Sites Inria

Version française

Inria thesis


Oysters give advance warning of pollution

Hafiz Ahmed - © Inria

A native of Dhaka, the capital of Bangladesh, Hafiz Ahmed has been a student in France since his master’s degree. He is currently a PhD student in the Non-A project team (Non-Asymptotic estimation for online systems ) at the Inria Lille - Nord Europe Centre. This is a joint project team with Centrale Lille, the CNRS, and the Lille 1 University*). He is working on the development of models to study the behavior of oysters in marine ecosystems in the Arcachon Basin with the aim of identifying disturbances to their biological cycles that may be indicators of water-borne pollution.

Hafiz Ahmed, how far are you along the path to your PhD?

I am in my second year in the Inria Non-A project team in Lille. I also lecture as a PhD student assistant at the Lille 1 University. My research is concerned with modeling and estimating the circadian and circatidal rhythms of oysters under the supervision of Efimov and Rosane Ushirobira in the Non-A project team and Damien Tran in the CNRS in Bordeaux. One application is to make use of disturbances in these rhythms to detect ‘hidden’ pollution in the water, that is pollution that cannot be seen and which could pass unnoticed.

Studying organisms that live under water imposes a number of constraints. The Non-A project team works on the development of mathematical models for data that cannot be measured directly. I also work in collaboration with colleagues in the EA (Aquatic ecotoxicology) project team, and the CNRS EPOC (Oceanic and continental environments and paleoenvironments) joint research unit in Bordeaux. The EA project team, led by Jean-Charles Massabuau (CNRS), are studying aquatic ecosystems in the presence of contaminants, together with the biological rhythms of aquatic fauna. They have developed an innovative measurement system based on high frequency valvometry.

What is the main difficulty in your work?

There are currently many sites monitoring oysters and other bivalves on a round the clock basis, and this generates vast amounts of data. We work on groups of sixteen Pacific oysters, Crassostrea gigas , the most widely cultivated species in the world. One of these groups has been placed under the Eyrac jetty in the Arcachon Basin. This site has been in operation since 2006, and we collect 54 000 data points every day from each oyster. That makes a total of 864 000 data points every day, and around 15 gigabits of data per year. That’s an awful lot of data to process.

What types of data do you collect from the oysters?

We record the opening and closing of their shells. This movement is the main factor controlling their feeding and it is strongly influenced by the tides and the positions of the Sun and Moon. We measure the movements by placing a tiny coil on each half of the shell. On each oyster, a sinusoidal signal is sent from one coil to the other every 1.6 seconds. The amplitude of this signal is inversely proportional to the distance between the two coils, and this enables us to record all the movements of the shell.

Parc à huîtres - (CC) jacme31

These oysters lie beneath three to seven meters of water depending on the tide. Getting the data to the EPOS research base is not easy as the information is transmitted from a computer that is required to work when submerged in seawater for periods of several months. The data is sent via GPRS (General Packet Radio Service), a second generation mobile phone standard that was in use prior to the explosion in 3G and 4G. In addition to the rhythms of the bivalves, the system records the positions of the Sun and Moon, and a number of other environmental parameters.

What tools do you use to process the data?

We use two well-known regression models, NARX and ARMAX. These are widely used by stock exchanges to estimate changes in share prices. ARMAX, an autoregressive–moving-average model with exogenous inputs, can be used to predict future values of a series by using previous values and a number of external factors having a linear relationship.

NARX, a non-linear autoregressive exogenous model is used when the relationships are non-linear, i.e. in a form other than a straight line such as a parabola, hyperbola, or sinusoid. The exogenous data used in either case are the positions of the Sun and Moon, the state of the tide, and precipitation. This enables us to predict the biological cycles of the oysters as a function of their environment.

Our initial work was presented at the European Control conference in 2014. We showed that a NARX model could be used to estimate the normal physiological rhythms of submerged oysters. Our team has also had a paper accepted for the 2015 conference where we intend to announce an algorithm to detect the laying period of the oysters automatically.

Why study these biological cycles?

We know from laboratory experiments that exposure to pollutants causes the abnormal closure of the shells of oysters. We are seeking to establish whether this is also the normal behavior of these bivalves in the field, so that we can recognize accurately when they have been disturbed. We cannot be sure that all unusual behavior in mollusks is associated with pollution, but it could be a very good indicator and act as an alarm signal.

The presence or otherwise of known pollutants can only be confirmed by analysis of the water or the tissue of the mollusks. However, this is an expensive process, and it is not possible to carry out this type of analysis on a continuous basis. The system on which I am working, and which was installed by EPOC, is in operation around the clock at a much lower cost. It should indicate exactly when analysis is required and samples should be taken. Satellite monitoring systems are currently in operation, but these are far more costly and only take account of the water at the surface rather than the organisms, like oysters, that actually occupy the environment.

Our method has the advantage of being suitable for use throughout the world on a continuous basis at low cost. We are currently working on the development of a commercial prototype, but I cannot say yet when it will be ready. It should be of interest to everyone monitoring pollution levels, especially the oyster growers themselves.

* as part of UMR 9189, CNRS-Centrale Lille-Lille1 University, CRIStAL.


Autoregressive model: A model in which future values in a series are derived from previous values rather than from external variables.

Circadian rhythm: A 24 hour rhythm in an organic function, generated by a molecular clock and synchronized to the changes between day and night.

Circatidal rhythm: A 12.4 hour rhythm in an organic function, synchronized to the rhythm of the tides.

High frequency valvometry: A method of measuring the activity of mollusks by means of a coil placed on each half of the shell. (For more details, see Molluscan-eye).

Keywords: Big data Biology Pollution detection Ecology Data processing