discussion / Acoustics  / 26 January 2016

Calculating species abundance from Acoustic Data

The question of how to calculate species abundance (rather than simply detecting presence/absence) from acoustic data seems to be a common thread with many of the people I've met through WILDLABS who are working in the acoustic monitoring space. It's my impression that community members seem to be at different stages along the continuum of being able to measure abundance - in some cases even working on similar species (e.g. sea birds). This suggests that a thread to discuss this issue might be beneficial in connecting people who share an interest in this question. 

So, are you working on determining species abundance from acoustic data? It would be great to hear about your work (e.g species focus, data you have access to), the challenges (if any) you are facing and if there are any questions you're mulling over - perhaps the someone else in the community can help you discover an answer. 




Great thread Stephanie! I have deep interest in this topic.

I have been using passive acoustic recorders to monitor primates in tropical forest, and one of my challenges is to know the best sampling design to gathering the data to measure abundance and density.

I would be grateful if someone in the community could help me on this topic.

Hi,

My colleagues and me have tried to estimate the size of a seabird breeding colony using acoustic monitoring, and one of my collaborators has now launched a company that specializes on the processing of acoustic data for monitoring.

The main challenges are: (1) quantifying the number of vocalisations from a stream of acoustic recordings; (2) scaling the number of vocalisations to population size [which is not necessarily linear across a large range - especially once a certain abundance is reached overlapping calls may no longer be recognisable and the count asymptotes]; (3) relating the estimated abundance to a certain area [i.e. over which area your acoustic signals are received by the recorder].

My take on this is that acoustic monitoring is great for monitoring, where you use the acoustic data as an index of population size that is comparable across time and space, but that estimating true abundance from acoustic monitoring alone will always be a bit error-prone.

cheers,

steffen

I have been in the process of setting up an acoustic monitoring network for tigers within India for several years now.

Mariane - I agree with you. One of my biggest challenges was trying to plan out how many recorders I would need for a given area. I consulted with several researchers who have studied big cats, specifically tigers, for most of their career. Several of them suggested applying the same guidelines used to determine how many camera traps you would need. An example of these guidelines can be found in the Panthera/WCS document available at https://goo.gl/NpY3M4. To summarize, the guidelines use home-range sizes for a given species to determine the maximum gap/straight-line seperation between camera traps...in our case, acoustic recorders.

For those of you who have went through this process before, do these guidelines seem okay to use with acoustic recorders? How did you go about determining how many recorders you needed for a given area? 

 

 

My biggest concern using acoustics monitoring to estimate the abudnace or density is the call rate problem, that is, how many calls are made for each animal in general? The uncertainty or variance with call rate then is propagated to be part of variance of the estiamted abundance. The problem (a big one) is then: when the variance of abundance is so large that you cannot compare abudnace from different times or locations.

I agree with Steffen's assessment that "estimating true abundance from acoustic monitoring alone will always be a bit error-prone". I lose the confidence on it since I cannot really know if the change of population trend is significant from statistical viewpoint. Using it as a population index we still need to accept the assumption that the uncertainty from the call rate and others (detection of probability and false alarm) is about the same over location and time. Be careful about it.

Yu Shiu

Postdoc, Cornell Univ.