Early warning detection of levee seepage through fiber‐optic temperature monitoring and statistical signal processing

Abstract This article presents a monitoring approach for the real‐time identification of seepage anomalies in river levees and foundations. The method is based on temperature measurements acquired via distributed temperature sensing (DTS), which provides continuous, high‐resolution thermal profiles over kilometer‐scale distances. To efficiently interpret the large datasets generated by DTS and isolate thermal signals specifically associated with seepage, the approach integrates principal component analysis (PCA) and independent component analysis (ICA). PCA reduces data dimensionality and highlights dominant patterns of thermal variability, while ICA isolates statistically independent components, some of which correspond to seepage‐induced heat transfer. The framework was validated on a levee section along the Adige River near Salorno (Italy), where DTS and conventional sensors were jointly installed in different vertical boreholes. During the October 2024 flood event, the method successfully detected advective thermal anomalies, consistent with visually observed seepage along the levee face and corresponding piezometric responses. The results confirm that the PCA–ICA technique, when coupled with DTS, offers a robust and scalable tool for early warning and continuous assessment of levee performance under hydraulic loading.