FaST work: smart new site-noise tool saves time, money & energy
The decibel levels produced by grounds maintenance machinery often exceed the thresholds observed to negatively affect animal behaviour and health. This throws up a critical question: are the dBA values we reference for noise levels an appropriate metric for assessing noise impacts on different species?
A-weighted decibel noise levels, or dBA values, are commonly used to measure sound levels in environments where human perception is the focus. The A-weighting filter mimics the human ear’s sensitivity to certain frequencies. However, it does not represent the frequency bias of other species, including those that are protected under law.
It’s helpful to consider examples.
Bats are known to hear and vocalise at ultrasonic frequencies, higher than those we are able to hear1, although the horseshoe bat for example 2, does hear some frequencies within our 20 Hz – 20 kHz hearing range. Reviewing noise impacts on bats based on human-centric metrics is widely understood to be nonsensical. Unfortunately, however, there are no sound assessment metrics for bats available in place of this, despite bats relying heavily on echolocation to navigate their surroundings 3 and despite evidence of noise impeding their hunting success 4.
There is a large amount of research on the likely impacts of anthropogenic noise, particularly road traffic noise 5, on birds. This research is centred around observed impacts at different dBA noise levels. However, the hearing biases of birds varies greatly 6 7 and in the presence of this large, complicated range, an average of some sort is typically taken to enable a conclusion to be drawn. The extent to which A-weighted noise levels is an inappropriate measure of sound energy for a bird, depends on the bird. It will also depend on the sound source, and its associated frequency biases. Research even shows that the nature of a bird’s song can influence how they are impacted by noise 8, which highlights the importance of sound frequency. Moreover, even if all these variables are compensated for to provide a like-for-like sound energy comparison, we cannot understand how different species of bird interact with sound. Even if we boldly assume that there is little variation in how birds of the same species interact with sound, it is likely that different species will have different conditioning, contexts, responses to stress, etc. We cannot ask a bird how a sound makes them feel. Therefore, if a bird flushes, i.e. flies away, in response to a sound, we can only assume that it is equivalent to another bird flushing in response to a sound, although research has considered whether noise, again dBA levels, is sufficiently high to provoke a bird to leave an entire area, rather than merely move to a safe place within the same area 9, which is a pragmatic approach.
An interesting bird to consider is an owl. Owls are notable sensitive to, and reliant on, sound. Anthropogenic noise can impair owl hunting 10, which is clearly an adverse impact to the owl, but is this bad overall? The owl’s prey certainty would not think so!
There are numerous protected species, such as the badger, that we do not have audiograms for, i.e. quantified representations of hearing thresholds at different frequencies. Our understanding of a badger’s hearing is limited to research on badger vocalisations. Consequently, we are not able to even estimate whether a sound is likely audible for a badger, never mind derive thresholds at which adverse noise impacts are likely to occur.
Defining how anthropogenic sounds influence the soundscape for wildlife, and how much that matters, is challenging, to put it mildly. There is a dearth of metrics for properly quantifying how sound environments influence the wellbeing of specific species. To really get an objective appreciation of the impacts of anthropogenic noise on wildlife, we need to move away from humanised measures of sound and think innovatively. Guidelines for long-term ecoacoustic monitoring 11 note that soundscapes are best quantified by choosing periods and frequencies applicable to the target taxa. A large proportion of existing research does not do this. However, recent research shows the potential of machine learning. Through analysing audio it has been discovered that elephants use name-like components to identify other elephants 12. By using species specific audio parameters and investing a wide range of potential metrics, there is serious potential for AI to enlighten us.
Until we properly understand how animals interact with sound, there is no knowing if, when, why and to what extent the noise we produce influences the wellbeing of wildlife, and how that influences local biodiversity. For humans, we know what sorts of noise levels correspond to adverse health outcomes 13, that a mix 14 of natural sound sources can promote restoration15 and relaxation 16; we also know that moderators such as what we see influence our perception of a soundscape 17. Perhaps one day, we can be informed enough to actively enhance soundscapes for specific species of wildlife.

Footnotes