Machine Learning Approaches for Predicting Harmful Algal Blooms in Freshwater and Marine Waters: A Review with Emphasis on Random Forest Models
Year: 2025
Type: article
Abstract: Harmful algal blooms (HABs) are an increasing problem in both freshwater and marine ecosystems, human health and socio-economic sustainability across the world. The rising intensity and frequency of the bloom events that are caused by nutrient enrichment, climate variations and hydrodynamic alterations have called limitations to the earlier empirical and process based predictive models. Over the past few years, machine learning (ML) methods have become significant alternatives to define the complex and nonlinear relationships governing the dynamics of HAB. The reviewed literature summarises the current developments in the field of ML-based prediction of HABs, specifically focusing on the Random Forest (RF) algorithm, as it is the strongest, interpretable, and fitting the characteristics of the heterogeneous environment conditions. We discuss the most important environmental drivers and data input used in the HAB modeling, exhaust the RF-based applications of freshwater and marine systems, and compare their performance with alternative ML-based and deep learning methods. Model inputs, prediction targets, and evaluation metrics, model ability to create ecologically meaningful insights in terms of feature importance analysis are paid special attention. The existing issues, namely data limitations, spatial transferability, and uncertainty in the conditions of climate change are argued seriously. Lastly, we present future research opportunities to combine RF with remote sensing and explainable artificial intelligence with hybrid modelling frameworks to facilitate the infrastructure of early-warning systems and sustainable aquatic ecosystem management. This review indicates that RF-based machine learning plays a central role in enhancing predictive machine learning of HABs and environmental sustainability in aquatic environments.
Author: J.Karthika
Language: English
Subfield: Environmental Chemistry
Field: Environmental Science
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W7219965231