Permafrost degradation in the High Arctic significantly impacts slope stability, infrastructure, and climate-related processes. However, long-term geophysical observations of subsurface mechanical changes remain scarce. We analyse more than three decades (1992–2025) of continuous recordings from the Spitsbergen Seismic Array (SPITS) near Longyearbyen, Svalbard, and interpret relative seismic velocity variations (<em>dv</em>/<em>v</em>) from ambient-noise interferometry as a path-averaged tracker of seasonal thermo-hydrological change in the active layer and uppermost permafrost. We begin by evaluating the quality and continuity of the SPITS data, accounting for changes in instrumentation and data availability across the array. The ambient noise field is characterized through probabilistic power spectral density analysis and frequency–wavenumber beamforming, which reveal seasonal variability and dominant directions of noise sources. Daily noise correlation functions are estimated between inter-station and single-station component pairs and stacked using several robust estimators to enhance signal stability under the variable noise conditions of the Arctic environment. Relative seismic velocity variations (<em>dv</em>/<em>v</em>) are then estimated from the coda of stacked correlations using the stretching method. Our analysis yields two key observations: (i) while the directional noise field varies seasonally, compatible with a mixture of far-field ocean-generated noise and regional environmental sources, it is more stable for the surface wave contribution used for <em>dv</em>/<em>v</em>, and (ii) seasonal <em>dv</em>/<em>v</em> fluctuations and long-term trends correspond closely with soil temperature and ground thermal conditions in the 2 m to 4 m range, particularly during freeze–thaw transitions. Long-term <em>dv</em>/<em>v</em> suggests a velocity decrease of 0.033 % decade<sup>−1</sup> since 1992, which is consistent with the recent Arctic warming. These results underscore the sensitivity of ambient noise interferometry for capturing long-term permafrost dynamics. Our study extends previous noise analyses of the SPITS array to its full operational period and to all available sensor components, and highlights the potential of seismic arrays for long-term monitoring of Arctic cryospheric processes in a changing climate.