Machine learning-based design and monitoring of algae blooms: Recent trends and future perspectives – A short review

Abstract

Machine learning (ML) models are widely used methods for analyzing data from sensors and satellites to monitor climate change, predict natural disasters, and protect wildlife. However, the application of these technologies for monitoring and managing algal blooms in freshwater environments is relatively new and novel. The commonly used models in algal blooms (ABS) so far are artificial neural networks (ANN), random forests (RF), support vector machine (SVM), data-driven modeling, and long short-term memory (LSTM). In the past, researchers have mostly worked on predicting the effluent parameters, nutrients, microculture, area and weather conditions, meteorological factors, ground waters, energy optimization, and metallic substances in algal blooms using ML models. Most of the studies have employed performance metrics like root mean squared error, mean squared error, peak signal, precision, and determination coefficient as their primary model performance measures for accuracy analysis, and the usage of transfer, and activation function. While there have been some studies on this topic, several research gaps are still to be addressed. The most significant gaps are related to the limited application of ML in different algae bloom scenarios, the interpretability of ML models, and the lack of integration with existing monitoring systems. Keeping these in mind, this review article has been methodically arranged to present an overview of the past studies, their limitations, and the way forward toward the application of ML in the prediction of ABS, thus benefitting future researchers in this area. This review aims to summarize the data that are available, including some benchmarking values.

Highlights

• Real-time monitoring of dynamics using ML is essential for mitigating algal blooms.

• Various complexities hinder applications of current ML algorithms in ABS.

• Activation and transfer functions can be used for selection of ML to predict ABS.

• Integrated ML algorithms can drive feature engineering to predict and control ABS.

Graphical Abstract

Keywords:

• Activation-function
• algae blooms
• monitoring
• machine learning
• performance metrics and prediction

HANDLING EDITORS:

• Hyunjung Kim and Scott Bradford

Disclosure statement

No potential conflict of interest was reported by the authors.