Modulation statistics of natural soundscapes

Modulation statistics of "natural soundscapes" were estimated by calculating the modulation power spectrum (MPS) of a database of acoustic samples recorded in nine pristine terrestrial habitats for four moments of the day and two contrasting periods differing in precipitation level. Following Singh and Theunissen (2003), a set of statistics estimating low-pass quality, starriness, separability, asymmetry, modulation depth and [Formula] temporal-modulation power-law relationships were calculated from the MPS of the samples and related to geographical, meteorological factors and diel variations. MPS were found to be generally low-pass in shape in the modulation domain, with most of their modulation power restricted to low temporal (<10-20 Hz) and spectral modulations (<0.5-1 cycles/kHz). Modulation statistics distinguished between habitats irrespective of moment of the day and precipitation period, with a greater role of Modulation depth and Starriness. Separability and Starriness were found to be related to the global biodiversity decrease from tropical to polar regions, suggesting that the lack of joint high spectral and fast temporal modulations, and MPS complexity are important features characterising "biophony", the collective sound produced by animals in a given habitat. These findings may help guide research on monitoring auditory behaviors and underlying mechanisms expected to exploit regularities of natural scenes.