Colour
The secret language of the Oceans

Let's start by laying out some fundamentals. The light that emanates from the sun and reaches our eyes is composed of multiple types of electromagnetic waves within the visible spectrum. Each type of wave is characterised by a different colour and wavelength, as well as a specific behaviour when it interacts with a surface. The variation in how different waves are absorbed by a surface creates an object's spectral signature.

When a light wave encounters a water surface, it's important to note that water has a relatively low reflectance across all wavelengths. In shallow depths, it absorbs long waves, which correspond to warm colours (for instance, red disappears before reaching a depth of 5 metres). In contrast, shorter waves, associated with cooler colours, penetrate deeper before being absorbed. As the wave ‘sinks,’ its amplitude diminishes, which in turn affects how we perceive colours underwater.

When the water is clear, its reflectance¹ is very low and blue is the last wavelength to be absorbed, which is why the sea appears blue. When the water is turbid, this indicates the presence of sediments. These sediments have their own unique spectral signature, meaning that reflectance increases depending on their composition, altering the spectral signature of the environment. As a result, the spectral signature of water combines both the intrinsic properties of the light wave and the characteristics of dissolved and suspended elements within the water column.

When a satellite zooms in on a marine area, the image will include the living micro-organisms that evolve on the surface, including phytoplankton. When surface waters contain sediment and a high concentration of phytoplankton, reflectance increases significantly in certain wavelengths and our perception of the colour of the water changes. The colour of the water will tend towards green, red or yellow, depending on the specific species of phytoplankton in that particular environment² and it is this variability in colour that the satellite sensor records and that is then analysed by scientists.

When you observe the sea via the proxy of a high-precision instrument at an altitude of several hundred kilometres, it is quite counterintuitive to realise that it is possible to discern living organisms measuring 2 μm. Antoine Mangin is still amazed: “You have to realise that the satellite is 800 km above the Earth, that the signal is 90% polluted when it passes through the atmosphere and yet we are able to distinguish between micro, nano and pico-phytoplankton solely on the basis of the variations in surface colour that the sensor picks up”. It's certainly a technological feat, but out of all proportion to our usual human scales. This almost naturally begs the question of why we are going to the trouble of committing so many resources to observing our Blue Planet from space instead of simply taking samples on site. For Antoine Mangin, the question is irrelevant: “The satellite offers an unrivalled synoptic view. If we wanted to do it with on-site measurements, we'd need tens of thousands every day, but today, the global archive of HPLC samples at sea comprises around ten thousand points, which is already fantastic.³”

We let ourselves be guided through the twists and turns of the multispectral.

Interpreting the spectral signature of the environment to anticipate environmental risk

It has long been documented that the increase in greenhouse gases in the atmosphere is having an upward impact on the quantity of phytoplankton in the seas. This means more cyanobacteria (a component of phytoplankton) and, by domino effect, a greater risk of toxic algal blooms. The episodes of toxic red and green tides visible from space in the media bear witness to this, and it's a real issue for fishing and aquaculture. It is interesting to note that some species of toxic algae are coloured, while others are colourless. When a bloom phenomenon is underway, to characterise the type(s) of algae involved, scientists generally look for its spectral signature. Each type of plant has its own spectral signature, with variations depending on its degree of growth and the environmental conditions in which it is found. Phytoplankton is no exception, and satellite images are analysed using existing spectral libraries.

A few years ago, Antoine Mangin worked on the problem of algal blooms. He explains the process: “There was a major algal bloom episode on the island of Chiloé in Chile, a site of large-scale aquaculture production, comparable to Norway in terms of volume. Diseases were reported in the farms and the presence of toxic algal blooms was confirmed. The species of algae that was harmful to the farms was unique in that it had no spectral signature in the visible spectrum. To monitor its evolution in the environment using satellite data, an index of the probability of its appearance was established, given that no visible trace could be observed. The index was constructed using other satellite data, combining chlorophyll maps that showed the presence of phytoplankton and temperature maps that allowed us to estimate the conditions under which eutrophication was occurring. The aim was to document the conditions that had led to the imbalance in the environment and that had favoured the appearance of toxic algae, so as to be able to anticipate them more effectively. The aim was to trace the dynamics of the spread of the bloom. When the index is between 0.4 and 1, we can be fairly sure that a bloom will occur. So we can be better prepared. »

Antoine gives another example of anticipating environmental risk. “In New Caledonia, we are observing the dynamics at sea of the plume of a river in the catchment area of nickel mines. This is a potentially major source of pollution and we are monitoring this plume over the long term. This is part of the monitoring role of marine protected areas. We are observing to anticipate any possible irreversible impact on coral reefs”.

Better management of the blue economy

ACRI-ST was recently approached by the European Union to work on a forward-looking food sovereignty project. In very specific terms, the company has been commissioned to produce a map of areas where it would be worth setting up aquaculture farms to grow algae and molluscs. The stated objective is ambitious in terms of volume, with a market of 10 million tonnes a year. The study was led by Philippe Bryère's team at ACRI-ST, with support from Marine Bretagnon, Aurélien Prat and Quentin Jutard. Antoine Mangin sums up the approach: “Using satellite data and algorithmic modelling, we produced a map showing the position of fish farms on European coasts. The idea was to pinpoint the areas where the conditions for this type of aquaculture would be ideal.”

While ACRI did this for algae and molluscs, ordered by the European Union, they also did it for other crops. In Algeria, for example, for bass. In this case, mechanical parameters are also taken into account in the modelling parameters. Swell height, for example. If the swell is more than 1m50, the cages are often destroyed. Current speed too. The wolf gets bored in a sea of oil... And as indicated above, the full range of possibilities offered by multispectral imaging was used to cross-reference these mechanical parameters with the colour ranges of the chlorophyll and the temperature of the environment. As Antoine explains: “Little by little, we refined the forbidden zones by adding layers of information. Until we have all the possible locations. We can also go further and add a growth model. This is what was done in Spain and Portugal in collaboration with local mussel farmers. We worked on the basis of multi-year chlorophyll and temperature maps, since we know that mussels can only be cultivated within a certain temperature range, and then we mixed this data with the wave heights at possible coastal implantation sites and with other parameters. One thing leading to another, we came to our conclusions and today I can clearly tell you which areas are the most suitable for mussel cultivation in the south of Spain and Portugal”. All that remains now is to find interested fish farmers.

In the fisheries sector, ACRI-ST has worked directly with Morocco's National Institute of Fishery Resources to try to understand a cyclical phenomenon of cold water upwelling that had a major impact on fishery resources at a certain period. By simply analysing colour variations from surface chlorophyll maps and temperature maps over several years and seasons, ACRI was able to detect certain anomalies in specific areas of the Moroccan coastal ecosystem. This data was passed on to the Moroccan Ministry responsible for fisheries, and very specific decisions were taken, particularly in terms of quotas and fishing zones for sardines, to help replenish fish stocks. This very specific use case is typical of the possible use of geoinformation derived from satellite remote sensing techniques. It all starts with a simple spectrum of colours.

The more we know, the less we say

The strength of the models developed by ACRI-ST is that they combine the cutting-edge technology of multispectral imaging with the analysis of physical and mechanical parameters based on long time series. What remains to be done is to estimate the degree of certainty of these models, or of uncertainty, depending on the outlook... And that is certainly the biggest unknown. We would all become Spanish mussel fishermen if it were so easy to succeed in mussel farming.

Antoine Mangin points out the bias inherent in long time series: “Instruments evolve over time. The thermometer that measured the temperature in Paris in 1850 is not the same as the one that measures the temperature in the same place today. The degree of accuracy has improved, but at the same time, this makes comparison more difficult, since the instruments change over time and alter the basis for comparison”. In the satellite sector, optical measurement instruments are evolving in the same way, and this inevitably has an impact on the way in which long time series are analysed. As Antoine sums up: “It's vital to be able to establish a precise degree of uncertainty. That's what determines the quality of the model. You have to be able to say that a given parameter is 15% inaccurate. Or that such and such a parameter is within 2%. It is only in this way that, little by little, we can begin to study the climate properly. Yet another illustration of how little we actually know...

At least now we have the courage to say so and admit it.”

1. Reflectance is the ratio of sunlight entering the ocean to the light backscattered from it.

2. More than 160,000 different species of phytoplankton have already been characterised. The satellite enables us to distinguish around ten classes (groupings), either in terms of size (pico, nano, micro, mentioned below), or in terms of functional groupings based on their role in the trophic chain.

3. HPLC - an acronym for high-performance liquid chromatography - is an analytical chemistry technique used to separate compounds in a chemical mixture. In very specific terms, this makes it possible to identify and quantify substances in a solution.