Performance of operational satellite bio-optical algorithms in different water types in the southeastern Arabian Sea

ORIGINAL

RESEARCH

ARTICLE

Performance

of

operational

satellite

bio-optical

algorithms

in

different

water

types

in

the

southeastern

Arabian

Sea

P.

Minu

,

Aneesh

A.

Lotliker

,

S.S.

Shaju

,

P.

Muhamed

Ashraf

*

,

T.

Srinivasa

Kumar

,

B.

Meenakumari

a

ICAR—CentralInstituteofFisheriesTechnology,Kochi,India

b_{IndianNationalCentreforOceanInformationServices(INCOIS),Hyderabad,India} c

IndianCouncilforAgriculturalResearch,NewDelhi,India Received31October2015;accepted30May2016

Availableonline16June2016

KEYWORDS Remotesensing reflectance; Algorithms; Chlorophyll-a; CDOM; ArabianSea; SATCORE

Summary Theinsituremotesensingreflectance(Rrs)andopticallyactivesubstances(OAS)

measuredusinghyperspectralradiometer,wereusedforopticalclassi_ficationofcoastalwatersin thesoutheasternArabianSea.The spectralRrs showedthreedistinctwatertypes,thatwere

associated withthe variability in OASsuch as chlorophyll-a (chl-a), chromophoric dissolved organicmatter(CDOM)andvolumescatteringfunctionat650nm(b650).Thewatertypeswere

classifiedasType-I,Type-IIandType-IIIrespectivelyforthethreeRrsspectra.TheType-Iwaters

showedthepeakRrsintheblueband(470nm),whereasinthecaseofType-IIandIIIwatersthe

peakRrswasat560and570nmrespectively.TheshiftingofthepeakRrsatthelongerwavelength

wasduetoanincreaseinconcentrationofOAS.Further,weevaluatedsixbio-opticalalgorithms (OC3C,OC4O,OC4,OC4E,OC3MandOC4O2)usedoperationallytoretrievechl-afromCoastal ZoneColourScanner(CZCS),OceanColourTemperatureScanner(OCTS),Sea-viewingWide Field-of-viewSensor(SeaWiFS),MEdiumResolutionImagingSpectrometer(MERIS),Moderate Resolu-tionImagingSpectroradiometer(MODIS)andOceanColourMonitor(OCM2).Forchl-a concentra-tiongreaterthan1.0mgm
3,algorithmsbasedonthereferencebandratios488/510/520nmto

PeerreviewundertheresponsibilityofInstituteofOceanologyofthePolishAcademyofSciences.

* Correspondingauthorat:ICAR_—CentralInstituteofFisheriesTechnology,MatsyapuriP.O.,Kochi682029,India.Tel.:+914842412300; fax:+914842666845.

E-mailaddress:[email protected](P.M.Ashraf).

Availableonlineatwww.sciencedirect.com

ScienceDirect

jo u rn al ho m e p age : w w w. jo ur na ls .e l se v i er.c o m / o ce an o lo g i a/

http://dx.doi.org/10.1016/j.oceano.2016.05.005

0078-3234/#2016InstituteofOceanologyofthePolishAcademyofSciences.ProductionandhostingbyElsevierSp.zo.o.Thisisanopen accessarticleundertheCCBY-NC-NDlicense(http://creativecommons.org/licenses/by-nc-nd/4.0/).

1.

Introduction

Oceancolourremotesensinghasbeenwidelyusedasatool todetermine thesurface chlorophyll-a (chl-a) concentra-tion which acts as a proxy for phytoplankton biomass. Severalempiricalandsemi-analyticalalgorithmshavebeen proposed for the retrieval of chl-a from satellite ocean colourdata.Thesealgorithmswerebasedonthenon-linear relationshipbetweenoceanicreflectanceandinsitu mea-suredchl-a,morepreciselytheratiosofreflectanceinblue andgreen bands or their combinations(Lee etal., 2002; O'Reilly et al., 1998, 2000). These algorithms using the spectral ratios of reflectance were mainly attributed to Case 1 waters where the optical properties were deter-mined mainly by phytoplankton and their accessory pig-ments.ForCase2waters,apartfromphytoplankton,other optically active substances (OAS) such as chromophoric dissolvedorganicmatter(CDOM)andtotalsuspended mat-ter(TSM)contributesignificantlytothereflectance(Morel andPrieur,1977).

Thedevelopmentofoceanchlorophyll2-band(OC2)and oceanchlorophyll 4-band(OC4) algorithmswasdone using SeaBAMdata set.TheOC2algorithm was revised(OC2v2) basedonanextensivedatasetof1174insituobservations andthereafterwiththeSIMBIOSdataset(McClainand Far-gion,1999).O'Reillyetal.(2000)updatedOC2andOC4with 2853insitudatasets(OC2v2andOC4v4)andsuggestedthe needto determineaccuracyof theserevisedalgorithmsin lowestchl-aconcentrations.

Thecharacteristicsofthereflectancespectrum,interms ofshapeandmagnitude,arelargelyinfluencedbythe pre-senceofOAS within the watercolumn (Minuetal.,2014; PiersonandStrombeck,2000).Inotherwords,theapparent opticalproperties (AOP)ofaquatic media, suchas remote sensingreflectance (Rrs),largelyaffectedbyOASand

geo-metryofambientlightfieldcanbequantifiedusinginherent opticalproperties (IOPs) of theOAS (Mobley, 1994;Morel, 1991).ThefundamentalIOPsinfluencingtheRrsare

absorp-tion(a),scattering(b)andvolume scatteringfunction(b). Thephytoplanktonpigment,chl-a,hasatendencytoabsorb lightintheblueandredbandsofthevisibleelectromagnetic spectrum(V-EMS)withtheformerbeing theprimarypeak. The CDOM also exhibits strong absorption in UV and the shorterwavelengthband of V-EMS whereas thesuspended matterusuallyscattersinthelongerwavelengthband.Apart fromthis,thewatermoleculesthemselvesabsorbstronglyin the red part of V-EMS. Further, the volume scattering describes the angular distribution of light scattered from an incident beam. In the absence of inelastic scattering, IOPofamediumiscompletelydeterminedbytheabsorption coefficientandb.Thesewhencombinedwiththeangularand

spectral distribution of the incident radiance field, just below the surface, the full radiative flux balance of the oceancanbesimulated(LeeandLewis,2003).Studiesusing hyperspectralradiometers,ontherelationshipbetweenIOP andconcentrationofOAS,indicatedthepresenceof identifi-ablesub-typesofcoastalwaterwithintheconventionalCase 2classification(MckeeandCunningham,2006).

Theratiobasedempiricalalgorithmsfortheretrievalof chl-afromCZCS,MOS-B,IRS-P4-OCMandSeaWiFShavebeen alreadycarriedoutinthesoutheasternArabianSea.Satheand Jadhav (2001) studied retrieval of chl-a using sea-leaving radiancefromMOS-BandshowedthatthesingleratioofCZCS algorithmfailsintheArabianSea.Theyalsoreportedthatthe two factor algorithmof SeaWIFS failsin 30%of the cases. Further, Nagamanietal. (2008) reportedthat OC4v4 algo-rithmoverestimateschl-ainthenorthernArabianSeawhen compared to MBR based OCM-2 algorithm. In addition to these,Chauhanetal.(2002)evaluatedtheaccuracy,precision andsuitabilityof differentoceancolouralgorithms forthe ArabianSea.Accordingtotheirstudy,OC2andOC4algorithms performedwellinCase1watersoftheArabianSea.Butboth algorithms failedto estimatechlorophyllin Trichodesmium dominatedwaters.Tilstoneetal.(2011)alsoassessedthree algorithms, OC4v6,Carderand OC5,for retrievingchl-a in coastalareasoftheBayofBengalandopenoceanareasofthe ArabianSea.Basedontheaccuracyassessment,they recom-mendedtheuseoftheOC5algorithmintheareaofstudy.Most ofthesestudieswereconfinedtooligotrophictomesotrophic waters.AssessmentofMODIS-aquachlorophyll-aalgorithmsin coastal and shelf waters of the southeastern Arabian Sea showed better performance of OC3M than GSM and GIOP (Tilstoneetal.,2013).However,ourstudydealswith meso-trophictoeutrophicwaters.Theproblemwithestimating chl-afromanoceancoloursatellitesensorinvolvestwocritical steps.Thefirststepistoeliminatetheeffectofthe atmo-sphericcontributionandthesecondstepistoselectasuitable bio-opticalalgorithm.Inthisstudy,wehavefocusedoninsitu bio-opticaldata,measuredusinghyperspectralradiometer,in theopticallycomplexwatersofthesoutheasternArabianSea, withthefollowingobjectives:

1.Analyzingthespectralremotesensingreflectanceinthe viewofdistributionofOAS.

2.Evaluationofoperationalempiricalalgorithmindifferent watertypes.

2.

Study

area

ApartofthesoutheasternArabianSeaformsthestudyarea (Fig.1).Theareahasastrongmonsoonalinfluenceresulting 547/550/555/560/565nm havetobe considered.Theassessmentof algorithms showedbetter performanceofOC3MandOC4.AllthealgorithmsexhibitedbetterperformanceinType-Iwaters. However,theperformancewaspoorinType-IIandType-IIIwaterswhichcouldbeattributedtothe significantco-varianceofchl-awithCDOM.

#2016Institute ofOceanology ofthePolish Academyof Sciences. Production and hosting by ElsevierSp.zo.o.ThisisanopenaccessarticleundertheCCBY-NC-ND license( http://creati-vecommons.org/licenses/by-nc-nd/4.0/).

inseasonalchangesinhydrographicconditionsinfluencedby river water dischargeand surface circulation. During pre-monsoon season (February—May) wind-induced upwelling alongwithanorthwardundercurrentandasouthward sur-faceflowassociatedwithstrongverticalmixingareobserved offKochiwaters(KumarandKumar,1996).Duringthe mon-soon period, freshwater discharge from adjoining rivers increasesthenutrientconcentrations ofthecoastalwaters alongwiththedetritalload.Theserapidchangesoftenlead to very high production at primary and secondary levels (Georgeetal.,2013).Upwelling processsupportedby the southerlycurrentisalsoobservedalongthecoastal waters duringmonsoonseason(JoshiandRao,2012).Theprevious studiesreportedthatfreshwaterdischargefromtheestuary andthehigh organicload ofthebottom sedimentare the potentialfactorsaffectingthebiogeochemistryofthestudy area(Iyeretal.,2003;SrinivasandDineshkumar,2006).

3.

Material

and

methods

3.1. Insitu data

Thedatausedforthepresentstudywasgeneratedasapart ofSATelliteCoastalandOceanographicREsearch(SATCORE) programmecoordinatedbyIndianNationalCentreforOcean InformationServices(INCOIS).Thedatawasgenerated,with monthly frequency, from eight stations, between April 2009and2011.Thesamplingwascarriedoutusing commer-cial purse seiner. SatlanticTM _{hyperspectral radiometer}

(HyperOCRII)wasoperatedinallthestations.Theinstrument containsopticalsensorswhichmeasuredownwellingsurface irradiance (Es) and profiles ofupwelling radiance (Lu) and

downwellingirradiance(Ed).Apartfromthesetheancillary

sensorsmeasurestilt,pressure,temperatureand conductiv-ity.TheradiometerwasalsoequippedwithECOtripletsensor (WetlabsTM_{ECOseries)whichmeasureschlorophyll}

fluores-cence, CDOM fluorescence and b at 650nm (b650). The

excitationandemissionwavelengthstomeasureCDOM fluor-escencewere370and460nmrespectivelywithasensitivity of0.28ppbofquininesulphatedihydrateequivalent(QSDE). Thevolumescatteringfunction,b,whenintegratedin back-warddirectionprovidesestimatesofbackscattering(Balch et al., 2001). The backscattering at longer wavelengths, especiallyinred,ismoresensitivetothesuspendedmatter.

Henceinthepresentstudybisconsideredasameasureof totalsuspendedmatter.

Utmostcare was taken to avoid tilt, maintain profiling velocity and avoid shadows from sources. The data from hyperspectralradiometerwasrecordedusingSatViewTM soft-wareandmulti-levelprocessing wascarried outusing Pro-softTMsoftware.

ThespectralRrswascomputedas

Rrsð0þ;lÞ¼

Lwð0þ;lÞ

Edð0þ;lÞ

; (1)

whereLw(0+,l) is water leavingradiance and Ed(0+, l) is

downwellingirradiance abovetheseasurface. Furtherthe upwellingradianceanddownwellingradiancewere comput-edasfollows:

Edð0þ;lÞ¼

Edð0
;lÞ

1
a ; (2)

whereaistheFresnelreflectionalbedoforirradiancefrom thesunandskyand

Lwð0þ;lÞ¼Luð0
;lÞ½1
rðl;uÞ

h2 wðlÞ

; (3)

wherer(l,u)istheFresnelreflectanceindexofseawaterand hw(l)istheFresnelrefractiveindexofseawater.

The spectral Rrs, measuredusing radiometer, was then

usedasaninputtoderivethechlorophyllconcentrationfrom respectivealgorithms.

3.2. Empiricalalgorithms

Six operational empirical algorithms (OC3C, OC4O, OC4, OC4E, OC3M, OC4O2) were selected for this study. These algorithmshavebeenoperationallyimplementedasdefault algorithms for Costal Zone Colour Scanner (CZCS), Ocean ColourandTemperature Scanner(OCTS),Sea-viewingWide Field-of-viewSensor(SeaWiFS),MediumResolutionImaging Spectrometer(MERIS),ModerateResolutionImaging Spectro-radiometer(MODIS)andOceanColourMonitor(OCM2).The functionalformsofthesealgorithmsaregiveninTable1.

Thesealgorithmshaveundergoneseveralrevisionsbased ontheinsitudatageneratedfromdifferentwatertypes.The OC2algorithm,originallydesignedforCZCS,hasundergone eightrevisions(OC2a,OC2b,OC2c,OC2d,OC2e,OC2,OC2v2 and OC2v4). The latest of which is OC2v4 which has a modified cubic polynomial form. The OC4O algorithm, designedforOCTS,hasundergonefourrevisionsandemerged as4thorder polynomial functionrelating themaximumof threebandratios.TheOC4algorithm,designedforSeaWiFS, hastwo versions(OC4andOC4v4).Both versionsuse max-imum ratio of four bands. The first version of OC4 was a modified cubicpolynomial and the current version uses a fourthorderpolynomialwithfivecoefficients(O'Reillyetal., 1998,2000).TheOC4Ealgorithm,designedforMERIS,isthe tunedversionofOC4v4andhasthefunctionalform of4th order cubic polynomial. The OC3 algorithm, designed for MODIShadundergonetwoversions(OC3d,andOC3e).Itis three banded maximum band ratio and uses fourth order polynomialfunction.ThedefaultalgorithmforOCM-2isthird ordermodifiedcubicpolynomial whichused themaximum ratiooffourbands(Nagamanietal.,2008).

4.

Results

Theresultsofthepresentstudyhavebeendescribedinthe following three sections. The first section deals with the optical classification of coastal waters based on AOP. The secondsectiondealswiththeassociationofOASwithspectral Rrsandthethirdsectionwiththeassessmentofoceancolour

algorithmsindifferentwatertypes.

4.1. Optical classificationusingremotesensing reflectance

Theentire datasets exhibitedthree different typesof Rrs

spectraandtheirspectralvariabilityisgiveninFig.2.Based onthevariabilityinRrsspectra,thecoastalwatersoffKochi

areopticallyclassified intothree types.Type-Iwaters had spectra(Fig.2a)withflattercurvebetween400and450nm whichthen increasedtopeak at482nm.After482nm, Rrs

decreasedgraduallytill608nmandthenthecurveflattened againtill700nmindicatingnoreflectanceintheregion.The

spectralRrsinType-IIwaters(Fig.2b)wasfoundtobedistinct

fromthatofType-I.TheaverageRrsincreasedgraduallyfrom

shorterwavelength(400nm)andshowedalmostaflatpeak between532and566nmwithamarginallyhighervalueat 560nm. Beyond 560nm, a steep valley was observed till 610nm.Afterwhichthedecreasewasgradualtill670nm. Asecondarymaximumwasalsoobservedat681nm.TheRrs

spectrainType-IIIwaters(Fig.2c)showedsimilaritytothatof theType-IIat thelongerwavelength. Inthistype, asteep increaseinRrswasobservedfrom400to570nm.ThepeakRrs

was more prominent. The spectral behaviour of Rrs from

570 to 700nm in this water type was similar to that of Type-II.However,thesecondarymaximumwasmore promi-nent(684nm)andhigherinmagnitude.

4.2. Associationofopticallyactive substances withdifferentwatertypes

Theanalysesofspatio-temporalvariabilityinOAS(chl-aand CDOM) andtheir IOP (b650) have been carried out for the

Table1 Thefunctionalformsofthealgorithmsusedtogeneratechl-afromCoastalZoneColourScanner(CZCS),OceanColourand TemperatureScanner(OCTS),Sea-viewingWideField-of-viewSensor(SeaWiFS),MediumResolutionImagingSpectrometer(MERIS), ModerateResolutionImagingSpectroradiometer(MODIS)andOceanColourMonitor2(OC4O2).

Sensor Algorithm Functionalform Modeltype

CZCS OC3C MBR R¼log10 max Rrs443 Rrs550   ; Rrs520 Rrs550       a¼½0:3330;
4:3770;7:6267;
7:1457;1:6673 C¼10ða0þa1Rþa2R2þa3R3þa4R4Þ

Fourthorderpolynomial

OCTS OC4O MBR R_¼log10 max Rrs443 Rrs565   ; Rrs490 Rrs565   ; Rrs520 Rrs565       a¼½0:2399;
2:0825;1:6126;
1:0848;
0:2083 C¼10ða0þa1Rþa2R2þa3R3þa4R4Þ

Fourthorderpolynomial

SeaWiFS OC4 MBR R¼log10 max Rrs443 Rrs555   ; Rrs490 Rrs555   ; Rrs510 Rrs555       a¼½0:2515;
2:3798;1:5823;
0:6372;
0:5692 C_¼10ða0þa1Rþa2R2þa3R3þa4R4Þ

Fourthorderpolynomial

MERIS OC4E MBR R¼log10 max Rrs443 Rrs560   ; Rrs490 Rrs560   ; Rrs510 Rrs560       a¼½0:2521;
2:2146;1:5193;
0:7702;
0:4291 C¼10ða0þa1Rþa2R2þa3R3þa4R4Þ

Fourthorderpolynomial

MODIS OC3M MBR R_¼log10 max Rrs443 Rrs547   ; Rrs488 Rrs547       a¼½0:2424;
2:7423;1:8017;
0:0015;
1:2280 C¼10ða0þa1Rþa2R2þa3R3þa4R4Þ

Fourthorderpolynomial

OCM2 OC4O2 MBR R¼log10 max Rrs443 Rrs555   ; Rrs490 Rrs555   ; Rrs510 Rrs555       a_¼½0_:475;
3:029;2_{:240;
1:253;
0:027} C_¼10ða0þa1Rþa2R2þa3R3þa4R4Þ

Thirdordermodi_fied cubicpolynomial

three water types based on the frequency distribution (Fig. 3).The frequencydistribution plotswerealso super-imposedwiththenormalGaussianfunctiontounderstandthe trend.Thefrequencydistributionshowedlargevariabilityin theOASforthethreewatertypes.Thechl-a,whichvaried betweenthethreewatertypes,was0.077and20.0mgm
3 (Fig.3a).Thechl-ainType-IandType-IIwatersrangedfrom 0.77to1.0mgm
3and1.0and2.0mgm
3respectively.The Type-IIIwaters,chl-arangedfrom1to20mgm
3whilethe frequencywaslessatlowerconcentration.

The CDOM varied between 0.39 and 2.25ppb QSDE (Fig. 3b). Type-I waters showed lower concentration of CDOM,between0.3and0.75ppbQSDE.TheCDOM concen-trationinType-IIandType-IIIwatersarevaried,from0.75to 1.25andHp1.25to2.25ppbQSDErespectively.The concen-trationofCDOMwascomparativelyhigherinType-IIwaters. b650, ranged between0.1610
3 and9.510
3m
1sr
1

(Fig.3c)withahigherfrequencyatthelowerrangebetween 0.1610
3and0.1910
3m
1sr
1.InType-IIandType-III waters, b650 was in the range from 1.310
3 to

6.110
3m
1sr
1 and 1.410
3 to 9.510
3m
1sr
1 respectively.The maximumvalue ofb650wasencountered

inType-IIIwaters.

4.3. Assessment ofoceanchlorophyllalgorithms

Thescatterplotsshowingrelationbetweeninsitumeasured chl-aandthatderivedusingOC4,OC3C,OC4O,OC4E,OC3M and OC4O2 algorithms is given in Fig. 4. The validation statistics have been computed using data from all water types.Thecorrespondingstatisticalindicatorsaregivenin Table2.TheR2_{wasbetterincaseofOC4(0.70),whichwas}

alsocomparablewiththatofOC3M(0.68).FurtherOC4and OC3M showed least Log10-RMSE (0.37). However, absolute

percentagedifference (APD)(35.9%)was betterin caseof OC4 whereas OC3M showed better slope (0.69), relative percentagedifference(RPD)(4.8%)andunbiasedpercentage difference(UPD)(17.1%).Theintercept(0.07)wasleastin thecaseofOC4Oand'r'wasclosertounityincaseofOC4O2. TheinversetransformratiosalsoshowedthatOC3Mperforms better at the median scale with Fmed close to unity. The

overallstatisticalanalysisshowed thatOC3MandOC4 pro-ducedcomparableresultshavingleastsignificantdifference betweenestimatedandmeasuredchl-a.

Althoughthevalidationhasbeencarriedoutfortheentire dataset,theaccuracyofthealgorithmswasalsoevaluatedin thedifferentwatertypes.Theassessmentofmeasuredand Figure 3 Frequency distribution of chlorophyll-a (chl-a), volume scattering function at 650nm (_b650) and concentration of

chromophoricdissolvedorganicmatter(CDOM)inquininesulphatedihydrateequivalent(QSDE)unitindifferentwatertypes.The solidline indicatesthenormaldistribution. TheX-axisrepresents concentrationof chl-a,_b650and CDOM.The Y-axisrepresents

frquencydistribution.

Figure2 Spectralvariabilityofremotesensingreflectance(Rrs)inthestudyarea.Thesimilarspectrahavebeenclubbedtogetheras

(a)Type-I,(b)Type-IIand(c)Type-III.Thesolidlinerepresentmeanandthedottedlinerepresentsthestandarddeviation.TheX-axis representswavelength[nm]andY-axisrepresentsRrs[sr
1]coefficients.

estimatedchl-ainType-Iwatersshowedagoodagreement withacorrelationcoefficient(R2_{)of0.93inthecaseofOC4O,}

OC4,OC4EandOC3M and0.89in thecase ofOC3C.Chl-a estimatedusing OC4O2 algorithm showed relatively lower correlation (R2=0.74). However, the measured chl-a was lowerinmagnitudethantheestimatedchl-ahavingaslope of1.06,0.60,0.68,0.67,0.71and0.68forOC3C,OC4O,OC4, OC4E,OC3MandOC4O2respectively.Theresultbasedonthe statisticalanalysisindicatedthatOC3Mperformsbetterthan otheralgorithmsin Type-Iwaters.InType-IIwaters,itwas observedthataclusterofpointshavingchl-aconcentration between9.0and13mgm
3showedapoorrelationbetween estimatedandmeasuredchl-a.Atthesepoints,b650was

ten-foldhigherthanthatinType-Iwaters.Inaddition,theCDOM washigherthantheaveragevalue(2.1—2.5ppbQSDE).Inthe absenceofthiscluster,itwasobservedthatestimatedand measuredchl-a hadmoderate agreement in caseof OC3C (R2_{=0.52), OC4 (R}2_{=0.67), OC4E (R}2_{=0.67), OC3M}

(R2=0.67)andOC4O2(R2=0.67).In thiswatertype, esti-matedchl-awaslowerinmagnitudethanthemeasuredwith aslope of0.59,0.50,0.63,0.63,0.90and0.88for OC3C, OC4O,OC4,OC4E,OC3MandOC4O2respectively.Basedon thestatisticalanalysis,comparableresultwasseenbetween OC3MandOC4O2.Theperformanceofboththesealgorithms wasbetterthanotheralgorithmsinType-IIwaters.InType-III waters,estimatedchl-awasfoundtobehigherinmagnitude ascomparedtomeasuredchl-a.Theslopeofregressionwas 1.46,1.45,1.39,1.65,1.44and1.84forOC3C,OC4O,OC4, OC4E,OC3MandOC4O2respectively.Althoughtheestimated andmeasuredchl-awasfoundtobevariableinmagnitude, the trend was ingood agreement. The R2 was 0.68,0.63, 0.69,0.66,0.70and0.70forOC3C,OC4O,OC4,OC4E,OC3M and OC4O2 respectively. The OC3M and OC4 algorithms performedbetterinType-IIIwaters.

Anattempthasbeenalsomadetounderstandexistenceof anycovariancebetweenCDOManderrorsassociatedwiththe Table2 Performanceindicesforrelativeerrorsbetweeninsitumeasuredandestimatedchl-afrominsituRrsusingOC3C,OC4O,

OC4,OC4E,OC3MandOC4O2algorithms.Theseindicesincludecorrelationcoef_ficient(R2_{),slope(S),intercept(I),sumationofratio}

ofmeasuredtoestimated(r),rootmeansquarederror(RMSE),absolutepercentagedifference(APD),relativepercentagedifference (RPD)andunbiasedpercentagedifference(UPD).Thegeometricmeanandone-sigmarangeoftheratio(F=Valuealg/Valuemeas)are

givenbyFmed,Fmin,andFmax,respectively.Thevaluescloserto1aremoreaccurate.Total42datapointswereusedfortheanalysis.

Algorithm R2 S I r RMSE APD RPD UPD Fmin Fmed Fmax

OC3C 0.66 0.65 0.11 1.74 0.39 41.0 17.2 29.8 0.50 1.21 2.94 OC4O 0.65 0.63 0.07 1.95 0.41 40.3 23.6 36.5 0.56 1.36 3.34 OC4 0.70 0.67 0.10 1.69 0.37 35.9 16.3 26.7 0.53 1.22 2.82 OC4E 0.67 0.67 0.10 1.70 0.38 37.2 16.6 27.7 0.50 1.20 2.88 OC3M 0.68 0.69 0.16 1.47 0.37 50.4 4.8 17.1 0.44 1.04 2.45 OC4O2 0.54 0.54 0.35 1.21 0.45 47.1 3.26 30.1 0.29 0.82 2.29

Figure4 Correlation betweeninsitu chlorophyll-a(chl-a) and thatderived usingOC4, OC3C,OC4O,OC4E,OC3Mand OC4O2 algorithmsinType-I(plus),II(triangle)andIII(circle)waters.Thesolidlineindicatethetrendandthedottedlinecorrespondsto1:1.

algorithmsin Type-IIand IIIwaters. Hence aratio of esti-mated chl-a,from variousalgorithms,to in situ chl-awas regressed with CDOM. The results showed that in Type-II waters, the covariance was 2.48, 1.42, 2.24, 2.48, 5.16 and 2.99% whereas in Type-III it was 39.64, 25.96, 27.09, 30.46, 25.41 and 35.86% respectively for OC3C, OC4O,OC4,OC4E,OC3MandOC4O2algorithms.

5.

Discussion

Theretrievalofgeophysicalproductsfromanoceancolour satellitesensorinvolvestwocriticalsteps.Thefirststepisto eliminatethe effectof theatmospheric contribution from thetotalradiancereceivedbythesatellitesensor(Huetal., 2004). The second step is to selecta suitable bio-optical algorithmwhichrelatesthesignalleavingthewatercolumn withtheconcentrationoftheOAS.Thepresentstudyaims towardsthesecondstep.Inthepresentcase,theempirical algorithms,usedoperationallytoretrievechl-afromCZCS, OCTS,SeaWiFS,MERIS,MODISandOCM2,wereappliedtothe insituRrsmeasuredusinghyperspectralradiometer.Further,

anattempthasbeenalsomadetoclassifythewatersbased onthereflectancecharacteristicsandtoevaluateimpactof OASonthesewatertypes.Subsequentlytheperformanceof thesealgorithmswastestedinthesewatertypes.

Thedatageneratedforthepresentstudyexhibitedthree distinctspectrafor Rrs. Thedistributionof OASwithin

dif-ferentwatertypesalsoshowedaremarkabledifference.In Type-Iwaters,Rrsspectrashowedthemaximumintheblue

(400—480nm) region andwas almost negligible inthe red region(beyond600nm).Thevariabilityinconcentrationof chl-a and CDOM was very low in this water type with a standard deviationof less than50% of the average value. Theb650alsoshowedlessvariabilityinthiswatertype.The

lowconcentrationindistributionofOASinType-Iwaterswas responsible for least absorptionin the shorterwavelength resulting in highRrs. However, thelower Rrs in thelonger

wavelengthmaybeattributedtothestrongabsorptiondueto water molecules.This indicatesthat the water molecules weretheprincipallightabsorbingcomponent inthiswater typealongwithchl-a.Thisisinagreementwiththeprevious studies carried outin oligotrophic watersof Lakshadweep Islands, hyperoligotrophicwaters in the South Pacific gyre andinthe ChesapeakeBaywheretheauthors have shown thatthewatermoleculesarethemajorcomponents respon-siblefortheabsorptioninthelongerwavelengthresultingin minimalvariabilityofRrs(Menon etal.,2005;Moreletal.,

2007;Tzortziouetal.,2007).ThespectralsignatureofRrsin

Type-IIwatersshowedsignificantdifferenceinshorter wave-lengths.InType-IIwaters,chl-aconcentrationandb650was

ten-foldhigherthaninType-Iwaters.Also,theCDOM con-centrationwasalmosttwiceinmagnitude.The phytoplank-tonpigment,chl-a,hastheprimaryabsorptionpeakinthe blue region (440nm). Apartfrom this, CDOMalso has a tendency for strong absorption in UVandthe blue region (Jorgenson,1999;Menon etal.,2005;Siegal etal.,2002). Therefore,therespectivesignatureofRrsinType-IIwaters

waspredominantlyduetothecombinedeffectofabsorption duetochl-aandCDOM.InType-IIIwaters,thepeakintheRrs

spectra shifted more towards the longer wavelength as compared to that in Type-II waters. In addition,the peak RrswasmoreprominentascomparedtoType-IandIIwaters.

Inthiswatertype,b650wascomparablewithType-IIwaters.

However, chl-a and CDOM concentration was found to be increasedby64%and76%respectively.Thisindicatesthatthe impactofchl-aandCDOMabsorptionhasfurtherincreased whichresultedinmuchlowerRrsattheshorterwavelength.

Menon et al.(2006) showed that CDOM significantly influ-encesthewaterleavingradianceattheshorterwavelengths andits impact canbe spread upto 650nm. Kutser etal. (2006)reportedan'abnormal'shapeoftheRrsspectrawhen

theconcentrationsofopticallyactivesubstanceswerehigh. ThestudycarriedoutbyCannizaroandCardershowedthat, in non-coastal oligotrophic waters, the peak Rrs was at

400nm whichwas shiftedto 490nm inhighlyreflective, opticallyshallow,mesotrophicwatersand560nmin opti-callydeepeutrophicwaters.OuillonandPetrenko(2005)also reportedpeakRrsat443and490nminCase1watersandat

560nminCase2waters.

Theabovediscussionclearlyindicatesthatdistinct spec-tral signatures of Rrs weremore attributed towards

varia-bilityinOAS.Further,thedistributionoftheseOAScouldbe controlledbyhydrographyofthestudyarea.Thestudyarea is subjected to coastal upwelling and heavy fresh water dischargefromtheestuaryofthePeriyarRiverinthe mon-soon period which enhances nutrient upload resulting in increased biological production (Jyothibabu et al., 2006; Nair et al., 1992). Studies by Srinivas et al. (2003) also reportedthatthecoastalwatersofthesoutheasternArabian Seaaretherecipientofapproximately1.91010m3offresh waterannuallyfrom theCochin backwaters.Wehave also analyzedtheinter-relationshipofchl-awithCDOMandb650in

differentwatertypes(Fig.5).Thechl-aandCDOMdidnot showanysignificantrelationinType-Iwaters.Thisindicates thatCDOMisindependentofthevariationsinchl-ainthese waters.HoweveralargeassociationwasseeninType-IIand Type-IIIwaters. Alsointhese waters,the concentrationof chl-aandCDOMwasveryhigh.Theco-varianceofchl-awith CDOMindicatesthatthesourceofnutrientsfor phytoplank-ton growth, and organic matter is the same. Since the primarysourceofCDOMisfromtheriver,itcanbeinferred thatriverdischargesareoneoftheprimarymechanismsfor distributionofchl-aandCDOMinthestudyarea.Theb650did

notshowanysignificantrelationshipwithchl-ainanyofthe watertypes.Thisindicatesthatchl-aandsuspendedmatter donotco-varyinwatertypes.Hence,itcanbeinferredthat river discharges are not the primary source of suspended matterin thestudyarea. Someof thepreviousstudiesby Jyothibabuetal.(2006),Srinivas etal.(2003)andThomas etal.(2004)alsoreportedthatcontinuousdredgingprocess occurringthroughouttheyearinthestudyareaincreasesthe nutrients and sediment loads into the estuary which is drainedintothecoastalwaters.

ThedistributionofOASthussignificantlyaltersthe spec-tralRrswhichistheinputfortheempiricalalgorithmforthe

retrievalofchl-a.Inthepresentstudy,chl-aalgorithmswere evaluatedindifferentwatertypeshavingsimilarvariability in spectral Rrs. Among all the algorithms, OC3M and OC4

performedbetterinthestudyarea.Ingeneral,allalgorithms showedthat,inType-Iwaters,themeasuredchl-awasless thanthatestimatedusingvariousalgorithms.However mea-suredandestimatedchl-ashowedthesametrendwithgood correlation.ThesewaterswerehavingtypicalCase1 char-acteristics.AlsotheRrssignalsdominatedin443nmband.In

Type-IIwaters,theclusterofpointswasfoundtobeweakly correlated.FurtherinType-IIwaters,thedominantsignalin Rrswasfromtheblue-greenband.InthecaseofOC4,OC4E

andOC4O2thedominantRrssignalwasfrom510nm,whereas

itwas520nminthecaseofOC3CandOC4O.ForOC3M,Rrs

signalwasdominatedbythe488nmband.Similarconditions wereobservedinType-IIIwaters.Theperformanceofallthe algorithmswas poorinType-IIwaterswhereasitwas com-parativelybetterinType-IIIwaters.AlthoughinType-IIand Type-IIIwatersconcentrationofchl-aandCDOMwasten-fold higherthanType-Iwaters,theco-variancebetweenchl-aand CDOMwassignificantlyhigherinType-IIIwaters.Desaetal. (2001)showedthattheratioofRrsat490—555nmisabetter

indicatorofchl-aintheeasternArabianSea.Theyillustrated significantimprovementinR2_{(0.93),slope(0.96)and}

inter-cept(0.26)bymodifyingSeaWiFSOC4v4coefficients. How-everShanmugam(2011)suggestedthatalthoughOC3reliably estimateschl-aintheopenoceanwaters,ittendstofailin thecoastalwatersoftheArabianSea.Thechl-aandCDOM competeforabsorptionoflightatalmostsimilarwavelengths intheblueregion.Asaresult,thesignalemergingfromthe watercolumncarriessignaturesofbothchl-aandCDOM.The empiricalalgorithmstaketheratioofthebluetothegreen band with an assumption that the water leaving radiance decreasesin theblue band with an increasein chl-a con-centration.However, ifthewatercolumn isdominatedby chl-aandCDOM,bothsignificantlycontributetothedecrease inthe water leavingradiance in the blueband. In such a scenario,theperformanceofaratiobasedalgorithms weak-enstheretrieval ofchl-a.Intheearlierstudies,Tzortziou etal.(2007)alsoreportedthatthefailureofMODISalgorithm ininshorewatersoftheChesapeakeBaywasduetothelarge contributionofnon-co-varyingCDOMandnon-algalparticles tototallightabsorptionintheblueregion.

6.

Conclusion

Thepresentstudyprimarilyfocusedontheevaluationofsix empiricalalgorithms (OC3C,OC4O, OC4, OC4E,OC3M and OC4O2), operationally used to retrieve chl-a from CZCS, OCTS, SeaWiFS, MERIS, MODIS and OCM-2. The algorithms were applied to the spectral Rrs measured in situ using

hyperspectral radiometer with an intention to assess the functional form and the coefficients. The effect of atmo-sphericcorrectionhasbeenignoredinthepresentstudy.The

Rrsshowedthreedistinctwatertypesbasedonthespectral

variability.Thesewatertypeswereattributedtothe varia-bility in theconcentration of OAS. In Type-Iwaters chl-a, CDOMandb650wereverylowindicatingthatthesewerepure

Case-IwatersandthepeakRrswasintheblueband.InType-II

waters, thepeak Rrsshifted to the greenband which was

attributed to elevated concentration of chl-a, CDOM and b650.InType-IIIwaters,thepeakRrsfurthershiftedtolonger

wavelengthsduetoanincreaseinchl-aandCDOM.Alsochl-a wasfoundtobeassociatedwithCDOMindicatingthatrivers are one of the primary sources for discharging essential parameters for the growth of phytoplankton. The overall statisticalanalysis showedthat theperformancesofOC3M and OC4 were better as compared to other algorithms. Further in the case of chl-a, at more than 1.0mgm
3 it wasfoundthattheratioofhigherwavelengths(488,510and 520nm)dominates.Theassessmentofalgorithmsin differ-ent water types indicated better performance of all the algorithmsin Type-I waters.The performance was poor in Type-IIandType-III waters.The errors associatedwith the estimationofchl-aweresignificantlyco-variedwithCDOMin Type-III waters. Therefore, in the regions where there is dominanceofOASotherthanchl-a,itisrequiredtodevelop IOP-basedalgorithmthattakesintoaccountabsorptionand scatteringduetoindividualOAS.

Acknowledgements

Thestudywas fundedby IndianNationalCentrefor Ocean InformationServices(INCOIS)(INCOIS:F&A:XII:D2:021), Min-istryofEarthSciencesunderSATelliteCoastalOceanographic REsearch(SATCORE)programme.Theauthorsarethankfulto theDirectorofINCOISforencouragement.Thanksarealso duetotheDirectorofICAR—CentralInstitute ofFisheries Technology,forsupportandencouragement.Theauthorsare also gratefulto themanagement andstaff of MV 'Bharath Darshan',fortheirunrestrictedsupportduringthecruises.

References

Balch, W.M.,Drapeau,D., Fritz,J., Bowler, B., Nolan, J., 2001. OpticalbackscatteringintheArabianSea_—continuousunderway measurementsofparticulateinorganicandorganiccarbon. Deep-SeaRes.Pt.I48(11), 2423—2452,http://dx.doi.org/10.1016/ S0967-0637(01)00025-5.

Figure5 Scatterplotshowingrelationbetween(a)insituchlorophyll-a(chl-a)andCDOMinQSDEunitand(b)insituchl-aandb.The plus(+)signcorrespondstoType-I,opentriangle(~)correspondstoType-IIandopencircles(*)correspondstoType-IIIwaters.

Chauhan, P., Mohan, M., Sarangi, R.K., Kumari, B., Nayak, S., Matondkar,S.G.P., 2002.Surfacechlorophyllestimationinthe ArabianSeausingIRS-P4OCMOceanColorMonitor(OCM)satellite data.Int.J.RemoteSens.23(8),1663—1676,http://dx.doi.org/ 10.1080/01431160110075866.

Desa,E.S.,Suresh,T.,Matondkar,S.G.P.,Desa,E.,2001.Seatruth validationofSeaWiFSoceancoloursensorinthecoastalwatersof theeasternArabianSea.Curr.Sci.India80(7),854_—860.

George,R.,Muraleedharan,K.R.,Martin,G.D.,Sabu,P.,Gerson,V. J.,Dineshkumar,P.K.,Nair,S.M.,Chandramohanakumar,N.,Nair, K.K.C.,2013.NutrientbiogeochemistryoftheeasternArabian Seaduringthesouthwestmonsoonretreat.Environ.EarthSci.68 (3),703_—718,http://dx.doi.org/10.1007/s12665-012-1772-2. Hu,C.,Montgomery,E.T.,Schmitt,R.W.,Muller-Karger,F.E.,2004.

The dispersalof theAmazon and Orinoco River waterin the tropical Atlantic and Caribbean Sea: observation from space and S-PALACE floats.Deep-SeaRes. Pt. II 51 (10—11), 1151— 1171,http://dx.doi.org/10.1016/j.dsr2.2004.04.001.

Iyer,P.C.S.,Sindhu,M.,Kulkarni,S.G.,Tambe,S.S.,Kulkarni,B.D., 2003. Statistical analysis of the physic-chemical data on the coastalwatersofCochin.J.Environ.Monitor.5(2),324—327,

http://dx.doi.org/10.1039/B209219K.

Jorgenson,S.V.,1999.StandardCase1algorithmsinDanishcoastal waters.Int.J.RemoteSens.20(7),1289—1301,http://dx.doi. org/10.1080/014311699212731.

Joshi,M.,Rao,A.D.,2012.ResponseofSouthwestmonsoonwindson shelfcirculationoffKeralacoast,India.Cont.ShelfRes.32,62_— 70,http://dx.doi.org/10.1016/j.csr.2011.10.015.

Jyothibabu,R.,Madhu,N.V.,Jayalakshmi,K.V.,Balachandran,K.K., Shiyas,C.A.,Martin,G.D.,Nair, K.K.C.,2006.Impactoffresh waterin_fluxonmicrozooplanktonmediatedfoodwebinatropical estuary(Cochinbackwaters_—India).Estuar.Coast.ShelfSci.69 (3—4),505—518,http://dx.doi.org/10.1016/j.ecss.2006.05.013. Kumar, H.P.V.,Kumar, M.N.,1996. Onthe flowand thermohaline structureoffCochinduringpre-monsoonseason.Cont.ShelfRes. 16(4),457—468,http://dx.doi.org/10.1016/j.ecss.2006.05.013. Kutser, T., Vahtma, E., Martin, G., 2006. Assessing suitability of multispectral satellites for mapping benthic macroalgal cover inturbidcoastalwatersbymeansofmodelsimulations.Estuar. Coast.ShelfSci.67(3),521—529,http://dx.doi.org/10.1016/j. ecss.2005.12.004.

Lee,M.E.,Lewis,M.R.,2003.Anewmethodforthemeasurementof theopticalvolumescattering functionintheupperocean.J. Atmos. Ocean. Tech. 20 (4), 563_—571, http://dx.doi.org/ 10.1175/1520-0426(2003)20_{<563:ANMFTM>2.0.CO;2}.

Lee,Z.,Carder,K.L.,Arnone,R.A.,2002.Derivinginherentoptical propertiesfromwatercolor:amultibandquasi-analytical algo-rithmforopticallydeepwaters.Appl.Opt.41(27),5755_—5772,

http://dx.doi.org/10.1364/AO.41.005755.

McClain,M.E.,Fargion,G.S.,1999.Simbiosprojectannualreport. DocumentID:20000013962,Report/PatentNumber: NASA/TM-1999-209486,NAS1.15:209486Rept-2000-00655.GoddardSpace FlightCenter,Greenbelt,Maryland,132pp.

Mckee,D.,Cunningham,A.,2006.Identificationand characteriza-tion oftwo optical watertypes inthe Irish Sea from insitu inherentopticalproperties and seawaterconstituents. Estuar. Coast.ShelfSci.68(1),305—316,http://dx.doi.org/10.1016/j. ecss.2006.02.010.

Menon,H.B.,Lotliker,A.A.,Moorthy,K.K.,Nayak,S.R.,2006. Vari-abilityofremotesensingre_flectanceandimplicationsforoptical remote sensing_—a studyalong theeasternand northeastern watersofArabianSea.Geophys.Res.Lett.33,L15602,http://dx. doi.org/10.1029/2006GL026026.

Menon,H.B., Lotliker,A.A.,Nayak,S.R.,2005. Pre-monsoon bio-opticalpropertiesinestuarine,coastalandLakshadweepwaters. Estuar. Coast. Shelf Sci. 63 (1), 211_—223, http://dx.doi.org/ 10.1016/j.ecss.2004.11.015.

Minu,P.,Lotliker,A.A.,Shaju,S.S.,SanthoshKumar,B.,Ashraf,P.M., Meenakumari,B.,2014.Effectofopticallyactivesubstancesand atmosphericcorrectionschemesonremote-sensingreflectance atacoastalsiteoffKochi.Int.J.RemoteSens.35(14),5434— 5447,http://dx.doi.org/10.1080/01431161.2014.926420. Mobley,C.D.,1994.Theopticalpropertiesofwater. In:Bass,M.

(Ed.),HandbookofOptics,2nded.,vol.I.McGraw-Hill,NewYork, 1248pp.

Morel,A.,1991.Lightandmarinephotosynthesis:aspectralmodel withgeo-chemicalandclimatologicalimplications.Prog. Ocea-nogr.26(3),263_—306,http://dx.doi.org/10.1016/0079-6611(91) 90004-6.

Morel, A., Claustre, H., Antoine, D., Gentili, B., 2007. Natural variabilityofbio-opticalpropertiesinCase1waters:attenuation andreflectancewithinthevisibleandnear-UVspectraldomains, asobservedinSouthPacificandMediterraneanwaters.Biogeosci. Discuss.4(4),2147—2178.

Morel,A.,Prieur, L.,1977. Analysis ofvariationsin ocean color. Limnol.Oceanogr.22(4),709—720.

Nagamani,P.V.,Chauhan,P.,Dwivedi,R.M.,2008.Developmentof chlorophyllaalgorithmforOceanColourMonitoronboard OCEAN-SAT-2satellite.IEEEGeosci.RemoteSens.Lett.5(3),527—531,

http://dx.doi.org/10.1109/LGRS.2008.923213.

Nair,V.E.,Devassy,V.P.,Madhupratap,M.,1992.Bloomsof phyto-planktonalongthecoastofIndiaassociatedwithnutrient enrich-mentandtheresponseofzooplankton.In:Vollenweider,R.A., Marchetti,R.,Viviani,R.(Eds.),MarineCoastalEutrophication. Elsevier,Amsterdam,London,NewYork,Tokyo,819_—828.

O'Reilly,J.E.,Maritorena,S.,Mitchell,B.G.,Siegel,D.A.,Carder,K. L.,Garver,S.A.,Kahru,M.,McClain,C.R.,1998.Oceancolour chlorophyllalgorithmsforSeaWiFS.J.Geophys.Res.103(C11), 24937_—24953,http://dx.doi.org/10.1029/98JC02160.

O'Reilly,J.E.,Maritorena,S., Siegel,D.,O'Brien,M.C.,Toole,D., Mitchell,B.G.,etal.,2000.Oceancolorchlorophyllaalgorithms forSeaWiFS,OC2,andOC4:Version4.In:Hooker,S.B.,Firestone, E.R.(Eds.),SeaWiFSpostlaunchtechnicalreportseries.SeaWiFS post-launch calibration and validation analyses, Part 3, vol. 11NASA/GSFC.9—23.

Ouillon,S.,Petrenko,A.,2005.Above-watermeasurementsof re-flectanceandchlorophyll-aalgorithmsintheGulfofLions,NW MediterraneanSea.Opt.Exp.13(7),2531—2548,http://dx.doi. org/10.1364/OPEX.13.002531.

Pierson,D.C.,Strombeck,N.,2000.Amodellingapproachto evalu-atepreliminaryremotesensingalgorithms:useofwaterquality datafromSwedishGreatLakes.Geophysica36(1_—2),177_—202.

Sathe,P.V.,Jadhav,N.,2001.Retrievalofchlorophyllfromthe sea-leavingradianceintheArabianSea.J.IndianSoc.RemoteSens. 29(1_—2),97_—106.

Shanmugam,P.,2011.Anewbio-optical algorithmfortheremote sensingofalgalbloomsincomplexoceanwaters.J.Geophys. Res.-Oceans116,C04016,http://dx.doi.org/10.1029/2010JC006796. Siegal,D.A.,Maritorena,S., Nelson, N.B., Hansel, D.A.,

Lorenzi-Kayser,M., 2002. Globaldistribution anddynamics ofcolored dissolvedanddetritalorganicmaterials.J.Geophys.Res.107, 3228,http://dx.doi.org/10.1029/2001JC000965.

Srinivas,K.,Dineshkumar, P.K.,2006.Atmospheric forcingonthe seasonalvariabilityofsealevelatCochin,SouthwestofIndia. Cont.ShelfRes.26(10),1113—1133,http://dx.doi.org/10.1016/ j.csr.2006.03.010.

Srinivas, K., Revichandran, C., Maheswaran, P.A., Ashraf, M.T.T., Murukesh, N., 2003. Propagation of tides in the Cochin estuarinesystem,SouthwestcoastofIndia.IndianJ.Mar.Sci. 32(1),14_—24.

Thomas,J.V.,Premlal,P.,Sreedevi,C.,Kurup,M.B.,2004. Immedi-ateeffectofbottomtrawlingonthephysicochemicalparameters intheinshorewaters(Cochin-Munambum)ofKerala.IndianJ. Fish.51(3),277_—286.

Tilstone,G.H.,Ingrid,M.,Benavides,A.,Pradhan,Y.,Shutler,J.D., Groom,S., Sathyendranath,S., 2011.Anassessmentof chloro-phyll-aalgorithmsavailableforSeaWiFSincoastalandopenareas oftheBayofBengalandArabianSea.RemoteSens.Environ.115 (9),2277—2291,http://dx.doi.org/10.1016/j.rse.2011.04.028. Tilstone,G.H.,Lotliker,A.,Miller, P.I.,Ashraf,P.M.,Kumar, T.S.,

Suresh,T., Ragavan, B.R., Menon,H.B., 2013. Assessment of MODIS-Aquachlorophyllaalgorithmsincoastalandshelfwaters

of the eastern Arabian Sea. Cont. Shelf Res. 65 (1), 14—26,

http://dx.doi.org/10.1016/j.csr.2013.06.003.

Tzortziou, M., Subramaniam, A., Herman, J.R., Gallegos, C.L., Neale,P.J.,HardingJr.,L.W.,2007.Remotesensingreflectance and inherent optical properties in the mid Chesapeake Bay. Estuar. Coast. Shelf Sci. 72 (1), 16—32, http://dx.doi.org/ 10.1016/j.ecss.2006.09.018.