Testing the ecoacoustics event detection and identification (EEDI) approach on Mediterranean soundscapes

The Ecoacoustic Event Detection and Identification (EEDI) system is a recent methodology to investigate structure and dynamic of soundscapes. In this study, EEDI has been re-engineered by adding to the Acoustic Complexity Index (ACI) calculated for each temporal step (ACIft) and to its evenness (ACIft evenness), the evenness of ACI calculated along each frequency bin (ACItf evenness), and upgrading the ecoacoustic code (EC) from a two-digit to a three-digit code, thus increasing the analytical resolution of EEDI. Furthermore, a new metric, the Acoustic Information (AI) index, is introduced which is obtained by multiplying the value of each EC by the number of Recording Minutes (RM) in which every EC occurs. The EEDI was tested in four localities within a Mediterranean landscape using Soundscape Explorer [Terrestrial] digital recorders. Based on 273,360 RMs, 436 ECs were detected from a field recording system active from December 2015 to November 2016. The number of ECs and AI values increased in late winter and early spring, followed by a reduction in late spring and summer. A second increase was observed in early autumn. Based on bird phenology, non-significant differences in the number of ECs were found between the breeding (395 ECs) and non-breeding season (417 ECs). The sequence of the ECs during night time resulted in significant differences from the sequence of the ECs during daytime, but the number of RMs and of AI values of ECs exclusive of day or night time resulted negligible. Hourly abundance of ECs and the associated AI showed a similar sequence for all the localities, confirming a common daily pattern in soundscape organization. According to the way in which ECs are distributed along the 24 h, the first hundred more frequent ECs were aggregated into seven major patterns: some ECs have a dominant distribution around sunrise, whereas others prevail at midday or at night. EEDI provides a visual representation of each recording day (240 RMs) in an Euclidean space delimited by ACIft (x), ACIft evenness (y) and ACItf evenness (z). An “asymptotical like” distribution that occurs during rain or wind events, and a “shower jet like” distribution that occurs when biophonies are dominant, are the two major patterns that were discovered. The ecoacoustic codes EC000, EC999, EC997, EC476, and EC271, have been utilized after their identification, to classify and analyze ecoacoustic events as an example of the capacity of EEDI. This holistic approach starting with detection of patterns delimited by the combination of ACI metrics associated with the AI, overcomes the usual procedure to first identify the sources and then determine the resulting patterns. The use of EEDI provides the potential to explore and to compare the characteristics of soundscapes with other environmental variables and represents an innovative and flexible ecological indicator.