Distributed fiber-optic temperature monitoring and PCA–ICA analysis for near-real-time seepage detection in river levees

Smart monitoring of geotechnical structures increasingly relies on distributed fiber-optic sensing (DFOS) to provide continuous, high-resolution measurements of temperature, strain, or pressure over long distances. Such systems enable the early identification of local anomalies that may compromise structural integrity under variable conditions. This study presents a methodology for the early detection of seepage anomalies in river levees by combining distributed temperature sensing (DTS) with multivariate statistical processing. The proposed framework integrates principal component analysis (PCA) to reduce data dimensionality and identify dominant thermal patterns, followed by independent component analysis (ICA) to isolate statistically independent signals associated with seepage-induced advective heat transport. The methodology was applied to a levee section along the Adige River (Northern Italy), where DTS monitoring was implemented together with conventional pressure and temperature sensors. During a flood event, the approach successfully identified transient thermal anomalies consistent with visually observed seepage zones and corresponding pore-pressure responses. The algorithm was further evaluated under near-real-time conditions by progressively analyzing partial datasets, demonstrating the capability to detect anomalies within a few hours after their onset. The results highlight the potential of integrating DFOS-based monitoring with advanced statistical data processing to support automated and scalable assessment of levee behavior, contributing to proactive flood-risk management and smart monitoring of hydraulic infrastructures.