We present a numerical model of the ocean that couples a three-stream
radiative transfer component with a marine biogeochemical–ecosystem component in
a dynamic three-dimensional physical framework. The radiative transfer
component resolves the penetration of spectral irradiance as it is absorbed and
scattered within the water column. We explicitly include the effect of several
optically important water constituents (different phytoplankton functional
types; detrital particles; and coloured dissolved organic matter, CDOM). The
model is evaluated against in situ-observed and satellite-derived products. In
particular we compare to concurrently measured biogeochemical, ecosystem, and
optical data along a meridional transect of the Atlantic Ocean. The simulation
captures the patterns and magnitudes of these data, and estimates surface
upwelling irradiance analogous to that observed by ocean colour satellite
instruments.
We find that incorporating the different optically important
constituents explicitly and including spectral irradiance was crucial to
capture the variability in the depth of the subsurface chlorophyll
Spectra for
Light is fundamental to phytoplankton and photosynthesis. Understanding ocean productivity therefore requires detailed knowledge of how light penetrates through seawater. Attenuation of light within the water column is an interaction of absorption and scattering by “optically important constituents”, including water molecules, detrital matter, coloured dissolved organic matter (CDOM), and the phytoplankton themselves.
Phytoplankton absorb light in the visible spectrum (400 and
700
Much is known about the optics of water (e.g. Pope and Fry, 1997; Smith and Baker, 1981; Morel, 1974; Zhang and Hu, 2009; Kirk, 1994). Although much is known about the distributions of CDOM (Nelson and Siegel, 2013), detritus (Loisel, 2002), and phytoplankton (IOCCG report 15, 2014) it remains unclear how their distributions feed back to phytoplankton community structure and biogeochemistry. Numerical models provide useful tools to explore these interactions, but to do so requires an appropriately detailed description of the photosynthetically available radiation (PAR).
Several recent models resolve the light spectrum and some of the
absorption and scattering properties of different constituents (e.g.
Mobley et al., 2009; Fujii et al., 2007; Gregg and Casey, 2007; Bisset
et al., 1999). Such models include fully coupled radiative transfer,
but differ in the levels of simplification for computational
efficiency (e.g. Fujii et al., 2007; Gregg and Casey, 2007) and
differ in which and how they treat the different water
constituents. For instance, CDOM is treated as uniform in Fujii
et al. (2007), and linked to chlorophyll
In Sect.
We perform a number of sensitivity experiments that explore the value
of the additional model complexity (Sect.
The biogeochemical–ecosystem model resolves the cycling of carbon,
phosphorus, nitrogen, silica, iron, and oxygen through inorganic,
living, dissolved, and particulate organic phases as discussed in
Follows et al. (2007), Dutkiewicz et al. (2009, 2012), and Hickman
et al. (2010). The biogeochemical and biological tracers are
transported and mixed by a the MIT general circulation model (MITgcm)
(Marshall et al., 1997). The physical framework is flexible, but here
we employ a global configuration which is constrained to be consistent
with altimetric and hydrographic observations (the ECCO-GODAE state
estimates; Wunsch and Heimbach, 2007). This three-dimensional
configuration has 1
Similar to several of these previous studies, we resolve several
phytoplankton types,
We provide complete model equations, description and parameter values
in Appendix Sect.
Fixed biogeochemical–ecosystem model parameters (1).
Irradiance just below the surface of the ocean is provided by the
Ocean–Atmosphere Spectral Irradiance Model (OASIM) (Gregg and Casey,
2009) in two downward streams: direct (
We parameterize this “three-stream” irradiance model following Aas (1987), Ackleson et al. (1994), and Gregg (2002). The model is
described by the simultaneous equations for the light streams in each
waveband (
This set of equations can be simplified following Aas (1987) by
approximating
We calculate total scalar irradiance,
We note that the radiative transfer component is a simplification from a full radiance model and, in particular, does not resolve the angular distribution of light or angular dependence of scattering. These assumptions have been shown to be small in terms of the needs for ecosystem models (Mobley et al., 2009). Though not a full radiative transfer model, our three-stream treatment does provide the relevant output for our objectives: the spectral light available for photosynthesis and an upwelling component that at the sea surface is similar to that seen by a satellite.
Since the model resolves an upwelling stream of irradiance, we can
calculate a surface reflectance (unitless):
To compare to remotely sensed reflectance (
Fixed biogeochemical–ecosystem model parameters (2).
The bidirectional function
Attenuation of irradiance results from absorption by water molecules
(
In the model we use absorption and scattering coefficients
(Fig.
We assume absorption by water molecules (
The model uses the absorption and scattering spectra for detrital
matter (Fig.
We note that in the optical community the term “non-algal particles”, or NAP, is frequently used for any non-phytoplankton particles. In this paper we specifically use the term “detritus” instead, as we link to the non-living organic matter pool and do not explicitly resolve other non-algal particles such as viruses and heterotrophic bacteria.
Phytoplankton-specific parameter description.
CDOM absorbs highly in the short wavelengths and absorption decreases
exponentially with increasing wavelength (Kitidis et al., 2006; Nelson
and Siegel, 2013). CDOM is not usually explicitly resolved in marine
ecosystem models (exceptions are Xiu and Chai, 2014, and Bissett
et al., 1999). Here we have resolved an explicit CDOM-like tracer
(denoted “CDOM”) similar to Bissett et al. (1999). The model CDOM
has units of concentration (
We parameterize
The absorption and scattering by phytoplankton is the net effect of
each phytoplankton type resolved in our model,
The Chl
Phytoplankton-specific parameter values. The abbreviations here refer to model analogues of other large eukaryotes (LgEuk),
Phytoplankton growth is modelled as a function of temperature,
irradiance, and nutrients as in Hickman et al. (2010) following Geider
et al. (1998). The growth rate is equal to the carbon-specific
photosynthesis rate:
Since some pigments are photoprotective, phytoplankton do not use all
the light that they absorb for photosynthesis. Similar to Hickman
et al. (2010) and Bisset et al. (1999), the total absorption spectra
are therefore greater than the photosynthetic absorption spectra,
We resolve nine phytoplankton “functional” types: these include
analogues of diatoms, other large eukaryotes, coccolithophores,
picoeukaryotes,
Zooplankton/grazing-specific parameter description.
Cell size governs many traits. Smaller phytoplankton have lower
nutrient half-saturation constants and sink more slowly. The maximum
growth rates are guided by observations, diatoms having the highest
rates and
In this model we treat the phytoplankton light absorption and
scattering explicitly (Sect.
Spectra for absorption by photosynthetic pigments
(
Zooplankton/grazing-specific parameter values.
We parameterize all phytoplankton to have the same maximum quantum
yield of carbon fixation (
We resolve two zooplankton classes (large and small) that graze on the
phytoplankton using a Holling III scheme (Holling, 1959). The large
class preys preferentially on the diatoms, coccolithophores, and
The inclusion of radiative transfer and spectral light, as well as capturing several
important optical constituents, is a significant development in the
model. However, this version of the model is not without limitations. One
major, though currently necessary, simplification is to assume constant
absorption and scattering spectra (Fig.
Scattering, particularly by detrital particles, remains the least well developed aspect
of the model. In particular, we neglect variations in detrital particle size
distributions, which are likely to be important (Stramski et al., 2001). Additionally, the
spectrum for
We additionally currently neglect other potentially important optical constituents such as minerals (e.g. Stramski et al., 2001), particulate inorganic carbon (e.g. Balch and Itgoff, 2009), colloids and bubbles (e.g. Stramski et al., 2004), and non-photosynthetic organisms including zooplankton, bacteria (e.g. Morel and Ahn, 1991), and viruses (e.g. Stramski et al., 2001). We felt that these are, as yet, not well enough constrained to include explicitly in the model.
The limitations list above should not, however, detract from the major enhancement to the model, and our assumptions are similar to those of other models (e.g. Fujii et al., 2007; Gregg and Casey, 2007). This new model provides a unique platform to examine global implication of optical properties to the phytoplankton ecosystem, feedbacks to the biogeochemistry, and links to satellite data that are not possible with limited observational data. Here we first validate the model in a standard “default” configuration. We then provide a series of studies exploring the significance of each of the optical constituents and our parameterization. Several studies in progress build on these results.
We initialize the macronutrient fields (nitrate, phosphate, and silicic acid) from World Ocean Atlas (Garcia et al., 2006) climatologies and the iron from previous model output. We also use previous model output to provide distribution of the ammonium, nitrite, and dissolved and particulate matter. The total phytoplankton biomass is initialized from previous model output, divided equally between groups, except for the diazotrophs, which are initialized at a much lower value so as not to flood the system with new nitrogen in the first few time steps. Zooplankton are similarly initialized with equal distribution in both groups.
The model time step is 3 h. We tested this against smaller time steps with almost identical results. We run the simulation forward for 10 years with a repeating generic “year” from the physical ECCO-GODAE products (Wunsch and Heimbach, 2007). Model results shown in this section are from the last year of the simulation. The phytoplankton establish a repeating pattern after about 3 years such that we can assume a “quasi-steady state” by year 10. A slow drift as deep water nutrient distributions adjust does not significantly change the results over the remaining time period.
We evaluate the model results against a range of in situ observations
and satellite-derived products. In particular, we focus on the unique
data set including biogeochemical, ecological and (some previously
unpublished) optical properties that were obtained as part of the
AMT-15 cruise.
Though there are other AMT cruises that include some similar and/or different
combinations of optical data (e.g. AMT-19, Dall'Olmo et al., 2012, Martinez-Vicente et al., 2013),
we chose to look at only a single transect for clarity. In particular, the
combination of data on spectral irradiance penetration,
The model broadly reproduces the horizontal gradients at the surface
but, importantly, also captures the subsurface Chl
The model also captures observed variability in
Satellite (MODIS)-derived Chl
Comparison of model output (right column, October mean) with data
collected during AMT-15 (left column, collected from late September to late
October):
We have used the AMT-15 measured downwelling irradiance and upwelling
zenith radiance together with the inverse-modelling procedure of
Gordon and Boynton (1997, 1998) to estimate the total absorption and
total backscattering in several wavelengths
(Fig.
Since the model realistically captures much of the variability in
optical constituents it also accurately resolves the penetration of
light through the water column (Fig.
The model captures intricate patterns of absorption and scattering
that develop from the interplay of different optical constituents and
suggests the importance of treating each constituent separately for
reproducing the in situ light field. We explore this further in
Sect.
Comparison of model output (right column, October mean) with data
collected during AMT-15 (left column):
Comparison of data collected along AMT-15
That the model captures much of the Chl
Relative to the composite of iron data (Tagliabue et al., 2012), we
also capture high iron in the Atlantic Ocean and lower iron over much
of the Pacific (Fig.
We find that the spatial standard deviation (between 0.85 and 1.15) and correlation
(greater than 0.9) of the model vs. observed nutrients are encouraging
(Fig.
The model ecosystem has distinctive seasonal cycles
(Fig.
A unique feature of this model is irradiance reflectance output, which we have
converted to remotely sensed reflectance (
Model- and satellite-derived products and climatologies of in situ
measurements for annual mean and biases:
Eight of the nine phytoplankton functional groups that we resolve have
distinct biogeography (Fig.
We compare simulated biomass of the picophytoplankton to observations
from AMT-15 (Fig.
The MAREDAT (MARine Ecosystem DATa; Buitenhuis et al., 2013)
compilation provides a comprehensive, though still sparse,
climatological distribution of several plankton functional
groups. Here we re-grid the MAREDAT compilation onto a 5
Taylor diagram showing correlation and normalized SD between annual
mean modelled Chl
Comparison of monthly model Chl
Given the sparsity of in situ measurements of phytoplankton types, it
is natural to attempt to capture aspects of biogeography from space
(IOCCG report 15, 2014; IOCCG report 9, 2009). Here we compare the
model output to the PHYSAT product (Alvain et al., 2008), which
empirically relates optical properties to specific (probably dominant)
phytoplankton types (Fig.
These global “observations” contain many uncertainties stemming mainly from the scarcity of in situ data, but the model does not disagree with their findings. The model captures key patterns of observed optical and ecological properties. It provides a tool to explore aspects of the ocean biogeochemistry and ecosystem that are not possible with models that do not explicitly resolve radiative transfer, spectral irradiance, and resolution of the different water optical properties. In the next section we explore the role of the various water constituents on the irradiance spectrum and how they impact biogeochemistry and ecosystem structures.
We conduct two sensitivity experiments to highlight the importance of the
extra levels of complexity of this new version of the model. In the first
experiment (designated EXP-V0) the biogeochemistry and ecosystem are the same
as in the default experiment described above (designated EXP0) but there is
only a single band of irradiation (400–700 nm, summed over the original
25 nm, so that total PAR is conserved); attenuation (
The results from EXP-V0 (Fig.
In experiment EXP-V1 we include all the optical constituents explicitly (as
in EXP0), though with only a single band of PAR (as in EXP-V0). We assume the
absorption and scattering coefficients for 500 nm in this experiment. This
experiment (Fig.
These sensitivity experiments suggest that explicitly capturing regional changes in all optical constituents is essential for the realistic variations in the depth of light penetration. Resolving the light spectrum further enhances the realism of the results. The addition of the radiative transfer code is essential for obtaining upwelling irradiance that can link to satellite products.
Optical constituents play varying roles in their effect on irradiance
attenuation (absorption and scattering). These roles have long been a topic
of interest; however many studies have included only limited observations and
been of highly localized in character (e.g. Jerlov, 1953; Chang and Dickey,
1999) but have nonetheless recognized that they vary regionally (e.g. Barnard
et al., 1998; Simeon et al., 2003). Targeted cruises have also provided
larger-scale observations indicating a wide range of values for each
constituent and altering importance in different regions (e.g. BIOSOPE;
Bricaud et al., 2010). Additionally, several attempts have been made to
construct algorithms to determine the relative contributions from more easily
measured quantities, including those from satellites (e.g. Maritorena et al.,
2002; Lee et al., 2002, 2007; Ciotti and Bricaud, 2006; Werdell et al., 2013;
Zheng and Stramski, 2013). Our model provides a unique global
three-dimensional perspective. Here our results focus on an (extended) AMT
transect (Figs.
Comparison of model with satellite (MODIS)-derived remotely sensed
reflectance,
Model annual mean biomass (
Absorption by water molecules is most important at longer wavebands (Pope and
Fry, 1997) but still has an impact at shorter wavebands
(Fig.
We perform a series of sensitivity experiments to explore the role of each constituent in setting the irradiance field in the ocean and on surface reflectance, and to see how changes to these constituents feed back to the ecosystem and biogeochemistry. The range of values for these experiments is designed to cover and go beyond the natural range of the absorption and scattering by the water constituents. We additionally explore how different assumptions and parameterizations for the optical constituents affect the simulation results.
Comparison of model output (October mean) with data collected along
AMT-15:
Comparison of model plankton-type biomass (
Observations have determined that detrital matter
does play a role in light attenuation, though with
varying regional importance (e.g. Jerlov, 1953; Bricaud et al., 2010).
We conduct several sensitivity studies to explore the relative
importance of EXP0: this is the default run where EXP-D1: we set EXP-D2: we set EXP-D3: we set EXP-D4: we set EXP-D5: as in Fujii et al. (2007) we represent where
Removing the detrital absorption (EXP-D1) leads to bluer
wavebands reaching to greater depth (Fig.
We observe distinct biogeochemical feedbacks. With lower absorption by
detritus (EXP-D1) the depth integrated phytoplankton biomass
in the high latitudes increases (Fig.
Comparison of model phytoplankton dominate type with dominant
type found from PHYSAT (Alvain et al., 2008) satellite-derived product for
Sensitivity experiments examining value of increased optical
complexity in model. Chl
The main attenuation of light with depth is through absorption, and as such
alterations to the backscattering by detrital matter (EXP-D3 and
EXP-D4) have little effect on the irradiance fields at depth
(Fig.
In EXP0,
CDOM and its contribution to light absorption are observed to vary in
different regions of the ocean (e.g. Jerlov, 1953; Bricaud, 1981; Nelson and
Seigel, 2013, Morel et al., 2010), and many studies have
attempted to empirically link
In all experiments,
In the series of experiments we make different assumptions on
EXP0: EXP0-C1: EXP-C2: EXP-C3: EXP-C4: EXP-C5:
Community structure shifts significantly in response to the amount of
irradiance that the CDOM absorbs (Fig.
The three alternative parameterizations of
These experiments illustrate that the parameterization of CDOM has a very significant impact on community structure and reflectance and suggest that it is crucial to explicitly resolve CDOM in models and learn more about its variability in the ocean (Morel et al., 2010; Nelson and Siegel, 2013).
Model output along extended AMT-15 transect (annual mean) of
Model output along extended AMT-15 transect (annual mean) of
Idealized experiments were also conducted to explore the sensitivity
due to phytoplankton absorption and scattering
(Fig. EXP0: this is the default run, with each phytoplankton
type having a specific absorption and scattering spectrum
(Fig. EXP-P1: we artificially set
EXP-P2: we artificially set
EXP-P3: we set EXP-P4: we assume all phytoplankton have the same
absorption properties (the mean, black lines, in Fig. EXP-P5: we assume all phytoplankton types have the same
scattering and backscattering properties (the mean, black line, in
Fig.
Altering the absorption by phytoplankton (EXP-P1 and
EXP-P2) has a similar impact to altering CDOM or detritus
(Fig.
As discussed above, the main attenuation of light is through absorption, and
thus when we assume no scattering by phytoplankton (EXP-P3) there is
almost no change in dominant functional type. However, since scattering does
substantially affect the upwelling light, there is some (though small) change
in reflectance compared to the default run (EXP0). An experiment with 4
times
In EXP-P4 and EXP-P5 we explore the importance of
the phytoplankton type-specific absorption and scattering spectra in
setting their biogeography and biogeochemical consequences. Total
Detritus sensitivity experiments.
CDOM sensitivity experiments.
Phytoplankton sensitivity experiments.
When assuming a mean scattering spectrum for all phytoplankton
(EXP-P5) we find, similar to EXP-P3, almost no difference to the
irradiance field, dominant functional type, or biogeography. There are,
however, small changes to the reflectance. Changes in the reflectance are
also apparent when the mean
In this paper we have presented a version of the MIT biogeochemistry–ecosystem model (the “Darwin Project” model) which now incorporates radiative transfer, spectrally resolved irradiance, and explicit representation of optically important water constituents. Our treatment of optical properties combines many features from prior studies (e.g. Gregg et al., 2007; Fujii et al., 2007; Mobley, 2011; Bissett et al., 1999, 2004) but is more comprehensive than most. In particular, we include a detailed absorption by several different types of phytoplankton as in Gregg and Casey (2007), explicitly resolve a CDOM-like tracer as in Xiu and Chai (2014) and Bisset et al. (1999), and also resolve detrital particulate matter in a similar manner to Fujii et al. (2007).
We have evaluated our model against a range of in situ observations and satellite-derived products. The model captures the large-scale biogeochemical, ecosystem, and optical characteristics as suggested by these data sets. In particular, we have used a unique data set collected during AMT-15 which includes concurrent optical, biogeochemical, and ecosystem measurements. The model captures the observed basin scale and vertical distribution. In many of the instances where the model does not compare well to the observations, we find that the physics of the model are at least partly responsible.
The model captures spatial light absorption by different optical
constituents and the relative magnitude of the scattering. However,
the scattering, particularly by detrital particles, remains the least
well constrained aspect (see Sect.
Each of the optical constituents resolved in the model (water, CDOM,
detrital particles, and phytoplankton) has an important role in
attenuating irradiance through the water column, but the relative
importance differs between regions, with depth, and with wavelength
(Fig.
Our sensitivity experiments suggest that models that neglect the
explicit and independently varying absorption by detrital particulate
matter and CDOM are missing important components that have
implications for the biogeochemistry and productivity of the
model. For instance, we find that the magnitude of the light absorption
of any of the water constituents that we resolve is important in
setting the penetration of irradiance in different wavebands.
The subsurface Chl
Changes
to the irradiance spectrum will have important ramifications for the
community structure. Lower absorption by the optical constituents
leads to deeper penetration of blue light and favours phytoplankton
which absorb better in the shorter wavelengths
(e.g.
An important product of the model is the surface irradiance
reflectance, which
provides a more direct comparison to satellite data than derived
products such as Chl
The absorption by any of the optical constituents strongly determines the amount of upwelling irradiance and consequently the surface reflectance. In particular, we found that the regional variations in CDOM are important in setting the patterns of reflectance (see EXP-C5). Though alterations to scattering appears to have little effect on the in-water optical fields, they have a significant impact on the surface reflectance fields. Even slight changes to the scattering by phytoplankton (see EXP-P5) have an effect on the reflectance. Such changes are important when attempting to retrieve information on the community structure from ocean colour satellite products (e.g. IOCCG report 15, 2014).
The amount and type of irradiance that penetrates through the water
column is an important issue when studying phytoplankton productivity
and community structure. And yet, ocean models routinely offer very
crude parameterizations of light attenuation and neglect the spectral
quality. We have improved the MITgcm ecosystem and biogeochemistry
model by incorporating spectral light, explicit radiative transfer, and
representations of several optical constituents. The model performed
well when compared to observations.
Capturing each of the optically important constituents explicitly and
including a spectrum of light was important for obtaining realistic
variability in depth of the subsurface Chl
The sensitivity studies were intentionally hypothetical to provide a wide range of responses. They provide evidence that capturing how each of the optical constituents absorbs and scatters irradiance has important ramifications for biogeochemistry and the phytoplankton community structure. This feedback between the light field and the biogeochemistry can only be captured by a fully three-dimensional coupled ecosystem–radiative transfer model.
The model provides a platform to explore the relative importance of different optical constituents for biogeography, biogeochemistry, and optical properties such as those measured by satellite. We believe that this model will be useful in examining the role of the irradiance spectrum and pigments in setting biogeography, in exploring how changes in irradiance and/or optical constituents will impact the future oceans, and in providing a laboratory to explore the use of water-leaving radiance as a marker of changes in the marine ecosystem.
The model equations are based on those of Follows et al. (2007),
Dutkiewicz et al. (2009, 2012), and Hickman et al. (2010). We consider
the cycling of phosphorus, nitrogen, silica, and iron, as well as carbon,
alkalinity, and dissolved oxygen (the latter three following Ullman
et al., 2009). We also resolve here explicit dynamic Chl
Several nutrients,
The source and sinks of each tracer,
Biological rates (plankton growth and the parameterization of
remineralization of organic matter) are represented as a function of
temperature, following the Arrhenius equation (Kooijman, 2000), similar to
Eppley (1972):
Phytoplankton growth is a function of temperature, irradiance, and
nutrients. We follow Hickman et al. (2010), which in turn follows
Geider et al. (1998), such that the growth rate is equal to the carbon-specific photosynthesis rate:
The light-saturated photosynthesis rate is a function of nutrients and
temperatures:
For each phytoplankton
The local Chl
The increase of Chl
Phytoplankton can be photoinhibited (following Hickman et al., 2010), such
that
Nutrient limitation is determined by the most limiting nutrient:
Limitations by
Nitrogen is available in three forms, of which ammonia is the preferred
type:
Zooplankton grazing is parameterized as
The maximum grazing
Zooplankton are assumed to have both a linear and quadratic loss term. The linear term represents excretion and mortality; the quadratic loss terms represent grazing by higher trophic levels (Steele and Henderson, 1992) that are not explicitly resolved in this model.
Phytoplankton take up DIN in three forms (
The oxidation of
Denitrification occurs when
The iron model we use is based on that of Parekh et al. (2004,
2005). We explicitly model the complexation of iron with an organic
ligand:
We assume that only the free iron (
The sedimentary source (
Air–sea exchange of
Dissolved inorganic carbon (DIC) carried in the model is made up of
carbon dioxide and carbonic acid and other carbonate species:
The radiance in the ocean in its most general form,
Here,
The attenuation coefficient
At the sea surface, the downward part of
where the downward scalar irradiance is
The downward diffuse and upward irradiance are defined as
The outer integral is split into contributions from
The effective backward scattering coefficients are defined as
corrections to
In terms of the effective backscattering coefficients,
Likewise,
In general,
We close the system of equations by making the following
assumptions (following Aas, 1987):
Equations (
The equation for
In contrast to Aas (1987), Ackelson et al. (1994) and Gregg (2002) we do not
make further approximations but instead solve the remaining equations
explicitly. We can write the remaining two equations (Eqs.
Following Kylling (1995) we write the inhomogeneous solution as
The eigenvalues of
Within a computational layer, the general solution can be written as
In the bottom layer,
In order to solve this coupled system of equations, we follow Kylling
et al. (1995) and Toon et al. (1989), who observed that it can be
transformed to tri-diagonal form by eliminating
Phytoplankton total light absorption spectra shown in
Fig.
Total phytoplankton light scattering was also taken for representative
phytoplankton types in culture, with every attempt to match types used for
absorption: Syn, generic
The absorption and scattering properties of the other optical constituents were also obtained from the literature, as outlined in the main text.
In order to estimate backscattering
Gordon and Boynton (1998) propose that
where
and
In order to solve for
For noisy data, this estimate of
We are grateful to NASA (NNX13AC34G), NERC (NE/H015930/2), NSF
(OCE1434007, OCE1155295), and the Gordon and Betty Moore Foundation for funding. The work on AMT-15 was supported by NERC through the AMT
consortium (NER/O/S/2001/00680). This is contribution number 262 of the AMT
programme. Participation of A. E. Hickman in AMT-15 was funded by a NERC PhD
studentship. We thank S. Alvain for making the PHYSAT data available to us.
We appreciate the NASA GSFC website for the MODIS level 3 data used here, as
well as the ocean productivity website at Oregon State University for primary
production data, and the efforts of the MAREDAT team for the phytoplankton
compilations. We would like to thank D. Stramski for phytoplankton absorption
and scattering data for phytoplankton and detritus, and D. Suggett and
L. Moore for additional phytoplankton absorption spectra. Thanks also to the
Atlantic Meridional Transect (AMT) community, in particular J. Heywood and
M. Zubkov for analytical flow cytometry measurements, L. Hay and G. Moore for
optics data, A. Stubbins for aCDOM data, M. Woodward and K. Chamberlain for
nutrient data, and A. Pattenden for assistance with Chl