@[@Eldridge2018] - The root-mean-square [RMS] of the raw audio signal gives a simple description of signal amplitude; RMS has been demonstrated to track ecologically-relevant temporal and spatial dynamics in [forest canopy][] [[Rodriguez2014][]], and shown to be strongly positively correlated with percentage of living coral cover in tropical reefs [[Bertucci2016][]], but has not been investigated in recent terrestrial correlation studies. Mean values are expected to increase with acoustic activity, variance may more accurately track avian vocalisations under the same logic as ACI.
@[@Eldridge2018] -Spectral centroid provides a measure of the spectral [centre of mass][]; it is widely used in machine listening tasks where is it recognized to have a robust connection with subjective measures of brightness. This and related spectral indices have been shown to be effective in automated recognition of environmental sounds in urban environments [[Devos, 2016][]].
@[@Eldridge2018] -Zero-crossing rate [ZCR] is one of the simplest time-domain features, which in essence reflects the rate at which sound waves impact on the microphone. ZCR provides a measure of noisiness [being high for noisy, low for tonal sounds] and is widely used in speech recognition and music [information retrieval][], for example as a key feature in the classification of percussive sounds [[Gouyon2002][]]. SC and ZCR have been demonstrated to be useful descriptors in classification of habitat type [[Bormpoudakis2013][]], but have yet to be evaluated as proxies for species diversity. We expect a negative association with avian activity for both: relative to the quiet broad-band noise of inactivity, avian vocalisations are predicted to be of lower frequency and more harmonic, resulting in a lower spectral centroid and zero-crossing rate. We might also expect the variance of each to positively track activity as the onsets of avian calls will cause rapid changes in values.