海洋生物地球化學(xué)性質(zhì)的水色遙感反演與應(yīng)用:方法和挑戰(zhàn)

李忠平，崔廷偉，孫 凌

(1．馬薩諸塞大學(xué)波士頓分校環(huán)境學(xué)院，波士頓，MA 02125，美國;2．國家海洋局第一海洋研究所，山東 青島266061;3．中國氣象局中國遙感衛(wèi)星輻射測量和定標(biāo)重點(diǎn)開放實(shí)驗(yàn)室，北京100081;4．中國氣象局國家衛(wèi)星氣象中心，北京100081)

1 Introduction

Ocean covers about 71%of the earth surface．The physical and biogeochemical processes in the ocean play critical roles in modulating the weather and climate of the earth system．In particular，photosynthesis by phytoplankton in the upper oceans provides the basic energy block to support the food web in the oceans and to draw down atmospheric CO2via the“biological pump”［1］．It is not only important but also critical to have accurate and adequate description of the spatial distributions of phytoplankton in the upper oceans and its temporal variations in a changing climate．

Because of the vast size of the oceans，it is impossible to adequately sample the ocean properties with ship surveys．Observation by satellites is the only feasible means to achieve adequate and repetitive basin－scale measurements of biogeochemical properties．And indeed，as demonstrated by the Coastal Zone Color Scanner(CZCS)，the first ever ocean color satellite，large scale variations of chlorophyll concentration(［Chl］，mg/m3)in the surface oceans can be well retrieved from the measurement of ocean color from the satellite altitude，which coined the term“ocean color remote sensing”．

The CZCSwas intended for an approval of the concept，and the eight years(1978－1986)of measurements revolutionized the study of biological oceanography in global scales．More ocean color satellite sensors were launched subsequently in the 1990s and 2000s，with more sophisticated sensors on the horizon or planned in the coming years．This short article provides an overview of a few latest sensors and the status in deriving biogeochemical products from the measurement of color，along with examples of their applications．

Tab．1 Ocean color satellite sensors and their major specs

2 Ocean color satellite sensors

Ocean color satellites，carrying radiometers to image the oceans in visible and near infrared bands，are designed for obtaining radiance originating from below the surface and emitting into space．This spectral radiance can be further used in various algorithms to estimate suspended sediment concentration，diffuse attenuation coefficient，and other bio－optical parameters．Table 1 provides a list of widely known satellites and ocean color sensors．

There have been several important milestones in the nearly four decades of ocean color satellite development．While CZCS was the first ocean color satellite sensor，SeaWiFS had significant advances in sensor technology and data processing．On the other hand，MODIS provides integrated observation of ocean，atmosphere and land．GOCI opened the era of ocean color measurements on a geostationary orbit．Hyperion and HICO achieved high spectral resolution and high spatial resolution at the same time．

Satellite Nation/Region Sensors Orbit information IRS－P3(1996/3～2004/5) India DLR(Germany)Resolution:500 m 18 bands(408～1 600 nm)Swath:200 km Sun－synchronous orbit Height:817 km Angle:99．05°ADEOS－Ⅰ(1996/8～1997/6) Japan OCTS Resolution:700 m 12 bands(408～1 600 nm)Swath:1 400 km POLDER(France)Resolution:6～7 km 15 bands(443～910 nm)Swath:2 400 km Sun－synchronous orbit Height:831 km Angle:98．6°Orbview－2(1997/9～2002/9) USA SeaWiFS 8 bands(402～885 nm)Resolution:1．1 km Swath:2 800 km Sun－synchronous orbit Height:705 km Angle:98．2°ROCSAT－1(1999/1～2004/6)Chinese Taiwan OCI Resolution:825 m 6 bands(433～12 500 nm)Swath:690 km Sun－synchronous orbit Height:600 km Angle:35°IRS－P4(OCEANSAT)(1999/5～) India OCM Resolution:350 m 8 bands(402～885 nm)Swath:1 420 km Sun－synchronous orbit Height:720 km Angle:98．8°KOMPSAT－1(1999/12～2008/1) Korea OSMI Resolution:850 m 6 bands(400～900 nm)Swath:800 km Sun－synchronous orbit Height:685 km Angle:98．1°TERRA(1999/12～)USA AQUA(2002/5～)MODIS 36 bands(410～14 385 nm)Resolution:1．0 km/band 8－36 500 m/band 3－7 250 m/band 1 and 2 Swath:2 330 km Sun－synchronous orbit Angle:98．2°Height:705 km ENVISAT(2002/5～2012/4) Europe MERIS 15 bands(412～900 nm)Resolution::300 m Swath:1 150 km Sun－synchronous orbit Height:800 km Angle:98°SZ－3(2002/3～2002/9) China CMODIS Resolution:400－500 m 34 bands(400～12 500 nm)Sun－synchronous orbit Height:330 km Angle:42．2°HY－1A(2002/5～2004/3)China HY－1B(2007/4～)COCTS Resolution:1．1 km 10 bands(402～12 500 nm)CCD Resolution:250 m Sun－synchronous orbit Height:798 km Angle:98．8°ADEOS－Ⅱ(2002/12～2003/10) Japan GLI Resolution:250/1 000 m 36 bands(375～12 500 nm)Swath:1 600 km POLDER－2(France)Resolution:6 000 m 9 bands(443～910 nm)Swath:2 400 km Sun－synchronous orbit Height:800 km Angle:98．6°

Satellite Nation/Region Sensors Orbit information Parasol(2004/12) France POLDER－3 Resolution:6 000 m 9 bands(443～1 020 nm)Swath:2 100 km Sun－synchronous orbit Height:750 km Angle:98．2°FY－3A(2008/5～)China FY－3B(2010/9～)MERSI 20 bands(400～12 500 nm)1．0 km/band 6～20 250 m/band 1－5 Sun－synchronous orbit Height:836．4 km Angle:98．8°ISS(2009/9～) USA HICO 400～900 nm Resolution:90 m Sun－synchronous orbit Height:343 km Angle:51．6°COMS－1(2010/6～) Korea GOCI 8 bands(412～865 nm)Resolution:500 m 8 images in a day Earth－synchronous orbit Height:35 800 km NPP(2011/10～) USA VIIRS Resolution:370/740 m 22 bands(402～11 800 nm)Swath:3 000 km Sun－synchronous orbit Height:824 km Angle:98．7°

3 Calibration and atmospheric correction

3．1 Vicarious calibration

Ocean color is measured by the spectrum of waterleaving radiance(Lw，W/m2/nm/sr)，i．e．light originated from below the surface emitting into space．Different from measurements made at the sea surface，Lwmakes just about 10%of the total radiance measured at the satellite altitude，where the other～90% resulted from atmospheric process are noises in this regard．Inversely，to achieve accurate estimation of biogeochemical properties from satellite ocean color measurements，it is critical to have an excellent sensor calibration as well as accurate correction of the atmospheric contributions［2］．

Ocean－color satellite sensors always go through dedicated and comprehensive radiometric characterization and calibration in the lab before sending to space．This labderived calibration factor，however，could not be maintained during the process of sending the instrument to space and due to the exposure to high－dose of solar radiation that includes extremely strong UV photons on top of the atmosphere．Therefore it is essential to conduct in－orbit calibration to monitor periodically the sensor’s sensitivity and to update the radiometric gain factors．This is achieved by comparing satellite sensor measured top－ofatmosphere radianceand the computed top－of－atmosphere radianceover the measurement target，a process termed as vicarious calibration［3，4］．In general，top－of－atmosphere radiance(Ltoa)can be expressed as［4，5］

HereLrandLarepresent contributions from Rayleigh scattering and aerosol(including Rayleigh－aerosol)scatterings［6，7］;whileLWCandLgrepresents contributions from the white cap and sun glint，both are surface effects．tis the diffuse transmittance of the atmosphere，andTis the direct transmittance［4］．

To vicariously calibrate an ocean－color satellite sensor，satellite measurements with minimum or no surface effects are necessary［8］，and in generalLWCis negligible for low－wind conditions［9］，so Eq．1 can be simplified to［5］，

and the vicarious gain factor is calculated as

Here the denominator on－the－right side of Eq．3 is measurement from a satellite sensor，while the nominator is calculated for the target viewed by the satellite sensor．And this vicarious gain，after it is matured and well established，is then applied to the entire satellite images for the derivation of bio－geochemical properties in the global oceans．In other words，is the“truth”for this calibration process，and the gain factor serves as an adjustment of the pre－launch radiometric calibration，“…thereby accounting for undetermined post－launch changes to the instrument．”［10］

Presently the industry standard to vicariously calibrate ocean－color satellite sensor is the measurements taken at the Marine Optical Buoy(MOBY)［11］，which is located～20 Km west off Island Lanai，Hawaii．All modern ocean－color satellite sensors that include SeaWiFS，MODIS，and VIIRSwere(are)vicariously calibrated with MOBY radiometric measurements．A similar buoy(BOUSSOLE;http://www．obs－vlfr．fr/Boussole/html/project/introduction．php)was deployed in the Mediterranean Sea by the European Space Agency to take measurements for calibrating ESA ocean－color satellite sensors．

MOBY has three radiometers placed at 1 m，5 m，and 9 m below the surface，respectively，to measure upwelling radiance(Lu)at these depths．An irradiance sensor is placed above surface to measure downwelling irradiance just above the surface(Es)for the calculation of remote－sensing reflectance(Rrs，ratio ofLwtoEs)．The sensors are hyperspectral，then theLwdata can be spectrally convolved for bands of the various satellite sensors for vicarious calibration．After obtaining enough(normally＞40)matchup measurements between satellite and MOBY，the calibration coefficient of the satellite sensor can be adjusted from these matchup data．

3．2 Atmospheric correction

As mentioned in Section 3．1，Ltoameasured by satellite sensor is dominated by the atmosphere，and ocean color(Lw)that carries useful information concerning biogeochemical properties of the water body makes only minor contributions．Accurate retrieval ofLwfrom TOA radiances requires precise removal of the atmospheric and ocean surface contributions，a process termed as atmospheric correction，a key step in ocean color data processing．

To obtainLw，the contribution of white cap，sun glint，scattering of atmospheric molecules，scattering and absorption of aerosols，transmittance(including effects due to gaseous absorption)must be determined and removed．In practice，observations under extreme conditions with significant contamination of white cap(due to strong wind)and sun glint(due to sun－sensor geometry)are avoided or excluded．In Eq．1，Rayleigh scattering can be precisely calculated given angles and sea surface pressure;white cap radiance can be estimated using an empirical model with wind speed;sun glint radiance can be estimated using rough surface model(e．g．，the Cox and Munk model［12］with wind and geometry)．Because of the large variation with time and location，aerosol properties must be determined in the atmospheric correction process to derive the contributions of aerosol radiance and transmittance．

For“case 1”clear waters［13，14］，because the water－leaving radiance in the near－infrared(NIR)can be neglected due to the strong absorption of water，the aerosol information(aerosol type and optical density)can be determined using measurements at two NIR bands，and then the aerosol radiance at other visible bands could be calculated．This dark pixel method was first presented by Gordon［15］，and subsequent operational atmospheric correction algorithms are based on this framework［16－18］．Look－up tables generated from pre－calculated radiative transfer simulations under various atmospheric and geometries are usually adopted for efficiency．For instance，to implement the extrapolation of aerosol contribution from NIR to the visible，SeaWiFSand MODISuse the ratio of aerosol single scattering reflectanceρa(bǔ)s(defined as(πLa)/(F0cos(θs))，withF0for extraterrestrial irradiance andθsfor solar zenith angle)［7］to determine the aerosol type．The dark pixel assumption and proper aerosol models are the two bases of operational atmospheric correction algorithms．

For turbid case 2 water，the dark pixel assumption of negligibleLwat NIR is invalid，even in“case 1”waters with chlorophyll concentration higher than 2 mg/m3［19］．In these waters，the water－leaving signal at NIR could result in over－correction of the atmospheric contributions and result in under－estimated water－leaving radiance，even with negative values，especially in the shorter wavelength(such as 412 and 443 nm)．Strong absorptive aerosol is another factor resulting in poor performance of NIR atmospheric correction［20］．

Many regionally modified algorithms have also been developed．Various spectral relationships of water，and/or aerosol，derived fromin－situmeasurements or simplified inherent optical property(IOP)［21］models are adopted to compensate the non－zeroLw［22－24］．IOP models are also included in the atmospheric correction process，thus properties of aerosol and water are determined simultaneously using iterative optimization［25－27］．Due to the strong absorption of water at short－wave infrared(SWIR)，SWIR instead of NIR bands are adopted for atmospheric correction［28，29］with a black pixel assumption at the SWIR bands for extremely turbid waters．Due to the fact that water－leaving radiance at ultraviolet(UV)can be neglected as compared with that at visible or NIR in most cases of highly turbid waters，UV band is also used to estimate the aerosol contribution［30］．In addition to iterative optimization，neural networks trained by simulated data is another way to achieve rapid atmospheric correction in complex environments［31，32］．

4 Ocean color products and algorithms

4．1 Concentration of chlorophyll-a

Starting from the era of the Coastal Zone Color Scanner(CZCS)［33］，historically，a key aspect of the satellite mission is to retrieve chlorophyll－a concentration(［Chl］)from the measured ocean color．This is based on the understanding that 1)［Chl］is an important index of the ecosystem status，and 2)［Chl］is a key parameter to scale－up discrete measurements of primary production to global estimates［34－36］．Because of such importance，［Chl］has been one of the key products of all satellite ocean color missions，with standard algorithms taking the blue－green ratio of measured ocean color for its calculation［2，37］．Many studies［2，38］have pointed out the caveats of such a simple approach in estimating［Chl］，which include the requirement of all optically active constituents co－vary with chlorophyll－a and a stable(universally fixed)relationship between the phytoplankton absorption coefficient and［Chl］．Numerous studies［39－42］have shown that these relationships are highly variable，thus intrinsic uncertainties are introduced into the ratio－derived［Chl］values［43，44］．

The blue－green band－ratio algorithm is also sensitive to atmosphere correction errors for waters with low［Chl］，whereasRrsat the green band is very small，thus the error due to imperfect atmospheric correction will be amplified［45］．To minimize this effect，Huet al．［45］developed a band－difference algorithm for low ［Chl］(［Chl］＜0．25 mg/m3，which covers more than～75%of the global oceans)waters，and significantly better image products are achieved．

4．2 Inherent optical properties

As discussed in detail inZaneveld et al．［46］，ocean color to the first order measures the variation of the IOPs，it is therefore critical to understand the relationships between the IOPs and ocean color and to characterize to what extent we can accurately retrieve the IOPs from ocean color．Such knowledge will not only advance the sciences of ocean optics but also improve the retrieval of biogeochemical properties．

The IOPs，which include the absorption and(back)scattering coefficients of the water constituents，are，in fact，valuable proxies for studying ocean biogeochemical processes［47－49］．For instance，the particle backscattering coefficient(bbp)is a good index for particulate organic carbon［POC］［50，51］and particle dynamics［52］;and the absorption coefficient of the colored dissolved organic matter(acdom)represents a big portion of the dissolved organic carbon in the oceans［53，54］．In addition，the bulk IOPs determine the propagation of solar radiation in the water［55－57］，which is important for the estimation of primary production［58－60］，heat budgets in the upper water column［61－63］，and the health of bottom substrates in shallow waters［64，65］．The phytoplankton absorption coefficient(aph)，on the other hand，is useful for studying phytoplankton dynamics［66］，the detection of harmful algae blooms［67，68］，and the estimation of primary production［69－71］．Furthermore，aphprovides information on phytoplankton pigments［72－74］，which are important for studying the biomass stocks and the biogeochemical cycles．

4．2．1 Algorithms for IOPs

In the past decades，a series of algorithms，both empirical and semi－analytical，have been developed for the retrieval of various IOPs from ocean color measurements［49，53］．The prototype IOP products derived for SeaWiFSand MODISincludea，bbp，aph，andacdmat various wavelengths．Hereacdmincludes the contributions from both the colored dissolved organic matter(CDOM)and the detritus．Because empirical algorithms are data driven and depend on regression analyses，which have limited association with the radiative transfer equation，the following descriptions focus on semi－analytical algorithms(SAAs)．

a．Fundamental relationships for SAAs

Semi－analytical algorithms are developed based on the relationships between the IOPs andRrs(λ)．Rrsis an apparent optical property(AOP)［21］，which provides a measure of ocean color，and it is a standard product of all satellite ocean color missions after the atmospheric correction process．ForRrs(Ω，λ)measured by a remote sensor at a sun－sensor angular geometry(Ω)，studies of the radiative transfer equation have shown that［75，76］

HereΩincludes information of the sun zenith angle，the sensor zenith angle，and the azimuth angle between the sensor and the Sun．rrsin Eq．4 is the subsurface remote－sensing reflectance corresponding to the angular geometryΩ，qrepresents the transmittance effect of the airsea interface，andΓincludes the effect of reflection when photons propagate from water to air．Both q andΓare independent of wavelength and water’s IOPs，and are weak functions ofΩ［77，78］．After values ofqandΓare determined through numerical simulations(e．g．，q≈0．52 andΓ≈1．7 for nadir－viewing measurements［76］)，rrscan be easily calculated from each measuredRrs．

Also based on the radiative transfer theory，it is found thatrrscan be expressed as a function of the bulk IOPs(aandbb)［75，79，80］，and the most commonly employed relationship in ocean color remote sensing practice is［75］

Hereg0andg1are model coefficients，which are independent of wavelength but vary with angular geometry［75，77］． For nadir－viewing measurements，Gordonet al．［75］found thatg0≈0．0949 sr－1，and g1≈0．0794 sr－1．

The bulk IOPs are the sum of the contributions from various constituents［81］，and commonly，and practically，expressed as［82，83］

Here subscript“w”represents pure seawater，andawandbbware generally considered global constants，although they vary slightly with temperature and salinity［84－86］．a(chǎn)phandacdmrepresent the absorption coefficient of phytoplankton pigments and the combination of detritus and CDOM，respectively;andbbpis the backscattering coefficient of suspended particles．These three are the primary component IOPs，which vary spatially and temporally．

To construct anRrsspectrum at any pixel in a satellite image，there are three unknown spectra(aph，acdm，andbbp)because angular geometry(Ω)is always known．That means there are more unknowns than known when using a measuredRrsspectrum to analytically solve for the optical properties，posing a perfect example of an ill－formed mathematical problem［87］．It is thus necessary to develop spectral models in order to make theRrsspectrum solvable［88］，and all SAAs work on different schemes on the spectral models for adequate solutions．Generally，two approaches have been developed for this goal［89］:

·Bottom up system(BUS)．BUS focuses on the component IOPs while at the same time derives the bulk properties．This scheme includes the spectral optimization approach(SOA)［90－96］，and the matrix inversion approach［88，97－99］．

·Top－down system(TDS)．TDStakes a step－wise approach，which derives the bulk IOPs first before it derives component IOPs．This scheme includes the quasianalytical algorithm(QAA)［100］，the system developed bySmyth et al［101］and the approach ofLoiselandStramski［102］．

Both BUSand TDSemploy the same radiative transfer model and assume similar spectral dependencies for the component IOPs，but a BUS requires more accurate spectral models for each component，while a TDS depends more on the reliability of measuredRrs(λ)．Because it is impossible to discuss all SAAs here，the SOA and the QAA are briefly reviewed below．

b．Spectral optimization algorithm(SOA)

Lab and field measurements have shown that the spectral variations of the three major component IOPs are more or less spectrally correlated．These IOPs can be generally described as

whereMph，cdm，bpare the IOPs of the three components at a designated wavelength(e．g．，443 nm)，SSph，cdm，bp(λ)represent their spectral shapes(normalized at this designated wavelength)and they are the spectral shape IOPs(SSIOP)．Applying these component models to Eqs．4－6，a modeledRrsspectrum can then be constructed［91，93］

Here“F”indicates thatRrsis a function of the parameter in the bracket“［］”．Further，an error function(ΔRrs)regarding the modeledRrsis defined as

withλifor wavelength at the ith band．Values ofMph，cdm，bpare then determined whenΔRrsreaches a minimum，which is also a state the modeledR～rs(λ)is optimized to match the measuredRrs(λ)spectrally．In essence，this approach solves forMph，cdm，bpnumerically from the measuredRrsspectrum，and it has a long history of being applied in ocean color remote sensing［90－93，103－105］．A key feature in this scheme is that it no longer assumes covariance among the three component IOPs，thus can potentially be applied to waters beyond Case－1［13，14］．

The key differences in the many SOA schemes are associated with the computer architecture［96］in efficiently obtaining the minimization(especially when processing large volume of satellite images)and the handling ofSSph，cdm，bp(λ)，as the spectral shapes do not remain constants globally．Specifically，SScdm(λ)is generally expressed as a simple exponential function ofλ［38，106］

and different values for the spectral slope(Scdm，nm－1)were adopted in the various SOA schemes(e．g．，0．0206 nm－1inMaritorena et al［93］，0．015 nm－1inLee et al［105］，etc．)．

SSbp(λ)is commonly expressed as a power－law function ofλ［2］

also different values for the angstromη(unitless)were used in different SOA schemes［92，93，105］．

SSph(λ)encountered the largest variations among the SOA practitioners．For instance:

Doerffer et al［90］，Bukata et al［103］，andRoesler and Perry［91］used differentSSph(λ)for different regions or ecosystems，withSSph(λ)of each region obtained from water sample measurements;

HOPE(Hyperspectral Optimization Processing Exemplar)［105，107］，Devred et al［92］，and GIOP(Generalized IOP)［94］:dynamically varyingSSph(λ);

根據(jù)具體實(shí)驗(yàn)室的情況，加強(qiáng)實(shí)驗(yàn)室安全管理制度建設(shè)，同時(shí)開展對(duì)制度的宣傳學(xué)習(xí)。要建立實(shí)驗(yàn)室安全教育的長效機(jī)制。從新生入學(xué)教育開展安全教育的綜合課程，到進(jìn)入實(shí)驗(yàn)室前的安全準(zhǔn)入制度學(xué)習(xí)，構(gòu)建實(shí)驗(yàn)室安全教育與實(shí)驗(yàn)教學(xué)、實(shí)驗(yàn)研究緊密聯(lián)系的長效機(jī)制[13]。

GSM(Garver－Siegel－Maritorena)［93］:one fixed multi－spectral shape for all waters．c．Algebraic algorithm(QAA)

Different from the SOAs where the component IOPs are derived simultaneously with the bulk IOPs，algorithms that separate the derivation of bulk and component IOPs from the measuredRrs(λ)have also been developed［100－102］．Here，using the QAA［100］as an example，the features of such algebraic algorithms are briefly summarized．

Fundamentally，for all waters，with the increase of wavelength，the total absorption of the water medium is generally dominated byThis means that we can always find a wavelength(λ0)in the longer range and obtain a good estimation ofa(λ0)．BecauseRrsis a function ofaandbb，the estimateda(λ0)and measuredRrs(λ0)enable the algebraic derivation ofbb(λ0)(and thenbbp(λ0)following Eq．6)through Eqs．4－5．Theoretically，we may choose the longest wavelength(e．g．，near－infrared or short wave)to ensure the highest accuracy in estimatinga(λ0)．In practice，however，becauseawin those bands are so high thatRrsat those bands are too low(or close to zero)to be accurately derived from a remote sensor，a compromise has to be made in order to ensure good estimation ofbb(λ0)．This resulted in that λ0is usually taken as～55x nm(547 nm for MODIS and 555 for SeaWiFS)for oceanic waters and～670 nm for coastal turbid waters(whenRrs(670)＞0．0015 sr－1)，and empirical algorithms have been developed for the estimation ofa(λ0)for global waters［100］(also see http://www．ioccg．org/groups/Software_OCA/QAA_v6．pdf)．

Aftera(λ0)is known，bbp(λ0)is calculated from

withu(λ0)(≡bb(λ0)/(a(λ0)+bb(λ0))，calculated fromrrs(λ0)，see Eq．5)．Thisbbp(λ0)is extrapolated to the shorter wavelengths by applyingSSbp(λ)，anda(λ)is then calculated from knownu(λ)and the derivedbbp(λ)，

In this process of deriving the total absorption and backscattering coefficients，the truly empirical component is the assumed or estimatedSSbp(λ)．The empirical determination ofa(λ0)could be changed to useaw(λ0)if there are high－quality measurements ofRrs(λ0)in the longer wavelengths．

Aftera(λ)is known，a(411)anda(443)are employed to obtainacdm(443)，and then to deriveaph(λ)with the help ofSScdm(λ):Hereζ=SSph(411)/SSph(443)andξ=SScdm(411)/SScdm(443)，and both are estimated empirically fromRrs(443)/Rrs(55x)in QAA(http://www．ioccg．org/groups/Software_OCA/QAA_v6．pdf)．

4．3 Attenuation coefficients

4．3．1 The“standard”Kd(490)product

The diffuse attenuation coefficient of the downwelling irradiance is an important optical property for the propagation of solar radiation in the upper water column．And，this property is a useful index for the classification of oceanic water types［109，110］and the evaluation of water clarity［111－113］．For nearly two decades，diffuse attenuation coefficient of the downwelling irradiance at 490 nm(Kd(490)，m－1)at the basin scale has been generated from satellite(e．g．，SeaWiFS，MODIS)ocean color measurements．Similar to the retrieval of［Chl］，the default(or“standard”)algorithm employed by NASA for the generation ofKd(490)product from satellite ocean color measurements(including SeaWiFS，MODIS，MERIS)takes the blue－green ratio(termed as BGR in the following)approach．For ocean color data provided by the MODIS sensor，the algorithm for theKd(490)product is(http://oceancolor．gsfc．nasa．gov/REPROCESSING/R2009/Kdv4/)

withζ0－4as－0．8813，－2．0584，2．5878，－3．4885，and－1．5061，respectively．

ThisKd(490)product，however，due to its default BGR approach(Eq．15)has significant caveats(see below for details discussions)，which limit its broad applications．

1)Questionable treatments of Kd(490)by the traditional BGR algorithm

Historically，AustinandPetzold［114］proposed a simple formula three decades ago for the estimation ofKd(490)(analogous to the retrieval of［Chl］)from a database of 88in situmeasurements made in various parts of the world:

Although this empirical algorithm forKd(490)was subsequently updated with more measurements［115－118］and was modified by incorporating different mathematical formulas and/or different band ratios(see Eq．16)，its data－driven and empirical nature remains the same．The algorithm coefficients(x，A，B，as well asζ0－4of Eq．15)are considered as global constants and are independent of solar elevation．In essence，all these empirical algorithms treatKd(490)as the same kind of property as［Chl］．However，based on the radiative transfer theory，Kd(490)is an AOP［119］．In other words，it is at least necessary to specify the solar elevation information for aKd(490)value，otherwise the representation of theKd(490)value is ambiguous．On the other hand，obviously，［Chl］is not an AOP，i．e．［Chl］is not a function of sun angle．Indeed，Kd(490)and［Chl］are fundamentally two different kinds of properties，and they have different relationships withLworRrs．

2)Lack of new information in the BGR-derived Kd(490)product

In addition to the computation ofKd(490)，Rrsis also the sole input to generate the［Chl］product from satellite ocean color measurements．For example，the operational algorithm(OC3m)for MODISis

withξ0－4as 0．2424，－2．7423，1．8017，0．0015，and－1．2280，respectively．

Therefore，because the BGR ofRrsis used as inputs for both［Chl］andKd(490)，the spatiotemporal information of the global oceans obtained by suchKd(490)product is nearly the same as that provided by［Chl］．In particular，for many coastal waters，it is exactly the same ratio used for both［Chl］andKd(490)，the spatial information from theKd(490)product is identical to that from the［Chl］product．This is contradictory to reality，because it is well known that coastal waters have complex relationships between chlorophyll and other optically active constituents(the commonly termed Case－2 waters［13，120］)，we should expect spatially differentKd(490)and［Chl］products for such regions．For other areas(mostly open oceans)whereRrs(443)＞Rrs(488)，although［Chl］uses the ratioRrs(443)/Rrs(551)as the input whileKd(490)uses the ratioRrs(488)/Rrs(551)，becauseRrs(443)andRrs(488)are highly correlated，the derived［Chl］andKd(490)would also be highly correlated．Therefore，compared to the［Chl］product，there is no(or very limited)new spatiotemporal information provided by the BGR－derivedKd(490)product，although they exhibit different quantitative values．

3)Not applicable to coastal turbid waters

The BGR algorithm was found working best for oceanic waters withKd(490)＜～0．2 m－1［117，118］． For turbid coastal waters，because the blue－green ratio(e．g．，Rrs(488)/Rrs(551)) will gradually plateau［121，122］，this BGR approach is then no longer sensitive to changes in in－water constituents that determine the value of the attenuation coefficient．This is well demonstrated in many studies［113，115，122－124］．

More importantly，becauseRrsis generally a function of the ratio of backscattering to absorption coefficients［75，77，80］，the ratio ofRrsat different bands to a large extent removes the information of backscattering［2］．On the other hand，Kd(490)is fundamentally a function of both absorption and backscattering coefficients［55－57，125］，thus the BGR－derivedKd(490)inherently under－represents the backscattering contribution to the attenuation coefficient，which can be significant for turbid coastal or river－plume waters［123，126］．

4)Extra uncertainty in the BGR-derived Kd(490)product

When ocean color satellite products are evaluated using match－up data between satellite andin situmeasurements，it is recommended to keep the time gap within 3 hours of satellite overpasses［127］．Depending on time and latitude of the sampling sites，however，this 3－h(huán)our gap covers a range of solar zenith angle of 0o－75o．Such an angular range results in more than 30%difference inKdfor the same water properties［57，128］;thus solar－elevation－related variation complicates the evaluation of the satellite－derivedKd(490)product when it is compared with field measurements．In fact，the empirically derivedKd(490)overestimatesKd(490)for observations with zenith angle under 30o，and underestimatesKd(490)for observations with zenith angle over 60o［129］．In essence，without matching sun angles ofKd(490)，comparingKd(490)values from two determinations is similar to comparing an apple with an orange．4．3．2 Theoretical model ofKd

There is a long history of studies to characterize and model the variation ofKd［55－57，130－133］，and results of these efforts show thatKdis a function of bothaandbb［57，134］．In general，following the radiative transfer theory，Kdcan be described as［57，134］

Here parametersm0andvdescribe the impact of the solar elevation as well as the relative weighting of absorption and backscattering coefficients toKd，respectively．Note that Eq．18 is exact as it is simply a re－arrangement of the radiative transfer equation(RTE)．

For practical applications，the above model is parameterized［57］via numerical simulations with Hydrolight［135］，

Hereθais the solar zenith angle in air．Recently，to explicitly account for the different effects of molecular and particle scatterings，the above model is updated to［129］

Where parameterηwis defined as［136］

which provides a measure of the relative contribution of the molecular scattering to the total scattering．This semianalyticalKdmodel is applicable for oceanic waters and coastal turbid waters．More importantly，with the solar elevation explicitly taken into account in such an RTEbased IOPs－Kdmodel，not only are uncertainties of the estimatedKdreduced［123，129］，but also exhibits a clear manifestation of the AOP nature ofKd．Furthermore，with such a radiative－transfer based model and the semianalytically retrieved IOPs fromRrs，not only canKdat 490 nm be retrieved，but alsoKdat wavelengths where there areRrsmeasurements．

4．4 Bottom properties

For many coastal regions(including inland lakes and remote islands)，the signal measured by satellites，after atmospheric correction，includes contributions not only from the water column but also from the shallow bottom．Waters with photons bounced back from the bottom and make not－negligible contribution toRrsare referred to as optically shallow waters(OSWs)．For the remote sensing of OSWs，bottom contribution complicates the retrieval of water－column properties．On the other hand，bottom contribution enables the possibility to retrieve bottom properties(bottom depth and reflectance)from spectralRrs［137］．

Over the past decades，there have been breakthroughs in both theory［76，138］and practice［139－144］for remote sensing of OSWs．In particular，the Hyperspectral Optimization Processing Exemplar(HOPE)has been developed and validated［105，145，146］to retrieve both water and bottom properties simultaneously from hyperspectral measurements．The HOPE algorithm has also produced promising results for deriving both water and bottom properties of OSWs when applied to MERISdata［147］．

Rrsover OSWs is a complex mix of processes for both the water column and the bottom，and in the last decade extensive studies have been carried out on forward and inverse modeling of OSWs［76，144］．Based on the RTE and numerical simulations，Rrscan be expressed as［76，138］

To derive bothρa(bǔ)ndHfrom Eq．22，the HOPE algorithm［105，139，145］has been developed and widely tested with hyperspectral data［146］．In short，in addition to model the component IOPs，bottom reflectance is described as

with B the bottom reflectance at 550 nm．The termis the spectral shape of bottom reflectance for a specific substrate(e．g．，sand or sea grass bottoms)．For each image pixel，ρ+i(λ)is selected based on the inputRrs(λ)［139］．

With the bio－optical models，Rrs(λ)is then expressed by a 5－parameter model:

P and G represent absorption coefficients of phytoplankton pigments and colored detrital matter at 440 nm，respectively，and X represents particle backscattering coefficient at 440 nm．HOPE varies their values until a best match is found between the modeled and measuredRrsspectra［105，139］．Unlike empirical approaches for OSWs［148，149］，the only required input for HOPE to process an image pixel is theRrsspectrum．

4．5 Properties on the water surface

In wide area of open ocean，coastal，or in－land waters，due to eutrophication and/or aggregation，there could be intense marine particulates floating on the seasurface．These organic materials include macro algae ofUlvaprolifera，cyanobacteriumTrichodesmiumspp．，macro algae ofSargassum spp．，and oil slicks［150－152］．It is important to identify and quantify these floating materials as they are not only an indicator of，but can also significantly affect，the quality of an ecosystem．In general，all these materials have enhanced optical signal in the NIR and even longer wavelengths，thus，although ocean color satellite missions are in general designed for the retrieval of properties in the upper water column，measurements in the NIR can be used for the detection and even quantification of such floating materials［151－154］．

For instance，it was found thatSargassumcould be detected and quantified from MERIS and MODIS measurements［153，154］．Further，a floating algae index(by including MODIS 859－nm band)was developed for the detection of cyanobacteria bloom in Taihu Lake，China［151］andUlva proliferain the Yellow Sea and East China Sea［152］． In addition，Subramaniamet al．［2002］found that surfaceTrichodesmiummats could also be detected from MERISNIR measurements．

Detection of oil slicks on the surface，however，is not relying on the spectral characteristics，rather on its effect on surface tension，thus resulting in different surface albedo between oil－covered surface and no－oil surface［150］．Such differences can be easily observed in the sun－glint spots of an ocean color image［150］．

5 Application of ocean color products

5．1 Primary production(PP)

Photosynthesis by phytoplankton is the primary mechanism of fixing organic carbon for the support of the marine ecosystem，and primary production provides a measure of this important process．Because of the vast size of the global oceans，F(xiàn)alkowski et al．［1］have concluded that“the estimation of primary production from satellite measurements of ocean color is critical for developing an understanding of how ocean biological processes affect，and are affected by，changes in atmospheric radiative budgets and global biogeochemical cycles．”Because of such importance，estimation of primary production of the global oceans is a central mission goal of ocean－color satellites［155］，and large efforts have been made worldwide to study oceanic PP［36，59，156－160］．

a)Traditional models to estimate PP of the global oceans

Mathematically，a wavelength－，depth－，and timeresolved model for the estimation of primary production is commonly expressed as［60，161］

HerePP(z)is primary production at depthz，its integration over the photic zone provides production of the water column(PPeu)．λ/hc converts energy to quanta(hc=1．5x10－16J nm quanta－1［82］)．φ(mol C/Ein)is the realized(in situ)photosynthetic quantum yieldis the phytoplankton absorption coefficient per unit chlorophyll(often referred to as the chlorophyllspecific absorption coefficient);E0is the scalar spectral irradiance and can be calculated fromEd(Morel［165］;Sathyendranath［158］)．

PP(z)can also be written in wavelength－integrated forms［60］

In Eqs．25 and 26，the photosynthetic parameter(φorφ)is not directly measureable from ocean－color remote sensing，but estimated through other properties(such as sea－surface temperature［35］)．When this parameter is known(see［60］)，the missing pieces for the estimation ofPP(z)are thenQPAR(orE0)andChl．PAR at the surface can be estimated through measurement of atmospheric properties［166］，and its propagation from surface to deeper waters can be modeled via radiative transfer theory［83］．As a result，it is then quite obvious that“Chl is a key parameter”for any satellite ocean－color missions［155］．Indeed，as concluded byBehrenfeldandFalkowski［35］，“Chl is the most important variable in estimating”PP，and various algorithms have then been developed for the retrieval of Chl from oceancolor measurements(see O'Reillyet al［37］)．Hence，this scheme of estimating PPis called the“Chl－based approach”［51］．There are two inherent and fundamental limitations，associated with one bio－optical propertyimbedded in this Chl－centric strategy(discussed below)that prevent an accurate estimation of PP from ocean－color measurements，which further diminishes the utility of such PPproducts to study its spatial and temporal variations and contributions to the carbon cycle．

1)Uncertainties inφ－－the Chl-specific quantum yield for PP

Earlier studies［35，167］have shown that PP can only be quantified with an accuracy of at best 50－60%even within situmeasured［Chl］．This is because whenever PP is modeled using［Chl］as an independent variable，it automatically requires the specification ofφor similar Chl－normalized photosynthetic parameter．And this Chlspecific photosynthetic parameter is directly related to

Note that the time and depth terms are omitted for convenience．The quantum yield(φ)andare two independent properties，and both vary spatially and temporally．Numerous studies［39，41，168，169］have pointed out that because of variations in pigment composition and the“package effect”［170］，vary significantly for a given［Chl］value．Even more problematic is the fact that there is no reliable method as yet to accurately estimate the spatial and temporal variation offrom satellite．This uncertainty，at least partially，explains a conclusion ofBehrenfeldandFalkowski［35］that“significant improvements in estimating oceanic primary production will not be forthcoming without considerable advance in our ability topredict temporal and spatial variability in”．This is because thatincludes the variation of

2)Uncertainties in satellite-derived［Chl］

To estimate PP using Eq．26，as mentioned above，a critical element is to acquire accurate estimates of［Chl］from satellite imagery［35，60］．Presently，standard satellite［Chl］product is derived from the band ratio ofRrs［171］．In theory，however，the near－universal and stable relationship is thatRrsis a function ofaandbb［75，79］．Only after empirically linking these optical properties with the concentration of water constituents such as［Chl］［2，14，172］，canRrsbe related to［Chl］．Therefore，the relationship between［Chl］and band－ratio ofRrscan be summarized as［2，173］:HereGChldescribes the relative amount of CDOM per unitChl;andλiandλjare wavelengths of bandiand bandj．Clearly，to get［Chl］fromRrsband ratios，bothGChlandneed to be known or both co－vary with［Chl］．However，extensive measurements in the field［38，106，174］have shown that there are no fixed values or relationships for bothGChlanda*pheven for oceanic waters．Consequently，the［Chl］product derived from the empirical algorithms shows regional bias in the global oceans［44］．

In summary，in the traditional Chl－based approach for PP estimation from ocean－color measurements the values ofare required，explicitly or implicitly．Because of the implicit involvement ofin two independent model－development processes，the errors associated withdo not cancel out each other，rather get enhanced，and then there are large uncertainties in the spatial and temporal distributions in the estimated globalPPeu．Although the Carder algorithm［175］explicitly includes a process to select the value ofa*phwhen deriving Chl fromRrs，but there is no process yet to accurately determinea*phvalue when quantifyingφ．Because of such complex impact froma*ph，large uncertainties in estimatedPPeuresulted．Siegel et al［176］found that“Site－specific and previously published global models of primary production both perform poorly and account for less than40%of the variance in∫PP．”The series of round robin reports［177－180］also highlighted the large difference in spatial distributions of global primary production from Chl－based models．Further，Saba et al［180，181］concluded that“ocean color models did not accurately estimate the magnitude of the trends of NPP over multidecadal time periods，and were even more challenged over shorter time periods，especially when the models used satellite－derived Chl－a”．

b)aph-based strategy to estimate PP of the global oceans

To avoid the engagement ofa*ph，another scheme is to take the absorption－based approach［71，182］．Considering quantum yield is wavelength independent，and becausea*phis a simple ratio ofaphtoChl，the wavelength－，depth－，and time－resolved model forPP(z)(Eq．25)can be written as

HereE0(t，z，λ)is the scalar irradiance and can be calculated fromEd(t，z，λ)as inMorel［165］andSathyendranath［158］．Note that bothEd(t，z，λ)andaph(λ)are available from ocean color remote sensing．

Historical studies［164，183－185］have found that light is the most important parameter in regulatingφ，and it was found thatφcan be modeled as［Kiefer and Mitchell］［164，186］:

Hereφmaxis the maximum quantum yield of photosynthesis［163，164］，andKφis an empirical model parameter with units asPAR．The exponential term accounts for photoinhibition just below the surface．Promising results ofaph－based PP estimation could be found in［69－71，187］．

5．2 Oceanic heat budget in the upper water column

In addition to the contribution to the carbon cycle via photosynthesis by phytoplankton，the change of transparency due to change of water constituents also affects the heat budget in the upper water column［61，62，188，189］．In the earlier days(e．g．，［62］)，the evaluation of biological impact on heat budget is largely based on the loosely defined water types［109］，where the boundary between water types is arbitrary．And，usually the models are one dimensional［188］，thus difficult to represent the dynamic changes resulted from vertical mixing and horizontal advection．In the past decade(e．g．，［190］;［191］)，with the mature of［Chl］products from the Sea－WiFSand MODIS，a more realistic parameterization considering the spatio－temporal variation of solar irradiance penetration due to changes in phytoplankton biomass has been established．When the chlorophyll－dependent penetration parameterization rather than the Jerlov water type is adopted，climate model simulations in the tropics are found to be significantly improved(e．g．，［192，193］)．It’s noted that the adopted relationship between the attenuation of solar radiation and［Chl］is empirical and may not be applicable to the ocean with regional characteristics，and a model based on the ocean optical properties［129，194］will probably ease the issue to large extent．In addition to the improvement in the parameterization，recent research focuses shift from the effects of solar penetration on the tropics to those in the extratropical oceans［195］，and from the impacts on the ocean to those on both the ocean and atmosphere circulation［191］．And the ecosystem importance in modulating ocean－atmosphere interaction has been emphasized［191］．

6 Looking forward

Ocean color satellite remote sensing is an indispensable means to observe and monitor basin－scale biogeochemical processes．Although there have been outstanding progresses in both hardware and software as well as product applications，there are still daunting challenges in obtaining consistent and accurate as well as desired ocean color products．For instance，because change ofRrsreflects changes in the bulk optical property(total absorption or backscattering coefficients)，the best retrieval in ocean color remote sensing is thus the bulk optical properties．The uncertainty is still quite high when decomposing the bulk property to various components，especially for coastal waters．It requires not only highly accurateRrsfrom ocean color satellite sensors，but also adequate and robust models for the spectral shapes of the various component IOPs in order to reduce the uncertainties．Further，systematic studies are required in order to accurately convert the retrieved components IOPs to biogeochemical properties，such as［Chl］or suspended sediments．On the estimation of primary production，it further requires to know how the quantum yield varies spatially and temporally，and how this physiological property changes among phytoplankton classes and species．Furthermore，fundamentally passive ocean color remote sensing provides a measure of vertically－weighed biogeochemical properties of the upper water column，but it is desired，and important，to develop adequate schemes and methodologies to provide vertically－resolved information．

Acknowledgement

Dr．Zhongping Lee is grateful for support from NASA and NOAA．

［1］ FALKOWSKI P G．Using satellite data to derive primary productivity in the world ocean，in SeaWiFS Technical Report Series，S．B．Hooker and E．R．Firestone，Editors．1998，NASA:Greenbelt，MD．

［2］ GORDONH R，MOREL A Y．Remote assessment of ocean color for interpretation of satellite visible imagery:a review［M］．New York:Springer－Verlag，1983:44

［3］ GORDONH R．Calibration requirements and methodology for remote sensors viewing the oceans in the visible［J］．Remote Sens Env，1987，22(1):103－126．

［4］ GORDON H R．In－orbit calibration strategy for ocean color sensors［J］．Remote Sensing of Environment，1998，63(3):265－278．

［5］ FRANZ B A，BAILEY SW，WERDELL J，et al．Sensor－independent approach to the vicarious calibration of satellite ocean color radiometry［J］． Applied Optics，2007，46(22):5068－5082．

［6］ GORDON H R，BROWN J W，EVANS R H．Exact Rayleigh scattering calculations for use with Nimbus－7 Coastal Zone Color Scanner［J］．Applied Optics，1988，27(5):862－871．

［7］ GORDON H R，WANG M．Retrieval of water－leaving radiance and aerosol optical thickness over oceans with SeaWiFS:a preliminary algorithm［J］．Applied Optics，1994，33(3):443－452．

［8］ EPLEE R E，ROBINSON W D，BAILEY S W，et al．Calibration of SeaWiFS．II．Vicarious techniques［J］．Applied Optics，2001，40(36):6701－6718．

［9］ MOORE K D，VOSS K J，GORDON H R．Spectral reflectance of whitecaps:Their contribution to water－leaving radiance［J］．J Geophys Res，2000，105(C3):6493－6499．

［10］ WERDELL P J，BAILEY S W，F(xiàn)RANZ B A，et al．On－orbit vicarious calibration of ocean color sensors using an ocean surface reflectance model［J］．Appl Opt，2007，46(23):5649－5666．

［11］ CLARK D K，YARBROUGH M A，F(xiàn)EINLOLZ M E，et al．MOBY，a radiometric buoy for performance monitoring and vicarious calibration of satellite ocean color sensors:measurement and data analysis protocols，in NASA Tech．Memo．2004－211621．2003，NASA，Goddard Space Flight Center:Greenbelt，MD

［12］ COX C，MUNK W．Measurement of the roughness of the seasurface from photographs of the Sun’s glitter［J］．J Opt Soc Am．1954，44(11):838－850．

［13］ MOREL A，PRIEUR L．Analysis of variations in ocean color［J］．Limnol Oceanogr，1977，22(4):709－722．

［14］ MOREL A，MARITORENA S．Bio－optical properties of oceanic waters:A reappraisal［J］．J Geophys Res，2001，106(C4):7163－7180．

［15］ GORDON H R．Removal of atmospheric effects from satellite imagery of the oceans［J］．Applied Optics，1978，17(10):1631－1636．

［16］ GORDON H R，VOSS K J．MODIS normalized water－leaving radiance algorithm theoretical basis document［J］．NASA Technical Report Series，NAS5－31363，1999．

［17］ ANTOINE D，MOREL A．A multiple scattering algorithm for atmospheric correction of remotely sensed ocean colour(MERIS instrument):principle and implementation for atmospheres carrying various aerosols including absorbing ones［J］．International Journal of Remote Sensing，1999，20(9):1875－1916．

［18］ SUN L，GUO M，ZHU J，et al．FY－3A/MERSI，ocean color algorithm，products and demonstrative applications［J］．Acta Oceanologica Sinica，2013，32(5):75－81．

［19］ SIEGEL D A，WANG M，MARITORENA S，et al．Atmospheric correction of satellite ocean color imagery:the black pixel assumption［J］．Applied Optics，2000，39(21):3582－3591．

［20］ SIEGEL D A，WANG M，MARITORENA S，et al．Atmospheric correction of satellite ocean color imagery:the black pixel assumption［J］．Applied Optics，2000，39(21):3582－3591．

［21］ PREISENDORFER R W，MOBLEY C D．Direct and inverse irradiance models in hydrologic optics［J］．Limnol Oceanogr，1984，29(5):903－929．

［22］ VIOLLIER M，STURM B．CZCSdata analysis in turbid coastal water［J］．JGeophys Res:Atmospheres(1984－2012)，1984，89(D4):4977－4985．

［23］ ARNONE R A，MARTINOLICH P，GOULD R W．Coastal optical properties using SeaWiFS，in Ocean Optics XIV［C］．in Ocean Optics XIV Conference．1998．Kailua－Kona，Hawaii．

［24］ RUDDICK K G，OVIDIO F，RIJKEBOER M．Atmospheric correction of SeaWiFSimagery for turbid coastal and inland waters［J］．Applied optics，2000，39(6):897－912．

［25］ FUKUSHIMA H，TORATANI M，KOBAYASHI H．Atmospheric correction algorithm for satellite ocean color dataover the Asian water［C］．in SPIE．2000．

［26］ MOORE GF，AIKEN J，LAVENDER S．The atmospheric correction of water colour and the quantitative retrieval of suspended particulate matter in Case II waters:application to MERIS［J］．International Journal of Remote Sensing，1999，20(9):1713－1733．

［27］ CHOMKO R M，GORDON H R，MARITORENA S，et al．Simultaneous retrieval of oceanic and atmospheric parameters for ocean color imagery by spectral optimization:a validation［J］．Remote Sensing of Environment，2003，84(2):208－220．

［28］ WANG M．Remote sensing of the ocean contributions from ultraviolet to near－infrared using the shortwave infrared bands:simulations［J］．Applied Optics，2007，46(9):1535－1547．

［29］ SUN L，GUO M，WANG X．Ocean color products retrieval and validation around China coast with MODIS［J］．Acta Oceanologica Sinica，2010，29(4):21－27．

［30］ HE X，BAI Y，PAN D，et al．，Atmospheric correction of satellite ocean color imagery using the ultraviolet wavelength for highly turbid waters［J］．Optics Express，2012，20(18):20754－20770．

［31］ SCHILLER H，DOERFFER R．Neural network for emulation of an inverse model operational derivation of Case II water properties from MERIS data［J］．International Journal of Remote Sensing，1999．20(9):1735－1746．

［32］ SUN L，ZHANG J，GUO M．Atmospheric Correction over Case 2 waters using neutral network［J］．Journal of Remote Sensing－Beijing，2007，11(3):398－405．

［33］ GORDON，H．R，CLARK D K，BROWN J W，et al．Phytoplankton pigment concentrations in the Middle Atlantic Bight:Comparison of ship determinations and CZCS estimates［J］．Applied Optics，1983，22(1):20－36．

［34］ ARRIGO K R，MATRAI P A，VAN DIJKEN G L．Primary productivity in the Arctic Ocean:Impacts of complex optical properties and subsurface chlorophyll maximaon large－scale estimates［J］．Journal of Geophysical Research－Oceans，2011，116(C11):11022－．

［35］ BEHRENFELD M J，F(xiàn)ALKOWSKI P G．Photosynthetic rates derived from satellite－based chlorophyll concentration［J］．Limnol．Oceanogr，1997，42(1):1－20．

［36］ PLATT T，SATHYENDRANATH S．Oceanic primary production:estimation by remote sensing at local and regional scales［J］．Science，1988，241(4873):1613－1620．

［37］ O’REILLY J E，MARITORENA S，MITCHELL B G，et al．Ocean color chlorophyll algorithms for SeaWiFS［J］．J Geophys Res，1998，103(C11):24937－24953．

［38］ CARDER K L，STEWARD R G，HARVEY G R，et al．Marine humic and fulvic acids:their effects on remote sensing of ocean chlorophyll［J］．Limnol Oceanogr，1989，34(1):68－81．

［39］ BRICAUD A，BABIN M，MOREL A，et al．Variability in the chlorophyll－specific absorption coefficients of natural phytoplankton:Analysis and parameterization［J］．J Geophys Res，1995，100(C7):13321－13332．

［40］ MOREL A，BRICAUD A．Theoretical results concerning light a bsorption in a discrete medium，and application to specific absorption of phytoplankton［J］．Deep－SeaRes Part A．Oceanographic Research Papers，1981，28(11):1375－1393．

［41］ SATHYENDRANATH S，LAZZARA L，PRIEUR L．Variations in the spectral values of specific absorption of phytoplankton［J］．Limnol Oceanogr，1987，32(2):403－415．

［42］ KAHRU M，MITCHELL B G．Seasonal and nonseasonal variability of satellite－derived chlorophyll and colored dissolved organic matter concentration in the California Current［J］．JGeophys Res，2001，106(C2):2517－2529．

［43］ BROWN C A，HUOT Y，WERDELL P J，et al．The origin and global distribution of second order variability in satellite ocean color and its potential applications to algorithm development［J］．Remote Sensing of Environment，2008，112(12):4186－4203．

［44］ SZETO M，WERDELL P J，MOORE T S，et al．Are the world’s oceans optically different?［J］．J Geophys Res，2011，116:C00H04．

［45］ HU C，LEE Z，F(xiàn)RANZ B．Chlorophyll a algorithms for oligotrophic oceans:a novel approach based on three－band reflectance difference［J］．J Geophys Res，2012，117:C01011．

［46］ ZANEVELD R，BARNARD A，LEE Z P．，Why are inherent optical properties needed in ocean－colour remote sensing，in Remote Sensing of Inherent Optical Properties:Fundamentals，Tests of Algorithms and Applications，Lee Z P，Editor．2006，International Ocean－Colour Coordinating Group:Dartmouth，NovaScotia，Canada．

［47］ OUBELKHEIR K，CLEMENTSON L A，WEBSTER IT，et al．Using inherent optical properties to investigate biogeochemical dynamics in a tropical macrotidal coastal system［J］．J Geophys Res，2006．111(C7):C07021．

［48］ BOSS E，SWIFT D，TAYLOR L，et al．Observations of pigment and particle distributions in the western North Atlantic from an autonomous float and ocean color satellite［J］．Limnol oceanogr，2008，53(5):2112－2122．

［49］ IOCCG，Remote sensing of inherent optical properties:fundamentals，tests of algorithms，and applications，in Reports of the International Ocean－Colour Coordinating Group，No．5，Z．－P．Lee，Editor．2006，IOCCG:Dartmouth，Canada．p．126．

［50］ STRAMSKI D，REYNOLDSR A，KAHRU M，et al．Estimation of particulate organic carbon in the ocean from satellite remote sensing［J］．Science，1999，285(5425):239－242．

［51］ BEHRENFELD M J，BOSS E，SIEGEL D A，et al．Carbonbased ocean productivity and phytoplankton physiology from space［J］．Global biogeochemical cycles，2005，19(1):1－14．

［52］ BOSSE，PEGAU W S，LEE M，et al．Particulate backscattering ratio at LEO 15 and its use to study particle composition and distribution［J］．J Geophys Res，2004，109(C1):C01014．

［53］ MANNINO A，RUSSM E，HOOKER SB．Algorithm development and validation for satellite－derived distributions of DOC and CDOM in the US Middle Atlantic Bight［J］．Journal of Geophysical Research－Oceans，2008．113(C7):C07051．

［54］ SIEGEL D A，MARITORENA S，NELSON N B，et al．Global distribution and dynamics of colored dissolved and detrital organic materials［J］．JGeophys Res，2002，107(C12):3228．

［55］ KIRK JT O．Dependence of relationship between inherent and apparent optical properties of water on solar altitude［J］．Limnol Oceanogr，1984，29(2):350－356．

［56］ GORDON H R．Can the Lambert－Beer law be applied to the diffuse attenuation coefficient of ocean water?［J］．Limnol．Oceanogr，1989，34(8):1389－1409．

［57］ LEE Z P，DU K P，ARNONE R．A model for the diffuse attenuation coefficient of downwelling irradiance［J］．J Geophys Res，2005，110:C02016．

［58］ SATHYENDRANATH S，PLATT T．Remote sensing of watercolumn primary production，in Measurement of primary production from the molecular to the global scale，W．K．W．Li and Maestrini，Editors．1995，ICESMarine Science Symposia:Copenhagen．p．236－243．

［59］ ARRIGO K R，VAN DIJKEN G L，BUSHINSKY S．Primary production in the Southern Ocean，1997－2006［J］．JGeophys Res，2008，113:C08004．

［60］ BEHRENFELD M J，F(xiàn)ALKOWSKI P G．A consumer’s guide to phytoplankton primary productivity models［J］．Limnol Oceanogr，1997，42(7):1479－1491．

［61］ LEWISM R，CARR M，F(xiàn)ELDMAN G C，et al．Influence of penetrating solar radiation on the heat budget of the equatorial pacific ocean［J］．Nature，1990，347(6293):543－545．

［62］ RONALD J，ZANEVELD V，KITCHEN J C，et al．The influence of optical water type on the heating rate of aconstant depth mixed layer［J］．J Geophys Res，1981，86(C7):6426－6428．

［63］ SIEGEL D A，OHLMANN J C，WASHBURN L，et al．Solar radiation，phytoplankton pigments and radiant heating of the equatorial Pacific warm pool［J］．J Geophys Res，1995，100(C3):4885－4891．

［64］ HOCHBERG E J，ATKINSON M J．Coral reef benthic productivity based on optical absorptance and light－use efficiency［J］．Coral Reefs，2008，27(1):49－59．

［65］ DUNNE R P，BROWN B E．Penetration of solar UVB radiation in shallow tropical waters and its potential biological effects on coral reefs;Results from the central Indian Ocean and Andaman Sea［J］．Marine Ecology－Progress Series，1996，144:109－118．

［66］ SHANG S，DONG Q，LEE Z，et al．MODIS observed phytoplankton dynamics in the Taiwan Strait:an absorption－based analysis［J］．Biogeosciences，2011，8:841－850．

［67］ CRAIG SE，LOHRENZ SE，LEE Z，et al．Use of hyperspectral remote sensing reflectance for detection and assessment of the harmful Alga，Kareniabrevis［J］．Applied Optics，2006，45(21):5414－5425．

［68］ CANNIZZARO J P，CARDER K，CHEN F R，et al．A novel technique for detection of the toxic dinoflagellate，Kareniabrevis，in the Gulf of Mexico from remotely sensed ocean color data［J］．Continental Shelf Research，2008，28(1):137－158．

［69］ LEE Z P，CARDER K L，MARRA J，et al．Estimating primary production at depth from remote sensing［J］．Applied Optics，1996，35(3):463－474．

［70］ LEE Z，LANCE V P，SHANG S，et al．An assessment of optical properties and primary production derived from remote sensing in the Southern Ocean(SO GasEx)［J］．J Geophys Res，2011，116(C4):C00F03．

［71］ MARRA J，TREESC C，O’REILLY J E．Phytoplankton pigment absorption:a strong predictor of primary productivity in the surface ocean［J］．Deep SeaRes Part I，2007，54(2):155－163．

［72］ BIDIGARE R R，MORROW JH，KIEFER D A．Derivative analysis of spectral absorption by photosynthetic pigments in the western Sargasso Sea［J］．J Mar Res，1989，47:323－341．

［73］ HOEPFFNER N，SATHYENDRANATH S．Determination of the major groups of phytoplankton pigments from the absorption spectraof total particulate matter［J］．JGeophy Res，1993，98(C12):22789－22803．

［74］ MISHRA S，MISHRA D R，LEE Z，et al．Quantifying cyanobacterial phycocyanin concentration in turbid productive waters:a quasi－analytical approach［J］．Remote Sensing of Environment，2013，133:141－151．

［75］ GORDON H R，BROWN O B，EVANSR H，et al．A semianalytic radiance model of ocean color［J］．J Geophys Res，1988，93(D9):10909－10924．

［76］ LEE Z，CARDER K L，MOBLEY C D，et al．Hyperspectral remote sensing for shallow waters．I．A semianalytical model［J］．Appl Opt，1998，37(27):6329－6338．

［77］ MOREL A，GENTILI B．Diffuse reflectance of oceanic waters．II Bi－directional aspects［J］．Appl Opt，1993，32(33):6864－6879．

［78］ GORDON H R．Normalized water－leaving radiance:revisiting the influence of surface roughness［J］．Appl Opt，2005，44(2):241－248．

［79］ SATHYENDRANATH S，PLATT T．Analytic model of ocean color［J］．Appl Opt，1997，36(12):2620－2629．

［80］ ZANEVELD J R V．A theoretical derivation of the dependence of the remotely sensed reflectance of the ocean on the inherent optical properties［J］．J Geophys Res，1995，100(C7):13135－13142．

［81］ STRAMSKI D，BOSSE，BOGUCKI D，et al．The role of seawater constituents in light backscattering in the ocean［J］．Progress in Oceanography，2004，61(1):27－56．

［82］ KIRK J T O．Light＆Photosynthesis in Aquatic Ecosystems［M］．1994，Cambridge:University Press．

［83］ MOBLEY C D．Light and Water:radiative transfer in natural waters［M］．1994，New York:Academic Press．

［84］ PEGAU W S，ZANEVELD JR V．Temperature－dependent absorption of water in the red and near－infrared portions of the spectrum［J］．Limnol Oceanogr，1993，38(1):188－192．

［85］ MOREL A．Optical properties of pure water and pure seawater，in Optical Aspects of Oceanography，Jerlov N G，Nielsen E S．Editor．1974，Academic:New York．p．1－24．

［86］ SULLIVAN J M，TWARDOWSKI M S，ZANEVELD J R，et al．Hyperspectral temperature and salt dependencies of absorption by water and heavy water in the 400－750 nm spectral range［J］．Appl Opt，2006，45(21):5294－5309．

［87］ DEFOIN－PLATEL M，CHAMI M．How ambiguous is the inverse problem of ocean color in coastal waters?［J］．JGeophys Res，2007，112(C3):C03004．

［88］ MOREL A．In－water and remote measurements of ocean color［J］．Boundary－Layer Meteorology，1980，18:177－201．

［89］ MOUW CB，et al．Optical remote sensing of coastal and inland waters:challenges and recommendations for future satellite missions［J］．Remote Sens．Env，2015，160:15－30．

［90］ DOERFFER R，F(xiàn)ISHER J．Concentrations of chlorophyll，suspended matter，and gelbstoff in case II waters derived from satellite coastal zone color scanner data with inverse modeling methods［J］．JGeophys Res，1994，99(C4):7475－7466．

［91］ ROESLER C S，PERRY M J．In situphytoplankton absorption，fluorescence emission，and particulate backscattering spectra determined from reflectance［J］．J Geophys Res，1995，100(C7):13279－13294．

［92］ DEVRED E，SATHYENDRANATH S，PLATT T．Inversion based on a semi－analytical reflectance model，in Remote Sensing of Inherent Optical Properties:Fundamentals，Tests of Algorithms and Applications，Z．－P．Lee，Editor．2006，GKSSResearch Centre:Geesthacht，Germany．p．87－94．

［93］ MARITORENA S，SIEGEL D A，PETERSON A R．Optimization of a semianalytical ocean color model for global－scale applications［J］．Appl Opt，2002，41(15):2705－2714．

［94］ WERDELL，PJ，F(xiàn)RANZ B A，BAILEY SW，et al．Generalized ocean color inversion model for retrieving marine inherent optical properties［J］．Appl Opt，2013，52(10):2019－2037．

［95］ CHAMI M，ROBILLIARD D．Inversion of oceanic constituents in case I and II waters with genetic programming algorithms［J］．Applied Optics，2002，41(30):6260－6275．

［96］ HUANG S，LI Y，SHANG S，et al．Impact of computational methods and spectral models on the retrieval of optical properties via spectral optimization［J］．Optics Express，2013，21(5):6257－6273．

［97］ SUGIHARA S，KISHINO M．An algorithm for estimating the water quality parameters from irradiance just below the seasurface［J］．J Geophys Res，1988，93(D9):10857－10862．

［98］ HOGE F E，LYON P E．Satellite retrieval of inherent optical properties by linear matrix inversion of oceanic radiance models:an analysis of model and radiance measurement errors［J］．J Geophys Res，1996，101(C7):16631－16648．

［99］ WANG P，BOSSE S，ROESLER C．Uncertainties of inherent optical properties obtained from semi－analytical inversions of ocean color［J］．Appl Opt，2005，44(19):4074－4085．

［100］ LEE Z，CARDERK L，ARNONE R A．Deriving inherent optical properties from water color:amulti－band quasi－analytical algorithm for optically deep waters［J］．Appl Opt，2002，41(27):5755－5772．

［101］ SMYTH T J，MOORE G F，HIRATA T，et al．Semianalytical model for the derivation of ocean color inherent optical properties:description，implementation，and performance assessment［J］．Appl Opt，2006，45(31):8116－8131．

［102］ LOISEL H，STRAMSKI D．Estimation of the inherent optical properties of natural waters from the irradiance attenuation coefficient and reflectance in the presence of Raman scattering［J］．Appl Opt，2000，39(18):3001－3011．

［103］ BUKATAR P，BRUTON J E，JEROME J H，et al．Optical water quality model of Lake Ontario．2:Determination of chlorophyll a and suspended mineral concentration of natural waters from submersible and low altitude optical sensors［J］．Appl Opt，1981，20(9):1704．

［104］ BUKATA R P，JEROME JH，KONDATYEV A S，et al．Optical Properties and Remote Sensing of Inland and Coastal Waters［M］．1995，Boca Raton，F(xiàn)L:CRC Press．

［105］ LEE，Z P，CARDER K L，MOBLEY CD，et al．Hyperspectral remote sensing for shallow waters:2．Deriving bottom depths and water properties by optimization［J］．Appl Opt，1999，38(18):3831－3843．

［106］ BRICAUD A，MOREL A，PRIEUR L．Absorption by dissolved organic matter of the sea(yellow substance)in the UV and visible domains［J］．Limnol Oceanogr，1981，26(1):43－53．

［107］ LEE Z P．Visible－infrared Remote－sensing Model and Applications for Ocean Waters，in Department of Marine Science．1994，The University of South Florida:St． Petersburg．p．160．

［108］ LEE Z P，CARDER K L．Absorption spectrum of phytoplankton pigments derived from hyperspectral remote－sensing reflectance［J］．Remote Sens Environ，2004，89(3):361－368．

［109］ JERLOV N G．Marine Optics［M］．1976，New York:Elsevier．

［110］ PHILLIPSD M，KIRK J T O．Study of the spectral variation of absorption and scattering in some Australian coastal waters［J］．Aust J Mar Freshwater Res，1984，35(6):635－644．

［111］ FEE E J，HECKY R E，KASIAN S E M，et al．Effects of lake size，water clarity，and climatic variability on mixing depths in Canadian Shield lakes［J］．Limnol Oceanogr，1996，41(5):912－920．

［112］ CHEN Z，MULLER－KARGERA F E，HU C．Remote sensing of water clarity in Tampa Bay［J］．Remote Sens Environ，2007，109(2):249－259．

［113］ WANG M，SON S，HARDING L W．Retrieval of diffuse attenuation coefficient in the Chesapeake Bay and turbid ocean regions for satellite ocean color applications［J］．J Geophys Res，2009，114:C10011．

［114］ AUSTIN，R W，PETZOLD T J．The determination of the diffuse attenuation coefficient of seawater using the coastal zone color scanner，in Oceanography from Space，Gower J F R．Editor．1981，Plenum Press:New York．239－256．

［115］ MUELLER J L，TREES C C．eds．Revised SeaWIFS prelaunch algorithm for diffuse attenuation coefficient K(490)．Case Studies for SeaWiFS Calibratin and Validation，NASA Tech Memo．104566，ed．Hooker S B．a(chǎn)nd Firestone E R．Vol．41．1997，NASA Goddard Space Flight Center:Greenbelt，MD．

［116］ MITCHELL B G，KAHRU M．Algorithms for SeaWiFSstandard products dataset［J］．Cal Coop Ocean Fish Invest Rep，1998，39:133－147．

［117］ MUELLER J L．SeaWiFSalgorithm for the diffuse attenuation coefficient，K(490)，using water－leaving radiances at 490 and 555 nm，in SeaWiFSPostlaunch Calibration and Validation Analyses，Part 3，Hooker SB，Editor，2000，p．24－27．

［118］ MUELLER JL，F(xiàn)argion G S，C．R．McClain．Ocean Optics Protocols For Satellite Ocean Color Sensor Validation，Revision 4［C］．2003，Goddard Space Flight Center，Greenbelt，MD:NASA．

［119］ PREISENDORFER R W．Hydrologic optics vol．1:introduction［M］．1976，Springfield:National Technical Information Service．Also available on CD，Office of Naval Research．

［120］ IOCCG．Remote Sensing of Ocean Colour in Coastal，and Other Optically－Complex，Waters，in Reports of the International Ocean－Colour Coordinating Group，No．3，S．Sathyendranath，Editor．2000，IOCCG:Dartmouth，Canada．

［121］ LEE Z，CARDER K L．Band－ratio or spectral－curvature algorithms for satellite remote sensing?［J］．Appl Opt，2000，39(24):4377－4380．

［122］ DARECKI M，STRAMSKI D．An evaluation of MODIS and SeaWiFS bio－optical algorithms in the Baltic Sea［J］．Remote Sens Environ，2004，89(3):326－350．

［123］ LEE Z P，DARECKI M，CARDER K L，et al．Diffuse attenuation coefficient of downwelling irradiance:An evaluation of remote sensing methods［J］．J Geophys Res，2005，110(C2):C02017．

［124］ JAMET C，LOISEL H，DESSAILLY D．Retrieval of the spectral diffuse attenuation coefficientKd(λ)in open and coastal ocean waters using aneural network inversion［J］．J Geophys Res，2012，117(C10):C10023．

［125］ KIRK J T O．The vertical attenuation of irradiance as a function of the optical properties of the water［J］．Limnol Oceanogr，2003，48(1):9－17．

［126］ MCKEE D，CUNNINGHAM A，DUDEK A．Optical water type discrimination and tuning remote sensing band－ratio algorithms:Application to retrieval of chlorophyll andKd(490)in the Irish and Celtic Seas［J］．Estuarine Coastal and Shelf Science，2007，73(3－4):827－834．

［127］ WERDELL P J，BAILEY SW．An improvedin－situbio－optical data set for ocean color algorithm development and satellite data product validation［J］．Remote Sens．Environ，2005，98:122－140．

［128］ MOREL A，GENTILI B．Radiation transport within oceanic(case 1)water［J］． J Geophys Res，2004，109(C6):C06008．

［129］ LEE Z，HU C，SHANG S，et al．Penetration of UV－Visible solar light in the global oceans:Insights from ocean color remote sensing［J］．J Geophys Res，2013，118(9):4241－4255．

［130］ TYLER JE．The Secchi disc［J］．Limnol Oceanogr，1968，13(1):1－6．

［131］ SMITH R C，BAKER K S．Optical properties of the clearest natural waters［J］．Applied Optics，1981，20(2):177－184．

［132］ KIRK J T O．Volume scattering function，average cosines，and the underwater light field［J］．Limnol Oceanogr，1991，36(3):455－467．

［133］ MOREL A，LOISEL H．Apparent optical properties of oceanic water:dependence on the molecular scattering contribution［J］．Applied Optics，1998，37(21):4765－4776．

［134］ MAFFIONE R．Theoretical developments on the optical properties of highly turbid waters and seaice［J］．Limnol Oceanogr，1998，43(1):29－33．

［135］ MOBLEY C D，SUNDMAN L K．HydroLight 5．2 User’s Guide［M］．2013，Bellevue，Washington:Sequoia Scientific，Inc．

［136］ MOREL A，GENTILI B．Diffuse reflectance of oceanic waters:its dependence on sun angle as influenced by the molecular scattering contribution［J］．Applied Optics，1991，30(30):4427－4438．

［137］ POLCYN F C，BROWN W L，SATTINGER I J．The measurement of water depth by remote－sensing techniques．1970，University of Michigan:Ann Arbor．

［138］ MARITORENA S，MOREL A，GENTILI B．Diffuse reflectance of oceanic shallow waters:influence of water depth and bottom albedo［J］．Limnol Oceanogr，1994，39(7):1689－1703．

［139］ LEE Z P，CARDER K，CHEN R F，et al．Properties of the water column and bottom derived from AVIRIS data［J］．J Geophys Res，2001，106(C6):11639－11652．

［140］ WETTLE M，DEKKER aG，BRANDO V E．Monitoring bleaching of tropical coral reefs from space:A feasibility study using a physics based remote sensing approach．2005，Inland＆Coastal Water Remote Sensing，CSIRO:Canberra，Australia．p．51．

［141］ WETTLE M，BRANDO V E．SAMBUCA:Semi－analytical model for bathymetry，un－mixing and concentration assessment．2006，CSIRO Land and Water:Canberra，Australia．

［142］ KLONOWSKI W M，et al．Bottom type classification using hyperspectral imagery［C］．in Ocean Optics XVII．2004．Perth，Australia．

［143］ GOODMAN J，USTIN S．Airborne hyperspectral analysis of coral reef ecosystems in the Hawaiian Islands［C］．in International Symposium on Remote Sensing of Environment．2003．ISRSE．

［144］ MOBLEY C D，SUNDMAN L K，DAVIS C O，et al．Interpretation of hyperspectral remote－sensing imagery by spectrum matching and look－up tables［J］．Appl Opt，2005，44(17):3576－3592．

［145］ LEE Z P，CASEY B，ARNONE R，et al．Water and bottom properties of a coastal environment derived from Hyperion data measured from the EO－1 spacecraft platform［J］．J Appl Remote Sens，2007，1(1):011502．

［146］ DEKKER A G，PHINN SR，ANSTEE J，et al．Intercomparison of shallow water bathymetry，hydro－optics，and benthos mapping techniques in Australian and Caribbean coastal environments［J］． Limnol Oceanogr Methods，2011，9:396－425．

［147］ LEE Z P，HU C，CASEY B，et al．Global shallow－water bathymetry from satellite ocean color data［J］．EOS Transactions，2010，91(46):429－430．

［148］ LYZENGA D R．Shallow－water bathymetry using combined lidar and passive multispectral scanner data［J］．Int J Remote Sensing，1985，6(1):115－125．

［149］ STUMPF R P，HOLDERIED K，SINCLAIR M．Determination of water depth with high－resolution satellite imagery over variable bottom types［J］．Limnol Oceanogr，2003，48(1):547－556．

［150］ HU C，LI X，PICHEL W G，et al．Detection of natural oil slicks in the NW Gulf of Mexico using MODIS imagery［J］．Geophys Res Lett，2008，36(1):L01604．

［151］ HU C，LEE Z，MA R，et al，Moderate Resolution Imaging Spectroradiometer(MODIS)observations of cyanobacteriablooms in Taihu Lake，China［J］．J Geophys Res，2010，115(C4):C04002．

［152］ HU C，LI D，CHEN C，et al．On the recurrentUlva proliferablooms in the Yellow Seaand East China Sea［J］．J Geophys Res，2010，115(C5):C05017．

［153］ HU C，CANNIZZARO J，CARDER K L，et al．Remote detection ofTrichodesmiumblooms in optically complex coastal waters:Examples with MODIS full－spectral data［J］．Remote Sens．Environ，2010，114(9):2048－2058．

［154］ GOWER J，HU C，BORSTAD G，et al．Ocean color satellites show extensive lines of floatingSargassumin the Gulf of Mexico［J］．IEEE Trans Geosci Remote Sens，2006，44(12):3619－3625．

［155］ MCCLAIN C R．A decade of satellite ocean color observations［J］．Ann Rev Mar Sci，2009，1:19－42．

［156］ ANTOINE D，MOREL A．Oceanic primary production 1．Adaptation of a spectral light－photosynthesis model in view of application to satellite chlorophyll observations［J］．Global biogeochemical cycles，1996．10(1):p．43－55．

［157］ BEHRENFELD M J．Toward a consensus productivity algorithm for SeaWiFS，in SeaWiFS Technical Report Series，S．B．Hooker and E．R．Firestone，Editors．1998，NASA:Greenbelt，MD．

［158］ SATHYENDRANATH S，PLATT T，HORNE E P W，et al．Estimation of new production in the ocean by compound remote sensing［J］．Nature，1991，353:129－133．

［159］ LONGHURST A，SATHYENDRANATH S，PLATT T，et al．An estimate of global primary production in the ocean from satellite radiometer data［J］．J Plankton Res，1995，17(6):1245－1271．

［160］ KAHRU M，KUDELA R，MANZANO－SARABIA M，et al．Trends in primary production in the California Current detected with satellite data［J］．J Geophys Res，2009，114:C02004．

［161］ SATHYENDRANATH S，PLATT T，CAVERHILL C M，et al．Remote sensing of oceanic primary production:computations using a spectral model［J］．Deep－Sea Res，1989，36(3):431－453．

［162］ CLEVELAND JS，PERRY M J，KIEFER D A，et al．Maximal quantum yield of photosynthesis in the northwestern Sargasso Sea［J］．JMarine Res，1989，47(4):869－886．

［163］ SMITH R C，PRéZELIN B B，BIDIGARE R R，et al．Biooptical modeling of photosynthetic production in coastal waters［J］．Limnol Oceanogr，1989，34(8):1524－1544．

［164］ KIEFER D A，MITCHELL B G．A simple，steady state description of phytoplankton growth based on absorption cross section and quantum efficiency［J］．Limnol Oceanogr，1983，28(4):770－776．

［165］ MOREL A，BERTHON J F．Surface pigments，algal biomass profiles，and potential production of the euphotic layer:relationships reinvestigated in review of remote－sensing applications［J］．Limnol Oceanogr，1989，34(8):1545－1562．

［166］ FROUIN R，LINGNER D W，GAUTIER C，et al．A simple analytical formula to compute clear sky total and photosynthetically available solar irradiance at the ocean surface［J］．J Geophys Res，1989，94(C7):9731－9742．

［167］ BALCH W，EVANSR，BROWN J，et al．The remote sensing of ocean primary productivity:use of a new data compilation to test satellite algorithms［J］．JGeophys Res，1992，97(C2):2279－2293．

［168］ MITCHELL B G，KIEFER D A．Chlorophyll a specific absorption and fluorescence excitation spectrafor light－limited phytoplankton［J］．Deep－Sea Res，1988，35(5):639－663．

［169］ SOSIK H M，MITCHELL B G．Light absorption by phytoplankton，photosynthetic pigments and detritus in the California Current System［J］．Deep Sea Research Part I:Oceanographic Research Papers，1995，42(10):1717－1748．

［170］ BRICAUD A，BéDHOMME A L，MOREL A．Optical properties of diverse phytoplanktonic species:experimental results and theoretical interpretation［J］．J Plankton Res，1988，10(5):851－873．

［171］ O’REILLY J E，and Coauthors．SeaWiFS Postlaunch Calibration and Validation Analyses，Part 3，in SeaWiFS Postlaunch Technical Report Series，S．B．Hooker and E．R．Firestone，Editors．2000，NASA Goddard Space Flight Center:Greenbelt，MD．p．58．

［172］ LOISEL H，MOREL A．Light scattering and chlorophyll concentration in Case 1 waters:A reexamination［J］．Limnol Oceanogr，1998，43(5):847－858．

［173］ SATHYENDRANATH S，COTA G，STUART V，et al．Remote sensing of phytoplankton pigments:A comparison of empirical and theoretical approaches［J］．Int J Remote Sens，2001，22(2－3):249－273．

［174］ DEL CASTILLO C E，GILBESF，COBLE P G，et al．On the dispersal of riverine colored dissolved organic matter over the West Florida Shelf［J］．Limnol Oceanogr，2000，45(6):1425－1432．

［175］ CARDER K L，CHEN F R，LEE Z P，et al．Semianalytic Moderate－Resolution Imaging Spectrometer algorithms for chlorophyll－a and absorption with bio－optical domains based on nitrate－depletion temperatures［J］．JGeophys Res，1999，104(C3):5403－5421．

［176］ SIEGEL D A，WESTBERRY T K，O’BRIEN M C，et al．Bio－optical modeling of primary production on regional scales:the Bermuda BioOptics project［J］．Deep－SeaResearch II，2001，48:1865－1896．

［177］ CAMPBELL J，ANTOINE D，ARMSTRONG R，et al．Comparison of algorithms for estimating ocean primary production from surface chlorophyll，temperature，and irradiance［J］．Global biogeochemical cycles，2002，16(3):9－1－9－15．

［178］ CARR M E，F(xiàn)RIEDRICHSM A M，SCHMELTZ M，et al．A comparison of global estimates of marine primary production from ocean color［J］．Deep－Sea Res．II，2006，53(5－7):741－770．

［179］ FRIEDRICHS M A M，CARR M E，BARBER R T，et al．Assessing the uncertainties of model estimates of primary productivity in the tropical Pacific Ocean［J］．Journal of Marine Systems，2009，76(1－2):113－133．

［180］ SABA V S，F(xiàn)RIEDRICHS M A M，ANTOINE D，et al．An evaluation of ocean color model estimates of marine primary productivity in coastal and pelagic regions across the globe［J］．Biogeosciences，2011，8:489－503．

［181］ SABA V S，F(xiàn)RIEDRICHSM A M，CARR M E，et al．Challenges of modeling depth integrated marine primary productivity over multiple decades:a case study at BATSand HOT［J］．Global Biogeochem．Cycle，2010，24(3):GB3020．

［182］ PERRY M J．Assessing marine primary production from space［J］．BioScience，1986，36(7):461－467．

［183］ MOREL A．Available，usable，and stored radiant energy in relation to marine photosynthesis［J］．Deep－SeaRes，1978，25(8):673－688．

［184］ CARDER K L，LEE Z P，MARRA J，et al．Calculated quantum yield of photosynthesis of phytoplankton in the Marine Light－Mixed Layers(50oN/21oW)［J］．J Geophy Res，1995，100(C4):6633－6655．

［185］ SOSIK H M．Bio－optical modeling of primary production:Consequences of variability in quantum yield and specific absorption［J］．Marine Ecology Progress Series，1996，143:225－238．

［186］ PLATT T，SATHYENDRANATH S．Fundamental issues in measurement of primary production，in Measurement of primary production from the molecular to the global scale，W．K．W．Li and S．Y．Maestrini，Editors．1993，ICESMarine Science Symposium．p．3－8．

［187］ HIRATA T，HARDMAN－MOUNTFORD N J，BARLOW R，et al．An inherent optical property approach to the estimation of size－specific photosynthetic rates in eastern boundary upwelling zones from satellite ocean colour:An initial assessment［J］．Progress in Oceanography，2009，83(1－4):393－397．

［188］ KIRK J T O．Solar heating of water bodies as influenced by their inherent optical properties［J］．J Geophys Res，1988，93(D9):10897－10908．

［189］ SATHYENDRANATH S，GOUVEIA A D，SHETYE S R，et al．Biological control of surface temperature in the Arabian Sea［J］．Nature，1991，349:54－56．

［190］ KARA A B，HURLBURT H E，ROCHFORD P A．The impact of water turbidity on the interannual seasurface temperature simulations in a layered global ocean model［J］．J Phys Oceanogr，2004，34(2):345－359．

［191］ LIANG X，WU L．Effects of solar penetration on the annual cycle of seasurface temperature in the North Pacific［J］．J Geophys Res，2013，118(6):2793－2801．

［192］ MURTUGUDDE R，BEAUCHAMPJ，MCCLAIN C R，et al．Effects of penetrative radiation on the upper tropical ocean circulation［J］．Journal of Climate，2002，15(5):470－486．

［193］ MILLER A J，ALEXANDER M A，BOER G J，et al．Potential feedbacks between Pacific Ocean ecosystems and interdecadal climate variations［J］．Bull Amer Meteor Soc，2003，84(5):617－633．

［194］ PENTA B，LEE Z，KUDELA R M，et al．An underwater light attenuation scheme for marine ecosystem models［J］．Optics Express，2008，16(21):16581－16591．

［195］ MANIZZA M，QUéRéCL，WATSON A J，et al．Bio－optical feedbacks among phytoplankton，upper ocean physics and seaice in a global model［J］．Geophys Res Lett，2005，32(5):L05603．