Visualisation and Modeling of Marine Ecosystem Using AI - A Way Forward for Ocean Sustainability: A Case of Flic en Flac Region, Mauritius

• Main Authors: Vincent, Agnes Nalini, Sakthidasan, K., Bagurubumwe, Uhoze, Laloo, Nassirah
• Format: Conference Proceeding
• Language: English
• Subjects: Artificial intelligence; benthic cover variables; Chemicals; clustering; Clustering algorithms; Data visualization; deep learning; fish assemblages; hard corals; machine learning; marine ecosystems; Mauritius; ocean salinity; ocean sustainability; prediction; Prediction algorithms; Recurrent neural networks; sea surface temperature; Sustainable development
• ISSN: 2687-7767
• DOI: 10.1109/UPCON59197.2023.10434714

The way people, communities, governments, and business entities perceive and react to ecological and climate change is changing because of data visualisation, predictive analytics and automated decision making. These techniques are enabled by disruptive technologies such as Artificial Intelligence (AI), Robotics and Internet of Things. As far as research and development in marine ecology is concerned, various kinds of AI techniques are being used to analyse the impact of climate change on marine life as this ecosystem is globally endangered. Thus, ocean sustainability has become the need of the hour as we need to sustain the seas and protect their ecosystem because they provide us with resources including food, ways to trade and transportation, energy, employment, leisure, and well-being. This research used physio chemical parameters as independent variables and benthic cover variables as dependent variables of Flic en Flac region of Republic of Mauritius. It analysed, visualised, and predicted the mean percentage benthic cover variables such as hard corals, fish assemblages and algae under the influence of physio chemical parameters sea surface temperature, pH, ocean salinity, and chemical oxygen demand. As a first finding this research revealed a correlation between benthic response variables and physio chemical parameters considered. The AI deep learning algorithm used in this research for prediction of benthic cover variables is Recurrent Neural Network (RNN). Clustering algorithms such as K-Means, Group KMeans and Self-Organising Map (SOM) were used to cluster the data. Based on the clustering performance indices, SOM was identified as a better algorithm and the pre-clustered data from the SOM algorithm was used for prediction by RNN algorithm. The evaluation metrics of RNN were determined. The visual analysis and predictive findings indicate a necessity to recognize and safeguard the marine ecosystem of region under consideration.