Monitoring marine phytoplankton seasonality from space

Remote Sensing of Environment xxx (2011) xxx–xxx

RSE-08090; No of Pages 12

Contents lists available at SciVerse ScienceDirect

Remote Sensing of Environment

j ourna l homepage: www.e lsev ie r .com/ locate / rse

Monitoring marine phytoplankton seasonality from space

Hervé Demarcq a,⁎, Gabriel Reygondeau a, Séverine Alvain b, Vincent Vantrepotte b

a IRD (Institut de Recherche pour le Développement), UMR EME 212, CRH (Centre de Recherche Halieutique Méditerranéenne et Tropicale), av. Jean Monnet, B.P. 171,34203 Sète cedex, Franceb Centre National de la Recherche Scientifique, Laboratoire d'Océanologie et de Géosciences’ UMR LOG CNRS 8187, Station Marine, Université des Sciences et Technologies de Lille – Lille 1,BP 80, 62930 Wimereux, France

⁎ Corresponding author. Tel.: +33 4 99 57 32 13; faxE-mail address: [email protected] (H. Demarcq)

0034-4257/$ – see front matter © 2011 Elsevier Inc. Alldoi:10.1016/j.rse.2011.09.019

Please cite this article as: Demarcq, H., et(2011), doi:10.1016/j.rse.2011.09.019

a b s t r a c t
a r t i c l e i n f o
Article history:Received 31 August 2010Received in revised form 23 August 2011Accepted 29 September 2011Available online xxxx

Keywords:PhytoplanktonSeasonalitySeasonal parametersBiogeographyRemote sensingSurface chlorophyll concentrationPhytoplankton Functional TypesWavelet transform

Remote sensing techniques are used to study the large scale patterns related to the seasonal modes of vari-ability of the marine phytoplankton. Ten years of monthly composite maps of sea surface chlorophyll-a con-centration and the PHYSAT database of four Phytoplanktonic Functional Types (PFTs), both from SeaWiFS, areused to investigate characteristics of phytoplankton seasonality in the trades and westerlies wind oceanic bi-omes, where data density is adequate. We use a combination of wavelet transform and statistical techniquesthat allow us to quantify both intensity and duration of the seasonal oscillation of chlorophyll-a concentra-tion and PFTs relative occurrence, and to map these relationships. Next, the seasonal oscillations detectedare related to four PFTs revealing six major global phytoplanktonic associations. Our results elucidate the in-tensity and duration of the seasonal dynamic of the chlorophyll-a concentration and of the relative occur-rence of four PFTs at a global scale. Thus, the typology of the different types of seasonality is investigated.Finally, an overall agreement between the results and the biogeochemical provinces partition proposed byLonghurst is found, revealing a strong environmental control on the seasonal oscillation of primary producersand a clear latitudinal organization in the succession of the phytoplankton types. Results provided in thisstudy quantify the seasonal oscillation of key structural parameters of the global ocean, and their potentialimplications for our understanding of ecosystem dynamics.

© 2011 Elsevier Inc. All rights reserved.

1. Introduction

In the global ocean, climatic conditions strongly influence the abi-otic factors that regulate the biological cycle of primary producers,and by extension, the timing, intensity and duration of their bloomingperiod (Nemani et al., 2003). Seasonal oscillations of marine vegeta-tion are crucial for marine ecosystems because of their impact on,and regulation of the biological parameters of many marine speciesincluding reproduction, migration and survival rate (Schwartz et al.,2006). Indeed, Edwards and Richardson (2004) have shown, usingthe Continuous Plankton Recorder dataset, that changes in the marineseasonal oscillation of primary producers may affect the associatedprimary consumers at a basin scale. The authors noticed a reorganiza-tion of the meso-zooplankton communities related to changes in phy-toplankton seasonality and hence, have assumed a deep restructuringof the throphodynamic of the North Atlantic ecosystems. However,owing to the inherent difficulty in exhaustively sampling the globalocean (Richardson and Poloczanska, 2008), only few advances inthe comprehension of global marine phytoplanktonic seasonality

: +33 4 99 57 32 95..

rights reserved.

al., Monitoring marine phy

have been made because of the lack of in situ exhaustive time series(Parmesan & Matthews, 2006).

Since the advent of remote sensing in biological oceanography asynoptic and dynamic picture of marine vegetation has become avail-able (Hovis et al., 1980). The very large amount of data that has beenprogressively gathered, has allowed for global and generic studies ofthe different temporal modes of chlorophyll-a concentration (Chla).This has included analysis of long term trends, decadal, inter annualand seasonal variability (Behrenfeld et al., 2006; Martinez et al.,2009; Vantrepotte & Mélin, 2011). Despite the fact that seasonal fluc-tuations of primary producers represent the main variability in globalChla patterns, only a few studies have focused on this type of varia-tion at local or global scale. Dandonneau et al. (2004) measured thestarting month, intensity and duration of the blooming period usingnumerical criteria that study the shape of Chla time series derivedfrom the global ocean color record. In a more statistical manner,Yoder and Kennelly (2003) characterized the global seasonal oscilla-tion of Chla using empirical orthogonal function (EOF) analysis andmapped the main seasonal patterns obtained. These results were sub-sequently refined by Vantrepotte and Mélin (2009) who used a Cen-sus methodology to describe the typology of the seasonal variation ofthe Chla. They were able to show a strong influence of environmentalconditions on Chla seasonal variation. At a local scale, Platt andSathyendranath (2008) put forward several empirical indices of the

toplankton seasonality from space, Remote Sensing of Environment

http://dx.doi.org/10.1016/j.rse.2011.09.019

mailto:[email protected]

Unlabelled image


Unlabelled image

http://www.sciencedirect.com/science/journal/00344257


2 H. Demarcq et al. / Remote Sensing of Environment xxx (2011) xxx–xxx

pelagic domain for studying the dynamics of marine ecosystems.These indices summarize the different types of marine vegetationseasonal blooms and include timing period, duration, and amplitudeof the bloom.

In the study reported herein, using remotely sensed data from theSea-viewing Wide Field-of-View Sensor (SeaWiFS) and Phytoplank-ton Functional Type (PFT) relative occurrence from the PHYSAT data-set (also derived from SeaWiFS data), we investigate the seasonaloscillations of marine primary producers. This is achieved throughthe application of a newly developed procedure, named SeasonalWavelet analysis Procedure (SWaP). The main goals of the studyare: (i) to determine, quantify and map the average intensity and du-ration of the phytoplankton seasonal oscillations using the Chla timeseries; (ii) to determine, quantify and map the average intensity andduration of the 4 PFTs (namely Diatoms, Synechococcus, Prochloro-coccus, and Nanoeucaryotes) extracted from the original PHYSATdataset (Alvain et al., 2008); (iii) to identify, map and describe themain global PFTs associations that contribute the most to the Chlaseasonal oscillation at a global scale, as well as their temporal succes-sion. Based on the results, the average seasonal parameters (intensityand duration) of Chla and PFTs occurrences are provided and dis-cussed to examine the different seasonal typologies in the globalocean. In addition, the latitudinal gradient of seasonal parameters ofthe Chla identified is discussed by relating the spatial distributionand temporal succession of the main PFTs associations detected, andthe influence of the environmental conditions.

2. Materials and methods

2.1. Remote sensing data

We use the most homogeneous and widely used global data set ofChla available from space, i.e. those processed from SeaWiFS from Oc-tober 1997 to December 2007. We use the 2009.1 reprocessing dataset made available from NASA at http://oceandata.sci.gsfc.nasa.gov/.

The data set is averaged at a monthly time step and at a spatialresolution of 1° of latitude and longitude. Because of the lack of obser-vations during the winter periods due to the light limitation and ahigh cloud cover, only latitudes between 50°N to 50°S are used inthe further analyses. To allow a worldwide comparison between theseasonality of the phytoplankton biomass and to account for thevery large range of values, the time series of each geographical cellare normalized according to Legendre and Legendre (1998).

In order to test the statistical robustness of the detection of theseasonal intensity, a second time series of fortnightly averages hasbeen generated.

2.2. Phytoplankton functional groups: PHYSAT

The PHYSAT approach is based on the identification of specific sig-natures in the water leaving radiance measurements spectra (nLw)from ocean color sensor measurements (Alvain et al., 2005, 2006).This empirical method is based on the comparison of two kinds of si-multaneous measurements: from remote sensing normalized waterradiance measurements and in situ measurements of specific phyto-plankton pigments. It has been shown that some phytoplanktongroups can be associated with specific normalized water leaving radi-ance (nLw*) when they are dominant (in terms of specific pigmentsconcentration). The nLw* are defined as the second order variabilityof the satellite signal, obtained by dividing the actual nLw by amean nLwref model which depends only on the standard Chla. A setof criteria has been defined to characterize each group sampled insitu, according to its nLw* spectrum. These criteria can be applied toglobal daily remote sensing observations to get global synthesis ofthe most frequent group of dominant phytoplankton (Alvain et al.,2008). When no group prevails over a month, pixels are classified as

Please cite this article as: Demarcq, H., et al., Monitoring marine phy(2011), doi:10.1016/j.rse.2011.09.019

‘unidentified’. This information can also be used to obtain monthlymaps of dominant group frequencies, used in this study. For eachgroup and pixel, group frequencies are determined by the relativenumber of daily detection to the number of daily detection detectableover the month.

In this study, the PHYSAT database is extracted from the work ofAlvain et al. (2008) and is used without any numerical transforma-tion. The database is used at the same resolution as the Chla, i.e.from 50°S to 50°N of latitude at a spatial resolution of 1° and at amonthly average time resolution, between October 1997 and Decem-ber 2007. Only the relative contribution of validated PFTs (more than50% similarity) in both spatial distribution and temporal fluctuationagainst in situ observation are considered in the study (Alvain et al.,2008). The Coccolithophorids genus group is not retained becauseits spatial distribution and temporal fluctuation are underestimatedby the PHYSAT methodology. The Phaeocystis-like group is notretained because it has yet to be validated against in situ observa-tions. Therefore, to avoid bias in the methodology, only the contribu-tions of four validated PFTs are considered: Diatoms (DIA),Nanoeucaryotes (NAN), Prochlorococcus (PRO) and the Cyanobacte-rium genus Synechococcus (SLC).

2.3. Seasonal Wavelet analysis Procedure: SWaP

The classic techniques used to identify the seasonality of phyto-plankton at regional scale have been mainly empirical (Platt &Sathyendranath, 2008). The numerical procedure of the study pro-posed is based on a wavelet analysis that enables study of the fre-quency composition of the time series (Torrence & Compo, 1998).The methodology can be summarized into 4 main steps, as shownon Fig. 1.

2.3.1. Wavelet transformation and power spectrumThe Continuous Wavelet Transform (CWT) analysis allows for

the description of the variability of a time series in both time andfrequency domains (Cazelles et al., 2008). The methodology cancope with aperiodic components, noise and transient dynamics(Daubechies, 1992; Lau & Weng, 1995; Torrence & Compo, 1998).The full methodology for CWT is fully described in Royer andFromentin (2006). The CWT is based on the convolution productbetween the time series and a mathematical function that is dilatedand translated onto the signal (i.e. time series). As usual in ecolog-ical applications, we used the Morlet wavelet, a continuous andcomplex wavelet adapted to wave-like signals, which allowsextracting time-dependent amplitude for a continuous range of fre-quencies (Cazelles et al., 2008; Menard et al., 2007). The relative im-portance of frequencies may be represented in the time–periodicityplane to form the wavelet power spectrum on a 2D plot, performedon each geographical cell of our time series (see Fig. 1).

2.3.2. Extraction of the intensity of the seasonal oscillationFor each 1° geographical cell, the wavelet power spectrum is used

to approximate the intensity of the seasonal oscillation of the param-eter studied (Fig. 1). As abundantly described in the literature, phyto-plankton presents annual and bi-annual oscillation patterns (Mann &Lazier, 1996; Tomczak & Godfrey, 2003). Therefore, the waveletpower spectrum is averaged, respectively from 0.5 yr±0.25 and1 yr±0.25 (Fig. 1) to obtain the mean wavelet power for the seasonalperiods of 6-month and 1-year respectively.

2.3.3. Extraction of the duration of the seasonal oscillationTo quantify the duration of a seasonal oscillation in each geographical

cell, surrogatesmethodology is applied to detect significant highly recur-ring patterns from thewavelet power spectrum (i.e. seasonal oscillation)(Fig. 1). The statistics (mean and variance of the temporal series, powerspectrum, and complete empirical distribution) are computed for the


http://oceandata.sci.gsfc.nasa.gov/


&

Fig. 1. Main steps of extraction of the intensity and duration of the chlorophyll-aconcentration seasonality using the SWaP method, as detailed in the material andmethods section.

3H. Demarcq et al. / Remote Sensing of Environment xxx (2011) xxx–xxx

original data and surrogate distributions are generated, thus allowing fornull hypothesis testing (no variation of the distribution). Here a nullmodel ‘Type 1 surrogates’ is selected due to the recurring reddenednoise pattern of the environmental data studied (Cuddington & Yodzis,1999; Royer & Fromentin, 2006). The model generates reddened noise,which is the convention in trying to capture endogenous fluctuationsin marine populations (Royama, 1992). In this case, the linear propertiesof the time series are fully described by the power spectrum. Surrogatedata are therefore built by adding random phases in [0,2π] to the com-ponents of the Fourier transform of the observed time series, and thencomputing their inverse Fourier transform (Theiler et al., 1992). Theresulting surrogates are Gaussian and have the same mean, varianceand power spectrum as the original data. The generation of a large num-ber of surrogates (in this case, 1000 repetitions) allows for the deriva-tion of a distribution of the recurrence plot based statistics (Royer &Fromentin, 2006). The significant level of periodicity in the waveletpower spectrum is assessed in a non-parametric manner, i.e. by count-ing up the fraction of realizations that produce a value greater than theobserved one. According toMarwan et al. (2003), the residual probabil-ity of false rejection (α) is set at 5% while the threshold of significativevalue is computed as (2α−1−1). Mean significant frequencies from0.5 yr±0.25 and 1 yr±0.25 are computed for each geographical celland are assumed to fall within the average seasonal duration of thetime series oscillation at 1 year and 6 months (Fig. 2).

2.3.4. Mapping and summarizingThe SWaP methodology is applied to the Chla data and each of the

relative occurrences (in percentage) of the four PFTs detected byPHYSAT (DIA, NAN, PRO and SLC). The spatial results for the Chla atthe 6-month and 1-year periods are mapped on Fig. 2 with the


relationship between intensity and duration for both periods. Thespatial distribution of the relations between the annual average Chlaand its seasonal intensity is mapped on Fig. 3 for both periods. Thespatial relationships between the seasonal intensity and durationare summarized (Fig. 4), and the typical associated time series arepresented. The relative occurrence of the four PFTs and the associatedseasonal parameters are represented in Fig. 5. The contribution of themain PFTs associations to the seasonal oscillation detected by ordi-nary statistical procedure (PCA) is mapped, as well as the seasonaldynamic of each association for both hemispheres (Fig. 6). Finally,and because the spatial variability of the seasonal oscillations aremainly latitudinally structured, the relationships between the aver-age concentration of Chla and the seasonality at both periods are plot-ted separately for the three main oceanic basins (Fig. 7).

2.4. Seasonal dominance of phytoplankton functional types

A principal component analysis (PCA, Jolliffe, 1986) is used to iden-tify the dominant contribution of the PFTs to the seasonal oscillation ofthe Chla. Values of the relative contribution of the four PFTs derivedfrom the PHYSAT dataset and the Chla from the SeaWiFS dataset arecomputed for each geographical cell for each month between October1997 and December 2007 in a matrix Xn,m,p (n=123 months, m=4PFTs and Chla, and p=259200 geographical cells). A PCA is performedusing PFTs time series for each cell using the corresponding matrix Xwith the associated Chla imposed as a supplementary vector in theprocedure.

To identify the main PFTs association contributing to the seasonaloscillation of Chla, each component resulting from the PCA is correlatedwith the supplementary vector. The component exhibiting the highestsignificant correlation is selected (76.1% on PC1, 21.7% on PC2 and2.2% on PC3). The PFTs that contribute at least to 40% to the selectedcomponent are identified. Six main PFTs associations have been foundby the PCA. The spatial distribution of these six PFTs associationswhich contribute the most to the seasonal cycle of the Chla is repre-sented on Fig. 6.

2.5. Temporal successions of the phytoplanktonic functional types

Considering the spatial distribution of each of the six major phyto-plankton associations that dominate the seasonal Chla oscillation, thetemporal successions of the four PFTs are examined for each PFT associ-ation. Firstly, for each geographical cell, themonthly average climatolo-gy of the four PFTs and associated Chla (transformed in log10 (x+1)) iscalculated using the PHYSAT and SeaWiFS databases fromOctober 1997to December 2007. Secondly, the average relative contributions, as wellas the Chla, are calculated for each of the six main associations by spa-tially aggregating the corresponding monthly climatologies of thePFTs and the Chla for each association. In order to consider large scaleeffects, and particularly those associated with the continental masses,the spatial aggregation over the sixmain PFTs associations is performedseparately for both hemispheres. The temporal successions of the fourPFTs and Chla of each main association are presented in Fig. 6 (lowerpanel).

3. Results

3.1. Seasonality of chlorophyll-a concentration

The spatial distribution of the seasonal oscillation of the phyto-plankton biomass, characterized by the intensity and duration of the6-month and 1-year periods, are examined for the global oceanusing the SWaP procedure (Fig. 2). The decomposition of the tempo-ral signal into its two main periodicities is performed to highlight re-gions with single (1-year) and double (6-month) blooming periods.The resulting map of yearly seasonal oscillation shows well defined


image of Fig.�1


b

a

Fig. 2. Intensity (a) and duration (b) of the productive season for the 1-year and the 6-month periods computed from the SeaWiFS chlorophyll-a concentration for the period1997–2007, from 50°S to 50°N. The boundaries corresponding to the Longhurst provinces (Longhurst, 2007) are superimposed in black lines. The scatterplot shows the relationshipbetween both variables for the yearly (green) and biannual (red) oscillations.

a

b

Fig. 3. SeaWiFS chlorophyll-a concentration versus its seasonal intensity for the1997–2007 period, respectively for the annual (a) and the biannual (b) seasonality.



spatial patterns, mainly structured by latitude (Fig. 2a and b, upperframes) corresponding to average and below average levels of Chla(Fig. 3). The regions of high yearly seasonal oscillation, associatedwith durations>3 months are the lower latitudinal parts of the tem-perate zones (between 25° to 40° of latitude in both hemispheres), inagreement with the results of Vantrepotte and Mélin (2009). Second-ary areas of relatively high seasonality are the North equatorial Pacificcurrent (10–15°N), the tropical band of the Atlantic or Indian Oceans(0–20°S), and to a lesser extent, the band between 40°S and 50°S inthe Atlantic and Indian Oceans. The biannual oscillation (two bloomsper year) depicted by the 6-month periodicity (Fig. 2a and b, lowerframe) is characteristic of the equatorial Atlantic upwelling and thewestern equatorial Indian Ocean.

In addition, the spatial variability of the relationship between theseasonal intensity and the average level of Chla (Fig. 3) is explored.On an annual scale (Fig. 3a), the latitudinal structure is primarilymarked by the negative relationship of the low average Chla/high 1-year variability class (orange color) at gyre borders and in the south-ern Atlantic and Indian Oceans. This structure is perturbed by the richand stable areas (high Chla/low 1-year seasonality, green color), i.e.coastal upwellings, large river discharges of the Congo and Amazonrivers, and the northern jet streams where the seasonality is mainlybi-annual (Fig. 3b) in the high Chla/high 6-month seasonality class(red color). Specifically, the coastal upwellings appear clearly. Theequatorial upwellings (particularly in the Atlantic) are characterizedby a similar high semi-annual seasonality (because of their summer


image of Fig.�2

image of Fig.�3


a

b

Fig. 4. Seasonal intensity versus seasonal duration computed from the SeaWiFS chlorophyll-a concentration for the 1997–2007 period respectively for the (a) annual and (b) biannualseasonality. The color intensity represents the strength of the relationship between both variables, from their average values (same relative scale for both seasonality). The time seriespresent typical single seasonal signals (1, 3 and 7), typical frequencies associations (2,4,5 and 6) or even a quasi absence of seasonal signal (4). The background colors are the samethan for the 2D plots.


maximum) although at lower than average levels of Chla (orangecolor).

The relationship between intensity and duration is investigatedand mapped for the two periodicities (Fig. 4a and b), using a simplepartitioning and a color gradient where intensity is proportional tothe stretch of the relationship. On an annual scale (Fig. 4a), the twoseasonal parameters are highly correlated (r=0.83, pb0.01). There-fore, the related map exhibits two main seasonal patterns with re-gions of high seasonal intensity and duration (in red) and regions oflow seasonal intensity and duration (in blue). The yearly signal, char-acteristic of the transition between tropical and temperate latitudes,shows a nearly linear relationship for most of its range, from 3 to4.5 months (Fig. 2), corresponding to a 3–8 range of intensity values.The asymptotic maximum duration recorded is 4.5 months. Thismeans that up to 75% of the theoretical illuminated period can be sig-nificantly productive. The lower seasonal intensity of the annual sig-nal (range 1–3) shows a drop to 2 months that correspond to adifferent response, adapted to the lower light variability of the tropi-cal regions. Nevertheless, some areas located in a band at 10–15°Nand in temperate bands at 40–45° North and South, exhibit a low sea-sonal intensity with a high duration of bloom. In addition, there is no-ticeable differences in both intensity and duration in the equatorialand tropical bands of the three basins, with the Pacific characterizedby a wide latitudinal band of low intensity and low duration, (inblue) in opposition to the Atlantic and Indian oceans, where the sea-sonal signal is of high intensity and duration (in red), specially southof 10°S.

At the 6-month periodicity (Fig. 4b), the relationship between theseasonal parameters is fuzzier due to a decoupling of the two


variables (r=0.51, pb0.05). Consequently, the map of the intensity-duration relationship (Fig. 4b) reveals more heterogeneous patternsthan for the yearly oscillation, with a lower range value (lightercolor on the map). Simultaneously, high values of the seasonal pa-rameters are mainly located in the Atlantic, western Indian equatorialOcean and in the temperate latitudes (40–45°N). Low values of theseasonal parameters are located in the tropics and in the gyres(blue color). Some regions exhibit a high seasonal duration and amoderate intensity (green or light red color), mainly in the northernhemisphere between 30°N and 38°N. Patterns of high seasonal inten-sity and moderate duration (light orange color) are detected mainlyin the southern hemisphere in the gyres.

3.2. Seasonality of the phytoplankton functional types

The seasonal oscillations of the relative occurrences of PFTsdetected by the PHYSAT method are investigated. The mean spatialdistribution and the seasonal intensity and duration of the occur-rences of each group are represented on Fig. 5.

In general, DIA is unevenly distributed in the high latitudinal re-gions (>35°N and S, and in latitudes not considered in the study)and mainly in the Southern Ocean and in coastal regions (Fig. 5a).High annual seasonal intensity is mainly restricted to the SouthernOcean and the subarctic regions of the Pacific and Atlantic Oceans.Nevertheless, DIA presence is always characterized by a long cycle(>3 months), even in low occurrence areas as in the tropics (ArabianSea and coastal upwellings). The NAN group globally coexists withDIA but also occurs in the equatorial Atlantic and Western Australia.It shows a high seasonal oscillation in the Southern Ocean, and in


image of Fig.�4


a b c

Fig. 5. Occurrence (a) and seasonality of the PFTs of the PHYSAT classification of Alvain et al. (2005), as computed by the SWaP method for the 1-year and 6-month periods. The seasonal signal is separated in its intensity (b), and duration(c) expressed in months.

6H.D

emarcq

etal./

Remote

SensingofEnvironm

entxxx

(2011)xxx–xxx

Pleasecite

thisarticle

as:Dem

arcq,H.,et

al.,Monitoring

marine

phytoplanktonseasonality

fromspace,

Remote

Sensingof

Environment

(2011),doi:10.1016/j.rse.2011.09.019

image of Fig.�5


25o

25o

oo

oo

Fig. 6. Spatial distribution of the main associations of Phytoplankton Functional Types that predominantly contribute to the seasonal oscillation of chlorophyll-a concentrationdetected by ordinary statistical procedure (PCA), as described in the material and methods section. The time series show the average monthly dynamic of each of the 4 Phytoplank-ton Functional Types and chlorophyll-a concentration for each association, separately for both hemispheres.


specific regions, namely the North Pacific polar front and the subtrop-ical fronts.

The Prochlorococcus and the Synechococcus are genii typical oftropical areas and particularly of the oceanic gyres. PRO dominatesthe Atlantic gyres and the tropical part of the other oceanic gyreswhile the presence of the SLC is mainly located in the higher latitudi-nal parts of the main oceanic gyres. Despite differences in their spatialdistribution (Fig. 5a), the seasonal oscillations of PRO and SLC repre-sented by the seasonal parameters (Fig. 5b and c) are very similar inboth intensity and location. The highest significant seasonal patternsare located in the external bounds of oceanic gyres, at 30° Northand South.

3.3. Contribution of the PFTs to the seasonality of the chlorophyll aconcentration

The spatial distribution of the main associations of PFTs contribut-ing the most to Chla seasonal oscillations is investigated using a PCA.Six major types of phytoplankton associations are detected (Fig. 6).Results reveal a poleward succession in the spatial distribution ofthe PFTs associations, varying between the different ocean basins.

In the equatorial regions of the ocean, the small size phytoplank-ton dominates. It is represented in the Equatorial Atlantic by the spa-tial succession of NAN, the NAN/SLC association, and the SLC. A


similar gradient is observed in the monsoon region of the IndianOcean. The PRO/SLC association dominates in the Pacific Ocean fromthe equatorial divergence to the warm pool, but also in the Northand South tropical gyres. This later association is detected in the trop-ical gyres of all oceans. At the boundary of all tropical gyres, a succes-sion characterized by SLC and SLC/NAN association is detected.Alongside these associations, at higher latitudes, NAN is either domi-nant or associated with PRO. At the highest latitudes examined in thisstudy (40–50°), combinations of NAN and NAN/DIA are identified.

3.4. Temporal succession of phytoplankton functional types

Phytoplankton temporal succession during an average year is in-vestigated using the same PHYSAT data at a monthly scale. Themonthly climatologies of each geographical cell of the four PFTs andassociated Chla are identified using the spatial distribution of thedominant groups detected in Fig. 6. The temporal successions ofeach PFT and Chla are plotted for the Northern and Southern hemi-sphere for each of the six PFTs associations.

The first group (red box, Fig. 6) located in latitudes>40° and inareas of high productivity is dominated by the DIA and NAN types.The temporal succession is characterized by an increase in the relativeoccurrence of DIA during spring (associated with PRO and SLC at alower level of occurrence), a dominance of DIA in summer (especially


image of Fig.�6


Chl

orop

hyll-

a (m

g m

-3)

Chl

orop

hyll-

a (m

g m

-3)

Chl

orop

hyll-

a (m

g m

-3)

Bio

mas

s se

ason

al in

tens

ity (

SW

aP)

Bio

mas

s se

ason

al in

tens

ity (

SW

aP)

Bio

mas

s se

ason

al in

tens

ity (

SW

aP)

a

b

c

Fig. 7. Mean latitudinal structure of the Phytoplankton biomass and seasonal oscillation for the three oceans from 50°S to 50°N and for the following parameters: chlorophyll-aaverage, seasonality at 1 year and 6-month (maps in Fig. 2a), as computed with the SWaP procedure. The values of the 6-month seasonality have been doubled to match thoseof the 1-year variability.


at the Southern hemisphere) and a decrease of DIA during autumn.Finally, DIA types are replaced by NAN, which dominate throughoutwinter. A similar temporal pattern was detected in the NAN group(yellow box, Fig. 6) which is located next to the DIA/NAN group, butat lower latitudes. However, in this last group, NAN dominates thewhole year, even during the increase of DIA during spring and sum-mer, nevertheless at a lower relative contribution than PRO and SLC.The NAN/PRO (orange box, Fig. 6) located between 25° to 40° of lati-tude and next to the two latest groups is dominated by PRO fromspring to autumn and by NAN in winter. In addition, a peak of SLC isdetected during spring and autumn in the southern hemisphere.

In less productive regions, mainly in the tropics and along theequator, SLC is generally detected alone or associated with NAN orPRO. The association between SLC and NAN (light blue box, Fig. 6),


mainly detected in subtropical regions, is characterized by a domi-nance of SLC during both Boreal and Austral summer, associatedwith a slight increase of DIA and a dominance of NAN during winter.The SLC group located at the boundary of the tropical gyres shows atemporal succession mainly driven by SLC and PRO with few varia-tions during the year. Indeed, the SLC group (dark blue box, Fig. 6)is dominated by SLC most of the year, except during boreal springwhen PRO dominates with a slightly higher relative occurrence, andduring austral winter when PRO and NAN dominate. In subtropicalgyres and Pacific tropical areas, SLC and PRO dominate whole yearwhereas few seasonal variations in DIA and NAN are detected. Differ-ences between both hemispheres are also found in the SLC/PROgroup: PRO dominates the Northern Atlantic gyres (Fig. 5) and con-tributes highly to other tropical gyres where SLC dominates.


image of Fig.�7



4. Discussion

4.1. Limitations of data and methods

The phytoplanktonic production of the ocean, based on the photo-synthetic efficiency of the phytoplanktonic biomass, is in tight equi-librium with the joint availability of light, nutrients and the stabilityof the water column (Sverdrup, 1953). Thus, the variability of theChla is usually used as a marker of potential changes of environmen-tal conditions (Ducklow, 2003). Phytoplankton biomass via its effecton the ocean color is the primary measurement made from spaceand provides the basis of most of the production models. Despitethe varying C:Chla ratio, the Chla time series derived from the Sea-WiFS sensor is considered as a reasonable index to approximate thetotal phytoplankton biomass in many global studies.

In this study, several biases of the remote sensing data have to beconsidered to apply our spectrally-based statistical techniques. First,data density limitation is extremely severe in high latitudes where thenumber of clear day drastically decreases in winter because of the con-jugate effects of light and cloud coverage (Demarcq, 2009). Given thatthemethodology used does not tolerate any interruption in the time se-ries, only geographical cells located between 50°N and 50°S are consid-ered in the study. Second, the main focus of the study is to investigatethe seasonal cycle of the Chla. Therefore, a comparison between resultsbased on fortnightly and monthly Chla time series has been performedto test if the seasonal cycle is equally detected. Results confirmed thatthe use of monthly average data set is of sufficient resolution to capturethe seasonal cycle at a global scale between 50°N and 50°S despite aweak gain in spatial coverage, particularly in the high latitudes, whencompared with data of fortnightly resolution. However, the spatial pat-terns of the seasonal parameters and their values are extremely similar.Themonthly composites are hence considered as an adequate time stepof observation for the study because they offer a satisfying compromisebetweendata density and the need of a global description of the season-al dynamics of the phytoplankton. Third, because of the high correlationbetween seasonal intensity computed on the original SeaWiFS databaseand the mean level of Chla (r=0.79 pb0.001), the Chla dataset hasbeen normalized (Legendre & Legendre, 1998) to capture the spatialgradient of the changes in seasonal parameters.

Despite the fact that no validation or reprocessing works has beenperformed on the PHYSAT methodology in the framework of thisstudy, it might be considered that a part of the limitation of someanalysis (see materials and methods, 3.2 to 3.4) can be attributed tothe PHYSAT inherent errors. Indeed, the original PHYSAT outputsused in this study have been confronted to in situ samples of phyto-plankton abundance from monthly maps at local and large scales(Alvain et al., 2008). The results have shown that at the monthlyand yearly frequencies spatial distributions and temporal successionsof PFTs appear to be in agreement with in situ observations and pre-vious studies (monthly average percentage of recognition in Alvain etal., 2008). In addition, further validations at a daily resolution haveshown a good recognition of DIA and NAN respectively at a level of82% and 87% (Alvain, pers. Com.). Limitation comes from PRO andSLC confusion owing to the close characteristics (geographical distri-bution, size and specific signal in PHYSAT) of these two groups (re-spectively 62 and 68% of good recognition, and 28% of confusionpixels between the two) (Alvain et al., 2008). This validation hasshown that PHYSAT leads to only a limited number of wrong identifi-cations. However, specific groups, for example the Coccolithophoridsgenus, exhibit a high wrong percentage of recognition. Indeed, theircells have a typical spherical shape made of calcium carbonate plates.At the end of a bloom, these plates are released and produce the well-known milky color in ocean color imagery. So, Coccolithophoridsblooms are theoretically easy to detect and have been added to thePHYSAT method. However, it turned out that the SeaWiFS data ar-chive used, i.e. Level 3 daily GAC products, are screened to remove


Coccolithophorids blooms because of a threshold on nLw measure-ments during the data processing itself. As a consequence, the Cocco-lithophorids genus detected by the PHYSAT method has beenremoved from our analysis.

4.2. Latitudinal associations

The main forcing factors governing the ocean primary productivityare light and nutrients, followed by temperature and mixed layerdepth (Behrenfeld, 2010; Sverdrup, 1953), all remainingmostly depen-dent on solar irradiation. Specific regional environmental conditionsoften disturb these major forcings through local winds in coastal up-welling areas, variation in mesoscale activity (i.e. eddies), streams,river discharges and other processes detectable through frontal struc-tures (e.g. Oceanic Polar Front, see Dietrich, 1964).

A direct consequence of the dominant role of the latitudinal distri-bution of the solar irradiation is the observed latitudinal structure ofthe phytoplanktonic biomass seasonality (Figs. 2a and 7), separatedby sharp transition areas. Because important spatial differences arefound between the different oceanic basins regarding the distributionof the phytoplankton biomass and of its main seasonal variabilitymodes (Figs. 2a and 7), patterns corresponding to each oceanicbasin are discussed here separately.

The Pacific Ocean probably shows the simplest spatial organiza-tion: the Chla based biomass (plain line, Fig. 7a) is characterized bythree maxima, well separated by the oceanic gyres, with a highlevel of symmetry from the equator to poleward areas, due to thewidth of the basin. The main difference between hemispheres is relat-ed to the highest surface of the continental shelves in the northernhemisphere that favors the global productivity, between 30°N and45°N. The intensity of the seasonal signal (1Y curve) shows largeand similar maxima between 20° and 40° of latitude in both hemi-spheres, that corresponds to the septentrional border of the oceanicgyres (centered around 20-25°). This pattern is primarily due to thehigh latitudinal gradient in the difference of incoming solar radiationbetween winter and summer that rapidly increases from the tropicsto poleward regions. Poleward of 35° of latitude, despite even higherseasonal light variability, the phytoplanktonic 1Y curve decreasesagain for light adaptation reasons. It is therefore not surprising to ob-serve that this area of pronounced variability also correspond to amajor boundary observed between PFTs occurrence at this latitudes(Fig. 5), especially between the PRO and NAN groups. More than thedecreasing light, that is favorable to the PRO group, the main reasonof this decrease in variability is the progressive higher nitrate avail-ability and probably the lower temperature, that favor the NANgroup (Partensky et al., 1999). Secondary peaks of variability are ob-served at 10° North and South, corresponding to the lower latitudelimits of the gyres. Their intensity is weaker because of the low sea-sonal variability of sun light in this latitude. The signatures of thecoastal upwellings are not visible for the Pacific because of their re-stricted areas. The 6-month variability of the Chla, that typicallycatches autumnal and spring peaks of production, is extremely lowin the Pacific with a smooth maximum at 40°S and 40°N.

The Atlantic Ocean (Fig. 7b) displays similar patterns of phyto-plankton biomass, at the same latitudes, with secondary peaks of var-iability due to the output of the major tropical rivers (e.g. the AmazonRiver at 4°S) as well as a more visible signature of the equatorial up-welling, around 5°N. The main difference with the Pacific Ocean interms of seasonal variability is the much larger spatial extent of theannual seasonality in the Southern Atlantic. Seasonality is very pro-nounced from the equator to 30°S and then decreases continuouslyup to 40°S. The latitudinal pattern of annual seasonality in the North-ern Atlantic is very similar to the Pacific one. In contrast, the AtlanticOcean is characterized by very high importance of the semi-annualseasonality associated with the northern and southern borders ofthe gyres, in accordance with the yearly seasonality. This is the




signature of the autumnal blooms, well known in the northern Atlan-tic, but also present with a similar intensity in the Southern Atlantic.

The Indian Ocean is characterized by a relatively low level of phy-toplankton biomass, similar to the Pacific, except in the Arabian Seaduring the Indian monsoon, where the high productivity signatureexponentially increases north of 10°N. The annual seasonal variabilityis very wide, from 5 to 35°S. However, the semi-annual seasonal var-iability is very low, except in the Arabian Sea, where the variability islikely under-estimated because of the cloud cover that prevents con-tinuous estimation.

4.3. Temporal and spatial phytoplankton successions linked to theenvironment

According to previous studies on phytoplankton abundance andspecies diversity (Boyce et al., 2010), environmental conditions di-rectly affect the metabolism of phytoplanktonic species (growth, pro-duction, reproduction) and hence, their spatial distributions andtemporal variations (Longhurst, 1995). Indeed, the dominant param-eters impacting phytoplankton abundance are nutrient concentrationand incoming light irradiance that control photosynthesis and there-fore new production (Platt et al., 1991). At a lower level of regulation,temperature and water column stability directly affect the growthand survival rates of phytoplanktonic species (Sathyendranath et al.,1995). The global spatial distribution and temporal fluctuation ofthese environmental conditions being highly variable, it is not sur-prising to observe a variation in the spatial distribution of theseasonal patterns of the phytoplanktonic biomass (Figs. 2 and 4),phytoplanktonic species association and temporal succession (Fig. 6).

The relatively good match between the spatial distribution of thebiogeochemical provinces (BGCP) of Longhurst (2007) and the sea-sonal patterns detected (Figs. 2and 4) confirmed the strong influenceof environmental conditions on the seasonal fluctuation of phyto-plankton. Clear overlaps are observed between areas with high sea-sonal biomass oscillations and highly seasonal subtropical or polarfront provinces (North Pacific Polar Front, North Atlantic Subtropicalprovinces or Subtropical Convergence areas), and low seasonal bio-mass oscillation and environmentally stable provinces (North andSouth Pacific subtropical gyres provinces and South Indian subtropi-cal gyre province) (Figs. 2 and 4). This relative spatial match needsto be attributed to the fact that each BGCP represents characteristicand relatively homogeneous environmental envelops (i.e. restrictedinterval of variation of environmental parameters affecting phyto-plankton growth, see Longhurst, 2007) where only a small numberof species can occur and fluctuate in accordance to their environmen-tal tolerance (Hutchinson, 1957; Longhurst, 2007). According to thevariation of the environmental parameters during an average yearin a given province, each PFT fluctuates around optimal values of sur-vival and growth in a form of seasonal succession of species (Fig. 6,lower panel). As each PFT is characterized by different cell sizes andChla content, their succession directly impacts the total Chla biomassobserved by remote sensing and drives the seasonal fluctuations ofChla, as detected in our analyses (Figs.2 and 4) (Alvain et al., 2008;Longhurst, 1995; Platt et al., 1991).

More precisely, starting from the equatorial areas of the oceanwhere equatorial currents and countercurrents occur (Tomczak &Godfrey, 2003), a low seasonality of the Chla is observed except in spe-cific areas such as the eastern Pacific and equatorial Atlantic and the Ara-bian sea where a high 6-month seasonality of Chla is found. Theseparticular areas are equatorial upwelling regions (Longhurst, 2007) char-acterized by a nutrient repletion caused either by the equatorial diver-gence of the surface water mass (in the Pacific and Atlantic) or by theinfluence of a particular wind regime (Indian ocean) caused by themon-soon flux (Banse, 1984) that enhance a strong upward flux of deep nutri-ent rich water and a destratification (Tomczak & Godfrey, 2003).According to the literature, these particular nutrient replete conditions


associated to high and constant irradiance allow for the growth ofmicro-phytoplankton and nanophytoplankton flagellates (e.g. DIA associatedwith a dominance of NAN, Fig. 5 and group NAN on Fig.6) that increasesthe total Chla during the upwelling period (Longhurst, 1993). Thus, theAtlantic and Indian equatorial upwelling are mainly characterized by adominance of NAN and a summer peak of DIA (Dugdale & Wilkerson,1998) while other adjacent Atlantic and Indian Ocean equatorial areasare mainly dominated by NAN group during winter and by the SLC dur-ing summer due to the increase of the stratification (Longhurst, 1993).However, the eastern Pacific equatorial divergence area is dominatedby the association of SLC and PRO (Fig. 6), with different balance oftheir respective occurrence levels according both hemispheres. This dif-ference in the PFTs dominance accordingbasins andhemisphere is in ac-cordance with the in situ observations of Chung et al. (1996) whoattribute the low abundance of micro and nano phytoplankton to theiron limited conditions and the quick grazing of zooplankton (Landryet al., 1997).

The central areas of the subtropical gyres, located at 15-25° of lati-tude, are well characterized by the PRO/SLC association (Fig. 6). Theseareas show weak seasonal variations with constant warm temperatureand incoming light irradiance. Moreover, owing to the permanent andstrong stratification caused by the low intensity of the local winds(high intensity of thermocline, see Reygondeau& Beaugrand, 2011) nu-trients replenishment from deep water is prevented and consequentlythe low nutrient concentration in the epipelagic layer limits the prolif-eration of primary producers, especially micro and nano phytoplanktonwhich have a growth highly restricted by the nutrient limited condi-tions. These particular environmental conditions are also characteristicsof the whole equatorial Pacific (i.e. the warm pool) characterized by animportant stratification with a deep and permanent nutricline causedby the currents and thermohalinemain regional features (see Le Borgneet al., 2002). Thus, this highly nitrate-limited conditions only allow theoccurrence of bacterial photosynthetic picophytoplankton species suchas PRO and SLC that benefits from recycled nutrients (Le Bouteiller et al.,1992; Agawin et al., 2000). As a result, these regions are characterizedby the lowest seasonal parameters (low intensity and low duration,Fig. 4) at both 1-year and 6-month periods, and no temporal speciessuccession is detected (Fig. 6, group SLC/PRO). Nevertheless, the differ-ences in specific nutrient limitations that occur between oceanic basins(especially iron limitation in the Pacific Ocean; Behrenfeld et al., 1996),lead to differences in the temporal domination of the area.While Pacificgyres are dominated by SLC whose growth is not limited by the lack ofiron (Johnson et al., 2010; Landry et al., 1997;Wells et al., 1994), the At-lantic gyres show a similar domination of the PRO group, which is high-ly iron limited (Fig. 5) and consequently more restricted to the centralpart of the gyres, especially in the North Atlantic, where the iron supplyfrom the Amazon river and the Saharan dust (Jickells et al., 2005) is veryimportant.

Poleward the anticyclonic subtropical and beyond the cyclonicpolar gyres lie areas gyres (between 25° and 40° N/S) characterizedby a high yearly seasonal oscillations (Figs. 2 and 4) which match ex-tremely well with the NAN/PRO association (Fig. 6). These regions arenamed transition zones (Longhurst, 2007; Roden, 1970) andmark theboundary where seasonal oscillations of incoming light irradiancepredominantly influence the oceanic environmental conditions suchas temperature and mixed layer depth (Tomczak & Godfrey, 2003).Thus , in winter when light and air temperature decrease, the watertemperature of the upper water column decreases, reducing the strat-ification (i.e. deepening of the mixed layer depth) and allowing weakmixing of the water column. During this period, only nanophyto-plankton (here NAN) occurs in these regions (Fig.6) due to the impor-tant nutrient repletion and their metabolic capacities to proliferate invarious light levels and weak stratification condition (Nommann &Kaasik, 1992). In spring, the temperature and the stratification raisesproportionally with the increasing irradiance. These conditions, asso-ciated with the high nutrients concentration brought to the surface




during the winter (i.e. nitrate and silicates), favor the occurrence ofmicrophytoplankton such as Diatoms, causing a significant increasein Chla (the “spring bloom”, Sverdrup, 1953). During summer, thewarm temperature combined with a strong stratification allowsonly cyanobacteria to occur due to the nutrient limitation (nitrateor phosphate) depleted by the NAN and DIA during the spring phyto-planktonic bloom. As PRO has a wider interval of tolerance to nutrientlimitation and temperature but also can proliferate in lower light con-ditions than SLC according to in situ observations and experimentaltests (Moore et al., 1995; Partensky et al., 1999), the group dominatesthese transition areas during this period.

Latitudinally beyond the previous transition area (between 40°and50° N/S), temperate and sub-polar areas are mostly characterized by6-month oscillation of Chla (Figs. 2 and 4). These areas exhibit colderwater and more pronounced but similar environmental oscillationsthan the previous transition area. Indeed, the seasonal variation ofincoming light irradiance is amplified and the effect of westerly windstress on the water column unstability is more pronounced allowinga deep winter mixing (Longhurst, 2007). Thus, these areas can becharacterized by a deep winter mixing (mixed layer depth can befind over 500 m; Reygondeau & Beaugrand, 2011; de Boyer Montégutet al., 2004) which brought to the surface a high concentration ofnutrient allowing the proliferation of nanophytoplankton whichsupport the important turbulence and the low light level conditions(Nommann & Kaasik, 1992). During spring, the increase of incominglight radiation combined with a progressive decrease of turbulenceallows the quick growth of microphytoplankton (Strass & Woods,1991). However, the relative occurrence of the DIA group is highlylinked to the silicate and nitrate concentration (Taylor et al., 1993).Due the regional nutrient specificity, this period is dominated eitherby NAN and SLC or the DIA group. This high amplitude spring bloomis quickly followed by a period of summer oligotrophy characterizedby a nutrient limitation and a strong stratification (Reygondeau &Beaugrand, 2011) where bacterioplankton (SLC and PRO) andnanophytoplankton (NAN) mostly contribute to the total Chla (Strass& Woods, 1991). During end of summer or beginning of autumn, dueto the increase of westerly winds and the progressive deepening ofthe mixed layer depth, an autumnal bloom can be detected duringthis period in the temperate and sub-polar area (Strass & Woods,1991). Indeed, contrary to the transition area, the temperate andsub-polar regions exhibit an autumnal destratification period thatenables high vertical nutrient fluxes associated to a second bloomingperiod of microphytoplankton and/or Coccolithophorids (Longhurst,2007; Reygondeau & Beaugrand, 2011). Therefore, the temperateand sub-polar areas are mainly characterized by the NAN/SLC associa-tion in regions with low silicate concentration and further polewardby the DIA/NAN association.

5. Conclusion

In the context of global climate change, the oceanic realm is in directinteraction with atmospheric radiation forcing and this might affect itsbiogeochemical composition. Expected changes of the oceanic environ-mental conditions (stratification, temperature and wind regimes)might result in variation in the timing, duration and intensity of phyto-plankton blooms caused by non-optimal growth conditions. Thechanges in the environmental conditions are expected to enhance pole-ward spatial shift in the abundance of some phytoplankton species thatmay proliferate in higher latitudes where environmental conditionsmay better fit with their optimal environmental tolerance ranges. Thereorganization of phytoplanktonic communities combined with thealteration of the synchrony between primary producers and associatedconsumers might lead to trophic mismatch and a possible modificationof the trophodynamic of several marine ecosystems. Thus, average sea-sonal parameters describing the blooming period (i.e. intensity, dura-tion, phytoplanktonic species composition and temporal successions)


provided in the study for each geographical cell between 50°S to 50°Nof latitude, are needed. This might be of interest to detect and under-stand any future anomaly in the seasonal oscillations of primary pro-ducers. Furthermore, the SWaP developed in this study can beapplied to different environmental parameters and marine organismsto characterize their seasonal oscillations during a period and detectchanges in seasonal parameters by comparing different periods.

Acknowledgements

We thank our colleagues Bernard Cazelles, Jean-Marc Fromentin andOlivier Maury for their helpful comments. We thank the NASA SeaWiFSproject and the NASA/GSFC/DAAC for producing and granting continuousaccess to the SeaWiFS data. Finally, the authors thank Lauren Biermannand Hayley Ever-King for their helpful contribution to the edition of themanuscript.

References

Agawin, N. S. R., Duarte, C. M., & Agusti, S. (2000). Nutrient and temperature controlof the contribution of picoplankton to phytoplankton biomass and production.Limnology and Oceanography, 45, 1891-1891.

Alvain, S., Moulin, C., Dandonneau, Y., & Bréon, F. M. (2005). Remote sensing ofphytoplankton groups in case 1 waters from global SeaWiFS imagery. Deep-SeaResearch I, 52(11), 1989–2004.

Alvain, S., Moulin, C., Dandonneau, Y., & Loisel, H. (2008). Seasonal distribution andsuccession of dominant phytoplankton groups in the global ocean: A satelliteview. Global Biogeochemical Cycles, 22(3), GB3001.

Alvain, S., Moulin, C., Dandonneau, Y., Loisel, H., & Bréon, F. M. (2006). A species-dependent bio-optical model of case I waters for global ocean color processing.Deep-Sea Research Part I, 53(5), 917–925.

Banse, K. (1984). Overview of the hydrography and associated biological phenomenain the Arabian Sea, off Pakistan. In B. U. Haq, & J. Milliman (Eds.), Marine Geologyand Oceanography of the Arabian Sea and Coastal Pakistan (pp. 271–303). NewYork: Van Nostrand Reinhold.

Behrenfeld, M. J. (2010). Abandoning Sverdrup's critical depth hypothesis on phyto-plankton blooms. Ecology, 91(4), 977–989.

Behrenfeld, M. J., Bale, A. J., Kolber, Z. S., Aiken, J., & Falkowski, P. G. (1996). Confirma-tion of iron limitation of phytoplankton photosynthesis in the equatorial PacificOcean. Nature, 383, 508–511.

Behrenfeld, M. J., O'Malley, R. T., Siegel, D. A., McClain, C. R., Sarmiento, J. L., Feldman, G. C.,et al. (2006). Climate-driven trends in contemporary ocean productivity. Nature, 444(7120), 752–755.

Boyce, D. G., Lewis, M. R., & Worm, B. (2010). Global phytoplankton decline over thepast century. Nature, 466(7306), 591–596.

Cazelles, B., Chavez, M., Berteaux, D., Menard, F., Vik, J. O., Jenouvrier, S., et al. (2008).Wavelet analysis of ecological time series. Oecologia, 156(2), 287–304.

Chung, S. P., Gardner, W. D., Richardson, M. J., Walsh, I. D., & Landry, M. R. (1996). Beamattenuation and micro-organisms: spatial and temporal variations in small particlesalong 140 degrees W during the 1992 JGOFS EqPac transects. Deep-Sea ResearchPart II, 43(4–6), 1205–1225.

Cuddington, K.M., & Yodzis, P. (1999). Black noise and population persistence. Proceedingsof the Royal Society of London - Series B: Biological Sciences, 266(1422), 969–973.

Dandonneau, Y., Deschamps, P. Y., Nicolas, J. M., Loisel, H., Blanchot, J., Montel, Y., et al.(2004). Seasonal and interannual variability of ocean color and composition ofphytoplankton communities in the North Atlantic, equatorial Pacific and SouthPacific. Deep-Sea Research Part II, 51(3), 303–318.

Daubechies, I. (1992). Ten lectures on wavelets, CBMS-NSF Regional Conference series inapplied mathematics. Philadelphia: Society for Industrial & Applied Math.

de Boyer Montégut, C., Madec, G., Fischer, A. S., Lazar, A., & Ludicone, D. (2004). Mixedlayer depth over the global ocean: an examination of profile data and a profile-based climatology. Journal of Geophysical Research C, 109, C12003.

Demarcq, H. (2009). Trends in primary production, sea surface temperature and windin upwelling systems (1998–2007). Progress in Oceanography, 83(1–4), 376–385.

Dietrich, G. (1964). Research in Geophysics: Oceanic polar front survey. Cambridge,Massachusetts USA: MIT Press.

Ducklow, H. W. (2003). Biogeochemical provinces: Towards a JGOFS synthesis. In M. J. R.Fasham (Ed.), Ocean Biogeochemistry: A Synthesis of the Joint Global Ocean Flux Study(pp. 3–17). Berlin Heidelberg: Springer-Verlag.

Dugdale, R. C., & Wilkerson, F. P. (1998). Silicate regulation of new production in theequatorial Pacific upwelling. Nature, 391(6664), 270–273.

Edwards, M., & Richardson, A. J. (2004). Impact of climate change on marine pelagicphenology and trophic mismatch. Nature, 430, 881–884.

Hovis,W. A., Clark, D. K., Anderson, F., Austin, R.W.,Wilson,W. H., Baker, E. T., et al. (1980).Nimbus-7 coastal zone color scanner: system description and initial imagery. Science,210(4465), 60–63.

Hutchinson, G. E. (1957). Concluding remarks. Cold Spring Harbor Symposium QuantitativeBiology, 22, 415–427.

Jickells, T. D., An, Z. S., Andersen, K. K., Baker, A. R., Bergametti, G., Brooks, N., et al.(2005). Global iron connections between desert dust, ocean biogeochemistry,and climate. Science, 308(5718), 67–71.




Johnson, Z. I., Shyam, R., Ritchie, A. E., Mioni, C., Lance, V. P., Murray, J. W., et al. (2010).The effect of iron- and light-limitation on phytoplankton communities of deepchlorophyll maxima of the western Pacific Ocean. Journal of Marine Research, 68(2), 283–308.

Jolliffe, I. T. (1986). Principal Component Analysis. New York: Springer-Verlag New YorkInc.

Landry, M. R., Barber, R. T., Bidigare, R. R., Chai, F., Coale, K. H., Dam, H. G., et al. (1997).Iron and grazing constraints on primary production in the central equatorial Pacific:an EqPac synthesis. Limnology and Oceanography, 42(3), 405–418.

Lau, K. M., & Weng, H. (1995). Climate signal detection using wavelet transform: howto make a time series sing. Bulletin of the American Meteorological Society, 76(12),2391–2402.

Le Borgne, R., Barber, R. T., Delcroix, T., Inoue, H. Y., Mackey, D. J., & Rodier, M. (2002).Pacific warm pool and divergence: temporal and zonal variations on the equatorand their effects on the biological pump. Deep-Sea Research Part II, 49(13–14),2471–2512.

Le Bouteiller, A., et al. (1992). Size distribution patterns of phytoplankton in the westernpacific: towards a generalization for the tropical Pacific Ocean. Deep Sea Research,39, 805–823.

Legendre, P., & Legendre, L. (1998). Numerical Ecology. The Netherlands: ElsevierScience B.V.

Longhurst, A. (1993). Seasonal cooling and blooming in tropical oceans. Deep SeaResearch Part I, 40(11–12), 2145–2165.

Longhurst, A. (1995). Seasonal cycles of pelagic production and consumption. Progressin Oceanography, 36(2), 77–167.

Longhurst, A. (2007). Ecological geography of the Sea. London: Academic Press.Mann, K. H., & Lazier, J. R. N. (1996). Dynamics of marine ecosystems: biological-physical

interactions in the oceans. Oxford: Blackwell Science, Incorporated.Martinez, E., Antoine, D., D'Ortenzio, F., & Gentili, B. (2009). Climate-driven basin-scale

decadal oscillations of oceanic phytoplankton. Science, 32(5957), 1253–1256.Marwan, N., Trauth, M. H., Vuille, M., & Kurths, J. (2003). Comparing modern and

Pleistocene ENSO-like influences in NW Argentina using nonlinear time seriesanalysis methods. Climate Dynamics, 21(3), 317–326.

Menard, F., Marsac, F., Bellier, E., & Cazelles, B. (2007). Climatic oscillations and tunacatch rates in the Indian Ocean: a wavelet approach to time series analysis. FisheriesOceanography, 16(1), 95–104.

Moore, L. R., Goericke, R., & Chisholm, S. W. (1995). Comparative physiology of syne-chococcus and prochlorococcus — influence of light and temperature on growth,pigments, fluorescence and absorptive properties. Marine Ecology Progress Series,116(1–3), 259–275.

Nemani, R. R., Keeling, C. D., Hashimoto, H., Jolly, W. M., Piper, S. C., Tucker, C. J., et al.(2003). Climate-driven increases in global terrestrial net primary productionfrom 1982 to 1999. Science, 300(5625), 1560–1563.

Nommann, S., & Kaasik, E. (1992). Hydrodynamical control of phytoplankton successionduring the vernal light-limited phase in the Baltic Sea.Marine Ecology Progress Series,84, 279–292.

Parmesan, C., & Matthews, J. (2006). Biological impacts of climate change. In M. J.Groom, G. K. Meffe, & C. R. Carroll (Eds.), Principles of conservation biology(pp. 333–360). Sunderland: Sinauer Associates, Inc.


Partensky, F., Blanchot, J., & Vaulot, D. (1999). Differential distribution and ecology ofProchlorococcus and Synechococcus in oceanic waters: a review. Bulletin de l'institutocéanographique de Monaco, 19, 457–476.

Platt, T., Caverhill, C., & Sathyendranath, S. (1991). Basin-scale estimates of oceanicprimary production by remote sensing: the North Atlantic. Journal of GeophysicalResearch, 96(C8), 15 147–15,159.

Platt, T., & Sathyendranath, S. (2008). Ecological indicators for the pelagic zone of theocean from remote sensing. Remote Sensing of Environment, 112(8), 3426–3436.

Reygondeau, G., & Beaugrand, G. (2011).Water column stability and Calanus finmarchicus.Journal of Plankton Research, 33(1), 119–136.

Richardson, A. J., & Poloczanska, E. S. (2008). Ocean Science: Under-resourced, underthreat. A policy forum recommending a global vision for improving and synthesisingclimate change impact studies in marine systems. Science, 320, 1294–1295.

Roden, G. I. (1970). Aspect of the mid-Pacific Transition Zone. Journal of GeophysicalResearch, 75, 1097–1109.

Royama, T. (1992). Analytical Population Dynamics. Chapman and Hall, London, UK:Kluwer Academic Publisher group.

Royer, F., & Fromentin, J. M. (2006). Recurrent and density-dependent patterns inlong-term fluctuations of Atlantic bluefin tuna trap catches. Marine Ecology ProgressSeries, 319, 237–249.

Sathyendranath, S., Longhurst, A., Caverhill, C. M., & Platt, T. (1995). Regionally andseasonally differentiated primary production in the North Atlantic. Deep-Sea ResearchPart I, 42(10), 1773–1802.

Schwartz, M. D., Ahas, R., & Aasa, A. (2006). Onset of spring starting earlier across theNorthern Hemisphere. Global Change Biology, 12(2), 343–351.

Strass, V. H., & Woods, J. D. (1991). New production in the summer revealed by the me-ridional slopeof thedeep chlorophyllmaximum.Deep-Sea Research Part I, 38(1), 35–56.

Sverdrup, H. U. (1953). On conditions for the vernal blooming of phytoplankton. ICESJournal of Marine Science, 18, 287–295.

Taylor, A. H., Harbour, D. S., Harris, R. P., Burkill, P. H., & Edwards, E. S. (1993). Seasonalsuccession in the pelagic ecosystem of the North Atlantic and the utilization ofnitrogen. Journal of Plankton Research, 15(8), 875–891.

Theiler, J., Eubank, S., Longtin, A., Galdrikian, B., & Farmer, J. D. (1992). Testing for non-linarity in time series: the method of surrogate data. Physica D, 58(1–4), 77–94.

Tomczak, M., & Godfrey, J. S. (2003). Regional Oceanography: An Introduction. Delhi:Daya Publishing House.

Torrence, C., & Compo, G. P. (1998). A practical guide to wavelet analysis. Bulletin of theAmerican Meteorological Society, 79(1), 61–78.

Vantrepotte, V., & Mélin, F. (2009). Temporal variability of 10-year global SeaWiFStime-series of phytoplankton chlorophyll a concentration. ICES Journal of MarineScience, 66, 1547–1556.

Vantrepotte, V., & Mélin, F. (2011). Inter-annual variations in the SeaWiFS globalchlorophyll a concentration (1997–2007). Deep Sea Research I, 58(4), 429–441.

Wells, M. L., Price, N. M., & Bruland, K. W. (1994). Iron limitation and the cyanobacteriumsynechococcus in equatorial Pacific waters. Limnology and Oceanography, 39(6),1481–1486.

Yoder, J. A., & Kennelly, M. A. (2003). Seasonal and ENSO variability in global oceanphytoplankton chlorophyll derived from 4 years of SeaWiFS measurements. GlobalBiogeochemical Cycles, 17(4), 1112–1126.



Documents

Monitoring marine phytoplankton seasonality from space