The growing complexity of environmental pollution, driven by rapid industrialisation and urbanisation, has highlighted the limitations of traditional approaches to environmental monitoring and control. In recent years, artificial intelligence (AI) and machine learning (ML) have emerged as powerful tools for environmental applications; however, most current implementations remain predominantly prediction-oriented, with limited interpretability, generalizability, and integration with underlying physical processes. This review provides a qualitative and critical synthesis of emerging AI paradigms, focusing on their applications, interconnections, and limitations in environmental pollution control. The review examines key paradigms, including physics-informed machine learning, explainable artificial intelligence, digital twin systems, edge intelligence, federated learning, and autonomous optimization approaches. These paradigms are analysed in terms of their capabilities to enhance model reliability, improve real-time decision-making, and support system-level integration. Applications across major environmental domains—air quality management, water and wastewater treatment, soil remediation, and solid waste management—are discussed to highlight both advancements and domain-specific challenges. Unlike previous reviews that focus on individual techniques, this work integrates multiple emerging paradigms within a unified framework, emphasising their complementary roles and potential for hybrid implementation. Particular attention is given to critical challenges, including data scarcity and heterogeneity, model transferability, computational requirements, reproducibility, and regulatory acceptance. The review also highlights that many current studies remain limited to simulation-based or site-specific implementations, underscoring the need for broader real-world validation. Key research gaps and future directions are identified, including the development of hybrid modeling frameworks, standardized evaluation metrics, improved uncertainty quantification, and responsible and transparent AI deployment. Emerging trends, such as decentralized and adaptive AI systems, are also discussed in the context of sustainable environmental management. In addition, the integration of AI with domain knowledge and environmental processes is essential to ensure reliability, transparency, and practical applicability in real-world systems. Overall, emerging AI paradigms are positioned as enabling tools for transitioning from prediction-centric models to integrated, interpretable, and adaptive systems for effective and sustainable environmental pollution control.
Aravind et al. (Wed,) studied this question.