582 / 2024-09-18 15:01:20
Red Tide Prediction in Coastal Waters of Zhangzhou, China Using Machine Learning Techniques
Machine Learning, Red Tide, BP Neural Network, RBF Neural Network, Prediction
Session 12 - Alleviating the impact of emerging Harmful Algal Blooms (HABs) to coastal ecosystems and seafood safety for a sustainable and healthy Ocean
Abstract Review Pending
Abstract:
With the advancement of computer technology and artificial intelligence, machine learning has become a key tool in early warning systems for red tides. The occurrence of red tides is driven by complex environmental factors, including water quality, meteorological conditions, and tidal dynamics, which often exhibit nonlinear interactions. This study aims to develop an efficient red tide prediction model by utilizing machine learning algorithms to address the challenges posed by multivariate and nonlinear data in red tide forecasting.
Historical data from Zhangzhou’s coastal waters, covering water quality, meteorological conditions, and tidal levels, were collected and subjected to strict quality control measures. Self-Organizing Maps (SOM) and Support Vector Machine (SVM) algorithms were employed for data preprocessing, generating a robust training set. Red tide forecasting models were then constructed using Back-Propagation (BP) and Radial Basis Function (RBF) neural networks. Validation of the models with two red tide events in May and December 2023 demonstrated a Probability of Detection (POD) of 70% and a Probability of Correct Results (POCR) of 88% in the 24-hour post-event predictions. These results highlight the potential of machine learning in improving the accuracy and timeliness of red tide predictions.
With the advancement of computer technology and artificial intelligence, machine learning has become a key tool in early warning systems for red tides. The occurrence of red tides is driven by complex environmental factors, including water quality, meteorological conditions, and tidal dynamics, which often exhibit nonlinear interactions. This study aims to develop an efficient red tide prediction model by utilizing machine learning algorithms to address the challenges posed by multivariate and nonlinear data in red tide forecasting.
Historical data from Zhangzhou’s coastal waters, covering water quality, meteorological conditions, and tidal levels, were collected and subjected to strict quality control measures. Self-Organizing Maps (SOM) and Support Vector Machine (SVM) algorithms were employed for data preprocessing, generating a robust training set. Red tide forecasting models were then constructed using Back-Propagation (BP) and Radial Basis Function (RBF) neural networks. Validation of the models with two red tide events in May and December 2023 demonstrated a Probability of Detection (POD) of 70% and a Probability of Correct Results (POCR) of 88% in the 24-hour post-event predictions. These results highlight the potential of machine learning in improving the accuracy and timeliness of red tide predictions.