Federated Learning for Environmental Monitoring: A Review of Applications, Challenges, and Future Directions

Federated learning (FL) is emerging as a pivotal paradigm for environmental monitoring, enabling decentralized model training across edge devices without exposing raw data. This review provides the first structured synthesis of 361 peer-reviewed studies, offering a comprehensive overview of how FL has been implemented across environmental domains such as air and water quality, climate modeling, smart agriculture, and biodiversity assessment. We further provide comparative insights into model architectures, energy-aware strategies, and edge-device trade-offs, elucidating how system design choices influence model stability, scalability, and sustainability. The analysis traces the technological evolution of FL from communication-efficient prototypes to robust, context-aware deployments that integrate domain knowledge, physical modeling, and ethical considerations. Persistent challenges remain, including data heterogeneity, limited benchmarking, and inequitable access to computational infrastructure. Addressing these requires advances in hybrid physics–AI frameworks, privacy-preserving sensing, and participatory governance. Overall, this review positions FL not merely as a technical mechanism but as a socio-technical shift—one that aligns distributed intelligence with the complexity, uncertainty, and urgency of contemporary environmental science.