Within this book you will find summary information about the three main types of measurement that are made by the Sentinel 3 satellite (see the following section) and make up the Copernicus Marine Data Stream from EUMETSAT.
What is ocean colour?
“Ocean colour” typically refers to radiometric measurements of either water leaving radiance (Lw) or remote sensing reflectance (Rrs) from satellite mounted spectrometers. The core of these measurements are made across the visible light region of the electromagnetic spectrum (see figure below), with additional measurements made in the ultraviolet (UV) and Infrared (IR). Where measurements are made exactly within this spectrum depends on the sensor configuration. Spectrometers that measure ocean colour mostly make measurements at a set of discrete wavebands that are designed to target area of the spectrum which contain specific information useful either for data processing or derivation of geophysical ocean properties.
Figure 1. The electromagnetic spectrum (Credit: From http://www.solarlightaustralia.com.au/2013/02/20/visible-light)
Ocean colour is a form of passive remote sensing. Light from the sun passes through the atmosphere and is absorbed and scattered by particles, gases and aerosols present, it then enters the ocean and is absorbed and scattered by various constituents including pure water, phytoplankton and other dissolved and particulate matter. A proportion of the incoming light is eventually ‘reflected’ back towards the satellite. To get to the water leaving radiance (that which contains the most information about the in water constituents) from the signal eventually received at the satellite it is necessary to subtract the effects of the atmosphere.
For more information on the derivation of reflectance, radiance, irradiance etc, atmospheric correction and other facets of ocean colour data, see the reports of the International Ocean Colour Coordinating Group (IOCCG) listed below.
Why do phytoplankton and other constituents change the colour of the water?
Phytoplankton absorb and scatter light as a result of the pigments they contain and the structure of their cells. For example, high concentrations of phytoplankton cells usually result in high concentrations of the primary photosynthetic pigment – Chlorophyll a. As part of photosynthesis this pigment absorbs light (primarily in the blue region of the spectrum). Simply, this leaves a larger proportion of incoming light at green wavelengths unabsorbed and more likely to be reflected, resulting in water looking less blue and more green the higher the concentration. However, it is also common to see spectral features in radiometric measurements at red and near infrared wavelengths as a result phytoplankton fluorescence (around 685 nm) and spectral features in phytoplankton absorption and scattering. These features can provide information that is related to phytoplankton as few other constituents have features in this spectral region. Fluorescence related products are now offered with ocean colour data and there is much research being conducted to see if fluorescence features can be linked to phytoplankton physiology.
Other pigments, which may play various roles in phytoplankton physiology can show specific absorption characteristics. In addition to the role of pigments, other properties of phytoplankton, including their size, shape and intracellular structures have been shown to influence their absorbing and scattering characteristics. Much research is underway to examine the extent to which these influences can be exploited to gather information about phytoplankton communities from satellite ocean colour data. Morel and Bricaud (1981 and 1986) provide a good introduction to understanding how phytoplankton interact with the light environment from a remote sensing perspective. A recent report from the IOCCG (2014) also summarises the progress made towards satellite based detection of phytoplankton functional types.
Beyond phytoplankton there are other constituents in the oceans that can affect its colour. Non algal particles such as sediments, as well as coloured dissolved organic matter (CDOM), influence the ocean colour and have different absorption and scattering properties, different from phytoplankton, and variable dependent on the sources of the sediments and dissolved matter. Understanding this component of the ocean colour signal is important for two main reasons:
The concentrations of these components can be useful for many applications e.g. understanding coastal sediment dynamics, quantifying coastal water quality, understanding the sources and impact of coastal pollution and various marine activties.
If we want to know about phytoplankton concentrations and characteristics, it’s important to be able to separate the phytoplankton component of the ocean colour signal from that associated with non-algal particles and dissolved matter.
How do we use this information to develop algorithms to quantify chlorophyll a concentrations, other information about phytoplankton assemblages, and concentrations of non-algal ocean properties?
Algorithms to derive chlorophyll a concentrations have historically been based on ratios between blue and green measurements of reflectance (see McClain et al., 2009 for an overview of the history of ocean colour measurements, algorithms and applications). These approaches have shown substantial success in waters where phytoplankton are the dominant source of variance and other constituents (e.g. coloured dissolved organic matter) co-vary with their concentration (commonly referred to as ‘case one’ waters). Ratios of reflectance at different wavelengths have been used in combination with in situ measurements to derive empirical relationships between ocean colour radiometry and [Chl a]. In more optically complex waters, these empirical relationships may fail due to signal covariance with varying concentrations of other substances (e.g. dissolved matter/sediments - these are ). In addition to this, a desire to understand how different components influence radiometric signals has prompted the development of semi-analytical models and inversion algorithms. These methods seek to link the absorption and scattering characteristics of the various constituents of the ocean waters (phytoplankton, sediments etc) with radiative transfer theory and ultimately with Rrs. Various mathematical methods can be employed in these techniques, including non-linear optimisation and neural networks. Several of the IOCCG reports provide an overview of the methods related to semi-analytical algorithm approaches (see references below for links).
The Copernicus Marine Data stream from EUMETSAT provides products which quantify the aforementioned ocean constituents using a variety of different methods (see section on data formats and processing for further details). The methods selected to derive these products builds on the heritage of those developed for the MERIS sensor aboard the ENVISAT satellite.
How do we assess the quality of satellite ocean colour measurements?
The quality of satellite ocean colour measurements can be looked at in two main ways:
Firstly, the quality and confidence flags associated with the data can be evaluated. The flags provide information on a pixel by pixel basis about a large variety of things e.g. whether the pixel represents land/sea/coastline/cloud. In addition to these classification flags, other flags are used to indicate when there are problems with the data, which may influence its validity i.e. it is partially cloudy or there is glint. A final set of flags suggest when the methods used to derive products may not perform well e.g. when the values estimated may be outside of the range for which the method was designed/calibrated etc.
Secondly, we can try and create a match up validation data set, where in situ radiometric and biogeochemical measurements are made in closely coincident time and space with satellite overpasses for comparison.
How do we collect in situ radiometric data for comparison to satellite measurements?
Measurements of radiometric quantities can be made using instruments, either above water or in-water, mounted on platforms/buoys or (in the case of in-water) by profiling instruments. Radiance and irradiance are commonly measured, and converted to either water leaving radiance (Lw) or Remote Sensing Reflectance (Rrs) for comparison to satellite data. This process can be quite complicated as we need to ensure highest accuracy and minimise/quantify errors in these measurements for satellite validation (and calibration). For more information on radiometers and deployment methods, see Hooker and Maritorena (2000) and Mueller et al., (2003).
References and resources:
Hooker, S.B. and Maritorena, S. (2000) An Evaluation of Oceanographic Radiometers and Deployment Methodologies. Journal of Atmospheric and Oceanic Technlogy, 17, pp 811-830.
McClain, C (2009) A decade of satellite ocean color observation. Annual Review of Marine Science, 1, pp 19-42.
Morel, A. and Bricaud, A (1986) Inherent optical properties of algal cells including picoplankton: Theoretical and experimental results. Can Bull Fish Aquat Sci. pp 521-559.
Mueller, J.L, et al., (2003) Ocean Optics Protocols For Satellite Ocean Color Sensor Validation, Revision 4, Volume III: Radiometric Measurements and Data Analysis Protocols. NASA/TM-2003-21621/Rev-Vol III. Available online at: http://oceancolor.gsfc.nasa.gov/DOCS/
IOCCG reports covering many topics – available online here.