Extracting subtle anomalies from distributed temperature sensing data for dam seepage investigations

Abstract

Fibre optic Distributed Temperature Sensing (DTS) systems are widely used in dams to measure temperature and detect leaks or concentrated seepage. Temperature anomalies caused by advective heat transport of water distort normal temperature patterns, signaling potential leakage. This research develops data visualization and processing techniques to analyze over 10 years of DTS data acquired at the Mactaquac Dam. The dataset shows strong seasonality from conductive heat exchange with air-exposed surfaces and the reservoir. Two approaches are applied to extract subtle anomalies. The first calculates the temperature gradient along the DTS cable, using median filtering to suppress noise and highlight sharp or short wavelength variations. The second, the DBM (deGante-Butler-MacQuarrie) algorithm, adapts the Karhunen-Loève Transform from seismic data processing to estimate and remove seasonal effects as a function of depth, revealing subtler patterns. New visualization methods, including temperature–depth and gradient-depth display inspired by seismic traces displays enhance interpretation.