In soundecology: Soundscape Ecology

Differences of ACI values between seewave and soundecology

The package seewave (Sueur et al. 2008) version 1.7.0 introduced the function ACI() to calculate the Acoustic Complexity Index (ACI). By then I was working on the last tests of my version of a function that would also calculate the ACI. So, I decided to keep my version in soundecology and let the community choose which version worked best for them.

The results of the functions do not match between the packages. I have received this question a few times, so I thought best to write a brief document about it since more people may wonder the same. After some tests, it seems that the differences are due to different assumptions or default values. The only way I can get the functions of both packages to agree was to set nbwindows in ACI() to a comparable value (duration of the file in seconds/seconds in each clump):

library(seewave)
library(soundecology)

data(tropicalsound)
duration <- length(tropicalsound@left)/tropicalsound@samp.rate # 20 seconds

duration

ACI(tropicalsound, nbwindows=(duration/5))

acoustic_complexity(tropicalsound) # j is set to 5 by default

Another example

I also tried with a 50 second-long file (not included in this vignette):

sound1<-readWave("LINE_2003-10-30_20_00_34.wav")
duration <- length(sound1@left)/sound1@samp.rate # 50 seconds

ACI(sound1, nbwindows=(duration/5))
> 1566.69

acoustic_complexity(sound1) # j is set to 5 by default
> 1567.414

Changing the cluster size j to 10 sec:

ACI(sound1, nbwindows=(duration/10))
>783.8312

acoustic_complexity(sound1, j=10)
> 784.0986

The difference between the results from each package seems to come from how each function uses the clumping argument, j, which is set to 5 seconds as a default in both the SoundscapeMeter software and soundecology.