Article Highlights

  • Ocean colour satellites are designed to investigate marine biota by their distinct optical signatures

  • Some common ocean-colour products are concentration of chlorophyll-a and marine primary production along with many other indicators of marine ecosystem status

  • Ocean colour has become a very useful tool in climate research and is recognized as an essential climate variable

1 Introduction

The global ocean covers two-thirds of the surface of our planet; it is about 3700 m deep on average; it is a highly dynamic fluid medium that experiences variability at multiple temporal and spatial scales. The ocean is the cradle of life as we know it and supports a complex network of species and ecosystems that have adapted to this fluid and at times hostile environment. There are many reasons why the ocean biota and their processes are essential for both humans and our green planet.

It is estimated that the oceans are home to some 2.2 million species, of which only 9% has been described (Mora et al. 2011). With such a small portion that has been observed and described, it is highly likely that we understand little about the diverse services that these organisms provide, to maintain the intricate and inter-connected properties that render the Earth a hospitable and pleasant environment for us humans; maintain the multitudes of life forms that make our planet beautiful and shape our landscape; and control the chemical composition of our atmosphere.

The ocean biota plays a significant role in the global biogeochemical cycles of this planet, notably the cycle of carbon, various nitrogen compounds (e.g., nitrate, nitrite, ammonia and nitrogen gas), phosphate, silica and iron. In fact, the role of marine phytoplankton (microscopic, free-floating, single-celled plants) as net primary producers of organic material in the ocean is estimated to be equivalent to that of terrestrial plants (Longhurst et al. 1995; Field et al. 1998). For this and other reasons, such as their role in absorbing sunlight and thereby modulating the depth distribution of solar heating of the upper ocean, phytoplankton are considered an important player in studies of the ocean carbon cycle.

Marine primary production, currently estimated to be about 50 Pg of carbon globally per annum (Kulk et al. 2020a, b, 2021), sustains all pelagic life in the oceans, including the fish and seafood that we rely on for nutritious, high-quality food (Budge et al. 2014). With aquaculture making an increasingly important contribution to the global supply of fish and seafood for human consumption, there are growing concerns on its potential adverse impact on the marine ecosystems and on water quality.

Whether it be from considerations of the impacts of human activities on water quality or the effect of climate change on marine ecosystems, or for understanding how to manage marine living resources in a sustainable manner, we have a need for observing the ocean biota globally, consistently, over long periods, and at various time scales commensurate with the intrinsic scales of biological processes and their variability.

Satellite-based observations have an important role to play in meeting this requirement. There are many factors that render remote sensing a suitable mode of observation. A particular advantage is the capability for observations at the global scale using a single instrument, in a consistent manner. Another is the ability of satellites to observe the oceans at a variety of scales that are relevant for biological processes in the ocean–at temporal scales ranging from hours to decades, and spatial scales ranging from a few meters to the global scale.

Of all the various types of satellite-borne sensors that are available to us now, it is the visible spectral radiometric observations, commonly referred to as ocean-colour measurements, that have the proven capability to interrogate ocean biota directly. But many others are also relevant, particularly for placing the biota in their environmental context.

In this paper, we examine the principles that underlie the use of ocean-colour sensors, and the ancillary information from other satellite sensors that can be used to enhance the applications of ocean-colour data. We recognise that satellites cannot satisfy all our observational requirements relating to the biological system; and so, we examine the value and importance of what is amenable to remote sensing, how to make the most use of the observations, sometimes in conjunction with other types of observations. We examine the challenges and identify the need and advantages of combining satellite-derived information with other modes of observation, to enable a holistic view of ocean biota and ocean ecosystems.

In this paper, we have attempted to present the myriad of ways in which ocean-colour data have contributed to the study of ocean biota and marine ecology and biodiversity. To present the case, we have relied heavily on our own work, and the paper is not designed to be an exhaustive review of the field.

2 Principles of Ocean Biology from Ocean Colour

It is a matter of everyday observation that the colour of natural bodies of water changes with location and with time (Fig. 1). We know instinctively that the colour is an indicator of water quality: we associate limpid, blue, transparent waters with purity; brown and reddish waters that are more opaque are linked with high content of mud; and we know that green waters likely indicate high plant content. But to progress from this type of casual observation to quantitative assessment of life in the oceans, we need to invoke the principles of marine optics.

Fig. 1
figure 1

Examples of colour of natural aquatic bodies at different locations and times (images from CEOS Report 2018, courtesy CSIRO). In this paper, we examine the optical processes responsible for these changes; whether we can quantify the responsible agents; what biological information we can garner from colour; and the applications of the information

Marine optics is built on a bedrock of two subdisciplines (https://www.oceanopticsbook.info/). One of them is devoted to the study of inherent optical properties of various types of material, whether particulate or dissolved, that are present in seawater. Inherent optical properties include spectrally-resolved absorption and scattering coefficients, and the volume scattering function that describes the angular distribution of scattered light. They represent the optical properties of the materials independent of the nature of the light field in which they are present. The second branch of marine optics deals with the study of radiative transfer in the oceans, and how the spectral and angular distribution of under-water light field (induced by sunlight or by artificial sources) is modified by the types and concentrations of material present in the water.

Ocean colour, as detected by a satellite, is defined by the spectral variations in remote-sensing reflectance, defined as the spectrally-resolved ratio of water-leaving radiance, normalised to the downward irradiance at the surface of the ocean:

$${R}_{rs}(\lambda ,\theta ,\phi ,0)=\frac{L\left(\lambda ,\theta ,\phi ,0\right)}{{E}_{d}(\lambda ,0)}$$
(1)

In the equation above, \({R}_{rs}(\lambda ,\theta ,\phi ,0)\) is the remote-sensing reflectance at wavelength \(\lambda\), at depth zero (just above the surface), zenith angle \(\theta\) and azimuth angle \(\phi ,\) with \(L\left(\lambda ,\theta ,\phi ,0\right)\) being the corresponding water-leaving radiance at the same wavelength, and \({E}_{d}(\lambda ,0)\) being the downwelling irradiance, also at wavelength \(\lambda\) and depth zero. Note that radiance is defined as flux per unit surface area and per unit solid angle, whereas irradiance is flux per unit area. It is typical when processing remote-sensing data to normalise \({R}_{rs}(\lambda )\) such that we deal with the special case of zenith angle of zero, to avoid any deviations in the signal arising from changes in the viewing angle. Intuitively, we can anticipate that reflectance will decrease as absorbing material increases in the water, and that it will increase with scattering, especially with back-scattering (or upward-scattering). In fact, various theoretical explorations have led to general solutions of the form (e.g., Morel and Prieur 1977; Sathyendranath and Platt 1997; 1998; Sathyendranath et al. 2001; Brewin et al. 2012):

$${R}_{rs}\left(\lambda \right)=f\frac{{b}_{b}\left(\lambda \right)}{a\left(\lambda \right)},$$
(2)

where \(f\) is a function of \({b}_{b}\left(\lambda \right),\) the back-scattering coefficient and \(a\left(\lambda \right),\) the absorption coefficient. Note that here, and in the following, we have dropped the arguments \(\theta ,\phi\) and zero, for simplicity.

The optical properties of pure water itself (with low absorption in the blue and green parts of the visible spectrum, and high scattering in the blue) render pure, clean water its characteristic deep blue colour. Typically, four major groups of material are responsible for changes in ocean colour. These are:

  • Phytoplankton that contain plant pigments–notably chlorophyll-a and its variants, and auxiliary pigments that have photosynthetic or photoprotective roles. Chlorophyll-a has an absorption maximum in the blue, a minimum in the green and a secondary maximum in the red, and a relatively flat backscattering spectrum that could be modulated in the vicinity of absorption bands.

  • Coloured dissolved organic matter (CDOM) or yellow substance, which is characterised by an absorption coefficient with low values in the red part of the spectrum, increasing exponentially towards shorter wavelengths. Because these are dissolved substances, their contribution to scattering is considered negligible.

  • Detrital material, which have absorption properties similar to those of CDOM; and

  • Suspended sediments, whose optical properties vary, depending on the type of sediments present: whether they be mud or sand, for example. Suspended sediments become important in certain coastal areas and in areas of river outflow, but are largely absent in the surface waters of the open ocean.

The optical properties of phytoplankton are such that the colour of water changes from blue to green with increasing concentrations of phytoplankton. CDOM and detrital material tend to impart colours to the water that range from a deep yellow to orange (often described as the colour of black tea), and suspended sediments yield muddy brown, blackish or turquoise coloured waters, depending on the type of sediment present. These transformations of the colour of water with the nature of substances present in it, is shown schematically in Fig. 2.

Fig. 2
figure 2

The optical properties of the principal constituents of seawater, and how they vary with wavelength in the visible portion of the electromagnetic spectrum. Photos of phytoplankton, non-algal and dissolved substances were adapted from Wernand et al. (2011). Photo of pure seawater taken in the centre of the oligotrophic North atlantic subtropical gyre, during an atlantic meridional transect

To extract quantitative information on these substances from ocean-colour data acquired by satellites, we express the absorption coefficient as the sum of absorptions by individual substances, as in:

$$a\left( \lambda \right) = a_{W} \left( \lambda \right) + Ba_{B}^{*} \left( \lambda \right) + Ya_{Y}^{*} \left( \lambda \right) + Da_{D}^{*} \left( \lambda \right) + Sa_{s}^{*} \left( \lambda \right) + \ldots$$
(3)

where the notations \(W\), \(B\), \(Y\), \(D\) and \(S\) stand respectively for water, phytoplankton biomass, yellow substances (CDOM), detritus and suspended material. Furthermore, for the latter four categories, the absorption is expressed as the product of their biomass-specific absorptions indicated by asterisks, and their concentrations \(B\), \(Y\), \(D\) and \(S\). We recognise that, when other substances are present, it may be essential to include additional terms (as indicated by the trailing dots in the equation). The phytoplankton biomass is usually taken to be the concentration of the main phytoplankton pigment, chlorophyll-a.

Similarly, the back-scattering coefficient \({b}_{b}(\lambda )\) can also be expressed as a sum of its components:

$$b_{b} \left( \lambda \right) = b_{{bW}} \left( \lambda \right) + Bb_{{bB}}^{*} \left( \lambda \right) + Db_{{bD}}^{*} \left( \lambda \right) + Sb_{{bs}}^{*} (\lambda ) + \ldots$$
(4)

where we have assumed there is no additional scattering from CDOM.

Once the theoretical framework is laid out, algorithm development in ocean-colour remote sensing becomes an inverse modelling problem of retrieving the concentrations of the constituents, given the remote-sensing reflectance \({R}_{rs}(\lambda )\). Alternatively, we can develop empirical algorithms relating \({R}_{rs}\left(\lambda \right)\) to the concentrations of the constituents and ensure that the empirical approach is consistent with the theory.

However, the remote-sensing reflectance at the sea surface is not what the satellites observe: they observe the radiance at the top of the atmosphere, which includes contributions to the signal from the atmosphere. The first step in the ocean-colour processing chain is therefore a procedure referred to as the atmospheric correction, designed to remove the atmospheric impact on the signal and to extract the water-leaving radiance. This is a critical step, since the atmospheric contribution to the signal at the top of the atmosphere is typically 80% or more of the overall signal (Fig. 3).

Fig. 3
figure 3

‘True Colour’ images of a region in the South Pacfic Gyre (top left) and the Baltic Sea (top right) derived from the VIIRS sensor aboard the Suomi NPP satellite. Lower panels show pixel extractions of radiances at pin locations. Lt, Lr, Lw and nLw are the top of atmosphere, Rayleigh, water leaving and normalised water leaving radiance respectively. Atmospheric correction was performed using the NASA l2gen processor (within SeaDAS version 7.5.3)

There are many sources of uncertainties in the retrieved remote-sensing products. These include uncertainties in the instrument calibration, errors introduced during the atmospheric correction procedure, uncertainties in pixel identification, and those in the algorithms used for retrieval of the in-water constituents. The overall uncertainties are established using formal error propagation analyses or through comparison of satellite products against field observations that are matched in time and location (Brewin et al. 2015a; Müller et al. 2015a; Jackson et al. 2017a; Sathyendranath et al. 2019; McKinna et al. 2019).

From this background, it becomes evident that the biological variable that is most readily retrieved from ocean-colour data is the concentration of the phytoplankton pigment, chlorophyll-a (see example in Fig. 4).

Fig. 4
figure 4

Example of an ocean-colour derived map of the chlorophyll concentration, from the ocean colour climate change initiative processing chain (version 5, dates 2018-08-29–2018-09-05)

3 Chlorophyll-a as an Index of Phytoplankton Biomass

Phytoplankton are complex organisms, and no single measure of its biomass will constitute a complete description of the organism. So, when choosing an index of biomass for routine estimations of phytoplankton biomass, it would be worthwhile to consider the ideal characteristics of such an index and examine to what extent the selected index meets those requirements. Holm Hansen (1973) listed the desired properties of indicators of biomass of a community of living organisms and concluded that ATP (adenosine triphosphate) would be a suitable index. We could use a similar set of criteria for an ideal index of phytoplankton biomass rather than for the entire community (Table 1) and examine the extent to which chlorophyll-a meets those requirements, and carry out a similar evaluation for phytoplankton carbon, which is also often considered a desirable index of phytoplankton biomass. The table shows that chlorophyll-a meets most of the requirements of an ideal index of biomass for routine measurements, more so than carbon.

Table 1 Requirements of an ideal indicator for routine monitoring of phytoplankton, and the extent to which two candidate indicators, chlorophyll-a and carbon, meet those requirements

Chlorophyll-a is also an important biological molecule because of the central role it plays in photosynthesis, on land and in the oceans. In photosynthesis, light energy is absorbed by a photosynthesising organism, which is then used to convert carbon dioxide and water into organic matter, and oxygen is released. In phytoplankton, as in all photosynthetic organisms, chlorophyll-a is at the heart of the photosynthetic unit and serves as the transducer that connects the supply of energy from the sun to the ecosystems of our planet. The photosynthesis equation can be written as follows:

(5)

Primary production is the rate of fixation of carbon dioxide in the presence of sunlight into organic matter, through photosynthesis. It is a rate, with dimensions [M(C)] [L]−3[T]−1 (Mass of Carbon per unit volume and unit time), or [M(C)][M]−2[T]−1 (Mass of Carbon per unit surface area, and per unit time). Any map of chlorophyll concentration (e.g., Fig. 4) shows how the strength of this energy–ecosystem coupling varies in the ocean: the higher the chlorophyll concentration, the stronger the coupling, and higher the potential for primary production. It is a most fundamental property of our ecosystem. Thus, in calculations of primary production, it is very useful to know the amount of chlorophyll-a that is available for the process. It is also possible to base primary production calculations using phytoplankton carbon as the state variable, rather than chlorophyll-a. But a comparison of the two indicators (Table 2) shows that chlorophyll-a has some advantages in this context, over carbon.

Table 2 Comparison of chlorophyll-a and carbon as the state variable in computations of primary production

All things considered, chlorophyll-a is an extremely useful variable for studies of marine ecosystems. It would be easy to make the argument that chlorophyll-a should be one of the top-most priorities for observations of the marine ecosystem, and it is fortunate that this property is accessible to remote sensing.

4 Ecological Indicators from Remote Sensing

Phytoplankton chlorophyll concentration is not the only piece of information on ocean ecology and ecosystems that is accessible to remote sensing. Over the years, a formidable range of ecological indicators have been developed from ocean-colour data (Table 3). Chlorophyll concentration underlies many of these products, though not necessarily all of them. For example, combining chlorophyll data with light available at the sea surface and photosynthesis models, it has been possible to compute marine primary production.

Table 3 Ecological indicators from remote sensing, adapted and updated from Platt and Sathyendranath (2008). Some key (and non-exhaustive) references are also given for further details

The spectrally-resolved radiance data have been used to develop algorithms for detection of phytoplankton functional types and phytoplankton size structure (IOCCG 2014). Information on other environmental variables have been used to improve ocean-colour algorithms; for example, light and temperature have been used to refine models for detection of phytoplankton size classes from space (Brewin et al. 2015b, 2019). Recent years have seen considerable progress in our efforts to detect various biological pools of carbon in the ocean (Brewin et al. 2021). Availability of methods to detect phytoplankton carbon from space implies that it is no longer essential to choose between carbon or chlorophyll as the ideal index of phytoplankton biomass: we can choose one or the other according to the problem at hand or personal preferences.

The utility of ocean colour for mapping large-scale structures in the marine ecosystem is closely linked to our efforts to map marine diversity from space. Large-scale structures in the marine ecosystem include variance fields in chlorophyll concentration and primary production, as well as the organisation of the marine ecosystems into ecological provinces or optical classes. The ecological provinces have formed the foundation of recent efforts to map biotic biogeography, species distribution and diversity of the pelagic ocean (Reygondeau and Dunn 2019). Though most organisms other than phytoplankton are not directly observable using ocean colour, exceptions exist: for example, Basedow et al. (2019) have used ocean-colour data to map swarms of a particular species (Calanus finmarchicus) of copepods (a type of zooplankton). Indirect methods (e.g., exploiting the size structure in the marine biota) have also been employed to map zooplankton populations in general (e.g., Strömberg et al. 2009).

Life under water is very much at the mercy of the underwater environment, with light and nutrients playing important parts in sustaining it. Ocean colour also enables the study of the underwater light environment, with algorithms to map various properties of the underwater light field, such as the diffuse attenuation coefficient (which determines the rate of penetration of solar light under water), the euphotic depth (at which light is reduced to 1% of its surface flux). turbidity, water colour and water clarity. Ocean colour data are also used to map concentrations of the suspended sediment load and coloured dissolved organic matter (IOCCG 2000; Sathyendranath et al. 2020b).

5 Ancillary Information from Satellites Needed to Probe Ocean Ecosystems

Whereas ocean-colour satellite sensors provide our only window into marine ecosystems at synoptic scales, we also rely on other satellite-derived physical variables as a complement to ocean-colour data. These include information on the optical thickness of aerosols in the atmosphere (essential for the atmospheric correction procedures), sea surface temperature, sea surface salinity and sea state (all of which influence phytoplankton community composition and dynamics).

6 Applications

Chlorophyll concentration and the suite of ecological indicators listed above have been revolutionary in biological oceanographic studies, providing global coverage, with unprecedented detail at fine spatial and temporal resolution. Satellite-based observations are the only means available to us today to sample an entire area defined by the pixel size of the sensor, and that too, using a non-invasive method. With all other observational tools, the sampled volume or area is always a woefully small fraction of the area we wish to investigate, and much effort has to be invested into methods to fill the gaps in between spot observations. This high coverage from satellites bestows a distinct advantage for studies aimed at obtaining the large-scale view.

Studies of marine primary production have benefited hugely from ocean-colour observations. With phytoplankton biomass (indexed either as chlorophyll-a or as carbon) derived from satellites, photosynthesis models have been developed that incorporate photosynthetically active radiation (PAR), also from satellites, and photosynthetic models, to compute marine primary production at large scales. Phytoplankton are at the base of the marine food web, and understanding marine primary production is key to all studies of the maintenance of the various trophic levels of the ecosystem, including zooplankton and fish larvae at the second trophic levels, all the way to fish and shellfish at higher trophic levels that we harvest as high-protein, high-quality food from the sea.

It has been argued that fisheries management plans should incorporate considerations of bottom-up control on fish recruitment, in addition to considerations of top-down control through fishing, to enable long-term, sustainable, ecosystem-based management of fisheries. In this context, it has been demonstrated that metrics of phytoplankton phenology derived from satellites can shed light on survival rate of important commercial fish (Platt et al. 2003; Koeller et al. 2009; Kassi et al. 2018a, b; Menon et al. 2019).

In general, there is a link between ecosystem health and human health. Especially for the millions of people who make their home in coastal locations, the sea is an important source of food and of a sense of well-being. Degraded ecosystems threaten this food supply as well as the sense of well-being. Industries such as tourism and transport are affected. Water quality also implies the microbial quality of water. Bacteria and viruses are natural components of marine ecosystems; but some of these are pathogenic in nature and are associated with water-borne diseases such as cholera. The Vibrio cholera bacteria, responsible for cholera, are found in free-floating form in water, and also in association with phytoplankton and zooplankton, and other hosts. This link between the bacteria and plankton has been exploited in many studies to develop predictive models that rely on satellite-derived chlorophyll-a data as input (Sathyendranath et al. 2020a; Anas et al. 2021). Thus, satellite observations are being used increasingly, to elucidate the link between ecosystem and human health.

Issues related to climate are all-encompassing and all consuming, and the many applications listed above come to the fore as an ensemble, in studies of the impact of climate change on marine ecosystems and on the feedbacks between various elements of our climate system. The Global Climate Observing System (GCOS) has in fact recognised ocean colour and phytoplankton as essential climate variables, and there is currently a concerted effort to produce satellite-based climate-quality data records of spectrally-resolved remote-sensing reflectances and chlorophyll concentrations, for climate studies. One of such efforts is led by the European Space Agency within the umbrella of its Climate Change Initiative.

7 The Challenges to Producing Climate Quality Time Series from Ocean-Colour Data

Preparing climate-quality data records brings stringent requirements over and above those that are required for most other applications. These additional requirements arise from the need to detect not just the state of the marine ecosystem or short-term changes and variability, but also long-term trends in the ecosystem status and processes. Since any putative climate-change related signal could be a small one embedded within a noisy environment that experiences high natural variability at multiple time scales, it becomes particularly important to ensure that the data are of the highest precision possible. Since the standard error in the mean decreases with increasing number of observations, high coverage and high number of repetitive measurements become important. Because we have to discount natural variability at all shorter scales to extract the climate signal, it also becomes of importance to avoid gaps in data. Another key requirement for climate data streams is that for a long time series, in which the data are acquired and processed in a consistent manner. For example, one has to avoid artefacts introduced into the time series because of systematic biases between sensors. Since satellite sensors have a finite life span, one could introduce spurious trends in the data when multiple sensors are merged to extend the time series, if inter-sensor bias is not corrected for. One also has to consider whether the in-water algorithms employed are appropriate for detection of changes in the marine ecosystem, whether they be in the biomass of phytoplankton present, as indexed by chlorophyll concentration, or in the community composition or in the phytoplankton phenology (Sathyendranath et al. 2017, 2019).

These considerations have led, within the ocean colour component of the Climate Change Initiative (CCI), to put in place dedicated, specialised processing chains (Fig. 5) for generation of multi-sensor, consistent, error-characterised, inter-sensor bias-corrected time series of remote-sensing reflectance and chlorophyll concentration (Sathyendranath et al. 2019). The philosophy that underpins the approach is explored in Sathyendranath et al. (2017). Some samples of ocean colour CCI chlorophyll products are shown in Fig. 6.

Fig. 5
figure 5

Processing chain used in the ocean colour climate change initiative to generate long-term time series of climate-quality ocean-colour products

Fig. 6
figure 6

Some examples of chlorophyll products from ocean colour climate change initiative (OC-CCI) showing daily, weekly, monthly and annual products, from an early version of OC-CCI. For further details about OC-CCI products, see Sathyendranath et al. 2019

The Ocean Colour Climate Change Initiative chlorophyll products and marine primary production derived from these products (Kulk et al. 2020a, b, 2021) can be used to map the 23-year climatology and linear trends (Fig. 7a, b).

Fig. 7
figure 7figure 7

Climatology and linear trends in a) chlorophyll-a and b) primary production for 1998–2020, derived from the ocean-colour climate change initiative data, version 4.2. The linear trends in chlorophyll and primary production (as a rate of change over time) are provided as maps, with non-significant trends in grey, and as histograms after area weighing, with all trends in light blue and significant trends (p < 0.05) in darker blue. Figures are adapted and updated (extended by 2 more years) from Kulk et al. (2020a, b, 2021), also see the IPCC AR6 working group I report, chapter 2

8 Synergy with Ecosystem Models

There has always been a close association between ecosystem models and ocean-colour data. Over the years, chlorophyll concentration derived from ocean-colour data has served to test outputs from ecosystem models. Conversely, models have helped fill the spatial and temporal gaps of the ocean-colour observations, as well as explain the mechanism that underlie their patterns. With the advent of data assimilation in ecosystem models, ocean-colour products became the primary source of biological and bio-optical data that are assimilated into those models. The incorporation of phytoplankton functional types into ecosystem models has paralleled development of algorithms for detection of phytoplankton types and size classes from space (IOCCG 2014). State-of-the-art models are now capable of assimilating ocean colour phytoplankton types for both ecosystems functioning studies and operational biogeochemistry predictions (Ciavatta et al. 2018; Skákala et al. 2018).

In the climate context, in the absence of long time series of ocean-colour products, one relied heavily on models for understanding threats to the marine ecosystems from climate change, and to predict future changes. But as the time series of ocean-colour data grows–for example, the ocean-colour time series from the Climate Change Initiative is now 24 years long–the satellite data are being used increasingly to test model outputs and to examine whether model structures are appropriate for detection of climate impacts on the ocean biota.

Until recently, quantitative comparisons of model outputs and satellite products have been hampered by the differences in the currencies used in the two approaches: ecosystem models use biogeochemically-important variables, such as carbon or nitrogen, as their main currency. On the other hand, the principal output from satellites is the chlorophyll concentration. This required a conversion between carbon and chlorophyll to enable comparisons between model outputs and satellite products. However, the recent developments in satellite oceanography, which have enabled computation of phytoplankton carbon from space using photo-acclimation models (Sathyendranath et al. 2020b and references therein) have facilitated model-data comparisons at a more fundamental level. More recent ecosystem models include chlorophyll as a state variable that depends on light availability, with the assimilation of ocean-colour chlorophyll leading to an impact on the simulation of the physiological status of phytoplankton (Ciavatta et al. 2011). One of the challenges of ecosystem modelling is in the assignment of suitable values for the many physiological and ecological rate parameters that are used in the models. Ocean-colour data assimilation into ecosystem models can also serve to improve our understanding of model parameters (Roy et al. 2012). Sathyendranath et al. (2020b) have argued that the focus of future research should be on understanding variability in model parameters that define rates of physiological processes, even more than on variability in the biological pools themselves.

Some of the recent ecosystem models include bio-optical modules, which expand the use of ocean colour to assimilate diffuse light attenuation products (Ciavatta et al. 2014), remote sensing reflectance (Jones et al. 2016), and light absorption of phytoplankton functional types (Skákala et al. 2020), in combination with water column observations from gliders in shelf-seas (Skákala et al. 2021) and biogeochemical-Argo floats in the ocean (Teruzzi et al., 2021).

A recent IOCCG report (IOCCG 2020) is devoted to the synergy between ocean colour and ecosystem models, which constitutes an excellent reference material on this topic.

9 The Need to Integrate Ocean Colour with Other Essential Variables

The strengths of satellite observations lie in their capability to provide synoptic views of the marine ecosystem; to provide global coverage using a single, or a handful of, sensors, ensuring consistency across all observations; in their capability to sample the entire area within each pixel of a satellite image; and to do this in a non-invasive manner, repeatedly and over long time-scales.

These strengths have to be balanced against their limitations: satellites only observe a finite layer of the surface ocean; ocean-colour sensors are unable to penetrate clouds; the need for stringent atmospheric correction further limits coverage under other adverse conditions (clouds, cloud shadow, high aerosol concentrations, low solar elevation, to name a few); the interpretation of the signal is complex, and may be associated with varying optical properties of phytoplankton (with changes in community structure and pigment composition) or with the presence of other substances in the water, obscuring the phytoplankton signal; and above all, by the limited number of ecosystem properties that are amenable to satellite observation. The list of observable properties can be extended by the use of proxies that include chlorophyll concentration and relevant environmental properties also amenable to remote sensing, such as photosynthetically active radiation, sea surface temperature, sea state and salinity. Yet, it would be hard to argue that satellites tell us everything we need to know about biology and ecosystems of the oceans.

Typically, one relies on meticulous process studies, in the laboratory or at sea, for new insights into the structure and function of marine ecosystems. Other modes of observations are essential to extrapolate surface observations from satellites through the water column, and to cover those properties and processes that are not observable through satellites. Furthermore, we rely on models of various sorts–conceptual, theoretical, and simulation models–to interpret and synthesise observations and understanding, and for predictions. Thus, it is only through an integrated approach that we can delve deeper into the mysteries of the ocean ecosystems.

Our understanding that biological systems on land and in the oceans are integral parts of our climate system is a relatively new one and was perhaps first revealed through paleontological studies that highlighted the links between biological processes, notably primary production, and changes in our planetary climate over geological time scales. But subsequent explorations of climate using Earth system approaches have led to new revelations on the inter-connectedness between various elements of the climate system. This leads to one of the grand challenges of the day: to explore the links between ocean biota and biogeochemical processes in relation to various components of the climate system, not only in the oceans, but also on land, atmosphere and cryosphere.

10 Concluding Remarks

Ocean colour from space has become the mainstay of biological oceanographic and marine biological research, making significant contributions to various subdisciplines of the fields. The achievements in ocean-colour research in the last few decades have been remarkable. However, it is by no means a closed chapter now: one anticipates further improvements in products and development of new products in the near future, keeping pace with technological developments, notably with the advent of hyperspectral optical sensors.