Abstract
Rainfall and other natural processes can be empirically monitored by analyzing characteristic spectral features in the ocean’s ambient sound. Previous work to detect and estimate rainfall from passive underwater acoustics used linear transformations of these spectral features; Ma and Nystuen [1] measured acoustic power at a few narrowband frequencies, later extended by Mallaryet al. [2] and Berg [3] to principal component analysis (PCA), which represents broadband spectra with a small number of linear coefficients [4]. This research proposes a new broadband detection scheme that constructs separate PCA subspaces by rain, wind and season. Separating the linear transformations before training produces subspaces more finely tuned for each class. PSDs are computed using Welch’s method and are separated into dry (< 1mm/hr ) and rainy (≥ 1mm/hr ) recordings for each season and wind speed category. A linear dimension reduction matrix is defined forthe dry PSDs of each season and wind category using eigenvectors of their covariance matrix (the principal components), while preserving over 99% of variance. Rainfall can then be detected using a likelihood ratio test of dimension-reduced PSDs for each class. Performance varies substantially by season and wind category, with detection ranging from 40% to 70% seasonally and 35% to65% by wind category at a 1% false alarm rate. Accounting for wind and season may improve rainfall detection and, more generally, monitoring of natural processes from underwater acoustic recordings.