrievals of euphotic depth (Zeu) and how different
Zeu retrievals influence the NPP estimated from satellite;
2) to evaluate and improve the satellite retrievals of diatoms abundance;
3) to investigate the mean pattern and interannual variability of diatom bloom phenology;
4) to examine the potential of ocean colour products to access environmental changes in the
Southern Ocean.
1.2 Outline of the thesis
Chapter 2 introduces the specific topics of the three main studies developed in this research
work. It starts by introducing satellite algorithms, moves to the SO and briefly explores the
dynamics of phytoplankton blooms. In addition, it gives an overview on two important climate
oscillations which influence the SO, ENSO and SAM. The next three chapters refer to the three
studies.
Chapter 3 addresses the uncertainties of the satellite derived euphotic depth (Zeu), which is a
common variable among different ocean colour NPP models. It examines retrievals from two
different methods (Chla and inherent optical properties of the water) and sensors (Sea-viewing
Wide Field-of-view Sensor - SeaWiFS and Moderate Resolution Imaging Spectroradiometer MODIS). It shows how the Zeu product can affect the retrievals of NPP. Furthermore, the NPP
model used in the study requires the knowledge of the phytoplankton absorption. For this reason,
the uncertainty of the satellite-derived phytoplankton absorption is also investigated. This chapter
was published in Remote Sensing of Environment (Soppa et al. 2013).
Chapter 4 is focused on the satellite retrieval of diatoms abundance. Most investigations have
been confined to global oceans. In this study, we highlight the importance of the SO in developing
a global algorithm for diatom using the Abundance Based Approach (ABA) of Hirata et al. (2011).
A large global in situ dataset of phytoplankton pigments is compiled, particularly with more
samples collected in the SO. A revision of the ABA is accomplished to take account of the
information on the penetration depth and to improve the relationship between diatoms and TChla1.
21
As result, the revised global model has improved the retrievals in the SO. Moreover, a regional
model, which further improves the retrievals of diatom abundance for the SO, is now available.
These findings were published in Remote Sensing (Soppa et al. 2014).
Chapter 5 investigates the diatom phenology in the SO applying the regional model to retrieve
the abundance of diatoms developed in the previous chapter. Using a novel merged satellite Chla
product applied to the regional algorithm, details are given to the diatom bloom phenology during
1997–2012. The different characteristics of the phenology (e.g. start, duration, biomass) are
investigated, as well as the interannual variability and trends. Finally, deep insights are given to
how the interannual variability of diatom bloom phenology could be modulated by the large scale
climate oscillations ENSO and SAM.
Chapter 6, the concluding chapter, draws together the main findings of the thesis and provides
an outlook for future research.
1
We refer to Chla (monovinyl chlorophyll-a plus chlorophyllid-a, allomers and epimers) as TChla (Chla plus divinyl
chlorophyll-a) since Prochlorococcus (which contain divinyl chlorophyll-a) is uncommon in the Southern Ocean south of
40°S. In chapter 4 we explicit refer to TChla because we consider the global distribution of phytoplankton.
22
Chapter 2
Scientific background
23
2
Scientific background
2.1 Ocean colour retrievals – an overview
The principle of ocean colour remote sensing is that the information of the radiance in the UVVIS (ultraviolet-visible) part of the spectrum can be used to infer the water optical properties. The
total radiance (LT, W m-2 sr-1) measured by the satellite sensor has different origins, but is the
water leaving radiance (Lw), originated from the in-water scattering, which contains the
information needed for the retrieval of the water properties (Figure 2.1). Within the water, the
solar radiation is absorbed and scattered by water molecules and water constituents. These
constituents, such as the phytoplankton, suspended non-algal particles and colored dissolved
organic matter (CDOM or gelbstoff), absorb and scatter the radiation at specific wavelengths and
with different intensities. The different spectral signatures and the fraction of each dissolved and
particulate constituents, together with the water molecules, control the backscattered radiation that
leaves the water and which carries information on the optical active constituents.
Figure 2.1. A schematic showing the components of the total upwelling radiance at the sensor:
water leaving radiance (Lw), atmospheric radiance (La) and surface-reflected radiance (Lr). Based
on Martin (2004).
The Lw represents up to 10% of the total upwelling radiance measured by the satellite sensor
(average for the visible spectrum, Kirk 2011). The other 90% originates from the scattering within
24
the atmosphere (La – atmosphere radiance, e.g. gases, aerosols, clouds) and the reflection at the
ocean/water surface (Lr – reflected radiance, e.g. sun glint, foam), but those contributions can be
mostly separated from the Lw by atmospheric correction algorithms (Kirk 2011). The LT can be
expressed as the sum of the different radiances (omitting the angular dependencies here and
further on):
‫ ் ܮ‬ሺߣሻ ൌ ‫ܮ‬௔ ሺߣሻ ൅ ‫ܮݐ‬௪ ሺߣሻ ൅ ‫ܮ‬௥ ሺߣሻ
(2.1)
where λ is the wavelength and t is an atmospheric transmittance factor to account for the
attenuation of Lw from the surface to the satellite. However, it is the reflected (ρ) upwelling
radiation from the ocean and passing through the atmosphere that is derived from ocean colour
sensors and that that is directly related to the water optical constituents (Kirk 2011). The ρ is
preferred to L because it can be more accurately estimated. Thus, Equation 2.1 can be re-written
as:
ߩ் ሺߣሻ ൌ ߩ௔ ሺߣሻ ൅ ‫ߩݐ‬௪ ሺߣሻ ൅ ߩ௥ ሺߣሻ
(2.2)
The reflectance of the ocean ρw (adimensional) can be expressed as:
ߩ௪ ሺߣሻ ൌ
గ௅ೢ ሺఒሻ
(2.3)
ிబ ሺఒሻ ୡ୭ୱ ఏ
where F0 is the extraterrestrial solar irradiance (top of atmosphere radiance, W m-2), θ is the solar
zenith angle and π converts the solar irradiance to units of radiance.
Rearranging the equation 2.3, Lw can be obtained as:
‫ܮ‬௪ ሺߣሻ ൌ
ிబ ሺఒሻ ୡ୭ୱ ఏ
గ
ߩ௪ ሺߣሻ
(2.4)
The Lw is often converted to normalized Lw (nLw, W m-2 sr-1) to remove the effects of the
atmosphere attenuation and solar orientation (Kirk 2011). The nLw is obtained as:
݊‫ܮ‬௪ ሺߣሻ ൌ
௅ೢ ሺఒሻ
(2.5)
ఌ೎ ୡ୭ୱ ఏ௧ವ ሺఏబ ǡఒሻ
where εc is a correction factor accounting for changes in the Earth-Sun distance and tD the
irradiance transmittance factor.
The nLw is often expressed as remote sensing reflectance (Rrs, sr-1) or radiance reflectance. The
Rrs is the standard input of many ocean colour algorithms, including the derivation of the Chla
concentration, obtained as:
ܴ௥௦ ሺߣሻ ൌ ௡௅ೢ ሺఒሻ
(2.6)
ிబ ሺఒሻ
25
The satellite signal is related to the optical properties of the water and before more details on
satellite algorithms are introduced, it is worth to define the two classifications of the optical
properties of the water, which are widely used in ocean optics: (i) the inherent optical properties
(IOPs) and the apparent optical properties (AOPs). Absorption and scattering of solar radiation are
IOPs. Their magnitude depends only on the concentration and type of the substance and not on the
illumination geometry or light distribution in the water (Kirk 2011). The AOPs (e.g. Rrs), on the
other hand, depend on the IOPs, as well as on the geometrical structure of the light field.
Thus, the magnitude and the spectral characteristics of the Rrs can be related to IOPs and can
be expressed as a function of absorption and backscattering coefficients as result of radiative
transfer models (Gordon et al. 1988):
ܴ௥௦ ሺߣሻ ൌ ቀ
௙ሺఒሻ
ொሺఒሻ
௧
௕ ሺఒሻ
್
ቁ ቀ௡మቁ ቀ௔ሺఒሻା௕
್ ሺఒሻ
ቁ
(2.7)
where f is an empirical factor that depends on the IOPs, incoming distribution of radiance and λ
(Reynolds et al. 2001) and Q is the ratio of the upwelling irradiance (Eu) to radiance (Lu). f/Q is
also specified as a shape factor describing the angular structure of the light field (Morel et al.
2002; Zaneveld 1995). t is the transmittance factor of the air-water interface, n is the real part of
the refractive index of the water, a is the total absorption coefficient, bb is the total backscattering
coefficient. The term “total” refers to the sum of the absorption coefficients of water, dissolved
matter and suspended particles (non-algal and algal particles). Likewise, bb is the sum of the
backscattering coefficients of the water and suspended particles; assuming no significant bb for
dissolved matter. In the next sections, the dependence on wavelength is further omitted for
simplicity. Most semi-analytical algorithms (e.g. Garver-Siegel-Maritorena Model, QuasiAnalytical Algorithm) used to retrieve Chla and IOPs are based on the relationship between Rrs
and IOPs.
2.1.1
Chlorophyll-a concentration
Phytoplankton contain several photosynthetic pigments which absorb light in different parts of
the spectrum from 400 to 700 nm. The spectral absorption characteristics of phytoplankton are
modulated by their pigment composition with different pigments exhibiting distinct absorption
features (IOCCG 2014). These pigments are one of the absorbers of the light in the water; together
with the coloured dissolved organic matter, non-algal particles and the water molecule itself. The
Chla is the major photosynthetic pigment present and shared by different phytoplankton groups
(except for Prochloroccocus sp.). For this reason, it is commonly used as indicator of the biomass
of phytoplankton.
The Chla has a distinct spectral characteristic in the visible spectrum, strongly absorbing at
blue and red bands, and less in the green region (Mobley 1994, Figure 2.2a). This spectral
signature makes the retrieval of Chla from satellite sensors possible and those were developed
with bands to exploit the information in the blue (443, 490, 510 nm) and green (550 or 555 nm)
regions. The standard SeaWiFS (OC4v.6) and MODIS (OC3M) Chla algorithms are empirically
26
derived from a large global dataset of in situ measurements of Chla concentration (hereafter
referred to as Chla) and the ratio of Rrs in different bands. The Chla is determined from a
polynomial function that relates the maximum band ratio (X) to the Chla, defined as:
݈‫݃݋‬ଵ଴ ሺ‫݈݄ܽܥ‬ሻ ൌ ܿ଴ ൅ ܿଵ ܺ ൅ ܿଶ ܺ ଶ ൅ ܿଷ ܺ ଷ ൅ ܿସ ܺ ସ
(2.8)
where for SeaWiFS,
ሺோೝೞ ሺସସଷሻǡோೝೞ ሺସଽ଴ሻǡோೝೞ ሺହଵ଴ሻ
ܺ ൌ ݈‫݃݋‬ଵ଴ ቂƒš ቀ
ோೝೞ ሺହହହሻ
ቁ ቃ
(2.9)
For MODIS two band ratios are used to replace the three band ratios in the SeaWiFS
algorithm: Rrs(443)/Rrs(550) and Rrs(490)/Rrs(550). The coefficients ܿ଴ǡ ܿଵ ǡ ܿଶ ǡ ܿଷ and ܿସ are 0.3272,
-2.9940, 2.7218, -1.2259 and -0.5683 for SeaWiFS and 0.283, -2.753, +1.457, +0.659 and -1.403
for MODIS (Feldman and McClain 2012; O'Reilly et al. 2000). The band ratio is used instead of
individual Rrs to reduce the uncertainties due to light propagation through the interface water-air
while the switch of Rrs-ratios preserves a high signal-to-noise ratio (SNR) (Martin 2004; O’Reilly
et al. 2000).
As Chla increases, the shape and magnitude of the reflectance spectrum changes (see Figure
2.2b). The region in the spectrum with higher Rrs shifts to higher wavelengths and the magnitude
decreases. In low Chla waters the SNR is higher at 443 than at 490 and 510 nm bands, but at
higher Chla the SNR is lower at 443 nm due to the stronger absorption (lower reflectance) in the
blue region and thus the other bands are used instead (Dierssen 2010; O'Reilly et al. 2000).
Figure 2.2. (a) Specific pigment absorption spectra for major phytoplankton pigments:
chlorophyll-a (Chla), chlorophyll-b (Chlb), chlorophyll-c (Chlc), photosynthetic carotenoids (PC)
and photoprotective carotenoids (PPC). Adapted from IOCCG (2014); originally from Bidigare et
al. (1990). (b) Spectral changes in the Rrs with respect to changes in Chla. Numbers above the
lines indicates the Chla concentration in mg m-3. Adapted from Dierssen (2010).
Nevertheless, these algorithms were developed considering that the optical properties of the
water are determined primarily by phytoplankton (case 1 waters), assuming that all other nonwater constituents vary closely with Chla (IOCCG 2006). Thus, when the optical properties of the
27
water are dominated by other constituents (case 2 waters), such as non-algal particles or CDOM,
the algorithms show higher uncertainties.
In situ measurements of Chla use mostly either fluorometric or the High Performance Liquid
Chromatography (HPLC) techniques. The advantages of fluorometric measurements are the lower
costs and it is less time consuming than HPLC. However, it is less accurate than HPLC due to the
spectral overlap of the fluorescence of other chlorophyll pigments. For example, when
chlorophyll-b is present in the sample, it leads to an underestimation of Chla (Arar and Collins
1997). Nowadays HPLC is the standard method to determine Chla concentration used to develop
and validate satellite Chla algorithms (IOCCG 2014). In the former days, fluorometric Chla data
were used for this purpose. Another advantage of the HLPC technique is that it enables the
simultaneous identification of a whole set of pigments.
2.1.2
Phytoplankton absorption
Chla is not the only pigment present in phytoplankton. There are other pigments that more
efficiently absorb the light at different wavelengths and with either photosynthetic or
photoprotective function (Kirk 2011). These pigments are also called accessory pigments. For
example: chlorophylls (chlorophyll-b), carotenoids and phycobiliproteins (phycocyanin). The
phytoplankton absorption spectrum results from the combination of the pigments present in the
phytoplankton. Thus, the spectral signature of the phytoplankton absorption (aph) varies with the
pigment composition, but also with pigment packaging.
Similarly to Chla, the phytoplankton absorption can also be derived from Rrs using semianalytical algorithms, such as the Quasi-Analytical Algorithm (QAA). The difference to band
ratio algorithms is that semi-analytical algorithms are based on theoretical models to relate Rrs to
IOPs, together with empirical models to describe the relationship between IOPs and the water
optical constituents (Martin 2004). Briefly, the QAA is an inversion algorithm that derives the
IOPs (absorption and backscattering coefficients at different wavelengths) from the Rrs using
empirical, analytical and semi-analytical approximations (Hirawake et al. 2011; Lee et al. 2009;
Lee et al. 2002). First, the total absorption coefficient (a) is calculated at a reference wavelength
(λ0, 555 nm for SeaWiFS and 550 nm for MODIS):
ܽሺߣ଴ ሻ ൌ ܽ௪ ሺߣ଴ ሻ ൅ ͳͲିଵǤଵସ଺ିଵǤଷ଺଺௑ି଴Ǥସ଺ଽ௑
మ
(2.10)
where aw(λo) is the absorption coefficient of pure seawater from Pope and Fry (1997) and,
ܺ ൌ ݈‫݃݋‬ଵ଴ ቈ
௥ೝೞ ሺସସଷሻା௥ೝೞ ሺସଽ଴ሻ
቉
ೝ ሺలలళሻ
௥ೝೞ ሺఒబ ሻାହ ೝೞ ሺరవబሻή௥ೝೞ ሺ଺଺଻ሻ
ೝೝೞ
(2.11)
where rrs is the Rrs just below the surface expressed as:
‫ݎ‬௥௦ ሺߣሻ ൌ
ோೝೞ ሺఒሻ
(2.12)
൫଴ǤହଶାଵǤ଻ோೝೞ ሺఒሻ൯
28
௕
The rrs can be modeled as a function of the ௔ା௕್ (represented by u):
್
‫ݑ‬ሺɉሻ െ
௕್ ሺ஛ሻ
௔ሺ஛ሻା௕್ ሺ஛ሻ
ൌ
ି଴Ǥ଴଼ଽହାඥ଴Ǥ଴଴଼ା଴Ǥସଽଽ௥ೝೞ ሺఒሻ
(2.13)
଴Ǥଶସଽ
Knowing a(λo) and u(λo), the particulate backscattering coefficient (bbp) at λo can be derived by:
ܾ௕௣ ሺɉ଴ ሻ ൌ ௨ሺ஛బ ሻ௔ሺ஛బ ሻ
൫ଵି௨ሺ஛బ ሻ൯
െ ܾ௕௪ ሺɉ଴ ሻ
(2.14)
where bbw is the backscattering coefficient of pure seawater from Morel (1974) at the λo.
Subsequently, the calculation is propagated to other wavelengths by:
஛
௒
ܾ௕ ሺߣሻ ൌ ܾ௕௪ ሺߣሻ ൅ ܾ௕௣ ሺɉ଴ ሻቀ బቁ
஛
(2.15)
where Y defines the spectral shape of bb(λ) and is defined as:
ܻ ൌ ʹ ቆͳ െ ͳǤʹ݁‫ ݌ݔ‬ቀെͲǤͻ
௥ೝೞ ሺସସଷሻ
௥ೝೞ ሺହହହሻ
ቁቇ
(2.16)
Then the total absorption a can be derived for the other wavelengths as:
ܽሺߣሻ ൌ ൫ଵି௨ሺఒሻ൯௕್ ሺఒሻ
(2.17)
௨ሺఒሻ
Knowing a(λ), the aph(λ) can be calculated by the following steps:
ሺସଵଶሻ
௔
Ƀ െ ௔೛೓ሺସସଷሻ ൌ ͲǤ͹Ͷ ൅
೛೓
Ɂെ
௔೒ ሺସଵଶሻ
௔೒ ሺସସଷሻ
଴Ǥଶ
ೝ ሺరరయሻ
଴Ǥ଼ା ೝೞ
(2.18)
ೝೝೞ ሺಓబ ሻ
ൌ െ‡š’ሺܵሺͶͶ͵ െ Ͷͳʹሻሻ
(2.19)
S is a parameter to describe the spectral shape of the absorption of gelbstoff and non-algal
particles (ag) calculated as:
ൌ ͲǤͲͳͷ ൅
଴Ǥ଴଴ଶ
ೝ ሺరరయሻ
଴Ǥ଺ା ೝೞ
(2.20)
ೝೝೞ ሺಓబ ሻ
ag(443) is determined as:
29
ܽ௚ ሺͶͶ͵ሻ ൌ
௔ሺସଵଶሻି஖ୟሺସସଷሻ
ஔି஖
െ
௔ೢ ሺସଵଶሻି஖௔ೢ ሺସସଷሻ
(2.21)
ஔି஖
For the other λ, ag is calculated as:
ܽ௚ ሺɉሻ ൌ ܽ௚ ሺͶͶ͵ሻ‡š’ሺെܵሺɉ െ ͶͶ͵ሻሻ
(2.22)
When the values of a(λ), aw(λ) and ag(λ) are known, aph(λ) can be derived as:
ܽ௣௛ ሺɉሻ ൌ ܽሺɉሻ െ ܽ௚ ሺɉሻ െ ܽ௪ ሺɉሻ
(2.23)
The standard in situ measurements of aph are carried out using the filter-pad method. In short,
water samples are filtered and placed in spectrophotometer which measures the light transmitted
through the filter. This process yields to the particulate absorption. To obtain the aph, the filter is
bleached with methanol to remove the pigments and the optical density is measured another time.
The aph is calculated as the difference between the measurements before and after the pigments
were removed. A detailed technical description can be found in Tassan and Ferrari (1995).
2.1.3
Euphotic Depth
Photosynthesis only occurs if light is available. The part of the water column with sufficient
light for supporting photosynthesis and thus NPP is called euphotic zone, or euphotic depth (Z eu)
(Falkowski and Raven, 2007; Kirk 2011). In biological terms, Zeu is the bottom of the euphotic
zone. In physical terms, Zeu is the depth where the downward photosynthetic available radiation
(PAR, EdPAR), the radiation in the spectral range of 400 – 700 nm, is reduced to 1% of its value at
surface (EdPAR(0)) (Morel and Berthon 1989).
In ocean colour remote sensing, Zeu is mostly estimated (i) empirically from the surface
chlorophyll-a concentration (Chla, Zeu-Chla) and (ii) semi-analytically from the inherent optical
properties of the water (IOPs, Zeu-IOP). The main difference between the two approaches is that
the derivation of Zeu from Chla assumes that the waters are classified as case 1. On the other hand,
the IOP approach determines the vertical distribution of light in the water from the IOPs and
therefore Zeu can be retrieved in optically complex waters as well, as shown by Lee et al. (2007)
and Shang et al. (2011b).
The relationship between Chla and Zeu can be expressed as (Morel, in Lee et al. 2007):
ܼ௘௨ ൌ ͵Ͷሺ‫݈݄ܽܥ‬ሻି଴Ǥଷଽ
(2.24)
From IOPs, the diffuse attenuation coefficient of PAR, KPAR, can be parameterized as a
function of a(490) and bb(490), which in turn are determined using the QAA as described in the
previous section (section 2.1.2). The vertical distribution of KPAR(z) is expressed as (Lee et al.
2005; Lee et al. 2007):
30
௄
‫ܭ‬௉஺ோ ሺ‫ݖ‬ሻ ൌ ‫ܭ‬ଵ ൅ ሺଵା௭ሻమ బǤఱ
(2.25)
where z is the depth and the coefficients K1 and K2 determined as,
‫ܭ‬ଵ ൌ ሾɖ଴ ൅ ɖଵ ሺܽሺͶͻͲሻሻ଴Ǥହ ൅ ɖଶ ܾ௕ ሺͶͻͲሻሿሺͳ ൅ Ƚ଴ •‹ሺߠሻሻ
(2.26)
where ɖ଴ , ɖଵ , ɖଶ and Ƚ଴ have values of -0.057, 0.482, 4.221 and 0.090, respectively, and,
‫ܭ‬ଶ ൌ ሾɖଷ ൅ ɖସ ܽሺͶͻͲሻ ൅ ɖହ ܾ௕ ሺͶͻͲሻሿሺȽଵ ൅ Ƚଶ …‘•ሺߠሻሻ
(2.27)
where ɖଷ , ɖସ , ɖହ , Ƚଵ and Ƚଶ have values of 0.183, 0.702, -2.567, 1.465 and -0.667, respectively
(Lee et al. 2005).
As mentioned above, Zeu is the depth where EdPAR(z) is 1% of EdPAR at surface. The
downward irradiance decreases exponentially with depth according to:
‫ܧ‬ௗ ሺœሻ ൌ ‫ܧ‬ௗ ሺͲሻ݁ ି௄ುಲೃ ሺ୸ሻǤ௭
(2.28)
Thus, Zeu can be determined as:
‫ܭ‬௉஺ோ ሺœሻܼ௘௨ ൌ ͶǤ͸Ͳͷ
(2.29)
In situ Zeu can be determined from vertical profiles of PAR measured with radiometers. The
sensors are mounted in a frame and lowered in the water avoiding the shadow of the ship. The
measurements have to be corrected for a series of uncertainties such as the self-shading, variation
of the incident sunlight during the deployment, bubbles and waves.
2.1.4
Phytoplankton functional types
As mentioned in Chapter 1, the intensity of the biological pump strongly depends on the size
and composition of phytoplankton cells, in addition to the structure of the trophic community. For
this reason it is important to distinguish the phytoplankton functional types (PFTs). Estimates of
PFTs from satellite can be also applied to support biogeochemical models in parameter estimation
and validating simulations for example (Robinson 2010). From satellite remote sensing PFTs can
be detected using different methods. In the following, satellite remote sensing algorithms that are
able to globally retrieve diatoms are introduced. We split them into abundance based and spectral
based approaches.
31
2.1.4.1 Abundance based approach
Abundance based approaches (ABA) are often used to retrieve global maps of phytoplankton
community (PFTs) and size structure. These algorithms are based on the concept that the
phytoplankton community structure changes with Chla. The Hirata et al. (2011) ABA method uses
satellite-derived Chla together with empirical relationships between Chla and PFTs (diatoms,
dinoflagellates, green algae, haptophytes, prokaryotes, pico-eukaryotes and Prochlorococcus sp.)
and is tuned using in situ phytoplankton pigment measurements. The ABA methods, that are tuned
using phytoplankton pigment data, are based on the assumption that different PFTs contain
different types of pigments, and that usually one major pigment (or several depending on the PFT)
can be attributed to a specific PFT (also called diagnostic pigment - DP, Table 2.1).
Table 2.1. Phytoplankton functional types and their respective diagnostic pigments (Hirata et al.
2011; IOCCG 2014; Vidussi et al. 2001).
Phytoplankton Functional Types
Diatoms
Dinoflagellates
Cryptophytes
Prymensiophytes and chrysophytes
Prymensiophytes
Cryptophytes
Green algae
Prochlorophytes
Cyanobacteria and prochlorophytes
Phytoplankton pigments / Diagnostic Pigments
Fucoxantin (Fuco)
Peridinin (Perid)
Alloxanthin (Allo)
19’-hexanoyloxyfucoxanthin (Hexfuco)
19’-butanoyloxyfucoxanthin (Butfuco)
Alloxanthin (Allo)
Monovinyl chlorophyll-b
Divinyl chlorophyll-b
Zeaxanthin (Zea)
The ABA starts by calculating the fraction (f) of Chla attributed to a specific PFT or PSC using
concentrations of DP from a large in situ database of phytoplankton pigment data (i.e. Diagnostic
Pigment Analysis – DPA, Vidussi et al. 2001, Uitz et al. 2006). According to Uitz et al. (2006), the
Chla can be expressed by the weighted sum of seven diagnostic pigments (DP) as:
‫ܲܦ̱݈݄ܽܥ‬௪ ൌ ܽଵ ‫ ݋ܿݑܨ‬൅ ܽଶ ܲ݁‫݀݅ݎ‬൅ܽଷ ‫݋ܿݑ݂ݔ݁ܪ‬൅ܽସ ‫ ݋ܿݑ݂ݐݑܤ‬൅ ܽହ ‫ ݋݈݈ܣ‬൅ ܽ଺ ܶ‫ ܾ݈݄ܥ‬൅ ܽ଻ ܼ݁ܽ
(2.30)
where DPw is the estimated Chla, ai are the partial coefficients (derived from multiple regression
analysis, Uitz et al. 2006). The summed terms on the right are the concentration of the diagnostic
pigments representing the main PFTs (Table 2.1).
The f-PFT/PSC is determined as the ratio of the DPs of a specific PFT/PSC to the sum of the
weighted concentration of the seven DPs (DPw). For example, the fraction of Chla that is attributed to diatoms (f-Diatom) is derived as:
݂ െ ‫ ݉݋ݐܽ݅ܦ‬ൌ
௔భ ி௨௖௢
(2.31)
஽௉ೢ
32
f-Diatom values lower than 0 and greater than 1 are set to 0 and 1, respectively. Table 2.2 presents
the DPA applied to calculate other PFTs and PSCs in Hirata et al. (2011).
Table 2.2. Equations used to calculate f-PFT/PSC (adapted from Hirata et al. 2011). DPw is the
sum of the weighted concentration of the DP and ai are the partial coefficients as in Uitz et al.
(2006).
PFTs/PSCs
Microplankton
Diatoms
Dinoflagellates
Nanoplankton
Haptophytes
Green Algae
Picoplankton
Prokaryotes
Pico-eukaryotes
Prochlorococcus sp.
Equation
(a1 Fuco + a2 Perid) / DPw
(a1 Fuco) / DPw
(a2 Perid) / DPw
(Xn a3 Hexfuco + a6 Chlb + a4 Butfuco + a5 Allo) / DPw
Nanoplankton - Green Algae
(a6 TChlb) / DPw
(Yp a3 Hexfuco + a7 Zea) / DPw
(a7 Zea) / DPw
Picoplankton – Prokaryotes
0.74 DvChla/Chla
Once the f-PFT/PSC has been determined (Table 2.2), the relationship between f-PFT/PSC and
Chla can be quantified using empirical equations (Figure 2.3a), with the empirical equation
varying according to the PFT/PSC (Table 2.3). With the knowledge of the empirical model, its
parameters and Chla, which is operationally produced as a satellite product, it is possible to
retrieve the f-PFT/PSC (Figure 2.3b). The Chla of each PFTs/PSC can be determined by
multiplying the f-PFT/PSC by the Chla. A limitation of retrieving PFTs from HPLC pigments is
the presence of a DP in more than one PFT (Hirata et al. 2011; Uitz et al. 2006). Thus,
uncertainties of ABA vary according to the PFT of choice and the best performance was obtained
for picoplankton (root mean squared error = 7.12 % Chla, see Table 4 in Hirata et al. 2011). For
diatoms the coefficient of determination was 0.73 and the root mean squared error 7.98 % Chla
(Hirata et al. 2011).
Table 2.3. Models and parameters used to estimate the f-PFT/PSC (adapted from Hirata et al.
2011).
PFTs/PSCs
Microplankton
Equation
[p0 + exp(p1 x + p2)]-1
p0
0.91
p1
-2.73
p2
0.4
Diatoms
[p0 + exp(p1 x + p2)]-1
1.33
-3.98
0.20
Dinoflagellates
Microplankton – Diatoms
Nanoplankton
1 - Microplankton – Picoplankton
Haptophytes
Nanoplankton - Green Algae
Green Algae
(p0 / y) exp[(p1 (x + p2)2]
0.25
-1.26
-0.55
Picoplankton
-[p0 + exp(p1 x + p2)]-1 + p3 x + p4
0.15
1.03
2
2
2
p3
p4
-1.56
-1.86
2.99
p5
p6
Prokaryotes
Pico-eukaryotes
(p0 / p1 / y) exp[p2 (x + p3) / p1 ] + p4 x + p5 x + p6
Picoplankton – Prokaryotes
0.007
0.62
-19.52
0.96
0.10
-0.12
0.06
Prochlorococcus sp.
(p0 / p1 / y) exp[p2 (x + p3)2 / p12] + p4 x2 + p5 x + p6
0.01
0.68
-8.6
0.97
0.007
-0.16
0.04
x=log10(Chla); y=Chla
33
Figure 2.3. (a) Relationship between Chla and Diatom (% Chla). To get % Chla, the f-Diatom is
multiplied by 100. The orange line represents the model and fitting parameters of Hirata et al.
(2011) for Diatoms, as presented in Table 3, red line represents a running mean of the in situ data.
Modified from Hirata et al. (2011). (b) Mean % of Chla of Diatoms over 1998-2010 for January.
Modified from IOCCG (2014).
2.1.4.2 Spectral based approach
Spectral approaches to retrieve the phytoplankton community structure are the algorithms of
Alvain et al. (2005; 2008) and Bracher et al. (2009). Alvain et al. (2005; 2008) developed an
algorithm called PHYSAT that determines the dominance of PFTs (diatoms, nano-eukaryotes,
Prochlorococcus, Synechocococus-like and Phaeocystis-like) by identifying their specific spectral
signatures in the nLw, more specifically in the satellite radiance anomaly (Ra) from SeaWiFS or
MODIS sensors. Ra is defined as the ratio of nLw and a reference nLw (nLwref) at five wavelengths
(e.g. 412, 443, 488, 531 and 555 nm for MODIS). The nLwref represents the average nLw spectra
for different Chla intervals from 0.02 to 3 mg m-3 (Ben Mustapha et al. 2014). The normalization
by the nLwref removes the first order effect of the Chla in the nLw (Alvain et al. 2005). A dataset of
satellite Ra spectra is matched with HPLC pigment data collected in different regions of the global
ocean to identify Ra spectra associated with a specific PFTs. Based on this information, Alvain et
al. (2005) established thresholds and shape criteria to distinguish between the spectral signatures
of different PFTs which were initially used to classify the satellite Ra spectra to obtain maps of the
PFTs dominance.
The PHYSAT was improved by Ben Mustapha et al. (2014) by using Self-Organizing Maps
(SOM). The SOM allows to automatically classify a larger number of satellite Ra spectra based on
their similarity in shape and amplitude, without the necessity of establishing thresholds. The Ra at
each pixel is assigned to a specific PFT if the Ra spectrum is between the mean ± one standard
deviation of the specific PFT Ra spectrum (or reference vector in the case of SOM). Otherwise,
the pixel is assigned to “unidentified phytoplankton assemblage” group (Figure 2.4). Validation
with in situ pigment data showed 83.3%, 66.7%, 58.1% and 66.7% of successful identification for
diatom, nano-eukaryotes like, Prochlorococcus and Synechocococus-like, respectively (Ben
Mustapha et al. 2014).
34
Figure 2.4. Climatology of the dominance of multiple PFTs for January (1997-2010) applied to
SeaWiFS. White areas represent missing data or unidentified phytoplankton type. Modified from
Ben Mustapha et al. (2014).
Like PHYSAT, PhytoDOAS is based on analyzing optical information and retrieves diatoms,
cyanobacteria, dinoflagellates and coccolithophores, by identifying their specific absorption in the
backscattered solar radiation. This is done by using the differential optical absorption
spectroscopy method (DOAS). First, the ratio of the backscattered solar radiation at the top of
atmosphere and the extraterrestrial irradiance, both measured by the sensor, is calculated.
Exploiting only the differential structures (a fitted low order polynomial is subtracted) all
contributions of oceanic and atmospheric constituents are fitted to this ratio, for example
phytoplankton groups, water vapour, water and atmospheric trace gases (e.g. ozone). The
contributions of constituents with low level of spectral structure, including Mie and Rayleigh
scattering and absorption of colored detrital matter, are approximated by a low order polynomial.
The DOAS method is applied to two spectral windows; one in the UV range of 340 nm to 385
nm and a second in the visible range of 429 nm to 521 nm. The UV window is used to retrieve the
spectral signature of the Vibrational Raman Scattering (VRS) of water molecules, a proxy for the
light penetration depth in the water (Dinter et al. 2015; Vountas et al. 2007). The DOAS-fit in the
visible window yields the specific phytoplankton absorption signatures. Thus, fitting is carried out
twice, first excluding the VRS part to derive the fit factor for the PFTs and, second, excluding the
PFT part to derive the fit factor for the VRS of water molecules, which is extrapolated to the
visible window. The concentration of each PFT is calculated by dividing the fit factor for the
specific PFT by the light penetration depth in the water (Bracher et al. 2009, Vountas et al. 2007).
The extended PhytoDOAS version, called multi-target fit, by Sadeghi et al. (2012)
simultaneously fits the absorption spectra of diatom, dinoflagellates and coccolithophores. Up to
now, PhytoDOAS is the only spectral algorithm applied to hyperspectral satellite data. Application
of PhytoDOAS is restricted to hyperspectral sensors and up to now it was applied to data from the
Scanning Imaging Absorption Spectrometer for Atmospheric Cartography (SCIAMACHY, level-1
top of atmosphere data) onboard the European Environmental Satellite (ENVISAT) (Figure 2.5).
A direct validation with in situ measurements is difficult owing to the large pixel size (about 30
km by 60 km). Nevertheless, preliminary validation by Bracher et al. (2009) indicated that
35
satellite-derived information on cyanobacteria and diatom distributions matched well with
corresponding in situ information based on pigment analyses of co-located water samples.
Figure 2.5. Monthly Chla for specific phytoplankton groups – October 2009. The coloured circles
are the Chla of the respective groups derived from HPLC pigment concentration and CHEMTAX
analysis of in situ samples taken during TransBrom Sonne cruise (9–23 October 2009) (IOCCG
2014).
2.1.4.3 Potentials and limitations of the methods
The potentials and limitations of the above mentioned algorithms are summarized in Table 2.4.
Some limitations are occasioned by the sensor and not directly by the algorithms. For example,
although PHYSAT can be applied on hyperspectral satellite data, currently there is no global
product of water leaving radiance from hyperspectral sensors, limiting the application of the
algorithm to multispectral sensors. Besides PhytoDOAS, all current global PFT products are
retrieved from multispectral sensors, which have a much higher spatial resolution (1 to 4.6 km)
and temporal coverage (1-3 days), when compared with current hyperspectral sensors. On the
other hand, the small number of wavelength bands and the broad band resolution of multispectral
sensors provide limited information on the phytoplankton absorption structures for spectral
algorithms (i.e. PHYSAT). As mentioned earlier, although different PFTs have different marker
pigments, they can also have some pigments in common. This is a limitation of retrieving PFTs
from HPLC pigments using the diagnostic pigment analysis, as in some abundance based
approaches (i.e. Brewin et al. 2010, Hirata et al. 2011). As shown by Brewin et al. (2014),
abundance-based approaches can be also parameterized using size-fractionated chlorophyll data
36
derived from filtration and in-vitro fluorometry for example.
Using PhytoDOAS, the specific optical signatures of different PFTs are separated by their
specific differential absorption spectra only visible in hyperspectral data. The drawback of
applying PhytoDOAS is the coarser spatial resolution of the ground scene of current global
hyperspectral sensors (e.g. SCIAMACHY, 30 km by 60 km) which is satisfactory for open oceans,
but limited close to the coasts and at the high latitudes where more factors can influence the
retrievals (e.g. CDOM, sea ice, different surface albedos). A second limitation of current global
hyperspectral sensors is the revisiting time which in the case of SCIAMACHY is every 6 days.
Especially for studies on phytoplankton bloom a data product with better temporal and spatial
resolution is required to meet the high spatial and temporal variability of blooms. This issue limits
up to now the use of PhytoDOAS products for phenological studies of phytoplankton blooms.
Table 2.4. Potential and limitations of the algorithms described in section 2.3. Based on IOCCG
(2014).
Algorithm
Potential
Limitations
PHYSAT
- detection of multiple PFTs
- spectral approach
- accounts for variability in the spectral
response within a PFT
- retrieves the dominant PFT and not concentration
- requires large number of match-up between in situ and satellite data for algorithm
development
- implemented for SeaWiFS and MODIS
- requires clear sky conditions
ABA
- detection of multiple PFTs and PSCs
- retrieves concentration
- applicable to any sensor that derives
Chla
- based on empirical relationships
- requires large in situ database for algorithm development
- products should not be treated as independent of Chla
PhytoDOAS
- detection of multiple PFTs
- retrieves concentration
- spectral approach
- does not account for variability in the absorption spectra within a PFT
- up to date applied only to SCIAMACHY
with low temporal and spatial resolution
2.2 The Southern Ocean dynamics and phytoplankton blooms
2.2.1
Southern Ocean circulation
The SO is of crucial importance for the global climate. The region is unique in that it contains
two important circulation features: an important part of the Meridional Overturning Circulation
(MOC) and the Antarctic Circumpolar Current (ACC) (Cunningham 2005). These two features are
responsible for storing and transporting heat, salt, CO2, nutrients and other substances, as well as
anomalies, around the globe.
The MOC connects all ocean basins through a large-scale system of surface and deep currents
(Schmittner et al. 2013). The role of the SO in this system includes the formation of the Antarctic
Bottom Water (AABW) and the wind-driven upwelling of deep waters (e.g. North Atlantic Deep
37
Water – NADW) (Talley 2013). The SO is the major region of the global oceans where the deep
waters are upwelled to the surface (Rintoul and Garabato 2013). A conceptual representation of
the MOC is presented in Figure 2.6a. The major water masses and pathways are next described
based on the works of Rintoul et al. (1999) and Talley (2013).
The NADW, Indian Deep Water (IDW) and Pacific Deep Water (PDW) rise into the mixed
layer in the SO where they form the Circumpolar Deep Waters (CDW). The IDW and PDW are
less dense than the NADW and upwell above and north of the NADW, being the main source of
the upper CDW (UCDW). Part of the UCDW is then diverted northward by Ekman transport
across the ACC and leads to the formation of the Subantarctic Mode Water (SAMW).
The SAMW is produced during deep winter convection (Rintoul et al. 1999) and is not only an
important source of nutrients for the global oceans (Sarmiento et al. 2004) but it is also rich in
anthropogenic CO2 by air-sea exchange (Sabine et al. 2004). The densest portion of the SAMW
further forms the Antarctic Intermediate Waters (AAIW) (Rintoul et al. 1999). The SAWM and
AAIW continue flowing northwards into the thermocline and are the main mechanism for
transporting anthropogenic CO2 to the ocean interior (Sabine et al. 2004).
The remaining fraction of the UCDW joins the upwelled lower CDW (LCDW), formed from
the NADW. They are then transported southward by eddies and will contribute to the formation of
the AABW. The cold winds blowing off Antarctica, over the continental shelves, cause the cooling
of ocean surface through heat loss as well as the brine release during the formation of sea ice.
These two processes produce dense waters that sink and form the AABW. Main regions for the
formation of AABW include the polynyas of the Weddell and Ross Seas. The AABW flows
northwards until it rises into IDW, PDW and NADW in the subtropics (Talley 2013).
The zonal circulation of the SO is driven by the ACC. The wind-driven ACC is the strongest
current of the global oceans (~ 137 Sv in the Drake Passage, Rintoul et al. 2010), flowing
eastward and connecting the southern part of the Atlantic, Indian and Pacific oceans. The ACC
transport is constrained into multiple strong narrow jets defined as fronts (Graham and De Boer
2013; Sokolov and Rintoul 2007). The position of the fronts is mainly controlled by the
interaction of the flow with the topography (Pollard et al. 2002). Such fronts act as boundaries that
split the SO in different zones of waters with distinct characteristics (Cunningham 2005; Sokolov
and Rintoul 2007). The main fronts observed in the ACC are: Subantarctic Front (SAF), Polar
Front (PF) and Southern Antarctic Circumpolar Front (SACCF) (Figure 2.6b). North of the ACC
there is the Subtropical Front (STF), the northern limit of the surface waters SAMW and AAIW
(Orsi et al. 1995).
38
Figure 2.6. (a) Two-dimensional view of the Southern Ocean part of the meridional overturning
circulation (MOC). NADW, North Atlantic Deep Water; Indian Deep Water – IDW; Pacific Deep
Water – PDW; Indonesian Throughflow – ITF; Upper Circumpolar Deep Water - UCDW; Lower
Circumpolar Deep Water - LCDW; Subantarctic Mode Water (SAMW), Antarctic Intermediate
Water (AAIW). Adapted from Talley (2013). (b) Southern Ocean bathymetry (m) overlaid with
the mean position of the maximum sea ice extent (1997-2012, solid white line, Fetterer et al.
2002), Subtropical Front (Orsi et al. 1995) and the Antarctic Circumpolar Current fronts (Orsi et
al. 1995, Salle et al. 2008). From north to south: Subtropical Front (STF, dashed line),
Subantarctic Front (SAF, solid line), Polar Front (PF, dashed line), and Southern Antarctic
Circumpolar Front (SACCF, solid line). SAZ, Subantarctic Zone; PFZ, Polar Front Zone; AZ,
Antarctic Zone.
39
2.2.2
Phytoplankton blooms
The SO is amongst the most productive regions of the global oceans (Comiso 2010; Rousseaux
and Gregg 2014). The high seasonal variation of the solar irradiance is the main control of the
seasonality of the phytoplankton photosynthesis, growth and distribution (Comiso 2010). Phytoplankton growth is limited in the winter by the reduced sunlight, lower water temperatures and
strong vertical mixing. Temperature affects metabolic processes of phytoplankton as well as the
stratification of the water column. The convective heat loss in the autumn and winter creates deep
mixed layers that dilute the phytoplankton population and rapidly moves phytoplankton through
the water column. As a consequence, phytoplankton receive significantly less light than if standing in the euphotic zone and also the grazing pressure is reduced (Behrenfeld and Boss 2014;
Franks 2014). Once the solar radiation increases again in spring, the convective mixing is reduced
and phytoplankton grow and form blooms, given that sufficient nutrients are available. The following decline of the blooms is controlled not only by the mechanisms above mentioned (light,
temperature, deep mixing) but also major factors are the depletion of nutrients and grazing.
Although open waters of the SO are generally characterized by high level of macronutrients,
the phytoplankton biomass remains low in many parts of the region, even in the summer when
light is normally not limited. The main reason is the lack of the micronutrient iron in the euphotic
zone, essential for electron transport in photosynthesis. The SO is a typical high nutrient low
chlorophyll (HNLC) system (Falkowski et al. 1998).
Nevertheless, the topography of the SO varies largely (Figure 2.6b) and when the flow
associated with the fronts interacts with topographic features, water rich in macro- and micronutrients rises to the surface and promotes phytoplankton growth and initiation of blooms. The
waters rich in nutrients and in phytoplankton are then advected horizontally with scales of
hundreds of kilometers (Sokolov and Rintoul 2007). Additional input of nutrients to the surface
occurs by the winter deep mixing, vertical diapycnal diffusion, Ekman upwelling, atmospheric
deposition and island’s shelf sediments (Borrione et al. 2014; Comiso 2010; Tagliabue et al.
2014). Whereas phytoplankton growth in open waters might be limited by nutrients, the
continental shelves of the SO have often higher biomass of phytoplankton. In these regions,
nutrients are supplied by river runoff, shelf sediments and melt of ice. In the seasonal ice zone, the
area between the minimum and maximum excursion of the sea ice, the increase in phytoplankton
biomass follows the retreat of sea ice. An important consequence of the melting of sea ice is the
increase of the vertical stratification of the water. At first, the stratification maintains
phytoplankton in the upper layers, in the euphotic zone, where nutrients and light promote
phytoplankton growth. Also important is that the partial ice cover reduces the mixing of the water
column by the winds (Taylor et al. 2013). As sea ice continues to melt, the area of open waters
expands. When sea ice has almost disappeared, the wind is able again to mix the water, which in
turn dilutes phytoplankton in the water column (Taylor et al. 2013).
2.2.2.1 Hypothesis on phytoplankton bloom initiation
To date, there is a general good understanding on the mechanisms controlling phytoplankton
40
blooms in high latitudes, but still the mechanisms leading to the start of the spring bloom are
discussed. The main hypotheses are briefly introduced below.
2.2.2.1.1 Sverdrup’s Critical Depth Hypothesis
The Sverdrup’s Critical Depth Hypothesis (Sverdrup 1953) is probably the most revisited one.
This hypothesis is based on the balance between the net primary production and the rate of losses
(respiration, grazing, sinking, mortality) which is controlled by the amount of available light
(Franks 2014). The depth where these two processes are equivalent is defined as the critical depth.
In the winter, phytoplankton population is diluted and mixed in the deep mixed layer. The
respiration exceeds primary production limiting phytoplankton growth and biomass accumulation
(Behrenfeld and Boss 2014). In spring, surface heating shoals the mixed layer; phytoplankton are
trapped in the surface and exposed to higher light levels. If mixed layer depth is above the critical
depth, the production is superior to the losses and concentration begins to rise.
Criticisms to the Critical Depth Hypothesis include the use of a constant loss rate and the
density-defined mixed layer depth. Classically, mixed layer depth is defined as the depth at which
a density changes by a given threshold value relative to the one at a near-surface (Montegut et al.
2004). Within this mixed layer the water properties are nearly homogeneous and there is a
continuous mixing. However, in the Critical Depth Hypothesis there is no differentiation on the
strength of the mixing or sinking of phytoplankton (Huisman et al. 2002).
2.2.2.1.2 Critical Turbulence Hypothesis
The Critical Turbulence Hypothesis (Huisman et al. 2002) is similar to the Critical Depth
Hypothesis, but it differentiates between mixed layer and turbulent mixed layer (Behrenfeld and
Boss 2014). The turbulence influences the light levels which phytoplankton are exposed to, as
well as the sinking of phytoplankton; hence the development of a bloom depends on the balance
in the turbulent mixing rates. If the turbulence is above a minimal threshold, critical turbulence,
phytoplankton are diluted and not exposed to light. On the other hand when the turbulence is low,
the sinking of phytoplankton out of the euphotic zone dominates and inhibits the development of
the bloom similarly. However, at intermediate turbulence phytoplankton are maintained at
adequate light levels, the production exceeds the losses and a bloom develops.
2.2.2.1.3 Disturbance-Recovery-Hypothesis
The Disturbance-Recovery Hypothesis (Behrenfeld et al. 2013) focus on the role of grazing on
controlling the balance between production and loss. According to this hypothesis, the decrease in
phytoplankton during autumn and winter is not only caused by the decrease in sunlight, deep
mixed layers along with the dilution of phytoplankton, but also by the dilution of zooplankton.
Accordingly, the phytoplankton biomass increases with the reduction of phytoplankton-grazers
encounters. Thus, the start of the bloom occurs before the shoaling of the mixed layer in spring,
41
even if the phytoplankton biomass is low due to the dilution by convective mixing (Behrenfeld
2014; Behrenfeld and Boss 2014).
2.3 Climate oscillations and the influence in the Southern Ocean
The phytoplankton biomass and the start, magnitude and duration of blooms vary each year
and studies in the SO have suggested that part of this variability is linked to the climate oscillations ENSO and SAM (Alvain et al. 2013; Arrigo and van Dijken 2004; Montes-Hugo et al. 2008;
Racault et al. 2012; Smith et al. 2008). Smith et al. (2008), for example, observed a later spring
sea-ice retreat and lower phytoplankton biomass offshore in the western Antarctic Peninsula in El
Niño or negative SAM events. ENSO and SAM dominate the climate variability on interannual
timescales in the tropics and in the SO, respectively. Whereas ENSO is an oscillation of the coupled ocean-atmosphere system, SAM is an oscillation of the atmospheric system.
2.3.1
El Niño Southern Oscillation - ENSO
The ENSO is an ocean-atmosphere phenomenon located in the tropical Pacific ocean that
occurs irregularly every 2 to 5 years (Clarke 2008). During neutral years, the surface atmospheric
pressure is low in the warmer waters of the western tropical Pacific and high in the colder waters
of the central and eastern tropical Pacific. The trade winds that blow eastward tend to pile up the
warm waters in the western Pacific forming the “warm pool” (Penland et al. 2013). The warm
water pool leads to strong atmospheric convection. The air in higher altitudes moves eastward,
sinks over the American continent and returns as easterly winds (Figure 2.7a, Robinson 2010).
Neutral years are also characterized by a deeper thermocline in the west side and shallower
thermocline in the east side of the Pacific.
In El Niño events anomalous high sea surface temperature (SST) are observed in the eastern
tropical Pacific (Figure 2.7b and 2.7d). These higher SSTs are accompanied by lower atmospheric
pressure and reduction of the easterly trade winds. With the weakening of the winds the waters of
the western Pacific are able to expand eastwards and the warm pool expands. This also creates an
anomalous deeper thermocline in the eastern Pacific and shallower in the west Pacific.
In La Niña events the system is reversed and the patterns are similar to normal years, however
intensified (Figure 2.7c). The easterly winds are stronger than normal. The SST in the east tropical
Pacific is colder than in normal years and the upwelling is intensified (Figure 2.7e).
42
Figure 2.7. On the left: ocean-atmosphere processes that occur during (a) normal years, (b) El
Niño event and (c) La Niña event. Modified from Robinson (2010). On the right: composite of
anomalies of Sea Surface Temperature from November to March in (d) El Niño (1965, 1972,
1982, 1987, 1991, 1993, 1994, 1997, 2002) and (e) La Niña (1950, 1955, 1956, 1964, 1971, 1974,
1988, 1998, 1999) events. The maps were produced from the data display pages of the
NOAA/ESRL
Physical
Sciences
Division,
Boulder
Colorado,
available
at
http://www.esrl.noaa.gov/psd/.
The ENSO strongly influences oceanic and atmospheric processes. One of the strongest El
Niño was observed in 1982. The upwelling in the coast of Peru was weaker than usual since the
thermocline in the region is deeper during El Niño events. As a consequence, less nutrients were
available that resulted in a decrease in the phytoplankton biomass with direct impact in the
anchovy, sea birds and seals population in the east Pacific region (Clarke 2008).
Although the ENSO is restricted to the tropical Pacific, its effects can be observed in different
regions of the globe. Induced changes in sea level pressure, surface air and water temperatures
and sea ice cover have been observed in different sectors of the SO (Kwok and Comiso 2002).
However, while the teleconnection mechanisms are well established in the tropical Pacific, they
are more complex in the SO. The heating in the tropical Pacific ocean and the deep convention
generate an atmospheric Rossby wave train that propagates until near Antarctica (Yeo and Kim
2015). In addition to that, the anomalous high SST changes the strength and position of the
atmospheric cells and jets that connect the tropics to the SO (Yuan 2004). More information on
this topic can be found in the works of Yuan (2004), L’Heureux and Thompson (2006), Ciasto et
43
al. (2015), Yeo and Kim (2015).
2.3.2
Southern Annular Mode - SAM
The SAM, also known as the Antarctic Oscillation, is the leading mode of atmospheric
variability south of 20°S (Pohl et al. 2010). It is characterized by differences in the atmospheric
pressure between mid-latitudes and the Antarctic region. A positive phase of SAM consists of
anomalous high pressure at mid-latitudes and anomalous low pressure at high latitudes (Figure
2.8a). This pressure difference strengthens and shift the westerly winds around Antarctica (Figure
2.8b). The opposite is observed during a negative phase of SAM.
The influence of SAM has been observed in distinct regions of the SO. The intensified winds
during a positive phase of SAM are associated with enhanced Ekman transport in the Antarctic
and Polar Frontal Zone (Lovenduski and Gruber 2005). Changes in the sea ice cover have also
been observed. Lefebvre et al. (2004) showed that during a positive SAM phase the
Bellingshausen and Weddell Seas are influenced by warm northerly winds and the sea ice cover
decreases. On the other hand, the Ross and Amundsen Seas are more affected by southerly winds,
which increase the sea ice cover.
Moreover, it has been shown that SAM presents a significant trend towards the positive phase
in the last decades (Pohl et al. 2010; Sallee et al. 2010). It has also been recognized that ENSO
and SAM are not linearly independent at interannual time scales (L'Heureux and Thompson 2006;
Pohl et al. 2010). Studies have also shown that El Niño yields to anomaly patterns in ocean and
atmosphere similar to a negative phase of SAM, and vice versa (Lovenduski 2007; Pohl et al.
2010). Fogt et al. (2011) observed that the teleconnection between ENSO and the SO is intensified
when ENSO co-occur with a weak SAM or when both oscillations coincide with opposing phases
(e.g. El Niño occurs with a negative SAM phase). When El Niño (La Niña) coincides with
positive (negative) SAM phase the magnitude of the teleconnection is reduced.
Figure 2.8. Regression of anomaly patterns of (a) atmospheric pressure at 700 mb and (b) wind
stress (dyne cm−2) onto the SAM index. Modified from Lovenduski (2007).
44
Chapter 3
Satellite derived euphotic depth in the Southern
Ocean: implication for primary production modeling
45
3
Study 1: Satellite derived euphotic depth in the Southern
Ocean: implication for primary production modeling
3.1 Motivation
The uncertainty of NPP models can be reduced by improving the input variables (Saba et al. 2011)
and a common one, shared by different models (Behrenfeld and Falkowski 1997; Hirawake et al.
2012; Hirawake et al. 2011; Westberry et al. 2008) is the Zeu. However, there is no detailed
evaluation of the satellite Zeu in the SO (defined here as the region south of 30°S). A comparison
of ocean colour sensor/retrievals with in situ measurements, as well as the further impact on the
NPP estimation is thus necessary. In this context, the main goal of this chapter is to investigate the
differences in estimating Zeu from satellite remote sensing using different approaches and sensors
in the SO. We compute Zeu from ocean colour products of (i) Chla and (ii) IOPs and validate those
using in situ measurements of Zeu. In addition, we compare Zeu derived from the MODIS and the
SeaWiFS sensors. The approaches and sensors are further examined in terms of the spatial
distribution of Zeu. Since aph data are used in the NPP calculation, we also examine the
uncertainties of MODIS and SeaWiFS aph derived with the Quasi-Analytical Algorithm (QAA,
Lee et al. 2002; Lee et al. 2009). Finally, we apply the absorption based primary production model
(ABPM, Hirawake et al. 2012, Hirawake et al. 2011) to investigate how different Zeu products
influence the estimation of NPP in the SO.
3.2 Material and Methods
3.2.1
In situ data
A dataset of in situ measurements of Chla (N=1032) and Zeu (N=1288) in the SO was built to
validate the satellite measurements. The dataset compiled measurements from 1997 to 2008 taken
by several investigators (Figure 3.1). The Chla data were restricted to Chla derived from HPLC
pigment analysis, within 12 m surface layer and taken within 3 hours of the Zeu in situ measurements. An average value of Chla was calculated if two or more samples were collected within the
surface layer. We used Zeu data provided in the databases that were calculated from in situ measurements of vertical profiles of PAR (N=977). In addition, vertical profiles of PAR were also
available in the SeaBASS database and those were used to calculate Zeu (N=311). A third dataset
of in situ measurements of ܽ௣௛ (N=465) was compiled to validate the ܽ௣௛ derived from satellite
Rrs. The ܽ௣௛ data are derived from filter pad measurements taken in the years 2007, 2008, 2010
and 2012. The ANT-XXVI/3 and ANT-XXVIII/3 data were measured according to the filter pad
method described in Taylor et al. (2011). Figure 3.1 presents the relative frequency distribution of
the Zeu, Chla and spectrally averaged ܽ௣௛ coefficient over 400–700 nm (ܽത௣௛ , see section 3.4) in
situ measurements that matched with SeaWiFS and MODIS data.
46
Figure 3.1. On the left, location of the in situ measurements in light grey and the matched ones
with satellite in black: (a) Zeu (1288), (b) Chla (1032) and (c)ܽ௣௛ ൫ܽത௣௛ ൯ (465). On the right, the
respective relative frequency distribution of the matched in situ measurements.
3.2.2
Satellite data
MODIS-Aqua (R2012.0) and SeaWiFS (R2010.0) level 3 products of Chla (CHL1), PAR, Rrs
were obtained at http://oceancolor.gsfc.nasa.gov/. The data are produced and distributed by the
NASA Goddard Space Flight Center's Ocean Data Processing System (ODPS). The SeaWiFS dataset has the longest time series; however, the data acquisition ended in December 2010. We used
MODIS and SeaWiFS data at 9 x 9 km2 spatial resolution. Satellite PAR and aph (see section
3.2.4) derived from Rrs were used in the NPP model. For the validation with in situ measurements
daily images were used; for spatial distribution analysis we used monthly data.
47
3.2.3
Zeu derived from ocean colour
Two approaches were used to derive Zeu from ocean colour products of: (i) Chla (Zeu-Chla) and
(ii) IOPs (Zeu-IOP), as presented in section 2.1.3. The QAA (version 5, Lee et al. 2009) was
applied to derive the absorption and backscattering coefficients at 490 nm (a490 and bb490) from
the satellite Rrs. The detailed QAA for the retrieval of a490 and bb490 and of Zeu was presented in
chapter 2 (section 2.1.2). The uncertainties of the IOPs retrieved from QAA are discussed in Lee
et al. (2005) and Lee et al. (2006; 2010).
3.2.4
Primary production model
The NPP was calculated using the Absorption Based Primary Production Model (ABPM,
Hirawake et al. 2012; Hirawake et al. 2011), an improved version of the Vertically Generalized
Production Model (Behrenfeld and Falkowski, 1997) for polar oceans. In the ABPM, the product
஻
of the chlorophyll-a normalized maximum photosynthetic rate in the water column ሺܲ௢௣௧
, mg C
(mg Chla)-1 h-1) and Chla (mg m-3) is replaced by a linear relation of the spectrally averaged ܽ௣௛
coefficient over 400–700 nm (ܽത௣௛ , m-1). This model eliminates uncertainties of the satellite Chla
୆
(Hirawake et al. 2011). The
product and the temperature effect on the estimation of the ܲ୭୮୲
ABPM is expressed as:
ܰܲܲ ൌ ൫ͳͲͻǤ͸͸ܽത௣௛ ሺͲି ሻ െ ͲǤͲʹ൯
଴Ǥ଺଺ଵଶହήாబ
ாబ ାସǤଵ
ܼ௘௨ ‫ܦ‬௜௥௥
(3.1)
where E0 is the daily integrated photosynthetic available radiation (PAR, Einsteins m -2 day-1) from
satellite
and
Dirr
is
the
photoperiod
(h)
calculated
as
described
in
http://orca.science.oregonstate.edu/faq01.php. The NPP estimated from Zeu-Chla is represented as
NPP-Zeu-Chla and from Zeu-IOP as NPP- Zeu-IOP.
The QAA was applied to derive the ܽ௣௛ at the SeaWiFS spectral bands of 412, 443, 490, 510
and 555 nm and MODIS spectral bands of 412, 443, 488, 531 and 555 nm. Satellite ܽതph were then
derived by adjusting the ܽ௣௛ integrated over the visible bands of SeaWiFS and MODIS to the in
situ ܽത௣௛ over the continuous visible range (400 – 700 nm) (Hirawake et al. 2012, Hirawake et al.
2011):
ܽത௣௛
ሺͲି ሻ
ൌ
ሺఒ ିఒ೔ ሻൗ
௔ σరಿసభ൤ቀ௔೛೓ ሺఒ೔శభ ሻା௔೛೓ ሺఒ೔ ሻቁቀ ೔శభ
ଶቁ൨
଻଴଴ିସ଴଴
(3.2)
where λ were the above mentioned spectral bands of SeaWiFS and MODIS. The parameter a
represents the slope of the regression of the satellite ܽത௣௛ to the in situ ܽത௣௛ and corresponded to
1.3656 for SeaWiFS and 1.5354 for MODIS.
48
3.2.5
Validation and statistical analysis
The MODIS and SeaWiFS match ups were obtained when the day, latitude and longitude of
the in situ observation fell within the limits of 1x1 pixel. The bias, average absolute percentage of
error (E), root-mean-square error (RMSE) and mean absolute error (MAE) were calculated to
evaluate the differences between the in situ Zeu and the satellite Zeu:
ଵ
݈‫݃݋‬ଵ଴ „‹ƒ• ൌ σே
௜ୀଵሺ݈‫݃݋‬ଵ଴ ሺܻ௜ ሻ െ ݈‫݃݋‬ଵ଴ ሺܺ௜ ሻሻ
(3.3)
ே
ଵ
௒೔ ି௑೔
ே
௑೔
‫ ܧ‬ൌ ቀ σே
௜ୀଵ ቚ
ቚቁ ͳͲͲΨ
(3.4)
ଵ
ܺ௜ ሻሻଶ
݈‫݃݋‬ଵ଴ ൌ ටே σே
௜ୀଵሺ݈‫݃݋‬ଵ଴ ሺܻ௜ ሻ െ ݈‫݃݋‬ଵ଴ ሺ
ଵ
݈‫݃݋‬ଵ଴ ൌ σே
ܻ௜ ሻ െ ݈‫݃݋‬ଵ଴ ሺܺ௜ ሻȁ
௜ୀଵȁሺ݈‫݃݋‬ଵ଴ ሺ
(3.5)
(3.6)
ே
where X was the in situ value, Y the satellite value and N is the number of matching pairs. The statistical indicators ݈‫݃݋‬ଵ଴ bias, E and ݈‫݃݋‬ଵ଴ RMSE were chosen based on the GlobColour Validation
Report (Durand 2007) and other literatures on satellite validation (Shang et al. 2011b; Zibordi et
al. 2006). The ݈‫݃݋‬ଵ଴ MAE was used as a statistical estimator of error for comparisons between the
sensors and ܽ௣௛ at different wavelengths, since N changes. Willmott and Matsuura (2005) showed
that RMSE is sensitive to the square root of N and MAE should be preferred instead. No outliers
were removed. For reference, a 1:1 line was included in the scatterplots to show how well the satellite and in situ data agree.
Monthly climatologies of Zeu and NPP in December, January and February, were computed to
investigate spatial differences. The climatology fields were calculated from monthly images for
the 2003-2009 period, excluding the year of 2008 when SeaWiFS did not acquire data. For each
pixel, the relative difference between the spatial fields was derived:
஺ି஻
ൌ ቀ
஻
ቁ ͳͲͲΨ
(3.7)
where A corresponded to Zeu-Chla, Zeu-SWF or NPP-Zeu-Chla and B to Zeu-IOP, Zeu-MODIS or
NPP-Zeu-IOP. We did not compare the spatial distribution of NPP between the sensors because
ܽത௣௛ , PAR and Chla might introduce differences in the NPP estimation.
3.3 Results
3.3.1
Comparison of satellite and in situ Zeu
49
Figure 3.2 presents the comparison between satellite and in situ Zeu. The overall statistics show
that the two approaches agree well with the in situ measurements. When Zeu-SWF was derived by
the IOP approach, the statistics are slightly better than Zeu-Chla improving the E in 3.5% (Figures
3.2a and b) and the regression line is close to the 1:1 line (dotted line). On the other hand, ZeuChla shows better results than Zeu-IOP for MODIS, reducing the E in 9.5% (Figures 3.3c and d).
Differences in log10MAE indicate that Zeu retrieved from SeaWiFS is more accurate than MODIS.
Negative biases are found for Zeu-MODIS and positive biases for Zeu-SWF.
Figure 3.2. Scatterplots of satellite Zeu against in situ Zeu. (a) and (c) Zeu is derived from Chla
approach (Zeu-Chla), (b) and (d) Zeu is derived from the IOP approach (Zeu-IOP). The solid line
represents the regression and the dotted line represents 1:1 line as reference.
Compared to collocated in situ HPLC Chla data of our validation dataset, the standard
SeaWiFS algorithm (OC4v.6) underestimates Chla (Figure 3.3). For MODIS, the OC3M
algorithm leads to under- and overestimation of Chla depending on the concentration of the in situ
Chla. For in situ Chla < 1.5 mg m-3, Chla was on average underestimated, whereas for higher
concentrations (> 1.5 mg m-3) the retrievals were overestimating the in situ values.
50
Figure 3.3. (a) Scatterplots of satellite and in situ Chla. The dotted line represents the 1:1 line as
reference. (b) Relative differences between satellite Chla and in situ Chla. The dotted line
represents the zero line.
3.3.2
Spatial distribution of Zeu-Chla and Zeu-IOP
Figure 3.4 presents the spatial distribution of the climatology of Zeu for February, using data
from 2003 to 2009. Deeper Zeu are associated with oligotrophic waters in the zonal band of 30° 40°S. Shallower Zeu are observed in the waters around the Antarctic continent, South America,
south and west part of South Africa and between 40° - 50°S, except for the eastern Pacific Sector.
Shallower Zeu are related to terrigenous influence (e.g. La Plata river plume in the Patagonian
Shelf region) and higher chlorophyll concentrations in upwelling regions (e.g. Benguela
upwelling), islands (e.g. Kerguelen islands) and continental shelves (e.g. Antarctic Peninsula). The
difference in calculating the climatology of Zeu from daily or monthly images was small. For
instance, the standard deviations of the difference between Zeu-Chla calculated from daily data
and monthly data in February 2003 are 1.22 m for SWF and 1.08 m for MODIS. For the IOP
approach the values are 0.91 m for SeaWiFS and 0.77 m for MODIS.
When Zeu-Chla was compared with Zeu-IOP, large differences were observed. While the range
of Zeu-Chla from SeaWiFS varies between 5.97 and 234.31 m (median = 65.50 m), using the IOP
approach this range is much narrower, from 2.5 to 150 m (median = 63.93 m). Similar for
MODIS, Zeu-Chla varies between 5.89 and 259.69 m (median = 65.50 m) and Zeu-IOP from 3.5 to
51
146.3 m (median = 62.37 m). On average, for the entire region, Zeu-Chla from SeaWiFS and
MODIS are 3.61 and 5.41% deeper than Zeu-IOP. These differences followed a zonal distribution.
The most notable difference was observed in the Pacific Sector within the 30°- 40°S zonal band,
corresponding to the South Pacific subtropical gyre, where Zeu-Chla is ~ 20 - 30% deeper than
Zeu-IOP. The spatial distribution maps also pointed out differences of about 10 - 15% south of
60°S, with Zeu-Chla usually deeper than Zeu-IOP; especially for MODIS. Regions corresponding
to deeper Zeu-IOP were also presented, but they were less abundant and only about ~ 10% deeper.
Figure 3.4. Spatial distribution of Zeu in the Southern Ocean (climatology of February). The white
pixels correspond to areas with no data.
Comparing the sensors, the spatial distribution of Zeu is similar in both approaches, with an
average difference (DIFF) of -0.005 and 2.68% for Zeu-Chla and Zeu-IOP, respectively in February
(Figure 3.5). However, the spatial differences are larger south of 60°S and more evident in ZeuIOP. A corresponding pattern was observed in December and January.
52
Figure 3.5. Spatial distribution of the relative percentage of difference between SeaWiFS and
MODIS. The white pixels correspond to areas with no data.
3.3.3
Primary Production
3.3.3.1 Validation of SeaWiFS and MODIS derived aph
The ocean colour NPP model used here is a function of ܽത௣௛ . The ܽത௣௛ can be determined
empirically through linear relations between in situ ܽത௣௛ and satellite ܽph integrated over the visible
spectral bands of SeaWiFS and MODIS. Hirawake et al. (2012; 2011) calculated these
relationships based on ܽ௣௛ derived from ship Rrs at the MODIS and SeaWiFS spectral bands,
using the QAA. However, within this study the satellite Rrs derived ܽ௣௛ were not validated due to
the insufficient number of collocations between satellite and in situ data. Furthermore, at the
current state of knowledge, there is no information on the performance of the QAA to derive ܽ௣௛
from satellite Rrs in the SO. Therefore, before we investigated the NPP, we briefly assessed the
quality of the ܽ௣௛ derived from SeaWiFS and MODIS Rrs using the QAA with in situ ܽ௣௛ . Results
are presented in Table 3.1.
The E of ܽ௣௛ -SWF increase for increasing wavelengths (except at 443 nm) and negative biases
indicate an underestimation of ܽ௣௛ . Results for MODIS show similar log10MAE at 412, 443 and
53
488 nm, increasing towards 550 nm. Negative ܽ௣௛ were retrieved at SeaWiFS bands 490, 510 and
555 and at MODIS bands 412 and 443 nm and lead to small but negative ܽത௣௛ Ǥ Those values were
removed before the calculation of the statistics presented in Table 3.1. Estimates of NPP on pixels
with negative ܽത௣௛ were removed as well.
Table 3.1. Statistical results of the comparison between QAA-satellite derived aph and in situ aph.
SeaWiFS (N=12)
Range
2
log10MAE log10bias
Wavelength (nm) r
E (%)
satellite
in situ
412 (N=12)
0.81
0.22
-0.20
38.26 0.002 - 0.15 0.001 - 0.110
443 (N=12)
0.57
0.20
-0.15
36.97 0.003 - 0.171 0.011 - 0.092
490 (N=11)
0.19
0.20
-0.12
44.71 0.008 - 0.119 0.011 - 0.056
510 (N=11)
0.06
0.40
-0.32
71.78 0.001 - 0.083 0.006 - 0.054
555 (N=11)
0.37
0.36
-0.29
60.15 0.0003 - 0.042 0.001 - 0.02
0.48
0.20
-0.15
39.29 0.005 - 0.076 0.006 - 0.04
ܽത௣௛ (N=11)
MODIS (N=36)
412 (N=34)
443 (N=34)
488 (N=36)
531 (N=36)
550 (N=36)
ܽത௣௛ (N=34)
0.48
0.49
0.52
0.48
0.29
0.58
0.17
0.15
0.16
0.23
0.41
0.14
-0.08
-0.04
-0.08
0.005
0.41
0.04
36.53
33.93
29.03
98.53
220.34
43.46
0.002
0.003
0.0005
0.001
0.009
0.002
-
0.066
0.079
0.05
0.025
0.025
0.038
0.001
0.002
0.001
0.0001
0.0
0.0005
-
0.056
0.064
0.047
0.025
0.014
0.029
3.3.3.2 Spatial distribution of NPP-Zeu-Chla and NPP-Zeu-IOP
Generally, higher NPP-Zeu-Chla than NPP-Zeu-IOP were observed using both sensors over the
SO (Figure 3.6). For SeaWiFS NPP-Zeu-Chla was 7% higher than NPP-Zeu-IOP and for MODIS
10.22% higher. The average of NPP-Zeu-Chla and NPP-Zeu-IOP were 321.18 and 283.84 mg C m-2
d-1 for SeaWiFS, respectively. Using MODIS data the NPP-Zeu-Chla and NPP-Zeu-IOP were
438.50 and 393.78 mg C m-2 d-1, respectively. Although these differences may not be significant
for studies focusing on the entire SO, for local comparisons they are relevant. For instance, in the
region south of 60°S (60°S – 80°S, 120°W – 160°W) NPP-Zeu-Chla was ~ 30% higher than NPPZeu-IOP.
54
Figure 3.6. Spatial distribution of net primary production (in the figure caption called PP) in the
Southern Ocean (climatology of February). NPP-Zeu-Chla (left), NPP-Zeu-IOP (right) and relative
percentage of difference between NPP-Zeu-Chla and NPP-Zeu-IOP (center). The white pixels
correspond to areas with no data.
3.4 Discussion
3.4.1
Validation of Zeu and Chla
This study investigated differences between two approaches to derive satellite Zeu: the first one
by Morel (in Lee et al. 2007) is empirical and based on Chla and the second one by Lee et al.
(2005) is semi-analytical and based on IOPs. We focused on the Chla approach because of its
simplicity, but also to investigate if the known inaccuracy of the standard satellite Chla products
in the SO would impact the Zeu retrieval. The SO is heterogeneous in terms of bio-optical
conditions. It comprises not only oligotrophic waters, but ultra-oligotrophic waters (e.g. South
Pacific Gyre), complex waters (e.g. high concentration of non-algal particles in the Patagonia
Shelf), upwelling regions (e.g. Benguela upwelling), polar fronts and coastal Antarctic waters (e.g.
Antarctic Peninsula). For this reason, we included a more complex approach in our investigation:
the IOP approach, which accounts for the vertical distribution of other in-water components that
also contribute to the light attenuation. The QAA can be applied globally, regardless of the optical
55
complexity of the water and has been widely used and cited in the literature.
Our validation dataset covered a wide range of bio-optical conditions (Figure 3.1); however
uncertainties in Zeu were only improved by the IOP approach for SeaWiFS (Figure 3.2). This observation agrees with Lee et al. (2007). The authors compared in situ Zeu with Zeu-Chla and ZeuIOP calculated from ship borne Rrs in the Monterey Bay, the Gulf of Mexico and the Arabian Sea
and reported improved Zeu from the IOP approach. In addition, Shang et al. (2011b) studied oligotrophic and coastal waters of the South China Sea using MODIS data and showed that Zeu-IOP
was more accurate than empirically deriving Zeu from Chla (Morel et al. 2007). Within MODIS
data, our Zeu estimation with the Chla approach yielded better results than the IOP approach.
Our results indicate that Zeu can be accurately estimated by both approaches and sensors with a
log10MAE within 0.10 and 0.23 m. The relative consistency observed between the sensors is related to the common processing schemes applied, such as the atmospheric correction and data binning, as already pointed out by Mélin (2011). Differences might be caused by the different coverage of areas, spectral bands, orbital characteristics and equator-crossing times. MODIS-Aqua
crosses the equator at 13:30 pm. For SeaWiFS the equator crossing time drifted throughout the
mission, from 12:00 to 14:20, but 12:30 pm was used for calculations. Additional sources of error
in the validation analysis include the in situ measurements, as the use of different field sensors
and data/sample processing.
Results of the Chla validation indicate that the satellite Chla products from SeaWiFS are more
accurate than from MODIS in the SO (Figure 3.3). Our MODIS validation dataset is, however,
biased towards high Chla waters (Figure 3.1); 95% of the in situ data had Chla > 1 mg m-3 where
the errors are generally higher as well. In contrast, the SeaWiFS validation dataset has only 65%
of samples at Chla > 1 mg m-3. For instance, the difference in the log10MAE between MODIS and
SeaWiFS for Chla < 1 mg m-3 is 0.02 mg m-3 (0.17 mg m-3 for MODIS and 0.15 mg m-3 for SeaWiFS); for higher concentrations this difference increases to 0.4 mg m-3 (0.61 mg m-3 for MODIS
and 0.21 mg m-3 for SeaWiFS). The observed underestimation of Chla by the operational SeaWiFS and MODIS algorithms (here only for Chla < 1.5 mg m-3) is in accordance with previous
studies that used earlier algorithm versions, indicating that this issue still persists in the SO
(Dierssen and Smith 2000; Garcia et al. 2005; Johnson et al. 2013; Kahru and Mitchell 2010;
Szeto et al. 2011). Further, it is important to mention that we used surface Chla instead of the
weighted Chla in the first optical depth. Our coincident in situ measurements of HPLC Chla profiles, Kd and Zeu were all concentrated in the Antarctic Peninsula region which represents a particular region of the SO, thus we used surface Chla values only. Moreover, we avoided the use of
fluorometric data in our study and used HPLC data. Marrari et al. (2006) showed that the chlorophyll fluorescence of accessory pigments (e.g. chlorophyll-b) interferes in the determination of
Chla by fluorometric methods in the SO.
Nevertheless, uncertainties of the satellite Chla have some but small influence on the Zeu-Chla,
which is in part linked to the nature of the power function that empirically relates Zeu to Chla. One
has to note that we used the Chla even in waters that hardly fit to the Case 1 assumption, for instance on the Patagonian shelf and around the Antarctic Peninsula (Dierssen and Smith 2000;
Garcia et al. 2005). The error in Zeu induced by the error in Chla depends on the in situ concentrations. A 100% error in lower Chla values has higher impact on Zeu than 100% error in high Chla
values. For instance, a 100% overestimation in the lowest and highest in situ Chla (0.05 mg m-3
56
and 9.98 mg m-3) of our SeaWiFS validation dataset would lead to an error of 26.79 and 2.82 m in
Zeu, respectively.
3.4.2
Zeu spatial distribution
The spatial distribution maps of Zeu-Chla and Zeu-IOP highlighted large differences in the
South Pacific subtropical gyre and south of 60°S (Figure 3.4). Morel et al. (2007) evaluated the
Chla approach for waters of the South Pacific subtropical gyre with data collected during the
BIOSOPE cruise and showed that an empirical relationship based on Chla (Morel and Gentili
2004) was valid to estimate Zeu in those waters. Thus, for this region the satellite Zeu-Chla may be
the better choice. Unfortunately, beside the data from the BIOSOPE cruise, there were no in situ
measurements of Zeu available from the South Pacific and other SO subtropical gyres for a
detailed investigation. Our comparison between satellite and in situ Zeu for data south of 60°S did
not show significant differences between the approaches for SeaWiFS and slightly better estimates
of Zeu-Chla for MODIS (Figure A1). Overall, Zeu-IOP was shallower than Zeu-Chla, as observed
by Lee et al. (2007) for other oceanic regions.
Although the spatial distribution of Zeu is consistent, it is important to mention that close to the
Antarctic continent the values might be impacted by ice contamination. Pixels contaminated by
cloud/ice and straylight are flagged in the Level-3 data. Nevertheless, Belanger et al. (2007) and
Wang and Shi (2009) showed that the standard SeaWiFS and MODIS flags may not remove all
pixels impacted by the adjacency effect, sub-pixel ice and mixed ice-water contamination. Based
on radiative transfer simulations Belanger et al. (2007) showed the significant impact of the
adjacency effect and sub-pixel ice contamination on the water leaving radiance and derived Chla
and IOP products. In general, the sub-pixel contamination leads to an overestimation of Chla and
the total absorption at 443 nm (ܽସସଷ ). The adjacency effect overestimates Chla in low Chla waters
(0.05 mg m-3) and for Chla > 0.5 mg m-3, ܽସସଷ and Chla retrievals are underestimated. Wang and
Shi (2009) observed that MODIS Chla is often overestimated in sea ice contaminated pixels.
Therefore, both shallower and deeper Zeu regions observed close to the Antarctic continent might
be biased.
In addition, when comparing the sensors, the spatial differences were larger close to the sea ice
edge and were likely related to the few pixels sampled at different times (Figure 3.5). These
differences were as large as 20% and more pronounced in the Zeu-IOP, which might be explained
by the following reasons. The IOP approach is probably more influenced by the atmospheric
correction since the QAA uses the 670 nm band to derive the total absorption at the reference
wavelength. The 670 nm band is important for the retrievals of IOPs from Rrs in high-absorption
waters (Lee at al. 2006, Lee at al. 2007). At 670 nm water absorption dominates and the signal to
noise ratio is low, which in turn leads to a high sensitivity to light conditions. This is also the most
likely reason for the large differences seen south of 60°S (Figures 3.4 and 3.5). Moreover,
differences between Zeu-SWF and Zeu-MODIS might be associated to changes in the QAA
depending on the sensor used. Examples of the QAA adjustment to sensors are the difference in
reference wavelength (555 nm for SeaWiFS and 550 nm for MODIS) and the constants used to
derive total absorption at the reference wavelength. These are based on relations to a different set
57
of collocations and the solar zenith angle. An alternative is the use of merged products (e.g.
GlobColour, OC-CCI), aimed to reduce discrepancies caused by the use of different sensors as
observed here.
3.4.3
Validation of aph
The assessment of MODIS and SeaWiFS QAA-derived ܽ௣௛ is presented in Table 3.1. The
comparison of Rrs-satellite and in situ ܽ௣௛ shows satisfactory results in terms of log10MAE for
both sensors at 412 and 443 nm, and at 488 and 531 for MODIS. The percentage differences are
higher than the values presented by Lee at al. (2011). Nevertheless, Lee at al. (2011) derived ܽ௣௛
from ship borne Rrs instead of satellite Rrs; larger uncertainties would be expected in satellite Rrs.
The E of ܽ௣௛ -MODIS at 412 and 443 is comparable to the values reported by Shang et al.
(2011a) when evaluating QAA-derived ܽ௣௛ from satellite MODIS Rrs in the Taiwan Strait as well.
At 488 nm the error is lower. Generally, ܽ௣௛ -MODIS showed lower log10bias and log10MAE than
ܽ௣௛ -SWF for the same wavelengths. Further, the comparison of log10MAE from the Chla and
ܽ௣௛ 443 validations suggested improvement of ܽ௣௛ 443 over Chla in the SO for MODIS, as observed by Shang et al. (2011a) in the Taiwan Strait.
Uncertainties in the validation of the satellite ܽ௣௛ could be introduced by error in the in situ
measurements of ܽ௣௛ , as well as in the satellite Rrs and in the estimation of gelbstoff absorption by
the QAA (Lee et al. 2006; Lee et al. 2011; Shang et al. 2011a). Unfortunately, it is beyond the
scope of this study to propose modifications in the algorithm for the SO. Hirawake et al. (2011)
modified the QAA based on underwater spectral radiation data and in situ ܽ௣௛ from the Indian
Sector of the SO. This modified QAA was also tested by us, but the results were less robust than
with the original QAA (results not shown). In part, regional differences across the SO, as discussed above and as seen in the Zeu, make it difficult to extrapolate local properties to the entire
region.
Because Zeu-IOP was calculated using the same approach asܽ௣௛ ͶͻͲ, we could also expect an
improvement of Zeu-IOP over Zeu-Chla; particularly for MODIS that showed a larger difference in
the validation between the two Zeu approaches and lower uncertainties of aph490 than Chla. However, this was not observed here and it is likely related to our validation datasets of Zeu, ܽ௣௛ and
Chla which greatly differ in number of samples and location. Moreover, our ܽ௣௛ validation dataset
is small, especially for SeaWiFS. From 271 in situ ܽ௣௛ collected between 2007 and 2010, 12
matched with SeaWiFS observations. Persistent cloudiness and high solar zenith angles limit the
satellite retrievals in the SO.
3.4.4
Primary Production
Finally, the impact of the Zeu products on the NPP was as expected; deeper Zeu led to an
increase in NPP as more light was available (Figure 3.6). Note that the classification of empirical
and semi-analytical used for Zeu is not valid for NPP since both NPP-Zeu-Chla and NPP-Zeu-IOP
58
used ܽ௣௛ derived from QAA. The spatial differences observed between Zeu-Chla and Zeu-IOP
strongly influenced the NPP estimation. In both NPP calculations we used the same set of input
data (PAR, Dirr, ܽത௣௛ ), except for Zeu, thus the observed differences can be directly attributed to Zeu.
In particular, NPP-Zeu-Chla estimates were much higher than NPP-Zeu-IOP in the western part of
the South Pacific subtropical gyre and south of 60°S. The latter region is of great importance in
the global carbon cycle, as pointed out by Arrigo et al. (2008) and Takahashi et al. (2009).
According to these authors, once the sea ice retreats in springtime, more light and nutrients
become available enhancing the development of phytoplankton blooms and leading to a strong
sink of atmospheric CO2. Accurate estimates of NPP are essential for a better understanding of the
role of the SO in the global carbon cycle.
From the results presented here it becomes clear that the uncertainties of Zeu should be
considered to improve the estimates of NPP. Saba et al. (2011) investigated how satellite derived
sea surface temperature, mixed layer depth, Chla and PAR affected the NPP estimates of 21 ocean
colour models. They found that when uncertainties of the Chla are accounted for in NPP models,
the root mean square difference is reduced by 44% in the Antarctic Polar Front Zone. They also
observed that biases in the ocean colour NPP estimates are related to the water column depth,
possibly due to uncertainties in the Zeu.
3.5 Conclusions
Here we provided a quality assessment of the Zeu derived from MODIS and SeaWiFS using a
large dataset of in situ measurements in the SO. In summary, satellite Zeu derived using the Chla
and IOP approaches are reliable in the region. Although uncertainties depend on the sensor and
approach used, the best results were obtained by the IOP approach using SeaWiFS data. Within
the MODIS data, Zeu estimation with the Chla approach generally yielded better results than the
IOP approach. When assessing the differences in the spatial distribution between Zeu-Chla and
Zeu-IOP, large discrepancies were observed over specific regions with significant impact on the
NPP retrievals. Those differences were not observed in the validation. Therefore, we emphasize
the importance of spatial studies together with the validation using in situ measurements for
comparing ocean colour satellite products retrieved from different sensors and approaches.
Further, we validated aph and found that MODIS data lead to lower uncertainties of ܽത௣௛ and
aph443 than SeaWiFS data.
To which extend these results are influenced by the lack of in situ measurements in our dataset
used for validation and/or by regional differences in the SO is still unclear. To look more deeply in
this issue and to address these differences found in the spatial distribution of Zeu and NPP, a more
representative dataset of simultaneous bio-optical and NPP data is necessary. The results
presented here can support future campaigns by prioritizing areas of disagreement between
approaches and poorly sampled regions to reduce uncertainty of NPP in regional and global
scales. In addition, special designed satellite missions using at least two quasi-polar orbits and
same optical sensor could be considered. In this case, earlier (later) equator crossing time in
descending (ascending) mode would increase signal to noise for the SO, thus reducing
uncertainties of NPP estimates.
59
Chapter 4
Global retrieval of diatoms abundance based on
phytoplankton pigments and satellite data
60
4
Study 2: Global retrieval of diatoms abundance based on phytoplankton pigments and satellite data
4.1 The role of diatoms in the Southern Ocean
Diatoms are the most diverse phytoplankton group of the global oceans (Armbrust 2009) and the
highest number of endemic diatoms species is found in the SO (Smetacek et al. 2004). These organisms occur in a wide range of environments due to several abilities. Under nutrient or light
stress, diatoms can: migrate in the water column by controlling their buoyancy, store nutrients in
the central vacuoles for later use, reduce iron requirements and maintain symbiosis with nitrogen
fixing cyanobacteria (Armbrust 2009; Kooistra et al. 2007). In addition, diatoms produce thick
cell walls, spines and toxins to avoid grazers (Kooistra et al. 2007; Smetacek 1999) and the rapid
mass sinking events are considered a seeding strategy to overcome periods adverse to growth
conditions (Kooistra et al. 2007; Smetacek 1985).
In iron-limited regions of the SO, the community of diatoms is dominated by large species
with thick silica shells and high silica to nitrogen ratio (>2, Smetacek et al. 2004) for grazer protection (Assmy et al. 2013). Key species in iron-limited regions include the endemic Fragilariopsis kerguelensis and Thalassiothrix antarctica (Smetacek et al. 2004). While the biomass is recycled in the surface by grazing, the sinking out of the thick shells sequesters silica, resulting in loss
of Si (silica sinkers) but retention of N and P at surface. Part of the frustules will dissolve while
sinking and accumulate as silicic acid in the Circumpolar Deep Water and part will be buried into
the sediments forming the major global biogenic silica accumulation (Smetacek 1999; 2004).
In iron-replete regions on the other hand, large-medium sized and weakly silicified diatoms
prevail. These diatoms have high growth rates and form high biomass blooms that drive the carbon pump (Smetacek et al. 2004). Common diatoms include Thalassioria antarctica and the genus Chaetoceros. As a result, diatoms shape the biogeochemistry of the oceans as carbon and silica sinkers (Assmy et al. 2013; Smetacek et al. 2004), in addition to their role in the marine food
web leading to intensive fisheries in coastal waters (Armbrust 2009).
4.2 Motivation
This second study examines the satellite retrievals of diatoms abundance derived by the ABA
of Hirata et al. (2011). Compared to optically based approaches, a great advantage of the ABA is
the smaller computational effort; even if the satellite data volume becomes larger with higher
temporal and spatial resolutions, the data processing load is not heavy and re-processing can also
be done relatively easily. The ABA can be applied to global level-2 or level-3 products of TChla,
which are freely available to the scientific community, as opposed to, for example, PhytoDOAS
method that uses the top of atmosphere radiance data (i.e. level-1 product) which is not freely
available. On the other hand, because ABA is an empirical model, it is recommended to re-
61
evaluate the approach in the light of additional in situ phytoplankton pigment data and/or satellite
data.
This study is based on two premises: (i) diatoms are the major primary producers and key
players in the carbon and silicon pump in the SO and (ii) 90% of the diffusely reflected irradiance
measured by ocean color sensors originates from the first optical depth, also referred to as the
penetration depth (Zpd) (Gordon and Mccluney 1975). These premises are also general limitations
of the existing ABA. Although Hirata et al. (2011) used a large global dataset of phytoplankton
pigments, new measurements, particularly in the SO (defined here as the region south of 50°S),
have become available since then. Main objectives of this study are:
(1). Compilation of a new and large global dataset of in situ phytoplankton pigment profiles,
including more measurements in the SO (Figure 4.1) which was not well covered previously, and
to investigate the relationship between fractional contribution of diatoms and TChla using the new
dataset in comparison to previous findings.
(2). Refinement of the ABA to account for the pigment information in the Zpd (ABAZpd). In
ABA (Hirata et al. 2011), the fractional contribution of diatoms to TChla was estimated based on
the previous work of Uitz et al. (2006), who used the phytoplankton pigment concentration integrated over the euphotic depth (Zeu). However, the pigment concentration estimated by the satellite sensor is an optically-weighted concentration in the Zpd, which is approximately 4.6 times
shallower than the Zeu (Hyde et al. 2007).
(3). Evaluation of the performance of the ABA (i.e. ABAZpd) for global oceans and for the SO
region.
Figure 4.1. Distribution of the quality controlled in situ measurements. The SO, region south of
50°S, is the portion of the global ocean presented in blue.
62
4.3 Data and Methods
4.3.1
In Situ Measurements of Phytoplankton Pigments
A dataset of phytoplankton pigment profiles measured with the HPLC technique was supplemented with data obtained from the SeaWiFS Bio-optical Archive and Storage System – SeaBASS (Werdell et al. 2003), Marine Ecosystem Data – MAREDAT (Peloquin et al. 2013), and
from the individual cruises KEOPS (Uitz et al. 2009b), Bonus Good Hope, ANT-XVIII/2 EisenEx, ANTXXI/3 – EIFEX (Smetacek et al. 2012), ANT XXVI/3, ANT XXVIII/3, Sonne
SO218 (Cheah et al. 2013), Merian 18-3, Meteor 55 and Meteor 60. The pigments from the cruises Meteor 55, Meteor 60, ANT XXVI/3 and ANT-XVIII/2 were measured in accordance with the
method described in Hoffmann et al. (2006) and for the cruises Merian 18-3 and ANT XXVIII/3
in accordance with Taylor et al. (2011).
The data were quality controlled in a way similar to the one used by Uitz et al. (2006) and Peloquin et al. (2013): (i) Samples with accessory pigment concentrations below 0.001 mg m-3 were
set to zero, (ii) samples with TChla below 0.001 mg m-3 and fewer than 4 accessory pigments
were excluded. To ensure that the profiles had a minimum vertical resolution, we restricted the
dataset to profiles with at least (i) one sample at the surface (0 to 12 m), (ii) one sample below the
surface, (iii) samples collected at four or more different depths, and (iv) with one sample within
the Zpd. The last quality control measure was based on the log10-linear relationship between TChlaZpd and the sum of all accessory pigments in the Zpd (TACCZpd). Data that fell outside the 95%
confidence interval were removed. The quality controlled dataset was corrected for Fuco to account for its co-existence in other PFTs, in accordance with Hirata et al. (2011).
In addition, samples located in coastal waters (< 200 m) were excluded using the ETOPO1 bathymetry (Amante and Eakins 2009). The final dataset contained 3988 samples, which were randomly split into work (70% of the data) and validation (30% of the data) subsets. While the whole
dataset was used to calculate the partial coefficients used for estimating f-DiatomZpd, the work and
validation subsets were used for model development and validation of the ABAZpd, respectively.
4.3.2
Satellite Data
Eleven years (2003–2013) of MODIS Aqua Level 3 4km binned TChla data (R2012.0) were
used. MODIS is a multispectral sensor on board of the Aqua satellite and with global coverage.
The data were obtained from http://oceancolor.gsfc.nasa.gov/ at daily temporal resolution. Monthly averages of diatoms abundance were calculated onto a 10 minute grid and used to derive climatological maps of diatoms abundance. To avoid coastal waters, where the retrieval of the ABA
was not intended, we removed grid cells located in waters shallower than 200 m using the
ETOPO1 bathymetry (Amante and Eakins 2009).
4.3.3
An Improved Abundance Based Approach
63
In previous approach (Hirata et al. 2011), the f-Diatom was calculated using the coefficients of
Uitz et al. (2006), which take account of the phytoplankton pigment integrated over the Zeu. Here,
we extended the ABA to take account of the information in the Zpd. For this purpose, we recalculated the coefficients ai (Eq. 2.30) using the updated global dataset of HPLC phytoplankton pigment profiles. The weighted pigment concentration in the Zpd (DPZpd) was calculated as described
in Gordon and Clark (1980) for TChla:
‫ܥ‬௣ௗ ൌ
ೋ೛೏
‫׬‬బ
஼ሺ௭ሻ௚ሺ௭ሻௗ௭
(4.1)
ೋ೛೏
௚ሺ௭ሻௗ௭
‫׬‬బ
where C is the TChla at a depth z and g is:
௭
݃ሺ‫ݖ‬ሻ ൌ ‡š’ൣെʹ ‫׬‬଴ ‫݀ܭ‬ሺ‫ ݖ‬ᇱ ሻ݀‫ݖ‬Ԣ൧
(4.2)
The same approach was applied to the other pigments. The light attenuation coefficient (Kd490,
m ) was estimated from profiles of chlorophyll-a concentration (Morel and Maritorena 2001):
-1
‫݀ܭ‬ସଽ଴ ሺ‫ݖ‬ሻ ൌ ‫݀ܭ‬௪ ൅ ͲǤͲ͹ʹͶʹ ‫݈݄ܽܥܶ כ‬ሺ‫ݖ‬ሻ଴Ǥ଺଼ଽହହ
(4.3)
where Kdw (m-1) is the attenuation coefficient for pure water (0.01660 at 490 nm). The Zpd was
computed as Zpd = Zeu/4.6 and Zeu was derived from the surface TChla as ܼ௘௨ ൌ ͵Ͷܶ‫ି ݈݄ܽܥ‬଴Ǥଷଽ
(Morel, in Lee et al. 2007). Profiles were interpolated with 1-m increments from the deepest sample to the sample closest to the surface before the calculation of DPZpd.
Nonlinear minimization was used to retrieve the partial coefficients, which represent the estimates of the TChla to the DP ratios (Uitz et al. 2006). The function to be minimized is expressed
as:
ԡ݁‫݌ݔ‬ሺܿሻ‫ܯ‬௓௣ௗ ԡ ՜ ݉݅݊
(4.4)
where c is a vector containing the seven coefficients which each correspond to each DP Zpd on the
log scale, and MZpd a matrix containing the seven DPZpd. The nonlinear minimization method requires an initial guess of c, which was obtained from the multiple linear regression analysis. The
standard deviation of the coefficients is given by the square root of the diagonal elements of the
inverse of the Hessian matrix.
Using the new coefficients, the f-DiatomZpd was calculated for each sample of the work and
validation subsets. The work subset was then sorted according to the TChla Zpd and smoothed with
a 5-point running mean filter to improve the signal-to-noise ratio (Hirata et al. 2008; Hirata et al.
2011). Next, the relationship between f-DiatomZpd and TChlaZpd was quantified using a nonlinear
least-square fit applied to the work subset and represented by a model and its fitting parameters.
Once the model has been defined, satellite-derived TChla data was applied to the model to obtain
64
the global distribution of f-DiatomZpd. Diatom abundance (DiatomZpd, mg m-3) is then obtained by
multiplying f-DiatomZpd by TChlaZpd.
The accuracy of the new model was tested using the validation subset. The uncertainties were
estimated by the mean absolute error (MAE, Willmott and Matsura 2005) between the modeled
and the measured (in situ) DiatomZpd. The models were compared by the difference between the
MAE of the original model and the new model, relative to the original model, and expressed in
percent (%). The data were log transformed prior to the calculation of the validation statistics. We
used log10(data + λ) where λ=0.00003, approximately one half of the smallest non-zero value of
the in situ DiatomZpd validation data, since the dataset contained zeroes. In addition, to investigate
whether using different partial coefficients results in significant changes in f-Diatom, we estimated f-Diatom using the coefficients of Uitz et al. (2006) and Brewin et al. ((2014)) and compared
the results based on the coefficient of determination. The processing steps of ABAZpd are summarized in Figure 4.2.
Figure 4.2. A flow chart of the processing steps conducted to retrieve diatom abundance using
ABAZpd.
4.3.3.1 A Regional Model for the SO
The main difference between the SO model and the global model is that the relationship between DiatomZpd and TChlaZpd is investigated not in terms of f-DiatomZpd, but instead in terms of
the concentration of TChlaZpd that is attributed to diatoms, similar to the approach adopted by
Brewin et al. (2010) to retrieve phytoplankton size classes. As in Brewin et al. (2010), the fit func65
tion was applied to log10-transformed data. To develop the regional model for the SO, we selected
the samples of the global work and validation datasets that were located in the SO, creating a SO
work and a validation dataset with 1069 and 460 samples, respectively. The relationship between
DiatomZpd and TChlaZpd was investigated and validated. Note that for the work dataset we applied
the running mean exclusively to the SO data.
4.3.4
Statistical Analysis of Trends
Linear trends were computed for February from monthly standardized anomalies over the
2003-2013 period in the SO using the regional model. To remove the seasonal cycle we calculated
the monthly anomalies in diatoms abundance for each grid cell by subtracting the climatological
mean from the corresponding monthly mean (e.g., February 2003 - climatology of February). The
monthly anomalies were divided by the corresponding climatological standard deviation (e.g.,
standard deviation of February) to enable the direct comparison of trends between different regions (grid cells). The trends were computed using the non-parametric Kendall’s tau test with
Sen’s method at the 95% confidence level and in grid cells with 100% temporal coverage.
4.4 Results and Discussion
4.4.1
The ABAZpd
Table 4.1 shows the partial regression coefficients, and their respective standard deviation, calculated with Eq. 4.4. For comparison, we also present the partial coefficients estimated by Uitz et
al. (2006), Brewin et al. (2014) and Fujiwara et al. (2014). Comparing our coefficients with those
from Uitz et al. (2006), there is a notable difference, except for the coefficients of Fuco and
TChlb. These differences result from the inclusion of more profiles, their geographical distribution, the adjustment of Fuco prior to the DPA analysis, and because we used the pigment concentration weighted in the Zpd, while Uitz et al. (2006) integrated the pigments over Zeu. When compared to the two other studies, where the partial coefficients were derived from surface measurements, our coefficients are more similar to those described in Brewin et al. (2014). Brewin et al.
(2014) included measurements of five Atlantic Meridional Transect (AMT) cruises in the Atlantic
Ocean, while Fujiwara et al. (2014) used measurements from three cruises in the Western Arctic
Ocean. Although our dataset includes measurements from these regions, the number of samples in
the Arctic region is fewer than that from the Atlantic (Figure 4.1).
66
Table 4.1. The partial regression coefficients and standard deviation (in brackets) where available.
The number of samples is indicated by N. The empty fields indicate that the coefficient is not
statistically significant.
Coefficients
Ocean
N
Present
study
Uitz et al.
(2006)
Brewin et al.
(2014)
Fujiwara et
al. (2014)*
Global
Perid
Hexfuco
Butfuco
Allo
TChlb
Zea
Atlantic
3988 1.554
(0.010)
2419 1.41
(0.02)
466 1.72
0.413
(0.568)
1.41
(0.10)
1.27
0.855
(0.068)
1.27
(0.02)
0.68
1.174
(0.145)
0.35
(0.15)
1.42
2.387
(0.099)
0.60
(0.16)
4.96
1.062
(0.070)
1.01
(0.10)
0.81
2.037
(0.040)
0.86
(0.09)
1.28
Arctic
76
1.49
1.74
5.88
1.31
3.54
Global
Fuco
1.85
* standard errors are less than 1.
Moreover, we have re-run the analysis taking into consideration the surface samples (< 12 m)
from our profiles and observed only a slight difference in the coefficient of Fuco (1.531) as compared to the weighted Zpd concentrations. Except for Perid and Hexfuco, the standard deviation of
our coefficients are lower than, or similar to, the ones obtained by Uitz et al. (2006).
Nonetheless, we observed very similar f-Diatom values when using the partial coefficients of
Uitz et al. (2006), Brewin et al. (2014) and ours. The coefficients of determination are higher than
0.98, suggesting the choice of partial coefficients has no influence on the retrievals of f-Diatom,
which is consistent with Brewin et al. (2014). Brewin et al. (2014) compared size-fractionated
chlorophyll (SFC) estimated from phytoplankton pigment data and calculated using Uitz et al.
(2006) partial coefficients and their own, with size-fractionated filtration (SFF) measurements.
They observed biases between SFC and SFF for nanoplankton and picoplankton size classes;
however, the variations in the partial coefficients did not influence the results significantly. The
high correlation between the TChlaZpd and ୵, with DPw calculated using Eq. 4.4 (r2 = 0.85, ୵
= 0.86 TChlaZpd + 0.074, N = 3988, p < 0.001), gives us confidence to use the partial coefficients
to determine the f-Diatom.
Figure 4.3 shows the change in the f-DiatomZpd with increasing TChlaZpd. The green and blue
lines represent the new model (ABAZpd) and the model of Hirata et al. (2011) (ABA*), respectively, parameterized with the DPZpd dataset. The red line represents the original model and fitting
parameters of Hirata et al. (2011) (ABA**). It can be seen that diatoms are dominant at high TChlaZpd (Figures 4.3a and b), which is consistent with previous studies (Hirata et al. 2011) even if a
significant number of new samples were added in our dataset. Moreover, we also observed unusually high f-DiatomZpd in low TChlaZpd waters (< 0.1 mg m-3, N = 670). Taking a closer look at the
profiles, in which FucoZpd corresponded to at least 50% of the TACCZpd, we observed that most of
the data (12 out of 16) are from samples taken in Antarctic, in the East Antarctic marginal ice
zone (BROKE cruise, Wright and van den Enden 2000). On average, the ratio of FucoZpd to TChlaZpd is 0.165 for the entire DPZpd dataset, 0.071 excluding the SO data, but 0.317 for the SO data,
indicating higher f-DiatomZpd values in low TChlaZpd waters in the SO.
67
Figure 4.3. Relationship between TChlaZpd and f-DiatomZpd: (a) Global dataset (N = 2806), (b)
global dataset excluding SO data (N = 1737) and (c) SO data (N = 1069). The datasets were
smoothed with a 5 point running mean to improve the signal-to-noise ratio (Hirata et al. 2011) The
green and blue lines represent the new model (ABAZpd) and the model of Hirata et al. (2011)
(ABA*) parameterized with the DPZpd dataset. The red line represents the original model and
fitting parameters of Hirata et al. (2011) (ABA**). The fitting parameters are presented in Table
4.2. The MAE values refer to the errors in terms of f-DiatomZpd. Note that we could not fit the
global models to the SO dataset exclusively. The cyan and green lines in (c) represent the regional
model for the SO and the ABAZpd plotted with the global fitting parameters as reference.
68
Table 4.2. Models of f-DiatomZpd as a function of TChlaZpd and their respective fitting parameters
used to plot the lines in Figure 4.3a and 4.3b. Note that we could not fit the global models to the
SO data exclusively. The fitting parameters of the original ABA model of Hirata et al. (2011)
(ABA**) do not change and therefore they are presented only once in the table.
Global dataset
Global dataset
excluding SO
data
f-DiatomZpd
ABAZpd
ABA*
ABA**
ABAZpd
ABA*
Model
a0 + a1sin(a2(x + a3))
[a0 + exp(a1x + a2)]-1
[a0 + exp(a1x + a2)]-1
a0 + a1sin(a2(x + a3))
[a0 + exp(a1x + a2)]-1
a0
0.4629
1.0733
1.3272
0.3909
1.5890
a1
0.3921
-2.0484
-3.9828
0.4131
-4.3778
a2
1.2214
0.1314
0.1953
1.3763
-0.1521
a3
-0.01412
-0.0114
-
x=log10(TChla)
*model of Hirata et al. (2011) parameterized with the DPZpd dataset.
**original model and fitting parameters of Hirata et al. (2011).
Considering our newer dataset, Hirata et al. (2011) considerably underestimates f-DiatomZpd in
almost the entire TChlaZpd range (Figure 4.3a - red line). This is partly due to the difference in the
dataset used. When we fit their model to the new dataset, the model is found to fit well to the data,
as indicated by the low errors (Table 4.3 and Figure 4.3a - blue line). However, it fails when predicting f-DiatomZpd in very low TChlaZpd waters, mostly for the SO. Thus, we test a new model, a
sinusoidal function to better fit this observed trend in the SO (Table 4.3, Figure 4.3a - green line).
The ABAZpd and ABA* produce almost identical curves for TChlaZpd above 0.065 mg m-3 and
similar fitting and validation statistics. The ABA** model provides accurate retrieval of diatoms
globally. However, it produces larger errors than the ABAZpd and the ABA* do for the SO. The
ABAZpd improves the MAE by 27.96% for the SO (Table 4.4 and Figure A2 in the Appendix).
Table 4.3. Statistical results of the fits for the global dataset and global excluding SO data using
the fitting parameters of Table 4.2. Note that we could not fit the global models to the SO data
exclusively. The fitting statistics for the SO dataset refer to the regional SO model (Figure 4.5).
The MAE is given in f-DiatomZpd for the global models and for the regional model in mg m-3
(log10-transformed data).
Fit
Global dataset
Global dataset
excluding SO data
SO dataset
ABAZpd
ABA*
ABA**
ABAZpd
ABA*
ABA**
Regional model
N
2806
2806
2806
1737
1737
1737
1069
*model of Hirata et al. (2011) parameterized with the DPZpd dataset.
**original model and fitting parameters of Hirata et al. (2011).
69
r2
0.71
0.70
0.66
0.89
0.88
0.88
0.95
p – value
=0
=0
=0
< 0.001
< 0.001
< 0.001
< 0.001
MAE
0.085
0.087
0.118
0.036
0.037
0.038
0.104
Table 4.4. Statistical results of the validation in terms of diatoms abundance. Note that we could
not fit the global models to the SO data exclusively. The results for the SO dataset correspond to
the global models using the global fitting parameters and the regional model. The MAE is given in
mg m-3. The statistics were calculated with log10-transformed data (e.g., log10(y+0.00003)).
Validation
N
r2
p - value
MAE
ABAZpd
1182
0.57
< 0.001
1.219
Global dataset
ABA*
1182
0.55
< 0.001
1.217
ABA**
1182
0.57
< 0.001
1.035
Global dataset
ABAZpd
722
0.59
< 0.001
0.883
excluding SO data
ABA*
722
0.68
< 0.001
1.195
ABA**
722
0.69
< 0.001
1.200
ABAZpd
460
0.40
< 0.001
0.559
SO dataset
ABA*
460
0.39
< 0.001
0.562
ABA**
460
0.39
< 0.001
0.776
Regional model
460
0.39
< 0.001
0.465
*model of Hirata et al. (2011) parameterized with the DPZpd dataset.
**original model and fitting parameters of Hirata et al. (2011).
To further investigate the influence of the SO data, we removed these data from the work
dataset (38% of the data), recalculated f-DiatomZpd, and fitted the models (Figure 4.3b and Table
4.2). The comparison of Figure 4.3a and Figure 4.3b shows clearly the influence of the SO data,
which is responsible for most of the data spread in Figure 4.3a as well as for the high f-DiatomZpd
in low TChlaZpd waters. When we exclude the SO data from the analysis, the fits improve greatly
the MAE decrease to values close to 0.04 (Table 4.3). In addition, it leads to a better
representation of the diatom abundance in oligotrophic waters, as well as to an underestimation of
the actual f-DiatomZpd in the SO, as shown in Figure 4.4. The advantage of including the SO data
is a more realistic retrieval of diatoms in the SO, but an overestimation in other regions of low
TChlaZpd. While the in situ data show that the f-DiatomZpd might be very low (~ 0) at very low
TChlaZpd (e.g. in oligotrophic gyres), the predicted f-DiatomZpd presents values higher than zero,
overestimating f-DiatomZpd in the oligotrophic gyres.
Figure 4.4. Monthly mean TChlaZpd (mg m-3) of diatoms for February 2003 using the ABAZpd
model parameterized with: (a) Global dataset (average = 0.060 mg m-3) and (b) global dataset
excluding SO data (average = 0.041 mg m-3). White areas correspond to waters with depths
70
shallower than 200 m or without satellite information.
It should be noted that the model used to retrieve f-DiatomZpd as a function of TChlaZpd was
empirically built upon in situ datasets, which showed that diatoms tend to be the dominant PFT at
high TChla. However, this may not be the case of blooms of mixed PFTs, or dominated by a different PFT as pointed out by Brewin et al. 2010. For example, the coccolithophore Emiliania huxleyi typically occur in the North Atlantic and can form massive blooms in the Bering Sea (Iida et
al. 2002). In the Ross Sea in the SO, blooms of the haptophyte Phaeocystis antarctica can exceed
15 mg m-3 and dominate the spring bloom, following by a later development of diatom blooms in
the summer (Smith et al. 2012). In such cases, additional information on PFTs derived from
methods that do not depend on this assumption (e.g. PhytoDOAS) may improve the knowledge on
the diatom abundance and their distribution pattern.
Moreover, we did not obtain significant results in fitting the two global models to the SO data
exclusively (Figure 4.3c, ABAZpd plotted as reference). The diatoms in the SO exhibit a variability
which is different from other oceanic regions (e.g., the North Pacific and the North Atlantic), and
there is a need for a regional SO model. Thus, we developed a regional model for the SO, and the
relationship between TChlaZpd and DiatomZpd can be expressed as: log10(y) = 1.1559log10(x) + (0.2901) (Figure 4.5). The validation results of the SO model show that the regional model is consistent and more appropriate than the global ABAZpd model for retrieving diatoms in the SO (Table 4.4 and Figure 4.5). The regional model improved by 17% the retrieval of diatoms abundance
in the SO compared with the ABAZpd.
Figure 4.5. On the left: relationship between TChlaZpd and DiatomZpd in the SO with the fit
function plotted in blue (log10 transformed data). On the right: validation calculated with both
log10 transformed data (e.g. log10(y+0.00003)). The red line represents the 1:1 line.
The ideal global retrieval of diatoms should apply the ABAZpd model parameterized with the
global dataset excluding SO data (Figure 4.3b green line) to the region north of 50°S, and the regional SO model for waters south of 50°S. These two models presented overall the lowest fitting
and validation errors for the corresponding regions. This approach would not only provide more
accurate retrievals of diatoms in the SO, but also overcome the overestimation of the global AB-
71
AZpd model in oligotrophic waters. However, applying two models generated a non-negligible offset between the SO and adjacent oceans (result not shown).
4.4.2
Satellite Retrieval of Diatoms using ABAZpd
Acknowledging the uncertainties of the satellite Chla product, we first assessed the difference
between the satellite retrievals of diatom abundance using the ABAZpd and the original ABA, for
the SO and global oceans. As expected from the previous findings (Figure 4.3a), we observed
that, on average, higher abundances of diatoms were retrieved with the ABAZpd than with the original ABA for the entire 2003-2013 period. For the SO, the concentration of diatoms calculated
using the global ABAZpd is 0.074 mg m-3 and for the global oceans 0.070 mg m-3. In contrast, estimates of diatoms with the original ABA are 0.049 and 0.050 mg m-3, respectively. For comparison, the concentration of diatoms using the regional SO model is 0.117 mg m-3. This evidence of
the enhanced abundance of diatoms retrieved from the global ABAZpd model and from the regional SO model suggests that the production and export of carbon to the deep ocean might be larger
than previously expected in the SO.
The new global climatology of diatom abundance is presented in Figure 4.6. The climatology
for the SO is presented in the Appendix (Figure A3). The general distribution of the global diatom
abundance is in line with current knowledge on the distribution of diatoms, i.e. higher concentrations of diatoms in the upwelling and coastal regions. Low concentrations of diatoms are observed
in oligotrophic waters of the subtropical gyres and in HNLC waters, such as regions in the SO
where waters are rich in macronutrients but are lacking in iron. There is also a clear seasonal cycle
in the polar regions, with diatoms reaching the highest concentrations during their respective
summer months, which is also observed in the climatology for the SO. Among other important
patterns is the increase in diatom concentration from January to March and again high concentrations in September in the Arabian Sea. These observed patterns are associated with the Northeast
and Southwest monsoons, respectively. According to Garrison et al. (2000), the monsoon seasons
are generally characterized by increased concentrations of diatoms, thus our result shows a consistency with the previous in situ study too.
72
Figure 4.6. Climatology of TChlaZpd of diatoms (mg m-3) for the months of January to December
based on the period 2003-2013 retrieved using the ABAZpd model. White areas correspond to
waters with depths shallower than 200 m or without satellite information.
The climatology mostly covers the spatial variability, within a limited temporal range, whereas
the trend gives information for a longer period, and both are important information for the understanding of ocean biogeochemistry. The spatial variability of the linear trends of diatom abundance in the SO is high, and no significant trend was observed for most of the sub regions of the
SO (results not shown). Overall the trend for the SO was 0.036 (year-1) (p-value = 0.019). Clearly,
a more detailed analysis is needed to investigate the main driving forces behind these trends.
4.4 Conclusions
In conclusion, we have shown that the original ABA underestimates the diatom abundance in
the SO. Our investigation revealed that diatoms in the SO might be more abundant than previous
thought, possibly because (1) the lack of in situ phytoplankton pigment data, and that (2) the relationship between Chla and the f-Diatom in the SO is distinct from the global relationship.
We have developed a new global and a regional ABAZpd that improves the uncertainties of the
retrievals of diatoms in the SO. The mean absolute error (MAE) declined from 0.776 to 0.559 using the global ABAZpd, improving by 28% the estimation of diatoms in the SO. The regional mod73
el further improved the MAE by 17% (MAE = 0.465) compared with the global ABA Zpd model.
This was achieved by re-evaluating the ABA using a large dataset of global phytoplankton pigment profiles spanning 24 years (1988–2012). Additionally, the ABA was further improved by
considering the information in the Zpd.
We have shown that the ideal global retrieval of diatoms combines the ABAZpd model fitted to
the dataset (excluding SO data, MAE = 0.883) with the regional SO model. However, applying
two models generates an offset between the oceans, thus selective use of the global and the SO
algorithms may be necessary depending on the objective of the application.
Satellite retrievals of PFTs are a useful tool for identifying and quantifying their presence in
the oceans and in this study we have advanced our knowledge on the retrieval of diatoms from
space by identifying limitations and developing improvements. Future studies should focus on
optimizing the ABA method also for other PFTs.
74
Chapter 5
Mean patterns and interannual variability of diatom
phenology in the Southern Ocean
75
5
Study 3: Mean patterns and interannual variability of diatom
phenology in the Southern Ocean
5.1 Motivation
In the previous study we have advanced our knowledge on the retrieval of diatoms abundance
from space by identifying limitations of the original method and developing a regional model for
the SO. However, the mean patterns and interannual variability of the diatom bloom phenology in
the SO remain unquantified. As the major primary producers in the region (Rousseaux and Gregg
2014), this study is focused on their specific phenology and variability. The aim of this chapter is
improve previous studies on phytoplankton bloom phenology by: (i) focusing on the SO, (ii) using
a new merged satellite Chla product with better spatial and temporal coverage than the datasets
used in previous studies, (iii) examining a longer time series (1997-2012), (iv) looking at the
concentration of Chla of diatoms using the regional algorithm of Soppa et al. (2014), (v)
examining the different characteristics of the phenology, (vi) investigating trends, and for
completeness, (vii) investigating if the interannual variability of the diatom phenology could be
modulated by the large scale climate oscillations ENSO and SAM.
5.2 Data and Methods
5.2.1
Satellite data
We analysed 15 years (September 1997 – April 2012) of the level 3 Chla data (ESACCI-OCL3S product, 4 km, version 1.0) from the Ocean Colour Climate Change Initiative (OC CCI). The
OC CCI project is a European effort to produce high quality ocean colour products combining the
MERIS, MODIS-Aqua and SeaWiFS sensors. Current data processing improves limitations of
ocean colour remote sensing in polar regions due to low solar elevation and frequent cloud cover.
Radiometric contaminations by sun glint, thin clouds or heavy aerosol plumes are removed from
the MERIS with the Polymer algorithm (Steinmetz et al. 2011), while the SeaWiFS and MODIS
data are processed for atmospheric correction with the algorithm of Gordon and Wang (1994).
Subsequently, the SeaWiFS OC4v6 algorithm is applied on the merged remote sensing reflectance
data to obtain the OC-CCI Chla product. The global validation of the Chla product with in situ
HPLC Chla have shown that the relative errors are lower than 30% for most of the Chla range,
except for concentrations lower than 0.1 mg m-3 (Krasemann et al. 2014). More details on the
project and processing steps can be found in http://www.esa-oceancolour-cci.org/, where also the
Chla data are available.
In our study, we calculated weekly averages of Chla from daily data onto a 15 minutes spatial
grid for the area south of 50°S. To avoid coastal waters, we removed the three closest grid cells to
the coastline. Diatom abundance was derived by applying the regional ABA Zpd SO model (Soppa
et al. 2014) to the weekly Chla data, hereinafter referred to as Diatom-Chlorophyll-a (Dia-Chla).
76
5.2.2 Polar fronts position
We used the weekly position of the Polar Front (PF), available at http://ctoh.legos.obsmip.fr/applications/mesoscale/southern-ocean-fronts, which is derived based on sea level anomalies observed in altimetry data and climatological mean sea level from historical data and ARGO
profiles (Sallée et al. 2008). The mean and the standard deviation of the PF position were calculated from 1997 to 2012, for the months of September to April, the same period that was used to
describe the phenology (see section 2.4). The standard deviation of the latitudinal position of the
PF was used as a proxy of the interannual variability of the PF position. In addition, we included
the mean position of the Southern Antarctic Circumpolar Current Front (SACCF) in our analysis.
The SACCF position is derived from historical hydrographic data of the SO until 1990 (Orsi et al.
1995).
5.2.3
Maximum Sea Ice Extent
We used monthly sea ice extent data (Fetterer et al. 2002) for the SO to delineate the seasonal
ice zone. The seasonal ice zone is the area delimited by the winter maximum of sea ice extent
(NSDC 2015) which occurs in September. Sea ice extent data is available by the National Snow
and
Ice
Data
Center
(NSIDC)
at
ftp://sidads.colorado.edu/DATASETS/NOAA/G02135/shapefiles/. The maximum sea ice extent
of September was binned into longitude bins of 1 degree. Coordinates were automatically extracted from the sea ice extent data and statistics (mean and standard deviation) of the latitude values
were calculated for each longitude bin to examine the interannual variability in the entire period
(1997 to 2012).
5.2.4
Climate indices
To investigate if the Dia-Chla phenology in the SO is influenced by ENSO and SAM, we used
two indices: the Multivariate El Niño Southern Oscillation index (MEI) and the Antarctic Oscillation (AAO) index. The MEI, available at http://www.esrl.noaa.gov/psd/enso/mei/#loadings, is
based on six variables (cloudiness, sea surface temperature, sea-level pressure, surface air temperature and the zonal and meridional components of the surface wind) over the tropical Pacific from
30°N to 30°S (Wolter and Timlin 1993). Positive MEI values characterize El Niño events while
negative values indicate La Niña events. The Antarctic Oscillation (AAO) index, available at
http://www.cpc.ncep.noaa.gov/products/precip/CWlink/daily_ao_index/aao/monthly.aao.index.b7
9.current.ascii.table, is based on the first principal component of monthly mean pressure anomalies at 700 mb for the region south of 20°S (Mo 2000). Positive/negative phases of the Southern
Annular Mode (SAM) are associated with positive/negative values of AAO, respectively. Annual
ENSO and SAM indices were calculated by averaging their respective indices from September of
77
the previous year to April of the following year, the same period used to estimate the phenological
indices (Figure 5.1).
Figure 5.1. Time series (adimensional) of annual Multivariate ENSO Index (MEI, solid line) and
Antarctic Oscillation index (SAM, dashed line).
5.2.5
Phenological indices
We assessed the diatom phenology using the threshold method of Siegel et al. (2002).
Phytoplankton blooms start (defining the bloom start date - BSD) when the Chla value exceeds
the value of 5% above the median (Siegel et al. 2002) and remains above this threshold for at least
two consecutive weeks (Thomalla et al. 2011). To isolate primary blooms from secondary blooms,
we first identified the maximum Chla of the time series and then looked backwards in time to find
the bloom start date (Brody et al. 2013). The bloom end date (BED) was determined as the first
week when Dia-Chla level fell below the threshold. The period between bloom start date and end
date defines the total bloom duration (BD). Within this period the Dia-Chla reaches a maximum
(CM) at the date of Dia-Chla maximum (CMD). The sub-periods before and after the maximum
determine the bloom growth duration (BGD) and bloom decline duration (BDD), respectively.
During the growth duration, the average (CAV) and integrated Dia-Chla values (CI) are
calculated. In addition, the amplitude of the bloom (CA) is determined as the difference between
maximum and threshold Dia-Chla value. The phenological indices are listed in Table 5.1 and
illustrated in Figure 5.2. Using these indices, we analyzed the phenology in the entire time series
(1997 to 2012), each year from September to April of the following year (e.g. September 2002 April 2003).
Before computing the phenological indices, the time series were linearly interpolated in time to
fill gaps less than 3 weeks in length (Henson and Thomas 2007). After the temporal interpolation,
if there were remaining gaps of more than two weeks between the date of Dia-Chla maximum and
the estimated bloom start or bloom end date, these phenological indices were not calculated to
avoid erroneous detection of the bloom timing. This led to slightly different data coverage of the
78
phenological indices. The best data coverage is achieved for date of Dia-Chla maximum, Dia-Chla
maximum and amplitude.
Table 5.1. Phenological indices.
Indices
Bloom Start Date
Date of Dia-Chla Maximum
Bloom End Date
Bloom Duration
Bloom Growth Duration
Bloom Decline Duration
Dia-Chla Amplitude
Dia-Chla Maximum
Dia-Chla averaged over BGD
Dia-Chla integrated over BGD
Abbreviation
BSD
CMD
BED
BD
BGD
BDD
CA
CM
CAV
CI
Units
Week
Week
Week
Week
Week
Week
mg m-3
mg m-3
mg m-3
mg m-3
Figure 5.2. Schematic of the indices used to describe the diatom phenology.
5.2.6
Statistical analysis
The mean spatial patterns were obtained by averaging the 15 years of phenological indices.
The interannual variability of the diatom phenology was examined by: (i) the relative standard
deviation (RSD) of the indices, (ii) trends, (iii) correlations and partial correlations with ENSO
and SAM and, (iv) composite maps of the anomaly of the indices. The relative standard deviation
is the standard deviation of the indices divided by the average over the 15 years. Trends, correlations and composite maps were calculated using the standardized anomaly data. Standardized
anomalies (adimensional and hereafter termed as anomalies) were produced by subtracting the
average (15-yr) from the annual phenology data (e.g. 2002-2003) and dividing by the standard
79
deviation (15-yr), pixel by pixel. Trends were investigated with non-parametric Kendall’s tau test
with Sen’s method at the 95% confidence level for each grid cell (only when 100% of data were
available). The correlation between the climate indices and anomalies of the phenological indices
was determined using Spearman correlation. Partial correlations were used to study the influence
of both oscillations separately, for example, by considering the relationship between SAM and
Dia-Chla maximum after removal of the variance of ENSO (Pohl et al. 2010). Composite maps of
the anomalies of the phenological indices were computed by averaging the anomalies from the
different phases (positive/negative) of the ENSO and SAM, as well as for amplified years (e.g. El
Niño coincided with negative phase of SAM). Using composite maps we investigated the dominant patterns of the anomalies associated with the different phases and oscillations (Kwok and
Comiso 2002).
5.3 Results and discussion
5.3.1
Mean spatial distribution of the phenological indices
The spatial patterns of the diatom phenological indices averaged over 15-yr of data are presented in Figures 5.3 to 5.5, together with the corresponding latitudinal variation in Figure 6. The
spatial patterns of the indices are generally associated to the location of the asymmetric contour of
the Southern Antarctic Circumpolar Current Front (SACCF) and of the maximum sea ice extent.
This association is strong for the start date, date of maximum (peak), growth duration and total
duration of the blooms. The connection between the indices and the fronts is clearer especially in
the western part of the SO.
The general pattern of the bloom start date is consistent with Thomalla et al. (2011). The
blooms start and the date of the maximum of Dia-Chla is reached earlier north of the SACCF outside the seasonal ice zone (Figure 5.3). The authors linked the start of the blooms south of
40°S to the light availability (indicated by PAR). The development of the blooms in the SO following the increase in the seasonal PAR was also described by Racault et al. (2012). On the other
hand, the end of the bloom of diatoms is more likely related to the exhaustion of nutrients
(Smetacek 1999; Smetacek et al. 2004; Smetacek 1985). Borrione and Schlitzer (2013) suggested
the exhaustion of silicate as limiting factor for the end of the spring bloom of diatoms in the South
Georgia region. Grazing pressure is thought to control the diatom species composition and biomass, rather than the end of the diatom blooms (Smetacek et al. 2004).
In the seasonal ice zone the start of the bloom is driven by light as well as water column stability; as the sea ice retreats, the melting of ice increases the stratification of the water column which
favors to maintain the phytoplankton in the euphotic zone. The end of the bloom occurs when the
mixed layer deeps caused by wind forcing, which dilutes the phytoplankton in the water column
(Taylor et al. 2013). Apart from melting of ice and wind forcing, changes in mixed layer depth in
the SO are also caused by heating from the atmosphere (Sallee et al. 2010).
Particularly notable is the early start of the blooms in the waters surrounding Antarctica (light
green), caused by the opening of areas free of ice around the continent. Arrigo et al. (2012)
80
showed that in the Amundsen polynya small areas free of ice occur throughout the year and that
their size increases with three factors: advection of sea ice offshore, increase in temperature and
melting of ice. These factors, combined with an increase in solar radiation and water column stability, as shown by Taylor et al. (2013), are linked to the earlier bloom start date in these waters
surrounding Antarctica as compared to waters northwards of the seasonal ice zone.
Figure 5.3. Spatial distribution of the mean diatom phenology in 1997 – 2012: (left) bloom start
date – BSD, (center) date of Dia-Chla maximum – CMD, (right) bloom end date - BED. Grey
areas represent missing data. Black solid lines show the mean position of the Polar Front (Sallee
et al. 2008) over 1997-2012. Dashed lines show the Southern Antarctic Circumpolar Front (Orsi et
al. 1995). Purple line displays the mean position of the maximum sea ice extent over 1997-2012
(Fetterer et al. 2002).
In general, the duration of the blooms is shorter south of the SACCF, in the seasonal ice zone,
and vice versa. Outside this region it forms a belt of higher values (longer duration) around the
Polar Front (PF), particularly between 30°W and 120°E (Figure 5.4). Compared to Racault et al.
(2012), the overall duration of the blooms is shorter. The average duration of the blooms for the
regions 50°S-60°S and 60°S-70°S was 8.3 and 6.5 weeks, while these authors reported 14 and 11
weeks, respectively. However, the authors used satellite Chla data to investigate the phytoplankton phenology, which includes all PFTs while in the present study we looked specifically at the
diatom Chla concentration.
As shown by Taylor et al. (2013), the duration of the bloom in the seasonal ice zone results
from a combination of factors influencing the growth and decline phases of the bloom, mainly
light and stability of the water column, while nutrients are less important. The belt of “longer lasting” blooms outside the seasonal ice zone is likely linked to a complex inter-play of different forcings: longer light periods and deeper mixed layers (Sallee et al. 2010) that enhance the supply of
nutrients at surface as well as reduce the grazing pressure by zooplankton (Behrenfeld et al.
2013). The mixed layer depth is deeper in the vicinity the fronts; around 100 m in the summer and
up to 400 m in the winter (Sallee et al. 2010).
81
The deepening of the mixed layer in the winter together with diapycnal diffusion replenishes
the surface with nutrients from subsurface waters, including iron (Tagliabue et al. 2014). It is
known that iron is a limiting nutrient in the surface waters of the SO controlling phytoplankton
growth, particularly in the open ocean. This micronutrient is rapidly depleted by spring blooms. In
late spring and summer, phytoplankton relies on the pelagic recycling until the following deepening of the mixed layer in autumn (Tagliabue et al. 2014). Open ocean diatoms have the ability to
reduce their requirement of iron (Armbrust 2009) which can help to sustain their blooms for longer periods.
In regions where the Antarctic Circumpolar Front (ACC) interacts with the topography, the nutrient supply is enhanced (Sokolov and Rintoul 2007) leading to higher Chla and consequently,
higher amplitude of the blooms (Figure 5.5). This occurs for example in the Pacific Antarctic
Ridge (see Figure 7 in Sokolov and Rintoul 2007). The enrichment from the coastal and shelf sediments close to islands (e.g. Kerguelen, Crozet and South Georgia Islands) are also important
sources of nutrients, especially iron (Blain et al. 2007; Borrione et al. 2014; Planquette et al.
2007). Other important factors controlling the duration of the bloom are the increasing in grazing
pressure and algal viruses (Behrenfeld et al. 2013; Smetacek et al. 2004).
Figure 5.4. Same as Figure 5.3, but for bloom growth duration (BGD), bloom decline duration
(BDD) and total duration (BD) of the blooms. Units are in week.
The relationship of the biomass indices with the fronts is not as evident as for the other indices
(Figure 5.5). The spatial distribution shows that more intense blooms (higher biomass) occur in
coastal regions, in the seasonal ice zone and in the Atlantic sector of the SO. Sokolov and Rintoul
(2007) have showed that at a broader scale the distribution of Chla is mainly controlled by the
upwelling of nutrients via Ekman transport while the upwelling associated with bathymetric
features is responsible for the magnitude and duration of the blooms. Blooms around Antarctica
can be considered as more efficient blooms, with short duration and high biomass.
82
Figure 5.5. Same as Figure 5.4, but for Dia-Chla maximum (CM), Dia-Chla amplitude (CA), DiaChla average (CAV) and Dia-Chla integrated over the growth duration (CI). Units are in mg m-3.
The latitudinal variability displays, from north to south, a progressive delay in the start, maximum and end date of blooms until about 73°S (Figure 5.6). The opposite is observed for the duration; there is a decrease in the growth, decline and total duration of the bloom from north to south.
South of 73°S the trend is reversed except by the growth duration which holds at about the same
duration. The biomass indices present similar latitudinal variations, but are rather small until the
first peak at ~ 67°S, followed by two steep peaks at ~ 72°S and at ~ 76°S, and then decreasing
towards the south.
Figure 5.6. Schematic representation of the latitudinal variability (longitudinal average) of the
phenological indices: bloom start date (BSD), date of Dia-Chla maximum (CMD), bloom end date
(BED), bloom growth duration (BGD), bloom decline duration (BDD), bloom duration (BD), DiaChla maximum (CM), Dia-Chla amplitude (CA), Dia-Chla averaged BGD (CAV), Dia-Chla
integrated over BGD (CI).
5.3.2
Interannual variability
The relative standard deviations of two of the phenological indices over the 15 years are displayed in Figure 5.7. To investigate whether the variability could be associated with a variation in
the mean position of the Polar Front and in maximum sea ice extent, their relative standard deviations are also shown (dashed lines). The spatial patterns of the relative standard deviation of the
bloom start date, date of Dia-Chla maximum and bloom end date are very similar to each other, as
well as among the duration or biomass indices. This is corroborated by the correlation maps
83
among the date, duration and biomass indices presented in Figure A4. Figure 5.7 shows the maps
with the highest variability of the relative standard deviation, namely bloom start date and DiaChla amplitude. The spatial variability of relative standard deviation of the duration indices does
not show a clear pattern and therefore is not presented.
Comparing the relative standard deviation maps of bloom start date and Dia-Chla amplitude,
the spatial distribution is different. The bloom start date presents higher variability in the vicinity
of the fronts, while the amplitude shows higher variability in the seasonal ice zone. This suggests
that the environmental forcings controlling the start date and the amplitude of the bloom are different. It is important to note that the variability of the indices is linked to the main position of the
PF and the maximum sea ice extent, rather than with their variability which is low (dashed lines).
The variability of the amplitude of the bloom seems to be linked to the interannual variability
in the sea ice concentration, which in turn controls the light availability and the stability of the
water column. On the other hand, the variability in the bloom start date is probably related with
the variability in the nutrient supply which is large where the seasonal amplitude of the mixed
layer depth is strong and small in the ice-influenced region.
Figure 5.7. Spatial distribution of the relative standard deviation (RSD) of the bloom start date
(BSD) and Dia-Chla amplitude (CA) for 15 years of data (1997-2012). Grey areas represent missing data. Black continuous lines represent the mean Polar Front position (Sallee et al. 2008) and
dashed black lines the standard deviation of the position over 1997-2012. Purple continuous lines
indicate mean of the maximum sea ice extent (Fetterer et al. 2002) and dashed purple lines the
standard deviation of the position for the years of 1997 to 2012. White line indicates the mean position of the Southern Antarctic Circumpolar Front (Orsi et al. 1995).
5.3.2.1 Trends
Most of the indices are not gap free and for this reason trends could only be determined for the
84
date of Dia-Chla maximum, Dia-Chla maximum and amplitude. For the same reason the anomaly
data were not detrended afterwards. Trends in the Dia-Chla amplitude are very similar to Dia-Chla
maximum and not shown.
Coherent patches of significant positive and negative trends were detected for the date of DiaChla maximum and Dia-Chla maximum (Figure 5.8). For example, in the region between the
Malvinas and South Georgia Islands (Figure 5.8, green star) there is a trend towards an earlier
maximum of the bloom leading to higher biomass (Dia-Chla maximum). However, the
relationship between the date of the maximum of the bloom and the biomass can be reverse as
well, as in the region south of 60°S and between 120°E and 150°E (Figure 5.8, black star) where a
later start of the bloom leads to an increase in biomass. Although we could not estimate trends in
the bloom start date and bloom end date, we can expect a similar pattern to the ones detected for
date of Dia-Chla maximum since these indices are highly correlated (Figure A4).
These observations combined with recent studies on the trends in sea surface temperature
(Maheshwari et al. 2013) and sea ice cover (Maksym et al. 2012) over the last three decades,
suggest a link between these two variables and the diatom phenology. For example, in the region
south of 60°S and from 60°E to 120°E (Figure 5.8, grey star) the earlier date of Dia-Chla
maximum and the increased Dia-Chla maximum coincide with the observed increase trend in SST
and decrease in sea ice cover (earlier sea ice melt).
Compared to literature, the spatial distribution of trends in Dia-Chla maximum are similar to
trends in total Chla from SeaWiFS reported by Henson et al. (2010) and Siegel et al. (2014) for
the 1997-2007 and 1997-2010 periods, respectively. The last study on global total Chla trends
(1998-2012), by Gregg and Rousseaux (2014), displays large areas in the SO with positive trends
as we observed here, but also positive trends between the 60°E and 150°E north of 60°S whereas
we observed negative trends. Different trends can be expected due to the different variables used
(total Chla instead of Dia-Chla), as well as data treatment (bias correction, data assimilation)
performed by the authors.
Figure 5.8. Trends of the standardized anomalies of date of Dia-Chla maximum (CMD) and DiaChla maximum (CM). Reddish colour indicates a positive trend and bluish indicates a negative
trend. Only statistically significant trends (p < 0.05) are shown. The stars highlight the regions
85
between Malvinas and South Georgia Islands (green) and south of 60°S between 120°E to 150°E
(black) and 60°E to 120°E (grey).
5.3.2.2 Relationships with ENSO and SAM
5.3.2.2.1 Correlation maps
To further explore the interannual variability of the diatom phenology, we examined the
relationship of phenological indices with ENSO and SAM (methods described in section 5.2.5 and
5.2.6). During the 1997-2012 period there were six El Niño years (1997/1998, 2002/2003,
2003/2004, 2004/2005, 2006/2007, 2009/2010), eight La Niña years (1998/1999, 1999/2000,
2000/2001, 2005/2006, 2007/2008, 2008/2009, 2010/2011, 2011/2012), seven years of positive
phase of SAM (1998/1999, 1999/2000, 2001/2002, 2007/2008, 2008/2009, 2010/2011,
2011/2012) and four of negative phase (2000/2001, 2002/2003, 2003/2004, 2009/2010). The
correlation maps are presented in Figure 5.9 for date of Dia-Chla maximum and Dia-Chla
maximum as representative of the date indices and biomass indices, respectively. Significant
positive (negative) correlations indicate that the anomalies are in phase with ENSO and SAM.
Coherent areas in the duration indices are less evident for the duration indices and not shown.
Several areas show significant correlation between ENSO/SAM and the diatom phenology.
The correlation coefficients for ENSO are opposite to that of SAM. For example, the date of DiaChla maximum in the sector of the seasonal ice zone between 120°W-150°W is negatively
correlated with ENSO (MEI index) and positively correlated with SAM (AAO index). Moreover,
the patterns in El Niño (La Niña) years and negative (positive) SAM are similar. These results are
in line with observations of the sea ice concentration, SST, Chla and wind speed and direction in
the SO (Lovenduski 2007; Pohl et al. 2010). Smith et al. (2008) also observed that high biomass
offshore the Western Antarctic Peninsula region was associated La Niña and/or positive SAM
events. Hence, we can expect the spatial patterns of the anomalies of phenological indices during
El Niño (La Niña) years and negative (positive) phase of SAM to resemble each other.
The most remarkable feature in the correlation maps of the date of Dia-Chla maximum can be
seen in the Pacific Sector (90°W to 150°W), north of the PF and south of the maximum sea ice
extent. This pattern is consistent with the results of Kwon and Comiso (2002) and Lefebvre et al.
(2004) for earlier periods, 1982-1998 and 1980-1999 and using satellite and model data,
respectively. For the same region, Kwon and Comiso (2002) observed an increase in SST and a
decline in sea ice concentration associated with El Niño. Lefebvre et al. (2004) showed that the
winter sea ice concentration decreases in negative SAM events. As a result, an earlier start,
maximum and end of the bloom can be expected in El Niño or negative SAM events.
The Dia-Chla maximum displays less significant correlations, but the general pattern of DiaChla maximum is consistent with the results of Lovenduski and Gruber (2005) for the relationship
between satellite Chla and SAM. In contrast, the general increase in diatom concentration
between 50°S and 70°S during positive SAM event showed by Hauck et al. (2013) was not
observed in our results. The authors used a coupled ecosystem – general circulation model and
86
lagged correlations (4 months) to investigate the relationship and these might be the cause for
disagreement between the results, as well as the different periods analysed in their study (19482010) and in the present study (1997-2012). In addition, the observed lower Dia-Chla maximum
at 60°E during El Niño (negative correlation) can be linked to lower SST in El Niño years, as
shown by Kwon and Comiso (2002) (see Figure 6 in Kwon and Comiso, 2002).
Because SAM and ENSO are not linearly independent at interannual time scales during the
austral summer season (L’Heureux and Thompson 2005; Pohl et al. 2010), we expected that some
of the variability we observed related to SAM may be biased by ENSO, or vice versa. This was in
part confirmed by the partial correlations (Figure A5), but the differences between the correlations
and partial correlations are in general small. Higher differences were observed between the date of
Dia-Chla maximum and MEI. The correlations with SAM and between the Dia-Chla maximum
and MEI/SAM did not change. One possible reason for not observing differences is that the short
time series used here might not allow distinguishing the respective influence of the oscillations.
Figure 5.9. Correlation coefficients of the standardized anomalies of date of Dia-Chla maximum
(CMD) and Dia-Chla maximum (CM) vs. ENSO (MEI) and SAM (AAO) indices. Only
statistically significant trends (p < 0.05) are shown. Black and purple lines indicate the mean
position of the Polar Front and maximum sea ice extent over 1997-2012, respectively.
87
5.3.2.2.2 Composite maps of anomalies
The composite maps of the anomalies of bloom start date, Dia-Chla maximum and bloom duration are shown in Figures 5.10, 5.11 and A6, respectively, and provide insight onto the magnitude of the anomalies during the ENSO and SAM events. In the seasonal ice zone there are two
regions with inverse patterns and high anomalies of bloom start date: the Weddell Sea region
(white dashed box) and the sector between 120°W and 180°W (white box), north of 70°S. In the
Weddell Sea, El Niño/negative SAM years are characterized by later start, shorter duration and
slightly higher biomass, which are likely a response of more extensive ice cover in these years
(Kwok and Comiso 2002; Lefebvre et al. 2004). In the sector between 120°W and 180°W the pattern is inversed.
Alvain et al. (2013) showed that, on average, the frequency of diatom dominance (derived using the PHYSAT) is higher in positive phases of SAM as a response of more intense winds and
nutrient supply. Their spatial pattern of differences between monthly mean diatom frequency
dominance during positive and negative SAM events reveal large areas with negative differences
as well (see Figure 3a in Alvain et al. 2013). To compare our results with those from Alvain et al.
(2013), we calculated the differences between the composite maps of Dia-Chla maximum anomalies during positive SAM and negative SAM phases (not shown). Our spatial pattern is very similar to that presented in their study for most of the region.
Figure 5.10. Composites of bloom start date (BSD) standardized anomalies during El Niño
88
(N=6), La Niña (N=8), positive SAM (N=7) and negative SAM (N=4) years. Grey areas represent
missing data. Black lines show the mean position of the Polar Front (Sallee et al. 2008) over
1997-2012. Purple line displays the mean position of the maximum sea ice extent (Fetterer et al.
2002) over 1997-2012. The white boxes depict the Weddell Sea region (dashed) and the sector
between 120°W and 180°W.
Figure 5.11. Same as Figure 5.10 but for Dia-Chla maximum.
5.3.2.2.2.1 Composite maps of anomalies in amplified years
In addition to these anomalous years influenced by ENSO and SAM, the effect of these oscillations can be amplified when El Niño (La Niña) coincides with negative (positive) phase of SAM.
During the period studied, three years (2002/2003, 2003/2004, and 2009/2010) were characterized
by El Niño co-occurring with negative phase of SAM, and six years (1998/1999, 1999/2000,
2007/2008, 2008/2009, 2010/2011, 2011/2012) of La Niña and positive phase of SAM (Figure
5.1). The composite maps of bloom start date and Dia-Chla maximum for the amplified years are
presented in the appendix (Figure A7). Unfortunately, because of the short length of our time series it was not possible to distinguish between amplified and non-amplified years.
89
5.4 Concluding remarks
We have used a solid and widely applied method to investigate phytoplankton phenology from
ocean colour data (Brody et al. 2013; Henson et al. 2009; Henson and Thomas 2007; Racault et al.
2012; Siegel et al. 2002; Thomalla et al. 2011). This method differs from Behrenfeld (2010),
where the bloom starts when the net phytoplankton growth rate becomes positive. Another difference is that Behrenfeld (2010) accounts for the dilution effect; when the mixed layer is deeper
than the Zeu the phytoplankton concentration is diluted and that can mask an increase in the net
growth rate (Behrenfeld 2010). Although Zeu can be accurately retrieved from satellite data in the
SO (Soppa et al. 2013), the uncertainties of the mixed layer depth datasets at weekly or monthly
resolution has still to be investigated. In addition, in our analysis the start of the bloom was defined as an increase of Chla over a threshold, after identifying the peak of biomass. An advantage
of this method is that it is more suitable for studying the mismatch hypothesis between phytoplankton and higher trophic levels (Brody et al. 2013) because it is based on the timing of the
maximum biomass. On the other hand, by this definition there will always be a bloom, even in
regions where the amplitude of the Chla is very low, as between 120°E and 150° E (Figure 5.5).
The advantage of using satellite data to study phytoplankton phenology is widely recognized.
The high temporal resolution allows investigating the full development of the bloom and the
spatial coverage enables to compare different regions simultaneously in time. However, there are
still knowledge gaps related to the ocean colour data that can affect the phenology studies and
where further investigations are needed. One limitation is that some regions of the SO present a
deep Chla maximum (~ 60 - 90 m) which is not seen by the sensor (e.g. southern Indian and
Pacific sectors of the SO, Holm-Hansen et al. 2005). This implies that subsurface blooms deeper
than the first optical depth are not accounted for in the satellite data. A second limitation is that
due to gaps in satellite data the date of the actual Chla maximum, for example, may be missed
(Kahru et al. 2011). In fact, Racault et al. (2014) showed that the error in the date of Chla
maximum arising from gappy datasets is nearly two months (global average). Data gaps also
reduce the length of the times series and the significance of the statistical tests. We have attempted
to minimize this error by filling the gaps by linear temporal interpolation and through the use of
merged satellite Chla product. For example, in January 2012 the average number of observations
in the SO improved by 45% using the merged satellite Chla data than if only using MODIS data.
Another way to overcome this issue is to apply techniques such as the Data Interpolating
Empirical Functions method (DINEOF, Alvera-Azcarate et al. 2007) to reconstruct the missing
data fields. Thus, a future goal is to reprocess the dataset of diatom concentration with the
DINEOF to obtain a gap-free dataset and to investigate the impact of missing data to determine
the phenological indices. This will make possible to estimate trends for the indices that remained
not analyzed because of the lack of gap-free time series and it will possibly increase the
significance of the correlations discussed in the section 5.3.2.2.
Phytoplankton blooms phenology has been studied before using satellite Chla data. However,
by looking at the total Chla provided by the satellite the contribution of all different PFTs are
taken into account. Using satellite-derived diatom concentration, we were able to look specifically
at the fraction or concentration of Chla that belongs to diatoms. We investigated the mean patterns
90
of the diatom phenology, but the most interesting results are related to the interannual variability.
We find a clear correspondence between ENSO and SAM and the phenology of diatoms, as
revealed by the correlation and the anomaly composite maps. These climate oscillations have
different effects among the regions of the SO. It is also evident that different phases in
ENSO/SAM have opposite effects in the diatom phenology. These results emphasize the influence
of climate oscillations over the SO and in the diatom phenology, which may be enhanced in
amplified years. For the first time these findings are described for the SO.
The mechanisms by which ENSO and SAM climate oscillations are associated with diatom
phenology will be investigated in a following study. From the literature we obtained the evidence
of the complex and strong link between these components through changes in the environment
(e.g. sea ice concentration, sea surface temperature, mixing); our next step is to quantify it. This
can be performed for example, by examining the composite maps of sea ice concentration during
the different phases of ENSO/SAM and also looking at correlations between the sea ice concentration and the phenological indices. Besides that, we did not found a literature which fully covers
the period we investigated.
In addition, such investigation needs a comprehensive dataset including not only information
on SST and PAR, as usual used in phytoplankton phenology studies since these variables are
freely available from remote sensing, but also water column mixing, sea ice concentration, dissolved iron and silicate for example. Combining remote sensing and model data can help to explain the missing link between climate oscillations, environmental anomalies and diatom phenology. A better understanding of diatom phenology also requires a consideration of the grazing
pressure, which unfortunately cannot be easily estimated, at least not at the temporal resolution
required, and probably for this reason its role on the diatom phenology is not well known.
Last, the knowledge of other PFT forming blooms in the SO, mainly haptophytes, is important
to understand phytoplankton community shift and the factors controlling it. E. huxleyi are known
to occur along the “Great Calcite Belt” in the SO and dense blooms are often observed at the shelf
break and off the Patagonian shelf after the spring bloom of diatoms. Moreover, while is generally
accepted that diatoms dominate the spring bloom in the SO (Smetacek 1985) this might not be
true everywhere as for the case of the spring bloom of P. antarctica in the Ross Sea (Smith et al.
2012).
91
Chapter 6
Synthesis and major outcomes
92
6
Synthesis and major outcomes
Changes in the SO are occurring, regardless if driven by natural or anthropogenic climate
variability. Among the most remarkable changes is the increase in the sea ice cover by 5% per
decade in the Ross Sea and the decrease by 7.1% per decade in the Bellingshausen/Amundsen Sea
(Comiso 2010). An important question which needs to be addressed in a changing SO is how
these changes are reflected in the phytoplankton composition, distribution and NPP and hence in
the biogeochemical cycles and ecosystem functioning.
The SO is, like the North Atlantic, a major region for uptake and long term storage of
anthropogenic atmospheric CO2 (Hauck and Völker 2015; Sabine et al. 2004; Takahashi et al.
2012). Part of the uptake is biology-driven. Maier-Reimer et al. (1996) addressed the importance
of the marine biological pump in the global atmospheric CO2 concentrations. Using a coupled
ocean-biogeochemical model, they found that without the biological uptake the CO 2 concentration
in the atmosphere would rise ~100 ppm in 50 years, which translates to an increase of 213 PgC.
For comparison, the preindustrial concentration was 280 ppm in the year 1800 and the current
concentration (March 2015) is 400.83 ppm (Dlugokencky and Tans 2015); that is an increase of
120 ppm (255 PgC) in 215 years. The current level is also considerably higher than the
concentration of 172-300 ppm for the last 800.000 years measured in ice cores from Antarctica
(Luthi et al. 2008).
The projected consequences of increasing anthropogenic atmospheric CO 2 in the SO include
strengthening of the westerly winds, ocean acidification, warming and freshening of surface
waters, enhanced stratification and decrease in sea ice extent (Bindoff et al. 2011; IPCC 2014). It
is argued that the enhanced stratification will reduce the vertical supply of nutrients to the surface
and lead to a community shift from large to smaller phytoplankton (Bopp et al. 2005; Finkel et al.
2010). This view is not supported by Kemp and Villareal (2013). The authors argue that diatoms,
which are in general medium and large phytoplankton, survive in a wide variety of conditions and
may adapt to a more stratified ocean due to their ability to grow in deeper layers under low light
conditions, to migrate in the water column by physiologically controlling their buoyancy, thus
enhancing the nutrient uptake. Further, open ocean diatoms can reduce their requirement of iron
under iron limiting conditions (Armbrust 2009).
Alongside anthropogenic-induced change, much of the variability in the SO has been associated to natural climate oscillations ENSO and SAM (Alvain et al. 2013; Arrigo and van Dijken
2004; Hauck et al. 2013; Kwok and Comiso 2002; Lovenduski 2007; Lovenduski and Gruber
2005). SAM has shown a positive trend towards positive events (Pohl et al. 2010; Sallee et al.
2010; Thompson et al. 2011) and the coupling with ENSO may strengthen the anomalies generated by SAM in the SO (Fogt et al. 2011).
Satellite observations provide the means for monitoring large scale changes in the SO
productivity; however, the separation of natural from anthropogenic influence is difficult using
ocean colour remote sensing. Short times series are currently available relative to the large
interannual variability intrinsic to phytoplankton (Beaulieu et al. 2013; Henson et al. 2010).
Henson et al. (2010) have shown that longer times series (~ 40 years) are required to detect
climate change trends on the satellite ocean productivity in the SO; a challenging task that
93
depends on the endurance of the satellite missions. The merging of data from different sensors as
performed by the ESA GlobColour and ESA OC CCI projects is a current effort to achieve this
aim which depends on simultaneous data acquisition, cross-calibration between sensors and data
merging (Beaulieu et al. 2013). Unfortunately, to date, only MODIS-Aqua, launched mid of 2002,
and the Visible Infrared Imager Radiometer Suite (VIIRS) launched end of 2011, are operational
at global scale, and there is a growing concern on MODIS-Aqua which is experiencing sensor
degradation since 2011.
Another important application of ocean colour is data assimilation and validation of marine
biogeochemical models (Robinson 2010). Furthermore, the use of satellite data in conjunction
with information derived from models can provide a more complete description of the
biogeochemical processes in the water column and on the export of carbon to deeper layers. In
situ data is also certainly very useful, but those are rather sparse and unable to solve the spatial
and temporal variability required to study these processes. A promising technology is the BioArgo float, which is another way of obtaining continuous ocean optical information. Argo floats
have now for 15 years measured profiles of temperature and salinity at daily temporal resolution
in the SO and recently, they have been equipped with several sensors to measure bio-optical
properties of the oceans (e.g. Chla, PAR, CDOM). These Bio-Argo floats are still experimental
but will potentially improve the number of bio-optical observations in the SO and together
provide information on the vertical structure of these properties (http://www.euroargoedu.org/explore/argoeu_2.php).
The synergistic use of ocean colour remote sensing, biogeochemical models and in situ observations provide complementary information and they have to walk together towards their improvement. From the ocean colour perspective, it implies that we do not only need sustained and
consistent ocean colour observations, but also to understand the typical patterns of phytoplankton
community composition and NPP in the SO. To assess the effects of unusual events on ocean biology we need first to understand its natural variability; a baseline against which to compare these
events (Henson 2014).
This thesis was set out to investigate ocean colour retrievals and phytoplankton dynamics in
the SO and it was developed as a multidisciplinary work using in situ and remote sensing data.
The studies developed here have moved forward our knowledge of ocean colour in the SO and
contributed to a better understanding of the ocean biogeochemical cycle from the ocean colour
perspective by adding new information on the uncertainties in the input terms of NPP models, on
the estimation of diatoms abundance and on the variability of diatoms phenology.
For the first time uncertainties of ocean colour retrievals of Zeu have been investigated in the
SO (Chapter 3; published in Soppa et al. 2013). Two Zeu retrievals (Zeu-Chla and Zeu-IOP) from
SeaWiFS and MODIS have been validated with in situ measurements of Zeu. The results showed
that both methods and sensors provide consistent estimates of satellite Zeu, although satellite
retrievals of Chla have substantial uncertainties. Spatial differences between Zeu satellite products
have been found and it is likely that these differences were not detected in the validation effort
due to the limited and uneven distribution of the in situ measurements. Therefore, we reinforce the
importance of looking at spatial patterns, together with in situ validation for comparing ocean
colour data retrieved from different approaches. Thus, the choice of the Zeu product led to
substantial differences in the satellite NPP over specific regions of the SO. In addition, a parallel
94
objective of this study was to examine the uncertainties of satellite aph as it is an input term in the
ABPM NPP model as well. As for Zeu, the validation of satellite aph was the first efforts to
investigate uncertainties in the SO and the results presented here can serve as a reference for
future studies.
To which extent one Zeu product leads to improved NPP retrievals in comparison to the other
one is a task for future research and could potentially suggest the more appropriate Zeu satellite
product for a given NPP model. A similar investigation could be conducted for aph and, in
addition, by examining the combined effect of the uncertainties of different input terms in the NPP
estimation, as performed by Saba et al. (2011). However, Zeu, aph and the ABPM were not
considered in their study; neither the use of different algorithms to retrieve the input terms of the
NPP models.
Moreover, the skill of satellite NPP models can be improved by separating the contribution of
different PFTs. Biogeochemical models (e.g. NOBM - NASA Ocean Biogeochemical Model,
Gregg and Casey 2007, PISCES - Pelagic Interaction Scheme for Carbon and Ecosystem Studies,
Aumont and Bopp 2006) frequently represent the community structure by including the information of one or more PFTs allowing to investigate their specific contribution to the NPP and biogeochemical cycles. Few ocean colour studies (Uitz et al. 2009a, 2010; Uitz et al. 2009b; Uitz et
al. 2012) have so far focused on this topic due to the difficulty in estimating the Chla and the photophysiological properties specific to each phytoplankton group (Uitz et al. 2010), but their importance is acknowledged. There is an increasing effort in the ocean colour community to develop
and improve PFTs retrievals using remote sensing. Equally important is the tuning of global satellite PFT approaches to regional application. Today global scale studies and parametrizations usually fail to fully capture important regional differences in the ocean properties which influence the
phytoplankton community composition (Brito et al. 2015).
Regarding this limitation, the second study, presented in Chapter 4 and published in Soppa et
al. (2014), has identified that the satellite derived diatom abundance using the ABA is underestimated in the SO. This conclusion has been achieved by revising the global ABA with a new in
situ dataset that included more phytoplankton pigment samples from the SO. It was observed that
the global relationship between Chla and the f-Diatom is not appropriate for the SO. A new global
function that accounts for the relatively high concentration of diatoms in low Chla waters was
proposed, improving the retrievals in the SO using the global parametrization, but leading to an
overestimation in other regions. Therefore, and for the first time, a regional ABA model for the
SO was developed which further improved estimation of the diatom abundance in the region.
Yet, the remoteness of the SO is still a limitation. Research cruises in the SO tend to focus on
regions with elevated phytoplankton concentration and close to the continents for logistic reasons
and scientific interests (e.g. sea ice, carbon export). Again, the development of new field
observation systems as Bio-Argo floats and gliders could help to obtain a better spatial and
temporal coverage in the SO. It is a long term aim to obtain a proper in situ bio-optical data set of
the SO that once achieved will allow to investigate the spatial differences observed in the first
study and will improve global and regional parametrizations of the relationship between Chla and
diatoms developed in the second study.
With the information of PFTs from satellite, the next stage is to investigate their dynamics. The
third study analyzed the diatom bloom phenology over 15-yr (Chapter 5) using the regional model
95
to retrieve the concentration of diatoms in the SO. Although important, phenology studies focused
on diatoms have not been conducted so far. The diatom bloom phenology was investigated using
ten indices that described the timing, duration and magnitude of the bloom at different stages. The
mean patterns revealed that diatom blooms are spatially and temporally heterogonous and are
associated with the position of the SACCF and the maximum sea ice extent. Earlier start and
maximum were observed in blooms north of the SACCF - outside the seasonal ice zone. Their
duration is also longer than blooms formed in the seasonal ice zone. On the other hand, blooms
around Antarctica are more intense blooms, with shorter duration and higher biomass, than
blooms outside this region.
In addition to the mean patterns, knowledge was gained in the relationship between large scale
climate oscillations (ENSO and SAM) and diatom bloom phenology. The robust signals of ENSO
and SAM observed in the phenological indices across different regions of the SO indicated
influence of these climate oscillations on the environment and diatom phenology. A follow-up
study should address in more detail the underlying mechanisms associated with changes in the
diatom phenology, for example, by looking at correlations and composite maps of sea ice
concentration or sea surface temperature and diatom phenology during ENSO and SAM events.
Once available, a longer time series of data should be employed likewise to examine trends and to
better distinguish between ENSO and SAM influence in the region.
If changes in the phytoplankton bloom phenology can be monitored, it may serve as indicator
for subsequent changes in the food web, productivity and ocean biogeochemistry. Future research
should focus on optimizing the ABA to other PFTs to expand the understanding on their dynamics
and interaction with other PFTs. However, since the ABA method is based on the assumption that
diatoms dominate at high Chla, it might incorrectly identify other PFTs occurring at high
concentrations. A combination with data provided by the PhytoDOAS method applied to
hyperspectral data, for example, can lead to a more objective characterization of the PFTs.
Recently
the
SynSenPFT
Project
founded
by
ESA
(http://www.awi.de/en/research/young_investigators/helmholtz_university_young_investigators_g
roups/phytooptics/projects/synsenpft/), under the Scientific Exploration of Operational Missions
Program, has been initiated. One of the main objectives is to develop improved PFT products by
the synergistic use of low spatial resolution hyperspectral data with high spatial resolution
multispectral data. Satellite retrievals of diatom, coccolithophores and cyanobacteria will be
developed using the ABA and PhytoDOAS methods. By the choice of these two methods, the high
spatial resolution of the PFTs derived with ABA and the high spectral resolution data used by
PhytoDOAS to derive PFTs can be exploited.
Ocean colour remote sensing has proven to be a valuable tool to adequately examine primary
production and the phytoplankton composition in a changing SO. In the future, long term high
quality time series will be available for further investigations in the topic of this thesis given the
continuity of the satellite space programs. The upcoming multispectral and hyperspectral sensors,
the Ocean Land and Color Instrument (OLCI) on Sentinel-3 and the Tropospheric Monitoring
Instrument (TROPOMI) on Sentinel-5P, are expected to be launched in October 2015 and 2016
respectively, as part of the ESA Sentinels mission. They will support the challenging task of
building a long and accurate global data record from ocean colour remote sensing.
96
Appendix
Figure A1. Scatterplots of satellite Zeu-Chla and Zeu-IOP against in situ Zeu south of 60°S. The
solid line represents the regression and the dotted line represents the 1:1 line as reference.
97
Figure A2. Scatterplot of the validation for the global DPZpd dataset (N= 1182): (a) new model
(ABAZpd), (b) model of Hirata et al. (2011) parameterized with the DPZpd dataset (ABA*) and (c)
original model and fitting parameters of Hirata et al. (2011) (ABA**). The samples located in the
SO are presented in grey (N = 460), together with the statistics of the validation. The red line
represents the 1:1 line. The statistics were calculated with log 10 transformed data (e.g.
log10(y+0.00003)).
98
Figure A3. Climatology of TChlaZpd of diatoms (mg m-3) using the regional algorithm for the SO
based on 2003-2013 period. The austral winter months of May, June, July and August are not
presented due to too few number observations available in these months. White areas correspond
to waters with depths shallower than 200 m or without satellite information.
99
Figure A4. Spearman correlation coefficients between the time series of phenological indices (15yr, 1997 – 2012): bloom start date (BSD), date of Dia-Chla maximum (CMD), bloom end date
(BED), bloom growth duration (BGD), bloom decline duration (BDD), bloom duration (BD), DiaChla maximum (CM), Dia-Chla amplitude (CA), Dia-Chla averaged BGD (CAV), Dia-Chla
integrated over BGD (CI). Only statistically significant trends (p < 0.05) are shown. White areas
correspond to non-significant correlations or missing data. Black and purple lines indicate the
mean position of the Polar Front and the mean position of the maximum sea ice extent over 19972012, respectively.
100
Figure A5. Partial correlation coefficients of the standardized anomalies of date of Dia-Chla
maximum (CMD) and Dia-Chla maximum (CM) vs. ENSO (MEI) and SAM (AAO) indices. Only
statistically significant trends (p < 0.05) are shown. Black and purple lines indicate the mean
position of the Polar Front and the mean position of the maximum sea ice extent over 1997-2012,
respectively.
101
Figure A6. Composites of bloom duration standardized anomalies during El Niño (N=6), La Niña
(N=8), positive SAM (N=7) and negative SAM (N=4) years. Grey areas represent missing data.
Black lines show the mean position of the Polar Front (Sallee et al. 2008) over 1997-2012. Purple
line displays the mean position of the maximum sea ice extent (Fetterer et al. 2002) over 19972012. The white boxes depict the Weddell Sea region (dashed) and the sector between 120°W and
180°W.
102
Figure A7. Composites of bloom start date (BSD) and Dia-Chla maximum (CM) standardized
anomalies during amplified years. Left plot: El Niño and negative SAM (N=3). Right plot: La
Niña and positive SAM (N=6). Grey areas represent missing data. Black and purple lines indicate
the mean position of the Polar Front and the mean position of the maximum sea ice extent over
1997-2012, respectively. The white boxes depict the Weddell Sea region (dashed) and the sector
between 120°W and 180°W.
103
References
Alvain, S., Le Quere, C., Bopp, L., Racault, M.F., Beaugrand, G., Dessailly, D., & Buitenhuis,
E.T. (2013). Rapid climatic driven shifts of diatoms at high latitudes. Remote Sensing of
Environment, 132, 195-201
Alvain, S., Moulin, C., Dandonneau, Y., & Breon, F.M. (2005). Remote sensing of
phytoplankton groups in case 1 waters from global SeaWiFS imagery. Deep-Sea Research
Part I-Oceanographic Research Papers, 52, 1989-2004
Alvain, S., Moulin, C., Dandonneau, Y., & Loisel, H. (2008). Seasonal distribution and
succession of dominant phytoplankton groups in the global ocean: A satellite view. Global
Biogeochemical Cycles, 22
Alvera-Azcarate, A., Barth, A., Beckers, J.M., & Weisberg, R.H. (2007). Multivariate
reconstruction of missing data in sea surface temperature, chlorophyll, and wind satellite
fields (vol 112, art no C03008, 2007). Journal of Geophysical Research-Oceans, 112
Amante, C., & Eakins, B.W. (2009). ETOPO1 1 Arc-Minute Global Relief Model:
Procedures, Data Sources and Analysis. NOAA Technical Memorandum NESDIS NGDC-24:
National Geophysical Data Center, NOAA
Arar, E.J., & Collins, G.B. (1997). In Vitro Determination of Chlorophyll a and Pheophytin a
in Marine and Freshwater Algae by Fluorescence. Cincinnati, Ohio: National Exposure
Research Laboratory, Office of Research and Development, U. S. Environmental Protection
Agency
Armbrust, E.V. (2009). The life of diatoms in the world's oceans. Nature, 459, 185-192
Arrigo, K.R., Lowry, K.E., & van Dijken, G.L. (2012). Annual changes in sea ice and
phytoplankton in polynyas of the Amundsen Sea, Antarctica. Deep Sea Research Part II:
Topical Studies in Oceanography, 71–76, 5-15
Arrigo, K.R., van Dijken, G., & Long, M. (2008). Coastal Southern Ocean: A strong
anthropogenic CO2 sink. Geophysical Research Letters, 35, L21602
Arrigo, K.R., & van Dijken, G.L. (2004). Annual changes in sea-ice, chlorophyll a, and
primary production in the Ross Sea, Antarctica. Deep-Sea Research Part Ii-Topical Studies in
Oceanography, 51, 117-138
Assmy, P., Smetacek, V., Montresor, M., Klaas, C., Henjes, J., Strass, V.H., Arrieta, J.M.,
Bathmann, U., Berg, G.M., Breitbarth, E., Cisewski, B., Friedrichs, L., Fuchs, N., Herndl,
G.J., Jansen, S., Kragefsky, S., Latasa, M., Peeken, I., Rottgers, R., Scharek, R., Schuller,
S.E., Steigenberger, S., Webb, A., & Wolf-Gladrow, D. (2013). Thick-shelled, grazerprotected diatoms decouple ocean carbon and silicon cycles in the iron-limited Antarctic
Circumpolar Current. Proceedings of the National Academy of Sciences of the United States
of America, 110, 20633-20638
Audran, S., Faure, B., Mortini, B., Aumont, C., Tiron, R., Zinck, C., Sanchez, Y., Fellous, C.,
Regolini, J., Reynard, J., Schlatter, G., & Hadziioanno, G. (2006). Study of dynamical
104
formation and shape of microlenses formed by the reflow method. Advances in Resist
Technology and Processing Xxiii, Pts 1 and 2, 6153, U1616-U1625
Beaulieu, C., Henson, S.A., Sarmiento, J.L., Dunne, J.P., Doney, S.C., Rykaczewski, R.R., &
Bopp, L. (2013). Factors challenging our ability to detect long-term trends in ocean
chlorophyll. Biogeosciences, 10, 2711-2724
Behrenfeld, M.J. (2010). Abandoning Sverdrup's Critical Depth Hypothesis on phytoplankton
blooms. Ecology, 91, 977-989
Behrenfeld, M.J. (2014). Climate-mediated dance of the plankton. Nature Climate Change, 4,
880-887
Behrenfeld, M.J., & Boss, E.S. (2014). Resurrecting the Ecological Underpinnings of Ocean
Plankton Blooms. Annual Review of Marine Science, Vol 6, 6, 167-U208
Behrenfeld, M.J., Doney, S.C., Lima, I., Boss, E.S., & Siegel, D.A. (2013). Annual cycles of
ecological disturbance and recovery underlying the subarctic Atlantic spring plankton bloom.
Global Biogeochemical Cycles, 27, 526-540
Behrenfeld, M.J., & Falkowski, P.G. (1997). Photosynthetic rates derived from satellite-based
chlorophyll concentration. Limnology and Oceanography, 42, 1-20
Bélanger, S., Ehn, J.K., & Babin, M. (2007). Impact of sea ice on the retrieval of waterleaving reflectance, chlorophyll a concentration and inherent optical properties from satellite
ocean color data. Remote Sensing of Environment, 111, 51-68
Ben Mustapha, Z., Alvain, S., Jamet, C., Loisel, H., & Dessailly, D. (2014). Automatic
classification of water-leaving radiance anomalies from global SeaWiFS imagery: Application
to the detection of phytoplankton groups in open ocean waters. Remote Sensing of
Environment, 146, 97-112
Bidigare, R.R., Ondrusek, M.E., Morrow, J.H., & Kiefer, D.A. (1990). Invivo Absorption
Properties of Algal Pigments. Ocean Optics X, 1302, 290-302
Bindoff, N.L., Rintoul, A.J., Haward, M., Allison, I., & Press, A.J. (2011). Climate change and
the Southern Ocean. ACE CRC Oceans Position Analysis: (p. 27). Hobart, Tasmania: The
Antarctic Climate & Ecosystems Cooperative Research Centre
Blain, S., Queguiner, B., Armand, L., Belviso, S., Bombled, B., Bopp, L., Bowie, A., Brunet,
C., Brussaard, C., Carlotti, F., Christaki, U., Corbiere, A., Durand, I., Ebersbach, F., Fuda,
J.L., Garcia, N., Gerringa, L., Griffiths, B., Guigue, C., Guillerm, C., Jacquet, S., Jeandel, C.,
Laan, P., Lefevre, D., Lo Monaco, C., Malits, A., Mosseri, J., Obernosterer, I., Park, Y.H.,
Picheral, M., Pondaven, P., Remenyi, T., Sandroni, V., Sarthou, G., Savoye, N., Scouarnec, L.,
Souhaut, M., Thuiller, D., Timmermans, K., Trull, T., Uitz, J., van Beek, P., Veldhuis, M.,
Vincent, D., Viollier, E., Vong, L., & Wagener, T. (2007). Effect of natural iron fertilization on
carbon sequestration in the Southern Ocean. Nature, 446, 1070-U1071
Bopp, L., Aumont, O., Cadule, P., Alvain, S., & Gehlen, M. (2005). Response of diatoms
distribution to global warming and potential implications: A global model study. Geophysical
Research Letters, 32
105
Borrione, I., Aumont, O., Nielsdottir, M.C., & Schlitzer, R. (2014). Sedimentary and
atmospheric sources of iron around South Georgia, Southern Ocean: a modelling perspective.
Biogeosciences, 11, 1981-2001
Borrione, I., & Schlitzer, R. (2013). Distribution and recurrence of phytoplankton blooms
around South Georgia, Southern Ocean. Biogeosciences, 10, 217-231
Bracher, A., Vountas, M., Dinter, T., Burrows, J.P., Rottgers, R., & Peeken, I. (2009).
Quantitative observation of cyanobacteria and diatoms from space using PhytoDOAS on
SCIAMACHY data. Biogeosciences, 6, 751-764
Brewin, R.J.W., Sathyendranath, S., Hirata, T., Lavender, S.J., Barciela, R.M., & HardmanMountford, N.J. (2010). A three-component model of phytoplankton size class for the Atlantic
Ocean. Ecological Modelling, 221, 1472-1483
Brewin, R.J.W., Sathyendranath, S., Lange, P.K., & Tilstone, G. (2014). Comparison of two
methods to derive the size-structure of natural populations of phytoplankton. Deep-Sea
Research Part I-Oceanographic Research Papers, 85, 72-79
Brito, A.C., Sa, C., Brotas, V., Brewin, R.J.W., Silva, T., Vitorino, J., Platt, T., &
Sathyendranath, S. (2015). Effect of phytoplankton size classes on bio-optical properties of
phytoplankton in the Western Iberian coast: Application of models. Remote Sensing of
Environment, 156, 537-550
Brody, S.R., Lozier, M.S., & Dunne, J.P. (2013). A comparison of methods to determine
phytoplankton bloom initiation. Journal of Geophysical Research-Oceans, 118, 2345-2357
Campbell, J., Antoine, D., Armstrong, R., Arrigo, K., Balch, W., Barber, R., Behrenfeld, M.,
Bidigare, R., Bishop, J., Carr, M.E., Esaias, W., Falkowski, P., Hoepffner, N., Iverson, R.,
Kiefer, D., Lohrenz, S., Marra, J., Morel, A., Ryan, J., Vedernikov, V., Waters, K., Yentsch, C.,
& Yoder, J. (2002). Comparison of algorithms for estimating ocean primary production from
surface chlorophyll, temperature, and irradiance. Global Biogeochemical Cycles, 16
Carr, M.E., Friedrichs, M.A.M., Schmeltz, M., Aita, M.N., Antoine, D., Arrigo, K.R.,
Asanuma, I., Aumont, O., Barber, R., Behrenfeld, M., Bidigare, R., Buitenhuis, E.T.,
Campbell, J., Ciotti, A., Dierssen, H., Dowell, M., Dunne, J., Esaias, W., Gentili, B., Gregg,
W., Groom, S., Hoepffner, N., Ishizaka, J., Kameda, T., Le Quere, C., Lohrenz, S., Marra, J.,
Melin, F., Moore, K., Morel, A., Reddy, T.E., Ryan, J., Scardi, M., Smyth, T., Turpie, K.,
Tilstone, G., Waters, K., & Yamanaka, Y. (2006). A comparison of global estimates of marine
primary production from ocean color. Deep-Sea Research Part Ii-Topical Studies in
Oceanography, 53, 741-770
Cheah, W., Taylor, B.B., Wiegmann, S., Raimund, S., Krahmann, G., Quack, B., & Bracher,
A. (2013). Photophysiological state of natural phytoplankton communities in the South China
Sea and Sulu Sea. Biogeosciences Discuss., 10, 12115-12153
Ciasto, L.M., Simpkins, G.R., & England, M.H. (2015). Teleconnections between Tropical
Pacific SST Anomalies and Extratropical Southern Hemisphere Climate. Journal of Climate,
28, 56-65
Clarke, A.J. (2008). An Introduction to the Dynamics of El Niño & the Southern
106
Oscillation. New York, NY.: Academic Press
Comiso, J. (2010). Polar Oceans from Space New York, NY.: Springer Publishing
Cunningham, S.A. (2005). Southern Ocean circulation. Archives of natural history, 32, 265280
Dierssen, H.M. (2010). Perspectives on empirical approaches for ocean color remote sensing
of chlorophyll in a changing climate. Proceedings of the National Academy of Sciences of the
United States of America, 107, 17073-17078
Dierssen, H.M., & Smith, R.C. (2000). Bio-optical properties and remote sensing ocean color
algorithms for Antarctic Peninsula waters. Journal of Geophysical Research: Oceans, 105,
26301-26312
Dinter, T., Rozanov, V.V., Burrows, J.P., & Bracher, A. (2015). Retrieving the availability of
light in the ocean utilising spectral signatures of Vibrational Raman Scattering in hyperspectral satellite measurements. Ocean Sci. Discuss., 12, 31-81
Dlugokencky, E., & Tans, P. (2015). Trends in Atmospheric Carbon Dioxide. In:
NOAA/ESRL
Duarte, C.M., & Cebrian, J. (1996). The fate of marine autotrophic production. Limnology
and Oceanography, 41, 1758-1766
Durand, D. (2007). Full Validation Report. In A. Mangin, & S. Pinnock (Eds.),
GlobCOLOUR : An EO based service supporting global ocean carbon cycle research
Edwards, M., & Richardson, A.J. (2004). Impact of climate change on marine pelagic
phenology and trophic mismatch. Nature, 430, 881-884
Falkowski, P.G., Barber, R.T., & Smetacek, V. (1998). Biogeochemical controls and feedbacks
on ocean primary production. Science, 281, 200-206
Falkowski, P.G., Katz, M.E., Knoll, A.H., Quigg, A., Raven, J.A., Schofield, O., & Taylor,
F.J.R. (2004). The Evolution of Modern Eukaryotic Phytoplankton. Science, 305, 354-360
Falkowski, P.G., & Ravenm, J.A. (2007). Aquatic Photosynthesis: Second Edition. (2 ed.).
Princeton, New Jersey: Princeton University Press
Feldman, G.C., & McClain, C.R. (2012). Ocean Color Chlorophyll (OC) v6, Ocean Color
WebOcean Color Chlorophyll (OC) v6, Ocean Color Web. In N. Kuring, S.W. Bailey, B.F.
Franz, G. Meister, P.J. Werdell, & R.E. Eplee (Eds.): NASA Goddard Space Flight Center
Fetterer, F., Knowles, K., Meier, W., & Savoie, M. (2002). Sea Ice Index. In. Boulder,
Colorado USA: National Snow and Ice Data Center
Field, C.B., Behrenfeld, M.J., Randerson, J.T., & Falkowski, P. (1998). Primary production of
the biosphere: Integrating terrestrial and oceanic components. Science, 281, 237-240
Finkel, Z.V., Beardall, J., Flynn, K.J., Quigg, A., Rees, T.A.V., & Raven, J.A. (2010).
Phytoplankton in a changing world: cell size and elemental stoichiometry. Journal of
Plankton Research, 32, 119-137
Fogt, R., Bromwich, D., & Hines, K. (2011). Understanding the SAM influence on the South
107
Pacific ENSO teleconnection. Climate Dynamics, 36, 1555-1576
Franks, P.J.S. (2014). Has Sverdrup's critical depth hypothesis been tested? Mixed layers vs.
turbulent layers. ICES Journal of Marine Science: Journal du Conseil
Fujiwara, A., Hirawake, T., Suzuki, K., Imai, I., & Saitoh, S.I. (2014). Timing of sea ice
retreat can alter phytoplankton community structure in the western Arctic Ocean.
Biogeosciences, 11, 1705-1716
Garcia, C.A.E., Garcia, V.M.T., & McClain, C.R. (2005). Evaluation of SeaWiFS chlorophyll
algorithms in the Southwestern Atlantic and Southern Oceans. Remote Sensing of
Environment, 95, 125-137
Garrison, D.L., Gowing, M.M., Hughes, M.P., Campbell, L., Caron, D.A., Dennett, M.R.,
Shalapyonok, A., Olson, R.J., Landry, M.R., Brown, S.L., Liu, H.-B., Azam, F., Steward, G.F.,
Ducklow, H.W., & Smith, D.C. (2000). Microbial food web structure in the Arabian Sea: a US
JGOFS study. Deep Sea Research Part II: Topical Studies in Oceanography, 47, 1387-1422
Gordon, H.R., Brown, O.B., Evans, R.H., Brown, J.W., Smith, R.C., Baker, K.S., & Clark,
D.K. (1988). A Semianalytic Radiance Model of Ocean Color. Journal of Geophysical
Research-Atmospheres, 93, 10909-10924
Gordon, H.R., & Clark, D.K. (1980). Remote-Sensing Optical-Properties of a Stratified Ocean
- an Improved Interpretation. Applied Optics, 19, 3428-3430
Gordon, H.R., & Mccluney, W.R. (1975). Estimation of Depth of Sunlight Penetration in Sea
for Remote-Sensing. Applied Optics, 14, 413-416
Gordon, H.R., & Wang, M.H. (1994). Retrieval of Water-Leaving Radiance and Aerosol
Optical-Thickness over the Oceans with Seawifs - a Preliminary Algorithm. Applied Optics,
33, 443-452
Graham, R.M., & De Boer, A.M. (2013). The Dynamical Subtropical Front. Journal of
Geophysical Research: Oceans, 118, 5676-5685
Gregg, W.W., & Casey, N.W. (2007). Modeling coccolithophores in the global oceans. DeepSea Research Part Ii-Topical Studies in Oceanography, 54, 447-477
Hauck, J., & Völker, C. (2015). Rising atmospheric CO2 leads to large impact of biology on
Southern Ocean CO2 uptake via changes of the Revelle factor. Geophysical Research Letters,
42, 1459-1464
Hauck, J., Völker, C., Wang, T., Hoppema, M., Losch, M., & Wolf-Gladrow, D.A. (2013).
Seasonally different carbon flux changes in the Southern Ocean in response to the southern
annular mode. Global Biogeochemical Cycles, 27, 1236-1245
Henson, S.A. (2014). Slow science: the value of long ocean biogeochemistry records.
Philosophical Transactions of the Royal Society a-Mathematical Physical and Engineering
Sciences, 372
Henson, S.A., Dunne, J.P., & Sarmiento, J.L. (2009). Decadal variability in North Atlantic
phytoplankton blooms. Journal of Geophysical Research: Oceans, 114, C04013
Henson, S.A., Sanders, R., Madsen, E., Morris, P.J., Le Moigne, F., & Quartly, G.D. (2011). A
108
reduced estimate of the strength of the ocean's biological carbon pump. Geophysical Research
Letters, 38
Henson, S.A., Sarmiento, J.L., Dunne, J.P., Bopp, L., Lima, I., Doney, S.C., John, J., &
Beaulieu, C. (2010). Detection of anthropogenic climate change in satellite records of ocean
chlorophyll and productivity. Biogeosciences, 7, 621-640
Henson, S.A., & Thomas, A.C. (2007). Interannual variability in timing of bloom initiation in
the California Current System. Journal of Geophysical Research: Oceans, 112, C08007
Hirata, T., Aiken, J., Hardman-Mountford, N., Smyth, T.J., & Barlow, R.G. (2008). An
absorption model to determine phytoplankton size classes from satellite ocean colour. Remote
Sensing of Environment, 112, 3153-3159
Hirata, T., Hardman-Mountford, N.J., Brewin, R.J.W., Aiken, J., Barlow, R., Suzuki, K., Isada,
T., Howell, E., Hashioka, T., Noguchi-Aita, M., & Yamanaka, Y. (2011). Synoptic
relationships between surface Chlorophyll-a and diagnostic pigments specific to
phytoplankton functional types. Biogeosciences, 8, 311-327
Hirawake, T., Shinmyo, K., Fujiwara, A., & Saitoh, S. (2012). Satellite remote sensing of
primary productivity in the Bering and Chukchi Seas using an absorption-based approach.
Ices Journal of Marine Science, 69, 1194-1204
Hirawake, T., Takao, S., Horimoto, N., Ishimaru, T., Yamaguchi, Y., & Fukuchi, M. (2011). A
phytoplankton absorption-based primary productivity model for remote sensing in the
Southern Ocean. Polar Biology, 34, 291-302
Hoffmann, L.J., Peeken, I., Lochte, K., Assmy, P., & Veldhuis, M. (2006). Different reactions
of Southern Ocean phytoplankton size classes to iron fertilization. Limnology and
Oceanography, 51, 1217-1229
Holm-Hansen, O., Kahru, M., & Hewes, C.D. (2005). Deep chlorophyll a maxima (DCMs) in
pelagic Antarctic waters. II. Relation to bathymetric features and dissolved iron
concentrations. Marine Ecology Progress Series, 297, 71-81
Huisman, J., Arrayás, M., Ebert, U., & Sommeijer, B. (2002). How do sinking phytoplankton
species manage to persist? The American Naturalist, 159, 245-254
Hyde, K.J.W., O’Reilly, J.E., & Oviatt, C.A. (2007). Validation of SeaWiFS chlorophyll a in
Massachusetts Bay. Continental Shelf Research, 27, 1677-1691
Iida, T., Saitoh, S.I., Miyamura, T., Toratani, M., Fukushima, H., & Shiga, N. (2002).
Temporal and spatial variability of coccolithophore blooms in the eastern Bering Sea, 19982001. Progress in Oceanography, 55, 165-175
IOCCG (2006). Remote Sensing of Inherent Optical Properties: Fundamentals,
TestsofAlgorithms, and Applications. In Z. Lee (Ed.), Reports of the International OceanColour Coordinating Group: IOCCG
IOCCG (2014). Phytoplankton Functional Types from Space. In S. Sathyendranath (Ed.),
Reports of the International Ocean-Colour Coordinating Group. Dartmouth, Canada
IPCC (2014). Synthesis Report. Contribution of Working Groups I, II and III to the Fifth
109
Assessment Report of the Intergovernmental Panel on Climate Change. In R.K. Pachauri, &
L.A. Meyer (Eds.), Climate Change 2014 (p. 151). Geneva, Switzerland: IPCC
Ito, A. (2011). A historical meta-analysis of global terrestrial net primary productivity: are
estimates converging? Global Change Biology, 17, 3161-3175
Johnson, R., Strutton, P.G., Wright, S.W., McMinn, A., & Meiners, K.M. (2013). Three
improved satellite chlorophyll algorithms for the Southern Ocean. Journal of Geophysical
Research: Oceans, 118, 3694-3703
Kahru, M., Brotas, V., Manzano-Sarabia, M., & Mitchell, B.G. (2011). Are phytoplankton
blooms occurring earlier in the Arctic? Global Change Biology, 17, 1733-1739
Kahru, M., & Mitchell, B.G. (2010). Blending of ocean colour algorithms applied to the
Southern Ocean. Remote Sensing Letters, 1, 119-124
Kemp, A.E.S., & Villareal, T.A. (2013). High diatom production and export in stratified
waters - A potential negative feedback to global warming. Progress in Oceanography, 119, 423
Kirk, J.T.O. (2011). Light and Photosynthesis in Aquatic Ecosystems. (3 ed.). New York:
Cambridge University Press
Kooistra, W., Gersonde, R., Medlin, L.K., & Mann, D.G. (2007). The origin and evolution of
the diatoms: their adaptation to a planktonic existence. Evolution of primary producers in the
sea, 207-249
Krasemann, H., Belo Couto, A., Brando, V., Brewin, R.J.W., Brockmann, C., Brotas, V.,
Doerffer, R., Feng, H., Froiun, R., Gould, R., Grant, M., Groom, S., Hooker, S., Jackson, T.,
Kahru, M., Kratzer, S., Melin, F., Mitchell, G., Morrison, R., Müller, D., Muller-Karger, F.,
Sathyendranath, S., Sosik, H., Steinmetz, F., Swinton, J., Valente, A., & Voss, K. (2014).
Product Validation and Intercomparison Report In J. Swinton, & S. Sathyendranath (Eds.),
Ocean Colour Climate Change Initiative (OC_CCI) - Phase One. Plymouth Plymouth Marine
Laboratory
Kwok, R., & Comiso, J. (2002). Southern Ocean climate and sea ice anomalies associated
with the Southern Oscillation. Journal of Climate, 15, 487-501
L'Heureux, M.L., & Thompson, D.W. (2006). Observed relationships between the El NinoSouthern Oscillation and the extratropical zonal-mean circulation. Journal of Climate, 19,
276-287
Leblanc, K., Arístegui, J., Armand, L., Assmy, P., Beker, B., Bode, A., Breton, E., Cornet, V.,
Gibson, J., Gosselin, M.P., Kopczynska, E., Marshall, H., Peloquin, J., Piontkovski, S.,
Poulton, A.J., Quéguiner, B., Schiebel, R., Shipe, R., Stefels, J., van Leeuwe, M.A., Varela,
M., Widdicombe, C., & Yallop, M. (2012). A global diatom database – abundance, biovolume
and biomass in the world ocean. Earth Syst. Sci. Data, 4, 149-165
Lee, Z., Du, K., Arnone, R., Liew, S., & Penta, B. (2005). Penetration of solar radiation in the
upper ocean: A numerical model for oceanic and coastal waters. Journal of Geophysical
Research: Oceans, 110, C09019
110
Lee, Z., & Hu, C. (2006). Global distribution of Case-1 waters: An analysis from SeaWiFS
measurements. Remote Sensing of Environment, 101, 270-276
Lee, Z., Lance, V.P., Shang, S., Vaillancourt, R., Freeman, S., Lubac, B., Hargreaves, B.R.,
Del Castillo, C., Miller, R., Twardowski, M., & Wei, G. (2011). An assessment of optical
properties and primary production derived from remote sensing in the Southern Ocean (SO
GasEx). Journal of Geophysical Research: Oceans, 116, C00F03
Lee, Z., Lubac, B., Werdell, J., & Arnone, R. (2009). An Update of the Quasi-An alytical
Algorithm.
Lee, Z., Shang, S., Hu, C., Lewis, M., Arnone, R., Li, Y., & Lubac, B. (2010). Time series of
bio-optical properties in a subtropical gyre: Implications for the evaluation of interannual
trends of biogeochemical properties. Journal of Geophysical Research: Oceans, 115, C09012
Lee, Z., Weidemann, A., Kindle, J., Arnone, R., Carder, K.L., & Davis, C. (2007). Euphotic
zone depth: Its derivation and implication to ocean-color remote sensing. Journal of
Geophysical Research: Oceans, 112, C03009
Lee, Z.P., Carder, K.L., & Arnone, R.A. (2002). Deriving inherent optical properties from
water color: a multiband quasi-analytical algorithm for optically deep waters. Applied Optics,
41, 5755-5772
Lefebvre, W., Goosse, H., Timmermann, R., & Fichefet, T. (2004). Influence of the Southern
Annular Mode on the sea ice–ocean system. Journal of Geophysical Research: Oceans, 109,
C09005
Lovenduski, N.S. (2007). Impact of the Southern Annular Mode on Southern Ocean
Circulation and Biogeochemistry. ProQuest
Lovenduski, N.S., & Gruber, N. (2005). Impact of the Southern Annular Mode on Southern
Ocean circulation and biology. Geophysical Research Letters, 32
Luthi, D., Le Floch, M., Bereiter, B., Blunier, T., Barnola, J.-M., Siegenthaler, U., Raynaud,
D., Jouzel, J., Fischer, H., Kawamura, K., & Stocker, T.F. (2008). High-resolution carbon
dioxide concentration record 650,000-800,000[thinsp]years before present. Nature, 453, 379382
Maheshwari, M., Singh, R.K., Oza, S.R., & Kumar, R. (2013). An Investigation of the
Southern Ocean Surface Temperature Variability Using Long-Term Optimum Interpolation
SST Data. ISRN Oceanography, 2013, 9
MaierReimer, E., Mikolajewicz, U., & Winguth, A. (1996). Future ocean uptake of CO2:
Interaction between ocean circulation and biology. Climate Dynamics, 12, 711-721
Maksym, T., Stammerjohn, S.E., Ackley, S., & Massom, R. (2012). Antarctic Sea Ice-A Polar
Opposite? Oceanography, 25, 140-151
Marrari, M., Hu, C., & Daly, K. (2006). Validation of SeaWiFS chlorophyll a concentrations
in the Southern Ocean: A revisit. Remote Sensing of Environment, 105, 367-375
Martin, S. (2004). An Introduction to Ocean Remote Sensing. Cambridge: Cambridge
University Press
111
Mélin, F. (2011). Comparison of SeaWiFS and MODIS time series of inherent optical
properties for the Adriatic Sea. Ocean Sci., 7, 351-361
Milutinovic, S., & Bertino, L. (2011). Assessment and propagation of uncertainties in input
terms through an ocean-color-based model of primary productivity. Remote Sensing of
Environment, 115, 1906-1917
Mo, K.C. (2000). Relationships between Low-Frequency Variability in the Southern
Hemisphere and Sea Surface Temperature Anomalies. Journal of Climate, 13, 3599-3610
Mobley, C.D. (1994). Light and Water: Radiative Transfer in Natural Waters. Academic Press
Montegut, C.D., Madec, G., Fischer, A.S., Lazar, A., & Iudicone, D. (2004). Mixed layer
depth over the global ocean: An examination of profile data and a profile-based climatology.
Journal of Geophysical Research-Oceans, 109
Montes-Hugo, M.A., Vernet, M., Martinson, D., Smith, R., & Iannuzzi, R. (2008). Variability
on phytoplankton size structure in the western Antarctic Peninsula (1997–2006). Deep Sea
Research Part II: Topical Studies in Oceanography, 55, 2106-2117
Morel, A. (1974). Optical properties of pure water and pure sea water. In N.G. Jerlov, & E.S.
Nielsen (Eds.), Optical Aspects of Oceanography (p. 24). New York: Academic
Morel, A., Antoine, D., & Gentili, B. (2002). Bidirectional reflectance of oceanic waters:
accounting for Raman emission and varying particle scattering phase function. Applied
Optics, 41, 6289-6306
Morel, A., & Berthon, J.-F. (1989). Surface pigments, algal biomass profiles, and potential
production of the euphotic layer: Relationships reinvestigated in view of remote-sensing
applications. Limnology and Oceanography, 34, 1545-1562
Morel, A., Claustre, H., Antoine, D., & Gentili, B. (2007). Natural variability of bio-optical
properties in Case 1 waters: attenuation and reflectance within the visible and near-UV
spectral domains, as observed in South Pacific and Mediterranean waters. Biogeosciences, 4,
913-925
Morel, A., & Gentili, B. (2004). Radiation transport within oceanic (case 1) water. Journal of
Geophysical Research: Oceans, 109, C06008
Morel, A., & Maritorena, S. (2001). Bio-optical properties of oceanic waters: A reappraisal.
Journal of Geophysical Research-Oceans, 106, 7163-7180
Nair, A., Sathyendranath, S., Platt, T., Morales, J., Stuart, V., Forget, M.H., Devred, E., &
Bouman, H. (2008). Remote sensing of phytoplankton functional types. Remote Sensing of
Environment, 112, 3366-3375
NSDC (2015). Cryosphere Glossary. In, https://nsidc.org/cryosphere/glossary/term/seasonalice-zone
O'Reilly, J.E., Maritorena, S., Siegel, D., O'Brien, M., Toole, D., Mitchell, B.G., Kahru, M.,
Chavez, F., Strutton, P., Cota, G., Hooker, S., McClain, C., Carder, K., Muller-Karger, F.,
Harding, L., Magnuson, A., Phinney, D., Moore, G., Aiken, J., Arrigo, K., Letelier, R., &
Culver, M. (2000). Ocean color chlorophyll a algorithms for SeaWiFS, OC2, and OC4:
112
Version 4. In E.R. Firestone, & S.B. Hooker (Eds.), SeaWiFS Postlaunch Calibration and
Validation Analyses, Part 3. Greenbelt, Maryland: NASA Goddard Space Flight Center
Peloquin, J., Swan, C., Gruber, N., Vogt, M., Claustre, H., Ras, J., Uitz, J., Barlow, R.,
Behrenfeld, M., Bidigare, R., Dierssen, H., Ditullio, G., Fernandez, E., Gallienne, C., Gibb,
S., Goericke, R., Harding, L., Head, E., Holligan, P., Hooker, S., Karl, D., Landry, M.,
Letelier, R., Llewellyn, C.A., Lomas, M., Lucas, M., Mannino, A., Marty, J.C., Mitchell,
B.G., Muller-Karger, F., Nelson, N., O'Brien, C., Prezelin, B., Repeta, D., Jr. Smith, W.O.,
Smythe-Wright, D., Stumpf, R., Subramaniam, A., Suzuki, K., Trees, C., Vernet, M.,
Wasmund, N., & Wright, S. (2013). The MAREDAT global database of high performance
liquid chromatography marine pigment measurements. Earth Syst. Sci. Data, 5, 109-123
Penland, C., Sun, D.-Z., Capotondi, A., & Vimont, D.J. (2013). A Brief Introduction to El
Niño and La Niña. Climate Dynamics: Why Does Climate Vary? (pp. 53-64): American
Geophysical Union
Planquette, H., Statham, P.J., Fones, G.R., Charette, M.A., Moore, C.M., Salter, I., Nedelec,
F.H., Taylor, S.L., French, M., Baker, A.R., Mahowald, N., & Jickells, T.D. (2007). Dissolved
iron in the vicinity of the Crozet Islands, Southern Ocean. Deep-Sea Research Part Ii-Topical
Studies in Oceanography, 54, 1999-2019
Pohl, B., Fauchereau, N., Reason, C.J.C., & Rouault, M. (2010). Relationships between the
Antarctic Oscillation, the Madden–Julian Oscillation, and ENSO, and Consequences for
Rainfall Analysis. Journal of Climate, 23, 238-254
Pollard, R.T., Lucas, M.I., & Read, J.F. (2002). Physical controls on biogeochemical zonation
in the Southern Ocean. Deep Sea Research Part II: Topical Studies in Oceanography, 49,
3289-3305
Pope, R.M., & Fry, E.S. (1997). Absorption spectrum (380?700 nm) of pure water. II.
Integrating cavity measurements. Applied Optics, 36, 8710-8723
Racault, M.F., Le Quere, C., Buitenhuis, E., Sathyendranath, S., & Platt, T. (2012).
Phytoplankton phenology in the global ocean. Ecological Indicators, 14, 152-163
Racault, M.F., Sathyendranath, S., & Platt, T. (2014). Impact of missing data on the estimation
of ecological indicators from satellite ocean-colour time-series. Remote Sensing of
Environment, 152, 15-28
Reynolds, R.A., Stramski, D., & Mitchell, B.G. (2001). A chlorophyll-dependent
semianalytical reflectance model derived from field measurements of absorption and
backscattering coefficients within the Southern Ocean. Journal of Geophysical Research:
Oceans, 106, 7125-7138
Rintoul, S. (2010). Antarctic Circumpolar Current. Ocean Currents: A Derivative of the
Encyclopedia of Ocean Sciences, 196
Rintoul, S.R., & Garabato, A.C.N. (2013). Dynamics of the Southern Ocean circulation. In G.
Siedler, S. Griffies, J. Gould, & J. Church (Eds.), Ocean Circulation and Climate: A 21st
century perspective (pp. 471–492): Academic Pres
Rintoul, S.R., Hughes, C., & Olbers, D. (1999). The Antarctic Circumpolar Current System.
113
Oceans and Climate, G. Siedler and J. Church,(eds.), Academic Press (submitted)
Robinson, I.S. (2010). Discovering the Ocean from Space. (1 ed.). Springer-Verlag Berlin
Heidelberg
Rousseaux, C.S., & Gregg, W.W. (2014). Interannual Variation in Phytoplankton Primary
Production at A Global Scale. Remote Sensing, 6, 1-19
Saba, V.S., Friedrichs, M.A.M., Antoine, D., Armstrong, R.A., Asanuma, I., Behrenfeld, M.J.,
Ciotti, A.M., Dowell, M., Hoepffner, N., Hyde, K.J.W., Ishizaka, J., Kameda, T., Marra, J.,
Melin, F., Morel, A., O'Reilly, J., Scardi, M., Smith, W.O., Smyth, T.J., Tang, S., Uitz, J.,
Waters, K., & Westberry, T.K. (2011). An evaluation of ocean color model estimates of marine
primary productivity in coastal and pelagic regions across the globe. Biogeosciences, 8, 489503
Sabine, C.L., Feely, R.A., Gruber, N., Key, R.M., Lee, K., Bullister, J.L., Wanninkhof, R.,
Wong, C.S., Wallace, D.W.R., Tilbrook, B., Millero, F.J., Peng, T.H., Kozyr, A., Ono, T., &
Rios, A.F. (2004). The oceanic sink for anthropogenic CO2. Science, 305, 367-371
Sadeghi, A., Dinter, T., Vountas, M., Taylor, B.B., Altenburg-Soppa, M., Peeken, I., &
Bracher, A. (2012). Improvement to the PhytoDOAS method for identification of
coccolithophores using hyper-spectral satellite data. Ocean Science, 8, 1055-1070
Sallée, J.B., Speer, K., & Morrow, R. (2008). Response of the Antarctic Circumpolar Current
to Atmospheric Variability. Journal of Climate, 21, 3020-3039
Sallee, J.B., Speer, K.G., & Rintoul, S.R. (2010). Zonally asymmetric response of the
Southern Ocean mixed-layer depth to the Southern Annular Mode. Nature Geoscience, 3,
273-279
Sarmiento, J.L., Gruber, N., Brzezinski, M.A., & Dunne, J.P. (2004). High-latitude controls of
thermocline nutrients and low latitude biological productivity. Nature, 427, 56-60
Schmittner, A., Chiang, J.C.H., & Hemming, S.R. (2013). Introduction: The Ocean's
Meridional Overturning Circulation. Ocean Circulation: Mechanisms and Impacts—Past and
Future Changes of Meridional Overturning (pp. 1-4): American Geophysical Union
Schwartz, M.D. (2013). Introduction. In M.D. Schwartz (Ed.), Phenology: An Integrative
Environmental Science (pp. 1-5): Springer Netherlands
Shang, S., Dong, Q., Lee, Z., Li, Y., Xie, Y., & Behrenfeld, M. (2011a). MODIS observed
phytoplankton dynamics in the Taiwan Strait: an absorption-based analysis. Biogeosciences,
8, 841-850
Shang, S., Lee, Z., & Wei, G. (2011b). Characterization of MODIS-derived euphotic zone
depth: Results for the China Sea. Remote Sensing of Environment, 115, 180-186
Siegel, D.A., Buesseler, K.O., Doney, S.C., Sailley, S.F., Behrenfeld, M.J., & Boyd, P.W.
(2014). Global assessment of ocean carbon export by combining satellite observations and
food-web models. Global Biogeochemical Cycles, 28, 181-196
Siegel, D.A., Doney, S.C., & Yoder, J.A. (2002). The North Atlantic spring phytoplankton
bloom and Sverdrup's critical depth hypothesis. Science, 296, 730-733
114
Smetacek, V. (1999). Diatoms and the ocean carbon cycle. Protist, 150, 25-32
Smetacek, V., Assmy, P., & Henjes, J. (2004). The role of grazing in structuring Southern
Ocean pelagic ecosystems and biogeochemical cycles. Antarctic Science, 16, 541-558
Smetacek, V., Klaas, C., Strass, V.H., Assmy, P., Montresor, M., Cisewski, B., Savoye, N.,
Webb, A., d'Ovidio, F., Arrieta, J.M., Bathmann, U., Bellerby, R., Berg, G.M., Croot, P.,
Gonzalez, S., Henjes, J., Herndl, G.J., Hoffmann, L.J., Leach, H., Losch, M., Mills, M.M.,
Neill, C., Peeken, I., Rottgers, R., Sachs, O., Sauter, E., Schmidt, M.M., Schwarz, J.,
Terbruggen, A., & Wolf-Gladrow, D. (2012). Deep carbon export from a Southern Ocean ironfertilized diatom bloom. Nature, 487, 313-319
Smetacek, V.S. (1985). Role of Sinking in Diatom Life-History Cycles - Ecological,
Evolutionary and Geological Significance. Marine Biology, 84, 239-251
Smith, R.C., Martinson, D.G., Stammerjohn, S.E., Iannuzzi, R.A., & Ireson, K. (2008).
Bellingshausen and western Antarctic Peninsula region: Pigment biomass and sea-ice
spatial/temporal distributions and interannual variabilty. Deep Sea Research Part II: Topical
Studies in Oceanography, 55, 1949-1963
Smith, W.O., Sedwick, P.N., Arrigo, K.R., Ainley, D.G., & Orsi, A.H. (2012). The Ross Sea in
a Sea of Change. Oceanography, 25, 90-103
Sokolov, S., & Rintoul, S.R. (2007). On the relationship between fronts of the Antarctic
Circumpolar Current and surface chlorophyll concentrations in the Southern Ocean. Journal
of Geophysical Research-Oceans, 112
Soppa, M.A., Dinter, T., Taylor, B.B., & Bracher, A. (2013). Satellite derived euphotic depth
in the Southern Ocean: Implications for primary production modelling. Remote Sensing of
Environment, 137, 198-211
Soppa, M.A., Hirata, T., Silva, B., Dinter, T., Peeken, I., Wiegmann, S., & Bracher, A. (2014).
Global Retrieval of Diatom Abundance Based on Phytoplankton Pigments and Satellite Data.
Remote Sensing, 6, 10089-10106
Sournia, A., Chretiennotdinet, M.J., & Ricard, M. (1991). Marine-Phytoplankton - How Many
Species in the World Ocean. Journal of Plankton Research, 13, 1093-1099
Steinmetz, F., Deschamps, P.Y., & Ramon, D. (2011). Atmospheric correction in presence of
sun glint: application to MERIS. Optics Express, 19, 9783-9800
Sverdrup, H.U. (1953). On Conditions for the Vernal Blooming of Phytoplankton. Journal du
Conseil, 18, 287-295
Szeto, M., Werdell, P.J., Moore, T.S., & Campbell, J.W. (2011). Are the world's oceans
optically different? Journal of Geophysical Research: Oceans, 116, C00H04
Tagliabue, A., Sallee, J.B., Bowie, A.R., Levy, M., Swart, S., & Boyd, P.W. (2014). Surfacewater iron supplies in the Southern Ocean sustained by deep winter mixing. Nature
Geoscience, 7, 314-320
Takahashi, T., Sutherland, S.C., Wanninkhof, R., Sweeney, C., Feely, R.A., Chipman, D.W.,
Hales, B., Friederich, G., Chavez, F., Sabine, C., Watson, A., Bakker, D.C.E., Schuster, U.,
115
Metzl, N., Yoshikawa-Inoue, H., Ishii, M., Midorikawa, T., Nojiri, Y., Kortzinger, A.,
Steinhoff, T., Hoppema, M., Olafsson, J., Arnarson, T.S., Tilbrook, B., Johannessen, T., Olsen,
A., Bellerby, R., Wong, C.S., Delille, B., Bates, N.R., & de Baar, H.J.W. (2009).
Climatological mean and decadal change in surface ocean pCO(2), and net sea-air CO2 flux
over the global oceans (vol 56, pg 554, 2009). Deep-Sea Research Part I-Oceanographic
Research Papers, 56, 2075-2076
Takahashi, T., Sweeney, C., Hales, B., Chipman, D.W., Newberger, T., Goddard, J.G.,
Iannuzzi, R.A., & Sutherland, S.C. (2012). The Changing Carbon Cycle in the Southern
Ocean. Oceanography, 25, 26-37
Talley, L.D. (2013). Closure of the Global Overturning Circulation Through the Indian,
Pacific, and Southern Oceans: Schematics and Transports. Oceanography, 26, 80-97
Tassan, S., & Ferrari, G.M. (1995). An alternative approach to absorption measurements of
aquatic particles retained on filters. Limnology and Oceanography, 40, 1358-1368
Taylor, B.B., Torrecilla, E., Bernhardt, A., Taylor, M.H., Peeken, I., Röttgers, R., Piera, J., &
Bracher, A. (2011). Bio-optical provinces in the eastern Atlantic Ocean and their
biogeographical relevance. Biogeosciences, 8, 3609-3629
Taylor, M.H., Losch, M., & Bracher, A. (2013). On the drivers of phytoplankton blooms in the
Antarctic marginal ice zone: A modeling approach. Journal of Geophysical Research: Oceans,
118, 63-75
Thomalla, S.J., Fauchereau, N., Swart, S., & Monteiro, P.M.S. (2011). Regional scale
characteristics of the seasonal cycle of chlorophyll in the Southern Ocean. Biogeosciences, 8,
2849-2866
Thompson, D.W.J., Solomon, S., Kushner, P.J., England, M.H., Grise, K.M., & Karoly, D.J.
(2011). Signatures of the Antarctic ozone hole in Southern Hemisphere surface climate
change. Nature Geoscience, 4, 741-749
Uitz, J., Claustre, H., Gentili, B., & Stramski, D. (2009a). Phytoplankton Class-Specific
Primary Production in the World's Oceans: An Estimate Based on Seawifs Time Series (19982007). Phycologia, 48, 134-134
Uitz, J., Claustre, H., Gentili, B., & Stramski, D. (2010). Phytoplankton class-specific primary
production in the world's oceans: Seasonal and interannual variability from satellite
observations. Global Biogeochemical Cycles, 24
Uitz, J., Claustre, H., Griffiths, F.B., Ras, J., Garcia, N., & Sandronie, V. (2009b). A
phytoplankton class-specific primary production model applied to the Kerguelen Islands
region (Southern Ocean). Deep-Sea Research Part I-Oceanographic Research Papers, 56,
541-560
Uitz, J., Claustre, H., Morel, A., & Hooker, S.B. (2006). Vertical distribution of phytoplankton
communities in open ocean: An assessment based on surface chlorophyll. Journal of
Geophysical Research-Oceans, 111
Uitz, J., Stramski, D., Gentili, B., D'Ortenzio, F., & Claustre, H. (2012). Estimates of
phytoplankton class-specific and total primary production in the Mediterranean Sea from
116
satellite ocean color observations. Global Biogeochemical Cycles, 26
Vidussi, F., Claustre, H., Manca, B.B., Luchetta, A., & Marty, J.C. (2001). Phytoplankton
pigment distribution in relation to upper thermocline circulation in the eastern Mediterranean
Sea during winter. Journal of Geophysical Research-Oceans, 106, 19939-19956
Vountas, M., Dinter, T., Bracher, A., Burrows, J.P., & Sierk, B. (2007). Spectral studies of
ocean water with space-borne sensor SCIAMACHY using differential optical absorption
Spectroscopy (DOAS). Ocean Science, 3, 429-440
Wang, M., & Shi, W. (2009). Detection of Ice and Mixed Ice-Water Pixels for MODIS Ocean
Color Data Processing. Geoscience and Remote Sensing, IEEE Transactions on, 47, 25102518
Werdell, P.J., Bailey, S.W., Fargion, G.S., Pietras, C., Knobelspiesse, K.D., Feldman, G.C., &
McClain, C.R. (2003). Unique data repository facilitates ocean color satellite validation. EOS
Trans. AGU, 84, 377
Westberry, T., Behrenfeld, M.J., Siegel, D.A., & Boss, E. (2008). Carbon-based primary
productivity modeling with vertically resolved photoacclimation. Global Biogeochemical
Cycles, 22, GB2024
Willmott, C.J., & Matsuura, K. (2005). Advantages of the mean absolute error (MAE) over
the root mean square error (RMSE) in assessing average model performance. Climate
Research, 30, 79-82
Wolter, K., & Timlin, M.S. (1993). Monitoring ENSO in COADS with a seasonally adjusted
principal component index. In, Proc. of the 17th Climate Diagnostics Workshop (pp. 52-57)
Wright, S.W., & van den Enden, R.L. (2000). Phytoplankton community structure and stocks
in the East Antarctic marginal ice zone (BROKE survey, January-March 1996) determined by
CHEMTAX analysis of HPLC pigment signatures. Deep-Sea Research Part Ii-Topical Studies
in Oceanography, 47, 2363-2400
Yao, X., & Schlitzer, R. (2013). Assimilating water column and satellite data for marine
export production estimation. Geoscientific Model Development, 6, 1575-1590
Yeo, S.-R., & Kim, K.-Y. (2015). Decadal changes in the Southern Hemisphere sea surface
temperature in association with El Niño–Southern Oscillation and Southern Annular Mode.
Climate Dynamics, 1-16
Yuan, X. (2004). ENSO-related impacts on Antarctic sea ice: a synthesis of phenomenon and
mechanisms. Antarctic Science, 16, 415-425
Zaneveld, J.R.V. (1995). A Theoretical Derivation of the Dependence of the Remotely-Sensed
Reflectance of the Ocean on the Inherent Optical-Properties. Journal of Geophysical
Research-Oceans, 100, 13135-13142
Zibordi, G., Melin, F., & Berthon, J.F. (2006). Comparison of SeaWiFS, MODIS and MERIS
radiometric products at a coastal site. Geophysical Research Letters, 33
117
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement