The spatial and temporal characteristics of outdoor wind noise are contrasted to those of unsteady acoustic signals through wavelet processing of data from a 49-microphone array. The results provide clear evidence that wind noise consists of non-planar, intermittent, and weakly coherent microgusts (small turbulent disturbances) propagating across the array. By applying beamforming algorithms to signal cross coherences calculated from complex continuous wavelet transforms (CWTs), the orientations of the microgusts and broadband acoustic signals can be deduced. Wavelet-based beamformers based on the conventional (Bartlett) steering method are shown to produce the most consistent results at low signal-to-noise ratios. A Gaussian-mixture-model classifier is formulated to distinguish between blasting sounds, unsteady but persistent sounds (music), and wind noise. Signal shape features reliably distinguish the blasting and music, whereas the relatively low coherence of the wind noise distinguishes it reliably from the acoustic signals.