Spectral and statistical analyses of ambient noise

Calving of glaciers generates sound audible underwater, useful in studying glacier processes and ice floes dynamics. Measurements of underwater ambient noise were carried out in two completely different Spitsbergen fjords - Hornsund, surrounded by melting glaciers, and Murchison, devoid of glaciers but full of floating ice floes, at frequencies from 20 Hz to 24 kHz using an omnidirectional hydrophone deployed at 18 meters deep. Calm weather during the experiments enabled measurements without noise coming from wind, rain or breaking waves. Statistical and spectral analyses of its variations were used to identify calving events and phenomena associated with the dynamics of ice floes. We hypothesise that, at low frequencies, the probability density distribution of the noise significantly differs from the normal distribution and gives clues about the number and diversity of contributing sources. Goodness-of-fit tests were performed across all frequency bands, and their results showed central-limit behavior in a high frequency range (>2.5 kHz). For low frequencies (below 1 kHz), the histograms of instantaneous noise spectrum level averaged over 1-minute intervals were compared to the theoretical Gaussian distribution, whose 0.9th quantile was determined as a threshold value in order to distinguish high energy spectral components of non-Gaussian origin.