Aquaculture 2022

February 28 - March 4, 2022

San Diego, California

RISK ANTICIPATION IN AQUACULTURE: FOUR DATA SOURCES TO SUPPORT OYSTER FARMING RISKS

Frederic Leroudier1*, Lucas Schaeffer1, Arthur Tré-Hardy1, Nicolas Prost1, Romain Pete2, Sébastien Mas3, Florence Bouillé-Chada4, Maxime Paris1

 

1 BIOCEANOR SAS, 1360 route des dolines, les Cardoulines B3, 06560, Valbonne, France

frederic.leroudier@bioceanor.com

2 SMBT, 328 Quai des Moulins, 34200 Sète, France

3 OSU-OREME, 2 Rue des Chantiers, 34200 Sète, France

4 CLS, 11, rue Hermès, Parc Technologique du Canal 31520 Ramonville Saint-Agne, France

 



Introduction

Since 2008, the worldwide shellfish and aquaculture industry has experienced major health crises, due to natural phenomena such as microbial infections or anoxic crises caused by oxygen depletion or Harmful Algal Bloom (HAB).  As one of the main oyster farming area in France, the Thau lagoon is no stranger to those risks. Beyond just shellfish farming, lagoon environments such as Thau are real biodiversity havens that are directly impacted by these harmful events. The SENSITHAU project aims to create a Lagoon Observation Network (ROL), implementing an ecosystem database for integrated ecological, health and production management within the lagoon. Bringing together key local and national players this network will allow enhanced monitoring thanks to a multitude of sensors capable of real time and high frequency measurements across a wide range of physicochemical parameters. The SENSITHAU project focuses on (1) deploying an in situ telemetry system to monitor lagoon water conditions, and (2) developing predictive algorithms to anticipate risks associated with anoxic crisis (called  “malaïgue”), microbial contamination, and algal blooms.

Material and methods

The project is conducted in the Thau lagoon, a semi-open environment in South Eastern France’s Mediterranean coast. Three datasets are created to gather a maximum of information: the first dataset includes data from physical and chemical sensors (including temperature, dissolved oxygen concentration, turbidity levels, salinity…) monitoring continuously at two depths (surface and bottom). The second database analyses water samples to collect biotic data, by sampling different strategic areas in the lagoon where events have occurred over time (including E. coli levels, phytoplankton diversity). Several other abiotic analysis (NO3, NH4, SO4, PO4) are also collected as a complement to those biotic parameters.

Figure 1: Global scheme of the construction of a large environmental data set

A third database consists of satellite images using Sentinel satellites 2 & 3 that examine chlorophyll concentrations along with suspended particulates measurements such as phytoplankton backscattering coefficient of suspended matter and suspended particulate matter. Finally, a fourth database includes other data such as wind, air temperature, rainfall level, flow rates of rivers and releases from wastewater treatment plants around the area.

The data is collected, harmonized and stored for analysis, then preprocessed to identify global behavior patterns of different parameters (i.e., seasonality, extreme values). We also analyze correlation between various parameters to have a complete data exploration at the end. At the end we build and compare models by using statistical methods as well as machine learning methods such as deep learning to determine the optimal performance.

Results

As shown in Figure 2, the deployment of sensors and data collection on April 2021 constituted the first step for a global monitoring. The devices are continuously connected to each other and send their readings through different means (sensitives areas, susceptible for anoxic event, bacterial contamination). Data is collected at many scales: continuously with autonomous devices; locally by sampling; as well as from satellites. Those analyses allow the identification of key parameters or variations over time before they occur - or after they have taken place. Machine learning will subsequently allow to anticipate events before they happen (Lafont et al. 2019). It can also identify correlations between in situ data and satellite images so that sentinel networks can be created to monitor lagoon conditions at multiple levels simultaneously.

Discussion and conclusion

The development of IoT (Internet of Things) technologies enables the efficient collection of large amounts and high frequency data, and their correlation with water sampling data. This information is crucial for industries dependent on water quality. Using the latest Machine Learning techniques with this data opens up the possibility of prediction, thus providing an advantage against risks that can be disastrous if not managed properly in their early stages (HAB - oxygen drop, bacterial contamination). The SENSITHAU project is designed to develop prediction tools against those risks so they can be mitigated before they happen. By improving the monitoring abilities from satellite and continuous measurements, it allows 24/7 surveying while automatically alerting end-users of potential risks. Artificial Intelligence and sensing technologies enable a wide scope of applications (including shellfish and oyster farming) around the World thanks to their usefulness in the development of sensitive areas where regular surveys or sample gathering are either impossible or prohibitively costly.