Introduction
Anoxia and harmful algae blooms (HAB) can lead to detrimental effects in the environment and in aquaculture. Anoxic events or periods of low dissolved oxygen (DO) can occur due to various factors (e.g. high water temperatures no/low currents and algae blooms). Algae can deplete the water of oxygen however some can also damage the fish gills. The IoT booming since several years allows high frequency and quality data monitoring. IoT improvement has allowed Bioceanor to develop specific predictive models. Through the example of this application in a shellfish farming sector subject to anoxic episodes, we have integrated different data sources (e.g. in situ monitoring, lab analysis and satellite images), to develop an operational tool used by shellfish farmers to anticipate dissolved oxygen concentration 48 hours in advance.
Results
All the sensors monitoring water quality in real-time (24/7, 20 min-1) in a lagoon, representing a sentinel network for risk visualization in real-time. Different parameters were measured in three ways, through in situ sensors, locally by sampling, and by satellite.
The collected data were first qualified and processed (elimination of outliers, treatment of erroneous data, integration of time series ...) to use only qualified data for the models. The data that have been collected was firstly analysed to help identify key parameters and their variation over time before, during, and after an event of interest like anoxia or HAB. Machine learning analyses were subsequently used to create algorithms that could predict these events. Bioceanor has been able to develop and to run an algorithm that predicts DO 48h in advance with a 4% error rate, using data from IoT devices. Algorithms to predict HAB using in situ measurements and satellite, under development, yield encouraging results.
Discussion and conclusion
The development of IoT simplifies the collection of large amounts of data, in real-time and at high frequency. It allows us to build larger and more robust data sets. Using these data with machine learning opens the world of forecasting. Several risks exist for the different industries dependent on water quality and being able to anticipate some of these events, like anoxia and HAB, can benefit aquaculture so the farmer can protect the farm and livestock. The development of this technology will be applicable all around the world and will be a benefit to many sensitive areas that are at risk. IoT development allows now massive data collection and open the era of forecasting for water quality and Aquaculture.