Dominant peak frequency bias modeling and boundary compensation in STFT-BOTDR with small BFS differences

Short-time Fourier transform (STFT)-based Brillouin optical time-domain reflectometry (BOTDR) utilizes a sliding window to estimate the Brillouin frequency shift (BFS). When the window spans the interface between a background region and an event region, the measured Brillouin gain spectrum (BGS) is manifested as a weighted superposition of two spectral components. For small BFS differences, the two components merged into a single distorted peak, and the extracted peak frequency was pulled, leading to a systematic bias in the estimated BFS. This bias dilated the boundary transition in the BFS profile, displaced the inferred start and end positions, and consequently overestimated the event length. In this study, an analytical model for dominant peak frequency pulling was developed, enabling a boundary correction method for small BFS differences. Experimental results demonstrate that the method reduces the boundary localization error to within 0.7 m and the event length error to within 0.4 m for BFS differences ranging from 8.24 to 36.19 MHz. Furthermore, across a broad FWHM range of 58-140 MHz, the event length error remained consistently below 0.2 m. By effectively mitigating boundary-induced BFS bias, this approach enhanced both boundary localization and event length estimation in cost-effective STFT-BOTDR systems for engineering applications.