Abstract
<jats:p>In this chapter, we introduce FedAqua-WQ, a federated IoT-based framework designed for real-time water quality prediction in aquaculture systems. With the growing challenges posed by climate variability, water pollution, and sustainable resource management, traditional centralized monitoring approaches face limitations in scalability, data privacy, and communication overhead. To address these challenges, FedAqua-WQ integrates IoT-enabled sensor networks, edge computing, and federated learning (FL) to provide an efficient and privacy-preserving solution. The framework enables distributed IoT devices to collaboratively train predictive models, sharing only model updates rather than raw data, thereby enhancing both security and efficiency. The findings confirm the superior generalization capability of the random forest algorithm, which maintains leading performance across all cross-validation metrics (MAE = 0.1019, MSE = 0.0189, RMSE = 0.1375, R = 0.6668). Moreover, FedAqua-WQ outperforms the centralized machine learning models; a configuration with eight clients provides the most balanced outcomes, as observed in turbidity (R = 0.6673) and nitrate prediction (R = 0.6270). This chapter highlights its role in improving adaptive decision-making, resource optimization, and sustainable aquaculture practices.</jats:p>