Hearing Their Pain: Acoustic Monitoring
Animals vocalize in all sorts of ways. Wolves howl, whales sing, birds chirp, snakes hiss, and lions roar. But what are they trying to communicate? Or are they communicating at all? Little is known so far about what animal sounds mean, particularly as they relate to wild animal welfare (WAW). This article examines our enormous knowledge gap and suggests ways remote acoustic monitoring could help to fill in some of these missing pieces.
Animal vocalizations likely evolved via natural selection. When used effectively, they can confer a survival advantage. In the wild, animal sounds convey information about the animal herself, other animals in the vicinity, and the environment. An animal may vocalize to warn others of impending danger, to stake a territory, or to advertise for a mate. Any of these sounds may hold clues to the animal’s affective, or emotional, state.
To gain more insight into WAW, remote sound capture could record vocalizations in the wild. Animals vocalize differently in response to changes in their nervous systems, which in turn affect their vocal output. These varying sounds may hold clues to their affective state. Scientists describe animal affect along two axes. The first is arousal, from high to low intensity. The second is valence, either positive or negative feeling. So far, researchers have found that highly aroused mammals produce longer vocalizations at faster rates. These are also louder and more variable in frequency. Researchers understand less about vocal characteristics that define valence.
Sound monitoring requires a variety of equipment. Devices such as microphones capture the vocalizations, which are then recorded for later analysis. Because small mammals such as mice and rats may produce sounds at frequencies unheard by the human ear, microphones must be attuned to wider ranges of sounds. Recordings must then be saved and transmitted wirelessly or put on an SD card for later retrieval. All of this technology must be able to function in the environment where the sounds are collected.
A setup with multiple microphones and recorders deployed in a large area over several months would capture an amount of information too large for human analysis. Thus, machine learning (ML) could be used to recognize species and identify affective states that signify good or bad welfare. ML has the potential to improve both the efficiency and accuracy of data analysis. Converting sound data into spectrograms, already a proven technology, could further automate the analysis process.
Another new technology that holds great promise is Convolutional Neural Networks (CNN). It is used for image classification and is more powerful and accurate than the techniques suggested above. A CNN can be programmed to filter and recognize welfare indicators, or it can learn itself. So far, CNNs have been used to identify farmed animal vocalizations with pigs and chickens. For WAW research, the CNN must learn to recognize potentially hundreds of species. This would likely require tens of thousands of recordings. Gathering this data could be time-intensive and costly, and more work is needed for this technology to live up to its promise. Yet a complete, accurate, and fully labeled dataset would hold tremendous value, akin to the sequencing of the human genome.
Once large-scale remote acoustic monitoring systems are in place, the benefits could be numerous. Such systems are non-invasive and don’t require the confounding effects of human presence. Potential uses include biodiversity surveys, population estimates, species discovery, and disease surveillance. Researchers could gather more details about animal behavior and identify dangerous locations, such as where animals cross roads. However, researchers should also take caution to make sure the technology isn’t exploited or used in unintended ways (e.g., by factory farms to prove their animals aren’t suffering, or by trophy hunters to identify their victims). Nevertheless, the opportunities for this technology should excite animal advocates. WAW is still a neglected area of research, and advocates can use this paper to support efforts to increase our collective knowledge of the wild animals who share our environment.