Top Banner
TECHNICAL NOTE The eect of phytoplankton pigment composition and packaging on the retrieval of chlorophyll-a concentration from satellite observations in the Southern Ocean Babula Jena Polar Sciences, ESSO, National Centre for Antarctic and Ocean Research, Ministry of Earth Sciences (MoES), Goa, India ABSTRACT The Antarctic waters are known to be optically unique and the standard empirical ocean colour algorithms applied to these waters may not address the regional bio-optical characteristics. This article sheds light on the performance of current empirical algorithms and a regionally optimized algorithm (ROA) for the retrieval of chlorophyll-a (chl-a) concentration from Aqua- Moderate Resolution Imaging Spectroradiometer (Aqua-MODIS) and Sea-viewing Wide Field-of-view Sensor (SeaWiFS) in the Indian Ocean Sector of Southern Ocean (IOSO). Analysis indicated that empirical algorithms used for the retrieval of chl-a concentra- tion from Aqua-MODIS and SeaWiFS underestimate by a factor varying from 2 to 2.9, resulting in underestimation when in situ chl-a exceeds about 0.3 mg m -3 . To explain these uncertainties, a study was carried out to understand the eect of phytoplankton pigment composition and pigment packaging on remote-sensing reectance (R rs,λ ), from the analysis of phytoplankton-specic absorption coecient (a ph, * λ ). The spatial variation of phytoplank- ton groups analysed using diagnostics pigments (DP) indicated shifting of the phytoplankton community structure from oshore to coastal Antarctic, with a signicant increasing trend for diatoms and a decreasing trend for haptophytes population. The diatom- dominated population exhibits lower a ph, * λ in the 405510 nm region (with relative attening in 443489 nm) compared with the a ph, * λ spectra of the haptophytes-dominated population that peaks near 443 nm. The attening of a ph, * λ spectra for the dia- tom-dominated population was attributed to its larger cell size, which leads to pigment packaging (intracellular shading) and in turn results in higher R rs,λ . The relationship between pigment composition (normalized by chl-a) and blue:green absorption band ratios (a ph, * 443 :a ph, * 555 and a ph, * 489 :a ph, * 555 ) corresponding to the Aqua-MODIS and SeaWiFS bands showed in-phase associa- tions with most of the pigments such as 19ʹ-hexanoyloxyfucox- anthin, 19ʹ-butanoyloxyfucoxanthin, peridinin, and zeaxanthin. In contrast, the out-of-phase association observed between the blue: green absorption ratios and fucoxanthin indicated apparent devia- tions from the general pigment retrieval algorithms, which assumes that blue:green ratios vary in a systematic form with chl-a. The out-of-phase correspondence suggests that the increas- ing trend of fucoxanthin pigments towards the Antarctic coast was ARTICLE HISTORY Received 24 August 2016 Accepted 9 March 2017 CONTACT Babula Jena [email protected] ESSO, National Centre for Antarctic and Ocean Research, Ministry of Earth Sciences (MoES), Headland Sada, Goa 403804, India INTERNATIONAL JOURNAL OF REMOTE SENSING, 2017 http://dx.doi.org/10.1080/01431161.2017.1308034 © 2017 Informa UK Limited, trading as Taylor & Francis Group
22

The effect of phytoplankton pigment composition and packaging …€¦ · The effect of phytoplankton pigment composition and packaging on the retrieval of chlorophyll-a concentration

Aug 20, 2020

Download

Documents

dariahiddleston
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Page 1: The effect of phytoplankton pigment composition and packaging …€¦ · The effect of phytoplankton pigment composition and packaging on the retrieval of chlorophyll-a concentration

TECHNICAL NOTE

The effect of phytoplankton pigment composition andpackaging on the retrieval of chlorophyll-a concentrationfrom satellite observations in the Southern OceanBabula Jena

Polar Sciences, ESSO, National Centre for Antarctic and Ocean Research, Ministry of Earth Sciences (MoES),Goa, India

ABSTRACTThe Antarctic waters are known to be optically unique and thestandard empirical ocean colour algorithms applied to thesewaters may not address the regional bio-optical characteristics.This article sheds light on the performance of current empiricalalgorithms and a regionally optimized algorithm (ROA) for theretrieval of chlorophyll-a (chl-a) concentration from Aqua-Moderate Resolution Imaging Spectroradiometer (Aqua-MODIS)and Sea-viewing Wide Field-of-view Sensor (SeaWiFS) in theIndian Ocean Sector of Southern Ocean (IOSO). Analysis indicatedthat empirical algorithms used for the retrieval of chl-a concentra-tion from Aqua-MODIS and SeaWiFS underestimate by a factorvarying from 2 to 2.9, resulting in underestimation when in situchl-a exceeds about 0.3 mg m−3. To explain these uncertainties, astudy was carried out to understand the effect of phytoplanktonpigment composition and pigment packaging on remote-sensingreflectance (Rrs,λ), from the analysis of phytoplankton-specificabsorption coefficient (aph,*λ). The spatial variation of phytoplank-ton groups analysed using diagnostics pigments (DP) indicatedshifting of the phytoplankton community structure from offshoreto coastal Antarctic, with a significant increasing trend for diatomsand a decreasing trend for haptophytes population. The diatom-dominated population exhibits lower aph,*λ in the 405–510 nmregion (with relative flattening in 443–489 nm) compared with theaph,*λ spectra of the haptophytes-dominated population thatpeaks near 443 nm. The flattening of aph,*λ spectra for the dia-tom-dominated population was attributed to its larger cell size,which leads to pigment packaging (intracellular shading) and inturn results in higher Rrs,λ. The relationship between pigmentcomposition (normalized by chl-a) and blue:green absorptionband ratios (aph,*443:aph,*555 and aph,*489:aph,*555) correspondingto the Aqua-MODIS and SeaWiFS bands showed in-phase associa-tions with most of the pigments such as 19ʹ-hexanoyloxyfucox-anthin, 19ʹ-butanoyloxyfucoxanthin, peridinin, and zeaxanthin. Incontrast, the out-of-phase association observed between the blue:green absorption ratios and fucoxanthin indicated apparent devia-tions from the general pigment retrieval algorithms, whichassumes that blue:green ratios vary in a systematic form withchl-a. The out-of-phase correspondence suggests that the increas-ing trend of fucoxanthin pigments towards the Antarctic coast was

ARTICLE HISTORYReceived 24 August 2016Accepted 9 March 2017

CONTACT Babula Jena [email protected] ESSO, National Centre for Antarctic and Ocean Research, Ministryof Earth Sciences (MoES), Headland Sada, Goa 403804, India

INTERNATIONAL JOURNAL OF REMOTE SENSING, 2017http://dx.doi.org/10.1080/01431161.2017.1308034

© 2017 Informa UK Limited, trading as Taylor & Francis Group

Page 2: The effect of phytoplankton pigment composition and packaging …€¦ · The effect of phytoplankton pigment composition and packaging on the retrieval of chlorophyll-a concentration

associated with the decreasing trend of blue:green absorptionratios and in turn results in higher Rrs,λ. Therefore, an increase inRrs,λ leads to underestimation of chl-a from Aqua-MODIS andSeaWiFS in the IOSO region.

1. Introduction

Sea-surface chlorophyll-a (chl-a) concentration is considered one of the essential para-meters for the estimation of primary production and study of carbon dioxide (CO2)dynamics. The observations are important to link the role of the Southern Ocean pelagicecosystem in the global biogeochemical cycle. In situ observations are sparse around theSouthern Ocean because of adverse weather condition, navigational hazards, remote loca-tion, and inhospitable environment with high sea states driven by strong winds. Satelliteremote sensing of the Southern Ocean colour provides synoptic and time-series coverage ofnear-surface chl-a concentration dynamics affected by seasonality (Johnson et al. 2013). Thesynoptic coverage of chl-a concentration is widely used to investigate the dynamics ofregional oceanographic features such as fronts, eddies, gyres, upwelling zones, plumes, andsurface current patterns. These oceanographic features are useful to study the SouthernOcean ecosystem, which supports large assemblages of phytoplankton, zooplankton, sea-birds, seals, and whales. This signifies the need for the retrieval of chl-a from ocean coloursensors with greater accuracy over the Southern Ocean. However, the retrieval is compli-cated at the regional and local scales as the spectral inherent optical properties (IOPs) of theocean influencing the ocean colour are complex (Sauer et al. 2012).

The satellite-derived chl-a concentration in the Indian Ocean of Southern Ocean(IOSO) may have large uncertainties owing to its different bio-optical characteristics, inaddition to sparse matchup observations (in situ and satellite), being used for algo-rithm development and validation. For example, in order to address the regional bio-optical characteristics of the IOSO region (60° E to 100° E and 40° S to the Antarcticcoast) for retrieval of chl-a concentration, only 15 in situ bio-optical profiles areavailable at National Aeronautics and Space Administration (NASA), bio-OpticalMarine Algorithm Data set (NOMAD). Mitchell and Holm-Hansen (1991) applied theCoastal Zone Colour Scanner (CZCS) global algorithm (GA; Gordon et al. 1983) in thecoastal waters of the western Antarctic Peninsula and the adjacent open ocean watersof Drake Passage. They noticed underestimation of CZCS-derived chl-a and the studyhighlighted the need for regional algorithm development and validation. In Ross Sea,Moore et al. (1999) validated Sea-viewing Wide Field-of-view Sensor (SeaWiFS)-derivedchl-a with in situ-extracted concentration for a range of 0.1–1.5 mg m−3 and the resultsrevealed an underestimation of the SeaWiFS-derived chl-a value. Carder et al. (2003)analysed the Southern Ocean chl-a concentration using the SeaWiFS OC4 algorithm(O’reilly 2000) and revealed significant underestimation of the values between 0.2 and10.0 mg m−3. Similarly, Korb, Whitehouse, and Ward (2004) reported that the SeaWiFSOC4 algorithm output values were only 87% of the in situ chl-aflu for concentrationslower than 1.0 mg m−3 and only 30% for concentrations above 5.0 mg m−3 in theSouth Georgia area, Southern Ocean.

2 B. JENA

Page 3: The effect of phytoplankton pigment composition and packaging …€¦ · The effect of phytoplankton pigment composition and packaging on the retrieval of chlorophyll-a concentration

In contrast to earlier studies, Marrari, Hu, and Daly (2006) reported that high-resolution(1 km) daily SeaWiFS chl-a data generated using the global OC4v4 algorithm was anaccurate measure over the waters of Antarctic Peninsula, if concurrent in situ data weregenerated using high-performance liquid chromatography (HPLC). Further investigationsby Szeto et al. (2011) showed that SeaWiFS chl-a using the global OC4v4 algorithmconsistently underestimates the chl-a value by 50%, although HPLC-based measurementswere considered as suggested by Marrari, Hu, and Daly (2006). A recent study by Johnsonet al. (2013) reported that the relationship between the maximum band ratio of remote-sensing reflectance (Rrs,λ) and in situ chl-a concentration in the Southern Ocean was poorlydescribed by global empirical algorithms and concluded that the current algorithmssignificantly underestimated the chl-a concentration. They attempted to develop site-specific algorithms over the entire Southern Ocean for Aqua-Moderate ResolutionImaging Spectroradiometer (Aqua-MODIS) and SeaWiFS, yet most of the matchup bio-optical observations were located approximately along the 150° E meridian. The developedalgorithms could improve the coefficients of determination (R2) from 0.26 to 0.51 for Aqua-MODIS and from 0.27 to 0.46 for SeaWiFS. The algorithm development process conductedby Johnson et al. (2013) has not addressed the issues associated with atmospheric correc-tions. In addition, the algorithm is yet to be validated in the IOSO waters where the regionaland seasonal variations in the environment may lead to different bio-optical characteristics.Hence, the present study was carried out over the IOSO waters (60° E to 100° E and 40° S toAntarctic coast, Figure 1) with the detailed objectives as follows: (i) employ the current GA(O’reilly 2000) and the southern ocean algorithm (SOA) (Johnson et al. 2013) for theretrieval and validation of chl-a concentration from SeaWiFS and Aqua-MODIS; (ii) analyse

Figure 1. Study area shows ETOPO 1 (Earth Topography One Arc-Minute Global Relief Model, 2009)bathymetric information along with locations of in situ bio-optical observations and HPLC phyto-plankton pigments from station 1 to station 10 (filled circles) collected on cruise on board the R/VRoger Revelle (I8SI9 N), used for the development of regional optimized algorithm (ROA). In situchlorophyll-a concentration collected on cruise on board the R/V Roger Revelle (filled triangles) andworld ocean database (squares) were used for validation of ROA.

INTERNATIONAL JOURNAL OF REMOTE SENSING 3

Page 4: The effect of phytoplankton pigment composition and packaging …€¦ · The effect of phytoplankton pigment composition and packaging on the retrieval of chlorophyll-a concentration

the possibility of developing a regionally optimized algorithm (ROA) for the estimation ofchl-a concentration; (iii) validate the robustness of the developed ROA, and (iv) diagnosethe possible causes of uncertainties in satellite retrieval, particularly the effect of phyto-plankton pigment composition on specific absorption coefficient (aph,*λ) through thepigment package effect.

2. Data analysis and methodology

2.1. Satellite measurements of chl-a concentration using current algorithms

Generating chl-a data from ocean colour satellites involves two major steps, namely theatmospheric correction of visible bands to derive the water-leaving radiance (Gordonand Wang 1994a, 1994B; Bailey, Franz, and Werdell 2010) and the development of asuitable bio-optical algorithm to estimate the chl-a concentration (O’reilly 2000). NASA’sOcean Biology Processing Group (OBPG) processes the global Level-0 ocean colourglobal data sets in near real-time mode and successively processes to generatescience-quality product of Rrs,λ. In this article, Level-3 global Standard Mapped Images(SMI) of weekly mean Rrs,λ from Aqua-MODIS and SeaWiFS were acquired from thewebsite (oceancolor.gsfc.nasa.gov). Afterwards, chl-a concentrations from SeaWiFS andAqua-MODIS are retrieved using various algorithms such as (i) GA (O’reilly 2000), (ii) SOA(Johnson et al. 2013), and (iii) ROA for the IOSO region. The forms of GA, SOA, and ROAMare described in Equations 1–14.

2.1.1. Global algorithms2.1.1.1. SeaWiFS, OC4v6

Chlorophyll chlað Þ ¼ 100:3272�2:9940Rþ 2:7218R2�1:2259R3�0:5683R4 ; (1)

R ¼ log10Rrs;443 > Rrs;490 > Rrs;510

Rrs;555

� �; (2)

where Rrs is the remote-sensing reflectance

2.1.1.2. Aqua-MODIS, OC3 M-547.

Chlorophyll chlað Þ ¼ 100:2424�2:7423Rþ 1:8017R2þ0:0015R3�1:2280R4 : (3)

R ¼ log10Rrs;443 > Rrs;490

Rrs;547

� �: (4)

The above algorithm (OC4v6) yields a strong correlation coefficient (r = 0.892) with insitu chl-a concentration on a global scale, which includes samples from all water types(O’reilly 2000). Nevertheless, the accuracy of this algorithm varies regionally. In addition,to address the regional differences, various band combinations and regression coeffi-cients were developed for different water types, with similar band-ratio forms. Forexample, empirical algorithms were proposed earlier to estimate chl-a concentrationsin the Southern Ocean (Johnson et al. 2013) as described next.

4 B. JENA

Page 5: The effect of phytoplankton pigment composition and packaging …€¦ · The effect of phytoplankton pigment composition and packaging on the retrieval of chlorophyll-a concentration

2.1.2. SOA by JOHNSON ET AL. (2013)2.1.2.1. SeaWiFS, OC4.

Chlorophyll chlað Þ ¼ 100:6736�2:0714R�0:4939R2 þ 0:4756R3 : (5)

R ¼ log10Rrs;443 > Rrs;490 > Rrs;510

Rrs;555

� �: (6)

2.1.2.2. Aqua-MODIS, OC3.

Chlorophyll chlað Þ ¼ 100:6994�2:0384R�0:4656R2 þ 0:4337R3 : (7)

R ¼ log10Rrs;443 > Rrs;490

Rrs;547

� �: (8)

The chl-a concentration retrieved in the Southern Ocean using the above-mentionedalgorithms yields R2 of 0.51 and 0.46 with in situ chl-a for Aqua-MODIS and SeaWiFS,respectively (Johnson et al. 2013). However, the performance of these proposed algo-rithms is yet to be validated in the IOSO region. In this article, uncertainties in theretrieval of chl-a concentration from GA and SOA were analysed initially by validatingwith fluorometrically determined surface in situ chl-aflu during the austral summer(January–February 2004; January–February–March 2006; and February–March 2007).High-quality chl-aflu observations were obtained from World Ocean Database (WOD),National Oceanographic Data Center (NODC), and NOMAD. Details of collection of watersamples, data generation, calibration, and quality control for WOD and NOMAD aregiven in Boyer et al. (2009) and Werdell and Bailey (2005), respectively. A total of 76 insitu chl-aflu observations were available in the study region (Figure 1). The matchupsbetween in situ and satellite observations were generated based on the optimumcombinations of spatial and temporal averaging strategy (3 × 3 pixel mean and 8 daysmean), a method adopted by several authors (Marrari, Hu, and Daly 2006; Johnson et al.2013). After implementation of this validation strategy, the number of matchupsreduced significantly (Table 1 and Figure 2).

Table 1. Validation of chlorophyll-a concentration (mg m−3) derived from Aqua-MODIS and SeaWiFSusing various ocean colour (OC) empirical algorithms such as global algorithm (GA), southern oceanalgorithm (SOA; Johnson et al. 2013), and regionally optimized algorithm (ROA).

Aqua-MODIS SeaWiFS

GA (OC3)SOA

(OC3) ROA (OC3) GA (OC4) SOA (OC4) ROA (OC2) ROA (OC4)

Coefficient of determination (R2) 0.889 0.42 0.84 0.996 0.98 0.99 0.99Standard error 0.27 0.10 0.27 0.03 0.11 0.04 0.06p-value 4.31–05 6.6–03 4.2–03 4.62–06 1.42–04 4.95–05 1.39–05

Slope 0.47 0.14 0.13 0.26 0.33 0.27 0.22Intercept 0.06 0.33 0.19 0.08 0.17 0.14 0.12No. of observations 10 10 10 06 06 06 06The rate of underestimation 2.30 2.35 2.46 2.90 2.01 2.32 2.71

INTERNATIONAL JOURNAL OF REMOTE SENSING 5

Page 6: The effect of phytoplankton pigment composition and packaging …€¦ · The effect of phytoplankton pigment composition and packaging on the retrieval of chlorophyll-a concentration

2.2. In situ bio-optical measurements for regional optimized algorithms(ROAs)

Considering the uncertainties associated with current algorithms (GA, SOA), it was attemptedto develop ROAs by utilizing NOMAD-based in situ bio-optical measurements collected oncruise on board the R/V Roger Revelle (I8SI9 N) (Figure 1). The measurements were sparse overthe study region having a combination of HPLC- (n=10) and Fluorometric (n=05)-determinedchl-a concentration corresponding to the period of 17 February 2007 to 6 March 2007. Chl-aHPLC is considered for the development of ROAs because HPLC data sets were largelycollocated with in situ optical observations and it also accomplishes the main scope of thearticle: studying the impact of phytoplankton pigment composition on satellite retrievalalgorithms. The coincident chl-aHPLC and in situ water-leaving radiance (Lw,λ) observationswere extracted for 10 observations. Since Rrs,λ is the standard input for the derivation ofgeophysical parameters, the Lw,λ converted into Rrs,λ by dividing with spectral solar irradiances(Es,λ) available forMODIS and SeaWiFS bands. Furthermore, empirical ROAswere developed byusing coincident Rrs,λ and in situ chl-aHPLC (Equations 9–14, Figure 3), for the retrieval of chl-aconcentrations from Aqua-MODIS and SeaWiFS. Earlier studies found that ocean coloursatellite retrievals were always underestimated irrespective of comparison with chl-aflu or

Figure 2. Observation on significant underestimation of chlorophyll-a concentration from (a) Aqua-MODIS and (b) SeaWiFS using the current empirical algorithms.

6 B. JENA

Page 7: The effect of phytoplankton pigment composition and packaging …€¦ · The effect of phytoplankton pigment composition and packaging on the retrieval of chlorophyll-a concentration

chl-aHPLC (Stuart et al. 2004; Szeto et al. 2011). Hence, the performances of ROAs wereevaluated using statistical and graphical criteria considering the in situ chl-aflu from WODand NOMAD, in the absence of an adequate number of chl-aHPLC observations for validation(Table 1 and Figure 4). In addition, the spatial pattern of the oceanographic features fromROA-derived chl-a concentration imageswas evaluated for its validity (Figure 5). The vertical profilesof in situ chl-a concentration and temperature data available with WOD were analysed toexplain the variability of surface oceanographic features fromROA-derived chl-a concentrationimages (Figure 5).

To explain the uncertainties in satellite retrievals, an analysis was carried out to under-stand the effect of phytoplankton pigment compositions and packaging on Rrs,λ throughaph,*λ. The phytoplankton groups were classified through HPLC-determined marker diag-nostics pigments (DP) after referring to the methodology given in Hirata et al. (2011)(Table 2), compiled from earlier works of Sieburth, Smetacek, and Lenz (1978), Uitz et al.(2006), and Brewin et al. (2010). The spatial variability of pigment compositions andphytoplankton groups is shown in Figure 6(a,b), respectively. Absorption by the phyto-plankton component (aph,λ) was obtained by subtracting the detrital absorption coefficient

Figure 3. Relationship between in situ chlorophyll-a concentration and the band ratios of in situremote-sensing reflectance (Rrs,λ) symbolized as dashed lines with ‘+’ representing regionallyoptimized algorithms (ROAs). The algorithms were applied to (a) Aqua-MODIS, (b) SeaWiFSOC2,and (c) SeaWiFSOC4 for the estimation of chlorophyll-a concentration. Circle represents the relation-ship between in situ chlorophyll-a concentration and Rrs,λ satellite products from Aqua-MODIS andSeaWiFS as processed by NASA’s Ocean Biology Processing Group (OBPG) .

INTERNATIONAL JOURNAL OF REMOTE SENSING 7

Page 8: The effect of phytoplankton pigment composition and packaging …€¦ · The effect of phytoplankton pigment composition and packaging on the retrieval of chlorophyll-a concentration

(ad,λ) fromparticulate absorption (ap,λ) (data source: NOMAD). The aph,*λwas then calculatedas a ratio of aph,λ to HPLC-determined chl-a concentration. The aph,*λ spectra of thedominant phytoplankton groups for various chl-a concentrations were analysed and com-pared with earlier studies by Ciotti, Lewis, and Cullen (2002) and Bricaud et al. (1995)(Figure 7). In addition, it was analysed to study the variability of aph,*443:aph,*555 and aph,*489:aph,*555 band ratios to determine their relationship with various pigments normalizedby chl-a concentration (Figure 8). The effect of pigment composition and pigment packa-ging on satellite-based Rrs,λ was described in Equation 15.

2.3. Additional data sets

To classify the sampling stations belonging to the onshore, offshore, and Kerguelen Plateauwaters, ETOPO1 (Earth Topography One Arc-Minute Global Relief Model, 2009) geo-refer-enced bathymetric data (21,601 by 10,801 cells) was acquired fromNational Geophysical DataCenter (NGDC) and subsetted for the study region (Figure 1). The northeast shifting of pixellocation in ETOPO1 was rectified as the method specified in Jena et al. (2012). The data onmean dynamic topography (MDT) and Southern Ocean fronts were analysed to compare thephysical features with the biological features (chl-a) derived from Aqua-MODIS and SeaWiFS(Figure 5(a,b)). The MDT of the sea surface was derived from the combined analysis of drifters,satellite altimetry, wind, and GRACE (Gravity Recovery and Climate Experiment) satellite-basedgeoid (Maximenko and Niiler 2005). Locations of various Southern Ocean fronts wereobtained from Australian Antarctic Data Centre (Orsi, Whitworth, and Nowlin 1995).

3. Results and discussion

3.1. Validation of chl-a retrieved using standard GA

In situ chl-a data sets during the austral summer (January–February 2004; January–February–March 2006; and February–March 2007) are presented in Figure 1. The chl-a

Figure 4. Chlorophyll-a concentration retrieved from (a) Aqua-MODIS and (b) SeaWiFS using variousempirical algorithms such as global algorithm (squares), southern ocean algorithm (circles), andregionally optimized algorithm, with (+ and) indicating significant underestimations for elevatedvalues (ellipses) owing to the pigment packaging effect.

8 B. JENA

Page 9: The effect of phytoplankton pigment composition and packaging …€¦ · The effect of phytoplankton pigment composition and packaging on the retrieval of chlorophyll-a concentration

concentrations from 76 observations range from about 0.012 to 19 mg m−3 in the studyregion. After implementation of the validation strategy, the number of matchupsbetween in situ and satellite observations reduced significantly, and the chl-a concen-trations range between 0.14 and 5.16 mg m−3 with a mean value of 0.455 mg m−3. Thecomparative result revealed that in situ measurements in the IOSO waters had significantin-phase correspondence with chl-a concentrations derived from Aqua-MODISGA(R2 = 0.889, p = 4.31–05) and SeaWiFSGA (R2 = 0.996, p = 4.62–06) (Table 1 and

Figure 5. Application of regionally optimized algorithm (ROA)-derived chlorophyll-a images from (a)Aqua-MODIS and (b) SeaWiFS for the period of 18–25 February 2006, indicating elevated values allalong the Antarctic coast, the Kerguelen plateau waters, and along the southern boundary of theAntarctic Circumpolar Current Front (sACCf). These features were evident in the WOD in situ profilesalong (c–d) 60° E, (e–f) 70° E, and (g–h) 80° E transects. The deep-sea chlorophyll-a maxima (DCM)varied from 50 m to 100 m and correspond well with the temperature minimum layer (TML), whichis known to be the winter remnant of the Antarctic Surface Water (AASW) that extends northwardsfrom the coastal Antarctic. The mean dynamic topography (MDT) isolines (solid lines) during 18–25February 2006 were overlaid to interpret the colour features associated with the circulation pattern.Locations of various major Southern Ocean fronts were represented as dashed lines (Orsi, Whitworth,and Nowlin 1995) .

INTERNATIONAL JOURNAL OF REMOTE SENSING 9

Page 10: The effect of phytoplankton pigment composition and packaging …€¦ · The effect of phytoplankton pigment composition and packaging on the retrieval of chlorophyll-a concentration

Table 2. A description on phytoplankton size classes (PSCs) and phytoplankton functional types(PFTs) represented by their pigments (Hirata et al. 2011).PSCs/PFTs Diagnostic pigments Estimation formula

Microplankton (>20 μm)b Fucoxanthin (Fuco), Peridinin (Perid) 1.41 (Fuco + Perid)/Σ(DP)b

Diatoms Fuco 1.41 (Fuco)/Σ(DP)b

Dinoflagellates Perid 1.41 (Perid)/Σ(DP)b

Nanoplankton (2–20 μm)a 19ʹ-Hexanoyloxyfucoxanthin (Hex)Chlorophyll-b (chl-b)Butanoyloxyfucoxanthin (But)Alloxanthin (Allo)

(Xn*1.27 Hex + 1.01 chl-b+ 0.35 But + 0.60 Allo)/Σ(DP)c

Green algae chl-b 1.01 (chl-b)/Σ(DP)b

Prymnesiophytesd

(Haptophytes)Hex, But

Picoplankton (0.2–2 μm)a Zeaxanthin (Zea), Hex, chl-b (0.86 Zea + Yp1.27 Hex)/Σ(DP)c

Prokaryotes Zea 0.86 (Zea)/Σ(DP)b

Pico-eukaryotese Hex, chl-bProchlorococcus sp. Divinyl chl-a (DVchl-a) 0.74 (DVchl-a)/(chl-a)

aSieburth, Smetacek, and Lenz (1978)bΣ(DP) = 1.41 Fuco + 1.41 Perid + 1.27 Hex + 0.6 Allo + 0.35 But + 1.01 chl-b + 0.86 Zea = chl-a (Uitz et al. 2006)cXn indicates a proportion of nanoplankton contribution in Hex. Similarly Yp indicates a proportion of picoplankton inHex (Brewin et al. 2010)

dGiven that the contributions of Allo to nanoplankton were only a few percentage points in our data set, haptophyteswere approximated to Nano minus Green Algae (see also Figure 2 caption of Hirata et al. 2011)

ePico-eukaryotes can be determined from picoplankton minus prokaryotes (see also Figure 2 caption of Hirata et al.2011).

Figure 6. Distribution of HPLC-determined (a) surface phytoplankton pigment compositions and (b)different phytoplankton community structures from onshore (station 1) to offshore region (station10) during 17 February to 6 March 2007, indicating shifting of the phytoplankton communitystructure with a significant increasing trend of diatom abundance from offshore to the Kerguelenplateau waters and coastal Antarctica. Station locations are depicted in Figure 1.

10 B. JENA

Page 11: The effect of phytoplankton pigment composition and packaging …€¦ · The effect of phytoplankton pigment composition and packaging on the retrieval of chlorophyll-a concentration

Figure 2). However, there was a systematic underestimation in case of chl-a concentra-tions derived from Aqua-MODISGA and SeaWiFSGA when the in situ measurementsexceeded about 0.3 mg m−3 (Figure 2). The overall magnitude of underestimation forAqua-MODISGA and SeaWiFSGA was roughly a factor of 2.3 and 2.9, respectively (Table 1).These results of underestimation by satellite observations were well supported by theearlier studies over the Southern Ocean (Moore et al. 1999; Carder et al. 2003; Korb,Whitehouse, and Ward 2004; Stuart et al. 2004; Szeto et al. 2011). To address this issue,Johnson et al. (2013) developed a site-specific algorithm for the entire Southern Oceanwith most of the bio-optical matchup observations located approximately along the150° E meridian. They reported that Aqua-MODISSOA and SeaWiFSSOA (Equations 5–8)were the improved algorithms for the retrieval of chl-a concentration over the SouthernOcean, compared with the operational existing GAs (Equations 1–4).

3.2. Validation of chl-a retrieved using SOA

To evaluate the SOA over the study region and validate the retrieved chl-a, Rrs,λ values fromAqua-MODIS and SeaWiFS were processed using Equations 5–8, which are further validatedwith in situ observations. Table 1 lists the statistics of comparative analysis. The R2 values forAqua-MODISGA and in situ-measured chl-a values (R2 = 0.889) were much better than thosemeasured from Aqua-MODISSOA and in situ (R2 = 0.42), indicating a better performance ofAqua-MODISGA in the study region. The overall rate of underestimation of chl-a concentrationfrom Aqua-MODISGA (factor of 2.3) remains unaffected even after employing Aqua-MODISSOA(factor of 2.35). The R2 between SeaWiFSSOA and in situ chl-ameasurements (R2 = 0.98) had no

Figure 7. Variability of specific absorption spectra of phytoplankton (aph,*λ) for two typical samplesin the Indian Ocean sector of Southern Ocean (dashed lines with circles) compared with earlierstudies by Ciotti, Lewis, and Cullen (2002) and Bricaud et al. (1995). The mean spectral shape of aph,*λ for the size fractionation of phytoplanktons (red, green, and blue colour solid lines) was analysedto assess the extent to which the phytoplankton cell size drives the spectral variation of aph,*λ. Thespectral characteristics of diatom-dominated phytoplankton populations exhibit lower aph,*λ in the405–510 nm region (with relative flattening at 443–489 nm) compared to the aph,*λ spectra ofhaptophytes, which peak near 443 nm. An inverse relationship was observed between chlorophyll-aconcentration and aph,*λ, meaning that when phytoplankton abundance increases, larger cell sizesare added incrementally to a background of smaller cells, which leads to pigment packaging(intracellular shading) and in turn aph,*λ decreases.

INTERNATIONAL JOURNAL OF REMOTE SENSING 11

Page 12: The effect of phytoplankton pigment composition and packaging …€¦ · The effect of phytoplankton pigment composition and packaging on the retrieval of chlorophyll-a concentration

improvement compared to SeaWiFSGA and in situ chl-a (R2 = 0.99). However, the rate ofunderestimation of chl-a concentration from SeaWiFSSOA was lesser (factor of 2.01) than thatof SeaWiFSGA (factor of 2.9). AlthoughSOAperformedbetter thanGA in terms of improvementin slope (Johnson et al. 2013), the SOA-derived chl-a concentration (both for Aqua-MODIS andfor SeaWiFS) indicated significant underestimation at high chl-a concentrations and over-estimation at low chl-a concentration in the IOSO region (Figures 2 and 4). Considering theuncertainty in GA and SOA, it was further attempted to narrow down the search for a suitablebio-optical algorithm for retrieval of chl-a concentration in the IOSO region.

Figure 8. (a) Evident indication of a decrease in the blue:green absorption ratios corresponding toSeaWiFS and Aqua-MODIS bands as associated with an increase in Fucoxanthin:chlorophyll-a pigment.In contrast, other pigments that showed in-phase correspondence with the absorption ratios were (b)19ʹ-Hexanoyloxyfucoxanthin, (c) 19ʹ-Butanoyloxyfucoxanthin, (d) Peridin, and (e) Zeaxanthin.

12 B. JENA

Page 13: The effect of phytoplankton pigment composition and packaging …€¦ · The effect of phytoplankton pigment composition and packaging on the retrieval of chlorophyll-a concentration

3.3. Development of ROA and its application

A variety of algorithms developed earlier for the retrieval of chl-a were based ontheoretical, empirical, analytical, semi-analytical, or combinations of theseapproaches, which need to be fine-tuned or improved. The main problem associatedwith algorithm development or model formulation for parameter retrieval is thecomplexity of physical processes involved and the uncertainties associated withthem (Jena, Swain, and Tyagi 2010). The development of a regional bio-opticalalgorithm requires a large number of in situ measurements, which is practicallydifficult to collect in the Southern Ocean considering the adverse weather conditionand navigational hazards. In the absence of large bio-optical data sets in the studyregion, it was experimented with limited observations (n = 10) from NOMAD toconstruct an ROA. The data set encompasses surface chl-a concentration between0.149 and 1.547 mg m−3 in the sampling stations from 1 to 10 (Figure 1). Typical chl-a concentrations between 0.05 and 1.50 mg m−3 were also reported earlier in theSouthern Ocean (Arrigo et al. 1998; El- Sayed 2005). Interaction of irradiance, deepmixing, iron, and grazing is known to limit phytoplankton growth all over theSouthern Ocean (Boyd 2002; Daly et al. 2001; Moline and Prézelin 1996; Venablesand Moore 2010). Elevated chl-a concentrations between 1 and 30 mg m−3 were alsoreported in regions of continental shelf and ice edge areas (El- Sayed 2005; Holm-Hansen et al. 1989; Moore and Abbott 2000).

The ocean colour algorithms based upon blue to green ratios of Rrs,λ have been usedwidely to derive chl-a concentration from ocean colour sensors (Carder et al. 2003;Gordon et al. 1988; O’reilly et al. 1998; O’reilly 2000). In this article, the NOMAD-based Rrs,λ versus in situ chl-aHPLC coincident data sets during 17 February 2007–6 March 2007were utilized to develop empirical ROAs for the retrieval of chl-a concentration fromAqua-MODIS and SeaWiFS. The ROAs are presented below in Equations 9–14, and as adashed line in Figure 3.

3.3.1. Aqua-MODISROA, OC3

Chlorophyll chlað Þ ¼ 100:080�3:705Rþ 18:41R2�42:41R3 þ 30:02R4 : (9)

R ¼ log10Rrs;443 > Rrs;490

Rrs;547

� �: (10)

3.3.2. SeaWiFS ROA, OC2

Chlorophyll chlað Þ ¼ 100:106�2:907Rþ 8:885R2�12:27R3 : (11)

R ¼ log10Rrs;490Rrs;555

� �: (12)

3.3.3. SeaWiFSROA, OC4

Chlorophyll chlað Þ ¼ 100:228�5:416Rþ 25:1R2�53:30R3 þ 36:34R4 : (13)

R ¼ log10Rrs;443 > Rrs;490 > Rrs;510

Rrs;555

� �: (14)

INTERNATIONAL JOURNAL OF REMOTE SENSING 13

Page 14: The effect of phytoplankton pigment composition and packaging …€¦ · The effect of phytoplankton pigment composition and packaging on the retrieval of chlorophyll-a concentration

The Aqua-MODISROA OC3 output yielded a strong R2 of 0.84, with in situ chl-a. Yet,the chl-a concentrations were underestimated by Aqua-MODISROA compared with insitu measurements (Figure 4). The SeaWiFSROA OC2 and OC4 outputs yielded a strongand one of the highest R2 with in situ chl-a (R2 = 0.99) (Table 1). The rate ofunderestimation and standard error were minimal in case of OC2 compared to thatin OC4 (Table 1). The performance of OC2 was better than OC4 for SeaWiFS-basedchl-a retrieval, because it employs the 490:555 nm band ratio and the atmosphericcorrection is much better in the 490 nm band compared to that in the 443 nm band(Carder et al. 2003).

To evaluate the spatial pattern of oceanographic features from the ROA-derived chl-aimages, the Rrs,λ from Aqua-MODIS and SeaWiFS was processed using Equations 9–14,for the period of 18–25 February 2006, and further compared with the WOD in situprofiles (Figure 5). Aqua-MODISROA and SeaWiFSROA chl-a images indicated elevatedvalues all along the Antarctic coast, Kerguelen Plateau waters, and along the southernboundary of Antarctic Circumpolar Current Front, sACCf (Figure 5(a,b)). The elevatedvalues of chl-a concentration from Aqua-MODISROA and SeaWiFSROA agreed well withthe WOD in situ data collected along the 60° E, 70° E, and 80° E meridional transects(Figure 5(c–h)). The depth of deep-sea chl-a maxima (DCM) varied from 50 m to 100 mand corresponds well with the temperature minimum layer (TML), which is known to bethe winter remnant of the Antarctic Surface Water (AASW) that extends northwards fromthe coastal Antarctic (Figure 5(c–h)) (Holm-Hansen, Kahru, and Hewes 2005). The surfaceoceanographic features such as fronts, eddies, and meanders were well captured byAqua-MODISROA- and SeaWiFSROA-retrieved chl-a images. The features agreed well withthe regional circulation pattern as inferred from the MDT data sets. Both Aqua-MODISROAand SeaWiFSROA images clearly indicated intense phytoplankton blooms above theKerguelen plateau waters and its downstream section along the polar front (PF). Thesedimentary source of iron from the island and the surrounding shallow plateau canfertilize the mixed layer, and possibly cause the bloom (Figure 5(a,b)). These blooms areimportant both as the base of the regional food web and for the carbon that theycapture and export to the deep oceans (Jena 2016; Mishra, Naik, and Anilkumar 2015).Albeit the surface water above the Kerguelen plateau is rich in chl-a, an exception of theoligotrophic condition dominated in between the north Kerguelen plateau (NKP) andthe south Kerguelen plateau (SKP) due to apparent intrusion of iron-limited highnutrient-low chlorophyll (HNLC)-regime waters from the Enderby basin through theFawn trough current (Figure 5). All these distinct features were also observed by thechl-a concentration images retrieved using GA and SOA (figures not shown), andsupported the validity of ROA in terms of the spatial patterns of features.

Although the oceanographic features were well captured from Aqua-MODIS andSeaWiFS chl-a images using all empirical algorithms, an accurate quantitative estimateof phytoplankton biomass in the Southern Ocean remained inconclusive and challen-ging. For example, evidence from our present analysis in the IOSO waters in addition toearlier studies suggested that all empirical algorithms (GA, SOA, and ROA) used for theretrieval of chl-a concentration from Aqua-MODIS and SeaWiFS were found to under-estimate by a factor of 2 at least and can be as large as 2.9 (Table 1). In particular, therewas a significant underestimation by Aqua-MODIS and SeaWiFS chl-a concentrationswhen the in situ measurements exceeded about 0.3 mg m−3 (Figure 2) even after

14 B. JENA

Page 15: The effect of phytoplankton pigment composition and packaging …€¦ · The effect of phytoplankton pigment composition and packaging on the retrieval of chlorophyll-a concentration

employing any empirical algorithms. To explain these observed uncertainties, an analysiswas carried out in the following section to understand the effect of phytoplanktonpigment compositions on the marine Rrs,λ from the analysis of IOPs. Besides, the possiblesources of uncertainties were also investigated when the standard atmospheric correc-tion scheme was applied to the IOSO region.

3.4. Diagnosis on the potential causes of observed uncertainties

The Antarctic waters are known to be optically unique and the standard empirical oceancolour algorithms applied to these waters may not address the regional bio-opticalcharacteristics. Being Case-1 water, the algorithms that relate blue to green Rrs,λ ratio(OC2, OC3, OC4, etc.) should not fail to measure the chl-a concentration in the Antarcticwaters. The reason for this failure is hypothesized first to the presence of optically complexwaters due to the influx of glacial and sea-ice melted water, which can increase particleconcentrations in the nearshore region and result in a conducive environment for adistinct phytoplankton assemblage with the dominance of large cell size planktons (e.g.large diatom cells). For example, few important studies by Smith and Nelson (1985) andArrigo et al. (2012) provided evidence on increased phytoplankton growth and accumula-tion within the marginal ice zone (MIZ) due to melt water production and stratification.Second, the mesoscale physical process induced by baroclinic instabilities is hypothesizedto alter the bio-optical characteristics of the Kerguelen Plateau waters and its downstreamregion after Antarctic Circumpolar Current (ACC) encounters the Kerguelen Plateau topo-graphy. The Kerguelen Ocean and Plateau compared Study (KEOPS) by Maraldi et al.(2009) showed an increase in chl-a concentration above the plateau and attributed it tothe availability of iron coming up from the seafloor after the interaction of strong currentswith the Kerguelen Plateau. The physical processes are known to enhance the particleconcentrations and trigger phytoplankton blooms (Figure 5(a,b)) in the euphotic zonethrough nutrient injection, often accompanied by large-sized cells (e.g. large diatom cells).The increase in phytoplankton pigment concentration and size can have a significantimpact on Rrs,λ through the variability of IOPs, and in turn the standard empirical algo-rithms may fail to accurately measure the chl-a concentration. These hypotheses wereexperimented further by determining the effect of phytoplankton pigment compositionson aλ and Rrs,λ.

Since Rrs,λ depends both on the variability of IOPs (e.g. phytoplankton absorption andscattering coefficients) and the geometric distribution of the light field, IOPs can be usedto model Rrs,λ. A brief summary of the Rrs,λ model used by Lee et al. (1994) and Carderet al. (2003), is described as follows:

Rrs;λ ¼ fλQλ

� �� t2

n2

� �� bb;λ

aλ þ bb;λ

� �; (15)

where f(λ) is an empirical factor. Q(λ) is calculated as the upwelling radiance (Lu) to thecorresponding upwelling irradiance Eu (Q = Eu/Lu), which describes the non-isotropiccharacter of the radiance field. Both f and Q factors are variable and strongly dependenton the solar zenith angle, geometric distribution of the light field, and IOPs. t is the

INTERNATIONAL JOURNAL OF REMOTE SENSING 15

Page 16: The effect of phytoplankton pigment composition and packaging …€¦ · The effect of phytoplankton pigment composition and packaging on the retrieval of chlorophyll-a concentration

transmittance of the air–sea interface, and n is the refractive index of sea water. Thevalue of t2/n2 is equivalent to 0.54, but can vary with sea-state condition and is relativelyindependent of wavelength (Carder et al. 2003). aλ and bb,λ are absorption and back-scattering coefficients, respectively. The spectral behaviour of bb,λ is not as dynamic asthat of aλ (Carder et al. 2003). Considering the variability of parameters described in theabove-mentioned equation, aλ was hypothesized to affect Rrs,λ dominantly. To investi-gate it, the impact of phytoplankton pigment compositions on aλ was studied. Theimpact of aλ on Rrs,λ can explain the observed underestimation of satellite-retrieved chl-aconcentration in the study region.

3.4.1. Phytoplankton pigment composition and spectral absorption propertiesFigure 6(a) shows the distribution of surface phytoplankton pigment composition duringaustral summer (17 February 2007 to 6 March 2007), indicating shifting of the phyto-plankton community structure from the coastal Antarctic (station 1) to the offshoreregion (station 10). The most dominant pigments were fucoxanthin and 19ʹ-hexanoylox-yfucoxanthin (19ʹ-HF), in addition to the presence of a smaller proportion of diadinox-anthin, 19ʹ-butanoyloxyfucoxanthin (19ʹ-BF), peridinin, zeaxanthin, and other pigments.The chl-aHPLC concentration during this period ranges from 0.149 to 1.547 mg m−3, withthe highest values observed near the Antarctic coast. The pigment compositions in thecoastal Antarctic (station 1) and Kerguelen Plateau waters (stations 2–4) were character-ized by the highest proportion of fucoxanthin pigments, indicating the presence ofdiatoms. In these waters, the phytoplankton populations were dominated by 53–78% ofdiatoms as determined using DP analysis (Table 2 and Figure 6(b)). Furthermore, theoffshore stations off Kerguelen Plateau (stations 5–10) were characterized by the highestproportion of 19ʹ-HF pigments (indicative of haptophytes), dominated by 48–69% ofphytoplankton populations (Figure 6(b)). Analysis indicated that there was a shifting ofphytoplankton community structure from offshore to coastal Antarctic, with a significantincreasing trend of diatoms population and a decreasing trend of haptophytes popula-tion (Figure 6(b)). The spectral absorption characteristics of these dominant phytoplank-ton groups were analysed further to determine its effect on Rrs,λ.

The spectral characteristics of aph,*λ are known to vary as a function of pigmentcomposition and pigment packaging, attributed to intracellular shading (Morel andBricaud 1981). The decreased absorption of pigments in cells compared with theabsorption potential of the same amount of pigments is being described as the pig-ment-packaging effect (Duysens 1956; Kirk 1976, 1994; Geider and Osborne 1987; Moreland Bricaud 1981; Sathyendranath, Lazzara, and Prieur 1987). Pigment packaging occurswhen the cell size increases (e.g. large diatoms) or the internal concentration of pig-ments increases (Kirk 1976; Morel and Bricaud 1981; Sosik and Mitchell 1994), whichyields flattening of absorption peaks in the blue spectrum (Dierssen and Smith 2000;Stuart et al. 2004). To examine the effect of pigment composition and packaging, theaph,*λ spectra of the dominant phytoplankton groups for various chl-a concentrationswere analysed and compared with earlier studies by Ciotti, Lewis, and Cullen (2002) andBricaud et al. (1995) (Figure 7). Analysis indicated that the spectral characteristics ofdiatom-dominated phytoplankton populations (diatoms = 74.1%; haptophytes = 21.7%)exhibit lower values of aph,*λ in the 405–510 nm region (relatively flattening at443–489 nm) compared to the aph,*λ spectra of haptophytes-dominated phytoplankton

16 B. JENA

Page 17: The effect of phytoplankton pigment composition and packaging …€¦ · The effect of phytoplankton pigment composition and packaging on the retrieval of chlorophyll-a concentration

populations (haptophytes = 69%, diatoms = 17.6%) that peak near 443 nm. Increase inchl-a concentration was associated with a decrease in aph,*λ, which indicated that whenthe phytoplankton abundance increases, larger cell sizes are added incrementally to abackground of smaller cells (Ciotti, Lewis, and Cullen 2002), leading to pigment packa-ging (intracellular shading) and in turn decreasing aph,*λ (Figure 7). To assess the extentto which the phytoplankton cell size drives the spectral variation of aph,*λ, the meanspectral shape of aph,*λ for the size fractionation of phytoplankton was studied by Ciotti,Lewis, and Cullen (2002). The study revealed the cell size (pico, nano, and microphytoplanktons) could explain more than 80% variability in the spectral shape of aph,*λ. The effect of pigment packaging is prominent for microphytoplanktons (with theflattening of the blue region aph,*λ spectra) compared with pico- and nanophytoplank-ton (Figure 7). The relative flattening of the aph,*λ spectra in the 443–489 nm region forthe diatom-dominated phytoplankton population can be attributed to pigment packa-ging (Figure 7), to the fact that the sizes of diatoms in the Antarctic waters are muchlarger (Brody et al. 1992). The observed flattening of the aph,*λ spectra (blue band) forthe diatom-dominant phytoplankton population and the elevated chl-a concentrationresults in greater Rrs,λ since absorption is inversely related to Rrs,λ (Equation 15).Therefore, an increase in Rrs,λ leads to an underestimation of satellite-retrieved chl-aconcentration for elevated in situ chl-a concentration.

Furthermore, the variability of aph,*443:aph,*555 and aph,*489:aph,*555 band ratios (corre-sponding to SeaWiFS and Aqua-MODIS bands) was analysed to determine their relation-ship with pigment composition normalized by chl-a concentration (Figure 8). In-phaseassociation was observed between blue:green band ratios and pigments such as 19ʹ-HF,19ʹ-BF, peridinin, and zeaxanthin (Figure 8(b–e)). The proportion of 19ʹ-HF to chl-a couldexplain about 76% of the variance (r = 0.87) in aph,*443:aph,*555, and for aph,*489:aph,*555, itwas about 73% (r = 0.85). In contrast, out-of-phase association was observed betweenblue:green band ratios and fucoxanthin pigments, indicating a decrease in the absorp-tion ratio as associated with increase in fucoxanthin pigments (Figure 8(a)). Analysisindicated that 72.9% of the variance (r = −0.85) in aph,*443:aph,*555 could be explained bythe proportion of fucoxanthin:chl-a, and for bands aph,*489:aph,*555 the varianceexplained was about 68% (r = −0.82). The out-of-phase correspondence suggests thatthe increasing trend of fucoxanthin pigments towards the Antarctic coast was associatedwith the decreasing trend of blue:green absorption ratios. The decrease in blue:greenabsorption ratio leads to greater Rrs,λ, which in turn underestimates the chl-a concentra-tion from various empirical algorithms used for Aqua-MODIS and SeaWiFS.

3.5. Possibility of improper atmospheric correction algorithm

In addition to the evidence of pigment composition and packaging effect on theempirical algorithms, the processes involved in atmospheric correction can be a majorsource of uncertainty for the derivation of Lw,λ and in turn effect the retrieval of chl-aconcentration from satellite observations. High concentrations of secondary organicaerosols like isoprene found over the coastal Antarctica, especially in the Prydz Bayand the fluxes, were reported up to 1200 μg m–2 day–1 (Hu et al. 2013). Investigationsare required to study the potentially complicated mixtures of different aerosol types and

INTERNATIONAL JOURNAL OF REMOTE SENSING 17

Page 18: The effect of phytoplankton pigment composition and packaging …€¦ · The effect of phytoplankton pigment composition and packaging on the retrieval of chlorophyll-a concentration

their spatio-temporal variations over the IOSO region, which can produce significanterror in the atmospheric correction procedure.

The classical atmospheric correction procedure assumes the Lw,λ is zero in the near-infrared part of the electromagnetic spectrum. This assumption is known to be invalidover a high concentration of water-scattering constituents and shallow bathymetry (caus-ing bottom reflectance). The recent atmospheric correction algorithms therefore accountfor the non-zero water-leaving signal in the Near Infrared (NIR) by applying a coupledapproach for the separation of atmospheric and aquatic contributions to the signal(Sathyendranath 2000). In case of Antarctic waters, the presence of sea ice and icebergsin the field-of-view of the sensor would also contaminate the surrounding desired Lw,λ.Gregg and Casey (2004) provided evidence on the contamination of ocean colour obser-vations by sea ice in their effort to validate the SeaWiFS chl-a concentration data. Similarly,Belanger, Ehn, and Babin (2007) and Wang and Shi (2009) examined the impact ofbackscattered photons from snow and sea ice on the desired Rrs,λ in the Arctic region,and indicated the impact can be several kilometres from the ice edge due to adjacencyeffect.

The large solar zenith angle in the study area can be another source of uncertaintywhile performing atmospheric correction. Both f and Q in Equation 15 are known to bedependent on the solar zenith angle. The Southern Ocean regions have large solar zenithangles and the proportion of diffuse sky light is high, especially in the blue part of thespectrum (Dierssen and Smith 1997). Morel and Gentili (1993) modelled the changingtrends in f:Q with latitudes and reported that the f:Q ratios are 5% greater in the polarregions than in the temperate regions. Since the f:Q ratio is directly proportional to Rrs,λ, anincrease in the f:Q ratio would result in a higher Rrs,λ value (Equation 15) and in turn canlead to lower chl-a concentration. The effect of f:Q ratio can have a propounded effect onRrs,λ in the study region considering the high latitudes, which range from 40° S to 41° S(Coastal Antarctic). Once all these issues are accounted for, the Lw,λ can be assessedprecisely to further retrieve the chl-a concentration from the satellite observations.

4. Conclusion

An accurate quantitative estimate of chl-a concentration from the satellite observa-tions in the southern IOSO region remained inconclusive and challenging. Forinstance, it is confirmed that the chl-a concentration retrieved from Aqua-MODISand SeaWiFS using empirical algorithms (GA, SOA, and ROA) indicated significantunderestimations for elevated in situ values owing to the dominance of fucoxanthinpigments (a biomarker of diatom) and pigment packaging effect. Hence, the utiliza-tion of these empirical algorithms could lead to large erroneous result while studyingthe long-term trends of chl-a concentration and phytoplankton productivity in thecontext of climate change.

An optimized bio-optical algorithm is required based on the absorption coefficient toaccount for the differences in the bio-optical characteristics of a diatom group ofphytoplanktons. The utility of the algorithm must be improved and validated with thecollection of more in situ bio-optical observations. The algorithm is essential for themonitoring of diatom-dominated phytoplankton blooms and for the precise estimationof the oceanic role in the biogeochemical cycle of carbon, specifically the importance of

18 B. JENA

Page 19: The effect of phytoplankton pigment composition and packaging …€¦ · The effect of phytoplankton pigment composition and packaging on the retrieval of chlorophyll-a concentration

the Southern Ocean as a sink for atmospheric CO2. The need for continuous long-termocean colour observations indeed exists for the regular monitoring of Antarctic marineecosystems.

Acknowledgements

Continuous support from Dr M. Ravichandran, Director, National Centre for Antarctic and OceanResearch (NCAOR), Dr Thamban Meloth (Group Director, Polar Sciences, NCAOR), and Dr AlvarinhoLuis (NCAOR), is gratefully acknowledged. MODIS and SeaWiFS data sets were provided by NationalAeronautics and Space Administration (NASA) Goddard Space Flight Center (NASA’s GSFC). Theauthors would like to acknowledge all scientists and crew members of R/V Revelle cruise (I8SI9N)who collected and analysed in situ bio-optical profiles and pigments that have gone into the NASAbio-Optical Marine Algorithm Data set (NOMAD). In situ chl-a and temperature profiles were obtainedfrom National Oceanographic Data Center (NODC). Earth Topography One Arc-Minute Global ReliefModel, 2009 (ETOPO1) bathymetric data was acquired from National Geophysical Data Center (NGDC).This is NCAOR contribution no 09/2017.

Disclosure statement

No potential conflict of interest was reported by the author.

Funding

Continuous support from Dr M. Ravichandran, Director, National Centre for Antarctic and OceanResearch (NCAOR), is gratefully acknowledged.

References

Arrigo, K. R., D. K. Perovich, R. S. Pickart, Z. W. Brown, G. L. Van Dijken, K. E. Lowry, M. M. Mills, et al.2012. “Massive Phytoplankton Blooms under Arctic Sea Ice.” Science 336 (6087): 1408.doi:10.1126/science.1215065.

Arrigo, K. R., D. Worthen, A. Schnell, and M. P. Lizotte. 1998. “Primary Production in SouthernOcean Waters.” Journal of Geophysical Research 103: 15587–15600. doi:10.1029/98JC00930.

Bailey, S. W., B. A. Franz, and P. J. Werdell. 2010. “Estimation of Near-Infrared Water-LeavingReflectance for Satellite Ocean Color Data Processing.” Optics Express 18: 7521–7527.doi:10.1364/OE.18.007521.

Belanger, S., J. K. Ehn, and M. Babin. 2007. “Impact of Sea Ice on the Retrieval of Water-LeavingReflectance, Chlorophyll-A Concentration and Inherent Optical Properties from Satellite OceanColor Data.” Remote Sensing of Environment 111: 51–68. doi:10.1016/j.rse.2007.03.013.

Boyd, P. W. 2002. “Environmental Factors Controlling Phytoplankton Processes in the SouthernOcean.” Journal of Phycology 38: 844–861. doi:10.1046/j.1529-8817.2002.t01-1-01203.x.

Boyer, T. P., J. I. Antonov, O. K. Baranova, H. E. Garcia, D. R. Johnson, R. A. Locarnini, A. V. Mishonov,et al. 2009. “World Ocean Database 2009.” In NOAA Atlas NESDIS 66, Ed. S. Levitus, 216. Wash., D.C.: U.S. Gov. Printing Office.

Brewin, R. J. W., S. Sathyendranath, T. Hirata, S. Lavender, R. M. Baraciela, and N. Hardman-Mountford. 2010. “A Three-Component Model of Phytoplankton Size Class for the AtlanticOcean.” Ecological Modelling 221 (11): 1472–1483. doi:10.1016/j.ecolmodel.2010.02.014.

Bricaud, A., M. Babin, A. Morel, and H. Claustre. 1995. “Variability in the Chlorophyll-SpecificAbsorption Coefficients of Natural Phytoplankton: Analysis and Parameterization.” Journal ofGeophysical Research 100: 13321–13332. doi:10.1029/95JC00463.

INTERNATIONAL JOURNAL OF REMOTE SENSING 19

Page 20: The effect of phytoplankton pigment composition and packaging …€¦ · The effect of phytoplankton pigment composition and packaging on the retrieval of chlorophyll-a concentration

Brody, E., B. G. Mitchell, O. Holm-Hansen, and M. Vernet. 1992. “Species-Dependent Variations ofthe Absorption Coefficient in the Gerlache Strait.” Antarctic Journal 27: 160–162.

Carder, K. L., F. R. Chen, Z. Lee, S. K. Hawes, and J. Cannizzaro. 2003. MODIS Ocean Science Team:Algorithm Theoretical Basis Document. ATBD 19 Case 2 Chlorophyll-a, version 7. http://modis.gsfc.nasa.gov/data/atbd/atbd_mod19.pdf.

Ciotti, A. M., M. R. Lewis, and J. J. Cullen. 2002. “Assessment of the Relationships betweenDominant Cell Size in Natural Phytoplankton Communities and the Spectral Shape of theAbsorption Coefficient.” Limnology and Oceanography 47 (2): 404–417. doi:10.4319/lo.2002.47.2.0404.

Daly, K. L., W. O. Smith, G. C. Johnson, G. R. Ditullio, D. R. Jones, and C. W. Mordy. 2001.“Hydrography, Nutrients, and Carbon Pools in the Pacific Sector of the Southern Ocean:Implications for Carbon Flux.” Journal of Geophysical Research 106: 7107–7124. doi:10.1029/1999JC000090.

Dierssen, H., and R. C. Smith. 1997. “Estimation of Irradiance Just below the Air-Water Interface.” InProc. SPIE 2963, Ocean Optics XIII, 204–209. doi:10.1117/12.266443.

Dierssen, H. M., and R. C. Smith. 2000. “Bio-Optical Properties and Remote Sensing Ocean ColorAlgorithms for Antarctic Peninsula Waters.” Journal of Geophysical Research 105: 26301–26312.doi:10.1029/1999JC000296.

Duysens, L. N. M. 1956. “The Flattening of the Absorption Spectrum of Suspensions, asCompared to that of Solutions.” Biochimica Et Biophysica Acta 19: 1–12. doi:10.1016/0006-3002(56)90380-8.

El- Sayed, S. Z. 2005. “History and Evolution of Primary Productivity Studies of the SouthernOcean.” Polar Biology 28: 423–438. doi:10.1007/s00300-004-0685-2.

Geider, R. J., and B. A. Osborne. 1987. “Light Absorption by a Marine Diatoms: ExperimentalObservations and Theoretical Calculations of the Package Effect in a Small ThalassiosiraSpecies.” Marine Biology 96: 299–308. doi:10.1007/BF00427030.

Gordon, H. R., O. B. Brown, R. H. Evans, J. W. Brown, R. C. Smith, K. S. Baker, and D. K. Clark. 1988. “ASemi-Analytic Model of Ocean Color.” Journal of Geophysical Research 93: 10909–10924.doi:10.1029/JD093iD09p10909.

Gordon, H. R., D. K. Clark, J. W. Brown, O. B. Brown, R. H. Evans, and W. W. Broenkow. 1983.“Phytoplankton Pigment Concentrations in the Middle Atlantic Bight: Comparison of ShipDeterminations and CZCS Estimates.” Applied Optics 22: 20–36. doi:10.1364/AO.22.000020.

Gordon, H. R., and M. Wang. 1994a. “Retrieval of Water-Leaving Radiance and Aerosol OpticalThickness over the Oceans with Seawifs: A Preliminary Algorithm.” Applied Optics 33: 443–452.doi:10.1364/AO.33.000443.

Gordon, H. R., and M. Wang. 1994B. “Influence of Oceanic White Caps on Atmospheric Correctionof Ocean-Color Sensor.” Applied Optics 33: 7754–7763. doi:10.1364/AO.33.007754.

Gregg, W. W., and N. W. Casey. 2004. “Global and Regional Evaluation of the Seawifs ChlorophyllData Set.” Remote Sensing Environment 93: 463–479. doi:10.1016/j.rse.2003.12.012.

Hirata, T., N. J. Hardman-Mountford, R. J. W. Brewin, J. Aiken, R. Barlow, K. Suzuki, T. Isada, et al.2011. “Synoptic Relationships between Surface Chlorophyll-A and Diagnostic Pigments Specificto Phytoplankton Functional Types.” Biogeosciences 8: 311–327. doi:10.5194/bg-8-311-2011.

Holm-Hansen, O., M. Kahru, and C. D. Hewes. 2005. “Deep Chlorophyll a Maxima in PelagicAntarctic Waters. II. Relation to Bathymetric Features and Dissolved Iron Concentrations.”Marine Ecology Progress Series 297: 71–81. doi:10.3354/meps297071.

Holm-Hansen, O., B. G. Mitchell, C. D. Hewes, and D. M. Karl. 1989. “Phytoplankton Blooms in theVicinity of Palmer Station, Antarctica.” Polar Biology 10: 49–57. doi:10.1007/BF00238290.

Hu, Q. H., Z. Q. Xie, X. M. Wang, H. Kang, Q. F. He, and P. Zhang. 2013. “Secondary Organic Aerosolsover Oceans via Oxidation of Isoprene and Monoterpenes from Arctic to Antarctic.” ScientificReports 3: 2280. doi:10.1038/srep02280.

Jena, B. 2016. “Satellite Remote Sensing of the Island Mass Effect on the Sub-Antarctic KerguelenPlateau, Southern Ocean.” Frontiers of Earth Science 10 (3): 479–486. doi:10.1007/s11707-016-0561-8.

20 B. JENA

Page 21: The effect of phytoplankton pigment composition and packaging …€¦ · The effect of phytoplankton pigment composition and packaging on the retrieval of chlorophyll-a concentration

Jena, B., P. J. Kurian, D. Swain, A. Tyagi, and R. Ravindra. 2012. “Prediction of Bathymetry fromSatellite Altimeter Based Gravity in the Arabian Sea: Mapping of Two Unnamed DeepSeamounts.” International Journal of Applied Earth Observation and Geoinformation 16: 1–4.doi:10.1016/j.jag.2011.11.008.

Jena, B., D. Swain, and A. Tyagi. 2010. “Application of Artificial Neural Networks for Sea SurfaceWind Speed Retrieval from IRS-P4 (MSMR) Brightness Temperature.” IEEE Geoscience and RemoteSensing Letters 7 (3): 567–571. doi:10.1109/LGRS.2010.2041632.

Johnson, R., P. G. Strutton, S. W. Wright, A. Mcminn, and K. M. Meiners. 2013. “Three ImprovedSatellite Chlorophyll Algorithms for the Southern Ocean.” Journal of Geophysical Research 118:3694–3703. doi:10.1002/jgrc.20270.

Kirk, J. T. O. 1976. “A Theoretical Analysis of the Contribution of Algal Cells to the Attenuation ofLight within Natural Waters. III. Cylindrical and Spheroidal Cells.” New Phytologist 77: 341–358.doi:10.1111/nph.1976.77.issue-2.

Kirk, J. T. O. 1994. Light and Photosynthesis in Aquatic Ecosystems, 509. Cambridge: CambridgeUniversity Press.

Korb, R. E., M. J. Whitehouse, and P. Ward. 2004. “Seawifs in the Southern Ocean: Spatial andTemporal Variability in Phytoplankton Biomass around South Georgia.” Deep Sea Research Part II51: 99–116. doi:10.1016/j.dsr2.2003.04.002.

Lee, Z. P., K. L. Carder, S. H. Hawes, R. G. Steward, T. G. Peacock, and C. O. Davis. 1994. “A Model forInterpretation of Hyperspectral Remote-Sensing Reflectance.” Applied Optics 33: 5721–5732.doi:10.1364/AO.33.005721.

Maraldi, C., M. Mongin, R. Coleman, and L. Testut. 2009. “The Influence of Lateral Mixing on aPhytoplankton Bloom: Distribution in the Kerguelen Plateau Region.” Deep Sea Research Part I56 (6): 963–973. doi:10.1016/j.dsr.2008.12.018.

Marrari, M., C. Hu, and K. Daly. 2006. “Validation of Seawifs Chlorophyll A Concentrations in theSouthern Ocean: A Revisit.” Remote Sensing of Environment 105: 367–375. doi:10.1016/j.rse.2006.07.008.

Maximenko, N. A., and P. P. Niiler. 2005. “Hybrid Decade-Mean Global Sea Level with MesoscaleResolution.” In Recent Advances in Marine Science and Technology, 2004, Ed. N. Saxena, 55–59,PACON International.

Mishra, R. K., R. K. Naik, and N. Anilkumar. 2015. “Adaptations of Phytoplankton in the Indian OceanSector of the Southern Ocean during Austral Summer of 1998—2014.” Frontiers of Earth Science9: 742–752. doi:10.1007/s11707-015-0541-4.

Mitchell, B., and B. Holm-Hansen. 1991. “Bio-Optical Properties of Antarctic Peninsula Waters:Differentiation from Temperate Ocean Models.” Deep Sea Research Part I 38: 1009–1028.doi:10.1016/0198-0149(91)90094-V.

Moline, M., and B. B. Prézelin. 1996. “Long-Term Monitoring and Analyses of Physical FactorsRegulating Variability in Coastal Antarctic Phytoplankton Biomass, in Situ Productivity, andTaxonomic Composition over Subseasonal, Seasonal and Interannual Time Scales.” MarineEcology Progress Series 145: 143–160. doi:10.3354/meps145143.

Moore, J. K., and M. R. Abbott. 2000. “Phytoplankton Chlorophyll Distributions and PrimaryProduction in the Southern Ocean.” Journal of Geophysical Research 105: 28709–28722.doi:10.1029/1999JC000043.

Moore, J. K., M. R. Abbott, J. G. Richman, W. O. Smith, T. J. Cowles, K. H. Coale, W. D. Gardner, and R.T. Barber. 1999. “Seawifs Satellite Ocean Color Data from the Southern Ocean.” GeophysicalResearch Letter 26: 1465–1468. doi:10.1029/1999GL900242.

Morel, A., and A. Bricaud. 1981. “Theoretical Results Concerning Light Absorption in a DiscreteMedium and Application to the Specific Absorption of Phytoplankton.” Deep Sea Research Part I28: 1357–1393.

Morel, A., and B. Gentili. 1993. “Diffuse Reflectance of Oceanic Waters. II Bidirectional Aspects.”Applied Optics 32 (33): 6864–6879. doi:10.1364/AO.32.006864.

O’reilly, J. E., 24 COAUTHORS. 2000. “Seawifs Post Launch Calibration and Validation Analyses, Part3.” Eds. S. B. Hooker and E. R. Firestone In NASA Technical Memorandum, Vol. 11, 49. MD, USA:NASA's Goddard Space Flight Center.

INTERNATIONAL JOURNAL OF REMOTE SENSING 21

Page 22: The effect of phytoplankton pigment composition and packaging …€¦ · The effect of phytoplankton pigment composition and packaging on the retrieval of chlorophyll-a concentration

O’reilly, J. E., S. Maritorena, B. G. Mitchell, D. A. Siegel, K. L. Carder, S. A. Garver, M. Kahru, and C.Mcclain. 1998. “Ocean Color Chlorophyll Algorithms for Seawifs.” Journal of Geophysical Research103 (C11): 24937–24953. doi:10.1029/98JC02160.

Orsi, A. H., T. Whitworth, and W. D. Nowlin. 1995. “On the Meridional Extent and Fronts of theAntarctic Circumpolar Current.” Deep Sea Research I 42 (5): 641–673. doi:10.1016/0967-0637(95)00021-W.

Sathyendranath, S. 2000. Remote Sensing of Ocean Colour in Coastal, and Other Optically-Complex,Waters, IOCCG Report, 140. Vol. 3. Dartmouth, Nova Scotia: IOCCG Project Office.

Sathyendranath, S., L. Lazzara, and L. Prieur. 1987. “Variations in the Spectral Values of SpecificAbsorption of Phytoplankton.” Limnology and Oceanography 32: 403–415. doi:10.4319/lo.1987.32.2.0403.

Sauer, M. J., C. S. Roesler, P. J. Werdell, and A. Barnard. 2012. “Under the Hood of Satellite EmpiricalChlorophyll a Algorithms: Revealing the Dependencies of Maximum Band Ratio Algorithms onInherent Optical Properties.” Optics Express 20: 20920–20933. doi:10.1364/OE.20.020920.

Sieburth, J. M., V. Smetacek, and J. Lenz. 1978. “Pelagic Ecosystem Structure: HeterotrophicCompartments of the Plankton and Their Relationship to Plankton Size Fractions.” Limnologyand Oceanography 23: 1256–1263. doi:10.4319/lo.1978.23.6.1256.

Smith, W. O. Jr, and D. M. Nelson. 1985. “Phytoplankton Bloom Produced by a Receding Ice Edge inthe Ross Sea: Spatial Coherence with the Density Field.” Science 227: 163–166. doi:10.1126/science.227.4683.163.

Sosik, H. M., and B. G. Mitchell. 1994. “Effects of Temperature on Growth, Light Absorption, andQuantum Yield in Dunaliella Tertiolecta (Chlorophyceae).” Journal of Phycology 30: 833–840.doi:10.1111/j.0022-3646.1994.00833.x.

Stuart, V., O. Ulloa, G. Alarcón, S. Sathyendranath, H. Major, E. Head, and T. Platt. 2004. “Bio-OpticalCharacteristics of Phytoplankton Populations in the Upwelling System off the Coast of Chile.”Revista Chilena De Historia Natural 77: 87–105. doi:10.4067/S0716-078X2004000100008.

Szeto, M., P. J. Werdell, T. S. Moore, and J. W. Campbell. 2011. “Are the Worlds Oceans OpticallyDifferent?” Journal of Geophysical Research 116 (C7): C00H04. doi:10.1029/2011JC007230.

Uitz, J., H. Claustre, A. Morel, and S. B. Hooker. 2006. “Vertical Distribution of PhytoplanktonCommunities in Open Ocean, an Assessment Based on Surface Chlorophyll.” Journal ofGeophysical Research 111: 2156–2202. doi:10.1029/2005JC003207.

Venables, H., and C. M. Moore. 2010. “Phytoplankton and Light Limitation in the Southern Ocean:Learning from High-Nutrient, High-Chlorophyll Areas.” Journal of Geophysical Research 115:C02015. doi:10.1029/2009JC005361.

Wang, M. H., and W. Shi. 2009. “Detection of Ice and Mixed Ice-Water Pixels for MODIS Ocean ColorData Processing.” IEEE Transactions on Geoscience and Remote Sensing 47: 2510–2518.doi:10.1109/TGRS.2009.2014365.

Werdell, P. J., and S. W. Bailey. 2005. “An Improved in Situ Bio-Optical Data Set for Ocean ColorAlgorithm Development and Satellite Data Product Validation.” Remote Sensing of Environment98: 122–140. doi:10.1016/j.rse.2005.07.001.

22 B. JENA