Auditory systems differ vastly across the animal kingdom (21, 22), from external mobile pinnae (“ears”) and internal ear canals to vibration-sensing bones and sensory bristles (21) to detect sound (Figure 2, panel 3). It is therefore important to perform background research on the auditory system of your study taxa or species [for example, within marine mammals there are five functionally different auditory groups; (23)]. Animals may also have certain hearing behaviors related to the orientation of the head and body, or the production of echolocation bursts in the case of bats and toothed whales (21). To complicate things further, the auditory sense is intrinsically linked to other senses; many species switch from hearing to sight or other sensory modalities when distracting sound impedes their ability to discriminate or recognize important sound (Figure 2, panel 3) (24).

A major challenge in bioacoustics is to measure sound in a meaningful way—to reflect what animals can hear, rather than what humans can hear. Specialized auditory tests have been performed on several species and give a baseline indication of what sound different taxa can hear (21). Broadly speaking, there is a negative relationship between body size and hearing capacity (highest audible frequency) (16). However, for many species, the specifics of their hearing capabilities remain unknown. Behavioral hearing tests involve playing a pure tone and training the animal to respond whenever they can hear it; the tone is gradually reduced in intensity and frequency until the animal no longer responds. However, behavioral hearing tests are biased toward animals that can be trained easily and have thus proved difficult for some taxa, e.g., herptiles (25). Alternatively, neural responses to sound can be measured using a technique called the auditory brainstem response (ABR). For this, a short pure tone or click is played (ideally in a sound-controlled booth), and electrical activity is recorded from electrodes on the animal's skin (26). Auditory brainstem responses can be conducted much faster than behavioral tests and do not require training, but it should be noted animals may respond neurologically to sound they cannot physically hear (25). An audiogram is a graphical representation of how well an animal can hear a sound at different frequencies, plotting decibels on the vertical y-axis against frequency on the horizontal x-axis. Audiograms are useful starting points to guide research, alongside the power spectrum of a particular sound (Section 2.4), taking into account what sound a species should be capable of hearing. However, species-level audiograms are not representative of normal hearing in “noisy” environments and also do not take into account individual variations, such as age-dependent hearing loss or damage (16, 17).

Auditory (hearing) ranges span several orders of magnitude across the animal kingdom (4), so it is not surprising bioacoustics is a vast field with distinct bodies of literature for various clades, particularly primates, bats, fish, and birds (27–29). The human ear best detects sound frequencies between 20Hz and 20 kHz, otherwise known as our audible range sounds. Sound above this frequency range is called ultrasound, including the echolocation signals of toothed whales and bats, and sound produced by many insects and amphibians. Sound below this frequency range is called infrasound and includes the vocalizations of large ungulates, pigeons, and some fish (30). Because ultrasound and infrasound are imperceptible to the human ear, they require specialist equipment for us to detect. It is important to remember ultrasound and infrasound are not only produced by animals, they are commonly produced by the environment in the form of thunder and waves, and artificially by HVAC (i.e., heating, ventilation, and air conditioning units) and other machinery (7). The vibrations we (and other animals) can feel as movements are often associated with low-frequency noise (31). While there is a high overlap in the hearing ranges of common laboratory species and humans (thus leading to a relatively high interest in sound we can collectively hear), there are large differences in upper and lower frequency hearing ranges across species (25). Many laboratory mammals have higher frequency hearing than humans; for example, the house mouse (Mus musculus musculus) can hear two octaves (frequency bands) higher than us but has poorer low-frequency hearing (25).

Animals purposely produce sound for many reasons, such as to communicate with conspecifics (mate attraction, territory defense, alarm calls), detect predators, forage, and navigate. The myriad ways animals produce sound, e.g., actively using specialized vocal anatomy or passively by locomoting, are reviewed elsewhere [see, (21), (30)]. We consider the production of incidental sound by animals, such as sound created by feeding, in Section 3.2.1.

A great deal of research has been undertaken on the detection, discrimination and recognition of vocalizations (Figure 2) in a range of species and contexts to understand their function [reviewed by (32)]. Once a researcher has determined the typical vocal variation or “feature space” of a species as its baseline, they can proceed to manipulate various variables affecting vocalization (9). This type of bioacoustics research may require recording spontaneous, naturally produced animal sound (Section 6.2). Or, you may need to artificially produce a sound so that it can be experimentally controlled. Playback experiments involve broadcasting natural or synthetic stimuli and recording the response of animals to these stimuli. Playback stimuli can be in any modality (visual, acoustic, etc.) but are most commonly associated with acoustic stimuli (33, 34). Playbacks can be used to answer many research questions relating to sound. For example, to investigate whether animals can perceive sound of various amplitude and frequencies [e.g., (35), (36)], discriminate between two or more sounds (37), recognize individuals or groups by their sound [e.g., (38)], or whether cognitive or behavioral performance is affected by the presence of sound [e.g., (39)].

There is a rich literature on acoustic monitoring of wildlife [reviewed by (10)]. Acoustic monitoring of wild animal presence, abundance and distribution are vital when visual monitoring is simply not possible (e.g., due to inaccessible terrain or harsh climate, or because species are rare, cryptic, or travel over hundreds or thousands of miles). Acoustic monitoring has received less attention in captivity than in the wild, due to the fact that captive environments do not have the aforementioned monitoring challenges. However, automated acoustic monitoring has immense value in intensive farm environments; facilities housing many thousands of animals under one roof are difficult to observe and have poor air quality for humans (40).

Acoustic monitoring is increasingly used in intensive farm and laboratory environments to track the identities and activities (rather than presence/absence) of animals when behavioral observation is difficult. In the laboratory, the ultrasonic vocalizations of house mice have been used to infer levels of social and sexual behaviors which would be very time-consuming to observe in person. In fisheries, acoustic monitoring has been used to detect incidental sound associated with mating, spawning, and feeding in fish and crustaceans (41, 42). Several studies have used sound (more specifically the sound of jaw movements or pecking) as a proxy for feeding in farmed ungulates and chickens [e.g., (43–46)], allowing farmers to monitor feed intake and general behavior patterns. The sex and genetic strain of intensively reared (broiler) chicks have been rapidly identified through their vocalizations, compared to very time-consuming and costly visual or genetic methods (47). The authors (47) found that the second formant, which is a specific acoustic measurement related to how the sound wave resonates, could reliably be used to identify chick sex and strain.

In practical terms, animal vocalizations need to be detectable against environmental background sound. It is possible to aurally detect vocalizations from sound recordings or to visually detect them on a spectrogram. For example, sound from a pod of zoo-housed killer whales (Orcinus orca) was collected using a hydrophone and then examined using spectrograms [Section 2.4, (48)]. More discrete sound produced by different individuals can be recorded using radio collars with onboard microphones, for example on African elephants (Loxodonta africana) (49) and laboratory-housed common marmosets (Callithrix jacchus) (50) but collars not practical for many species or contexts. Instead, the ability to detect and classify animal vocalizations within large groups has developed considerably with advancements in computer technology. It is now possible to use computer algorithms to pinpoint and only record vocalizations of interest, rather than recording hours of indiscriminate sound which must then be sifted manually (10). Furthermore, artificial intelligence can be used to classify vocalizations. For example, farm-housed goats (Capra hircus) vocalizations were categorized by individual identity, group membership, and age using a specialized computer program designed to mimic the decision-making process of the human brain (thus called an artificial neural network) (51).

Because decibels are a measure of sound energy, decibel level can also be used as a proxy for the animal's distance from the sound logger, taking into account the aforementioned differential absorption and reflection of sound due to physical features in the environment (Section 2.3). Interestingly, researchers recently developed a method to detect and discriminate wild elephant presence and behavior based on seismic data (i.e., ground vibrations) generated by the animals, providing another perspective to sound measurement (52).

The uptake of acoustics in captive animal health and welfare has been slow, despite evidence that animal vocalizations can reliably indicate emotional and physiological states (40, 53). Research in this area can be divided into the diagnosis of: (1) physical health problems; and (2) emotion as a welfare indicator.

The ability to quickly and reliably detect disease in intensively-reared animals has been well-studied because it has high commercial value. Sound-based precision farming techniques are appealing because they can be used when visibility is poor (e.g., in high animal densities, at nighttime), are non-invasive, and are robust against temperature changes (54). Most research in this area has been directed toward the detection of respiratory disease, for example in chickens (55, 56), cattle (57), and pigs (54, 58). The first step is to detect the presence of any vocalization and the second is to classify the type (such as the characteristic “rale” in chickens or “cough” in ungulates), thus leading to a disease diagnosis. For example, researchers extracted 23 acoustic features from recordings of chicken vocalizations and used discriminant function analysis, which is a multivariate statistical test of differences between groups, to identify the five best acoustic features for detecting disease (56). The authors then used an artificial neural network (Section 3.2) to detect healthy and unhealthy chickens from these acoustic features. Vocalization can also indicate reproductive health; whistle production in a zoo-housed female slow loris (Nycticebus sp.) reliably corresponded to estrus state (59). The authors used a method called pulse train analysis to automatically count the number of distinct calls in recordings. Estrus detection through vocalization has also been performed in farmed animals (60, 61).

Research on the connection between vocalization, emotional state, and captive animal welfare typically involves recording vocalizations under conditions of known “valence” (i.e., what we as humans believe are relatively positive or negative conditions for animals), and validating these against other welfare indicators such as behavior or heart rate (53, 62). Negative emotional states have been indicated by the vocalizations of chickens (63), goats, pigs (64, 65) and horses (66). Positive vocalizations in the form of contented “murmurs” have been postulated for cattle (67), but remain relatively understudied (62). Outside the farm, there has been considerably less work on bioacoustic welfare assessment (68, 69), presumably due to a lack of commercial value. Several studies point toward a relationship between emotional state and some characteristic(s) of vocalization, for example by calls becoming more intense or more irregular. Researchers recorded the barking of dogs in different emotional contexts and revealed barks could be classified by their emotional context, even within individual dogs (70). Similarly, reliable differences in laboratory common marmoset vocalizations were detected in situations of positive, neutral, and negative affective state, showing that recording vocalizations in a group environment is more naturalistic than lone testing (50). The “rumble” vocalizations of African elephants in a zoo during times of low and high social interaction (taken to infer relatively negative and positive emotional context, respectively) were compared, finding differences in amplitude, frequency, and duration between the two contexts. However, there was stronger evidence for vocalization indicating the intensity of emotion rather than whether it was relatively negative or positive (71, 72). “Non-linear phenomena” caused by irregular vibrations of the vocal anatomy are thought to be indicative of high emotional arousal, as identified on the spectrograms of infant giant pandas [Ailuropoda melanoleuca; (73)]. A study compared the barks of healthy and unhealthy dogs (the latter being housed in a veterinary clinic), calculating the harmonic-to-noise ratio (which compares regular to irregular vibrations in the call) showing healthy dogs have more regular harmonics (74). “Shimmers” and “jitters” (fluctuations in amplitude and frequency, respectively) are used to infer levels of negative stress or anxiety in human speech (75), and have also been applied to zoo-housed African elephants showing broadly similar findings (76).

For practical application, welfare-indicative vocalizations can be used to create a real-time captive animal welfare monitoring system (62), but this is rare outside of highly commercial farm environments. Beluga whales (Delphinapterus leucas) were less vocal several weeks after being transported to a new zoo enclosure, based on recordings made several times per day or per week (77). Rather than monitor vocalization, a long-term (i.e., over several years) behavior and welfare monitoring program at one zoo took into account daily decibel levels, thus acting as an early warning system during events or construction (78). Recently, an automated, real-time whistle (i.e., abnormal vocalization indicative of distress) detection system was devised for bottlenose dolphins (Tursiops truncatus) housed at a research facility (68). This allowed continuous monitoring from a hydrophone array on the floor of the enclosure and saved researchers over 6 days of manual labor per month processing acoustic data (68). Other reports of acoustic monitoring in captive wildlife have been short-term e.g., for several days or weeks but often over limited hours [e.g., galagos, Galago spp., (79)], rather than having permanent systems in place. Time-restricted acoustic monitoring could underestimate vocal activity, and miss rare and/or sporadic sounds with high biological significance.

Captive animal facilities vary greatly, but sources of background sound can intuitively be categorized into: (1) permanent heating, ventilation, and air conditioning (HVAC) systems, and “life support” water management systems in aquaria; (2) temporary equipment for cleaning, gardening, etc., and public announcement and music systems; (3) human speech and footfall; and (4) sounds produced by animals. In addition, substrates inside animal enclosures and the surrounding area will affect sound (Section 2.3), so it is important to take sound measurements wherever animals will be living (4). The active acoustic space is the distance an animal can detect or produce sound (82) and will most likely be derived from pure bioacoustic research. This active space is artificially restricted in captivity, meaning that sound detection and production in captivity can be very different from an animal's evolved capabilities. Reverberation (the persistence of sound due to the reflection of sound waves from non-absorbent materials) is problematic in enclosures with hard and smooth surfaces (2), which are regularly encountered in captivity.

A branch of bioacoustics research that has relevance to our discussion of background sound has examined the effect of background sound on acoustic communication (22, 83). A phenomenon called the Lombard Effect exists, whereby animals increase the amplitude of their vocalization in response to an increase in background noise (84). The Lombard effect has been found in a wide range of vertebrates in both wild and captive environments [reviewed by (85)]. For example, the amplitude of vocalizations from common marmosets was positively correlated to background white noise. Animals may also restrict their calls to periods of silence; in other words, changing the timing of vocalizations to avoid them being masked by other sound in the environment (86).

The effect of laboratory HVAC sound has been investigated (87), with increased interest over the past two decades due to the potential negative impact of sound on the validity of laboratory animal models (88–90). Several studies have continuously monitored decibel levels inside laboratory animal housing (the cages themselves or communal rooms), focusing on peak and average decibel levels and attributing peak levels to cleaning equipment or worker activity [e.g., (91–93)]. Given taxonomic differences in hearing, it is advisable to record decibels at both low and high frequencies if undertaking multi-species research (91, 93). Reporting how often decibel levels exceed an arbitrary threshold [e.g., (91, 94)] might have some value, as long as the threshold is in some way meaningful for animals (for example based on prior research findings of hearing or responses). The effects of vibrations (i.e., non-audible, solid-borne sound) associated with high-intensity construction or animal transport have been also investigated in farms and laboratories. Vibrations inside the cages of laboratory rodents have been measured and compared to pre-existing reference ranges to extrapolate how vibrations would resonate inside the bodies of humans and rodents (95). Vibrations experienced by animals in transport vehicles have been replicated experimentally by placing animals onto a vibration machine capable of different frequencies and accelerations [poultry: (96); cattle: (97); pigs: (98)].

The consensus from research in aquatic environments (i.e., aquariums and zoo marine mammal enclosures), where sound waves travel about four times faster than in air, is that loud sound comes from a wide range of sources including life support systems (e.g., water pumping and filtration equipment), wave machines, cleaning equipment, visitors, and amplified music or tannoys (99, 100). Hydrophone recordings and power spectra (Section 2.4) from 15 marine mammal facilities showed a large variation in ambient sound between concrete tanks and naturalistic (e.g., penned lagoon) enclosures (99). A separate study found that loud enclosure sound did not significantly overlap with the hearing thresholds of bottlenose dolphins; the life support system produced primarily low-frequency sound, whereas dolphins have high-frequency hearing (101, 102). Taking continuous sound recordings from a dolphin pool has been used demonstrate the contribution of one piece of cleaning equipment, and how dolphins whistled less when it was being used (103). Some elegant experiments on laboratory zebrafish (Danio rerio) have involved creating a sound pressure gradient across the tank and investigating how fish are attracted or repelled by different levels [and how placing the tank on sound-absorbing foam can minimize uncontrollable background sound from the facility; (104)].

Sound is commonly cited as an integral aspect of the “visitor effect” in zoos, but it has proved difficult to parse the effect of visitor-generated ambient noise from other connected factors, such as crowd size and behavior (105, 106). Research in zoos has focused on audible ambient sound levels (generated by zoo visitors and other sources), summarizing maximum and average decibel levels (7, 78). One study measured the sound of visitors knocking on aquarium glass, finding a typical knock was 125 dB (relative to 1μ pa in water). Decibel readers placed in visitor areas rather than enclosures themselves [e.g., (106)], can then be difficult to interpret due to sound attenuation (Section 2.3), i.e., the noise level in the visitor area may not be the same as that experienced by the animal a few meters away.

Interestingly, several studies report animal vocalizations are a major source of sound in captive environments. For example, most sound within the optimal hearing range of toothed whales (40–100 kHz) in captive facilities is generated by the whales themselves (99). A dog bark can exceed 100 dB [measured in unweighted decibels but no reference distance of the reading was given, (107)], and barks contribute significantly to the noise level of kennels (93, 107). In a cross-taxa analysis of vocal animals, researchers found that animal taxonomic groups contained species that could produce sound above 100 dB in air (20 μPa at 1m), and a few species of mammal and bird produced sound up to 125 dB (108). This causes a real challenge for animal caregivers; attempting to prevent animals from vocalizing has negative ethical and welfare connotations, but alternatively, changing the acoustics of a captive environment may be unfavorable to management.

We define a “sound event” in captive animal environments as temporary noise (i.e., sound that has no positive function or value for the animal, Section 2.3). This may include construction work, extreme weather events (storm, thunder), out-of-hours events held in zoos, or wider community events such as festivals, airshows, and fireworks. In these cases, researchers must be prepared (often at short notice) to record sound that is out of their control. This being said, it may also be possible to artificially replicate a real sound event [e.g., construction sound playback, (109)] and experimentally evaluate the effect, assuming ethical approval is granted. Like chronic background sound, sound events are important to study because manmade sound can mask biologically important signals or cues (17), and in captivity, there may be restricted opportunity for an animal to escape aversive sound.

One study compared the performance of Japanese macaques (Macaca fuscata), chimpanzees (Pan troglodytes), and Western lowland gorillas (Gorilla gorilla gorilla) during baseline ambient sound and an air show lasting several days (thus the authors could not control the noise condition themselves) (110). In that study, jets flying overhead were brief noise events identifiable from spectrograms (Section 2.4), but in more chronic noise events it can be difficult to ascertain exactly what characteristic or duration of noise is problematic (if at all) for animals. Construction is difficult to place as either a chronic or acute sound event; it can be fairly sporadic and unpredictable or occur regularly for several weeks or months. Evidence for the effect of construction sound in zoos has been relatively well-reported, but uses a variety of acoustic methods [e.g., (109, 111–113). In one zoo, sound contour maps for each enclosure were created, taking into account the location of speakers (playing back experimental construction sound), topography, and ear height which could then be used to help monitor actual construction sound (109). Another study (101) reported that music from an evening event in an aquarium was detectable in a nearby beluga whale tank. The authors modeled the propagation of music from the aquarium's ballroom (air) to the beluga tank (water) via an acrylic viewing window. In contrast, several other published reports of evening events in zoos (which vary greatly by duration, visitor type, music type, fireworks and/or music, as well as species studied) monitor animal responses before, during, and after events; but these events are presumed to be, rather than quantified as, noisy [e.g., (114–117)]. Researchers recently performed a multi-species comparison of the effects of concerts on zoo animal behavior, comparing pre-, post- and during-event median decibel levels and animal behavior (118).

There are currently no guidelines for sound thresholds for animals in various captive environments, and most studies use human occupational health standards e.g., the National Institute for Occupational Safety and Health (119) as a guide. Maximum noise levels for workplaces, according to the World Health Organization (120), are LA_eq = 85 dB and LC_peak = 135 dB. These are roughly translatable to animals' hearing ranges similar to ours (such as great apes) but not for animals with ultrasonic and infrasonic hearing. The varying hearing sensitivities of different species found in laboratories, farms, pet shelters, and zoos make determining standard thresholds within or between environments very unrealistic. In these situations, stating the signal-to-noise ratio threshold may be more relevant. The signal-to-noise ratio refers to the difference in amplitude between the sound of interest and the background sound. These amplitudes should be as different as possible so that the latter does not mask (cover up) the former. A signal-to-noise ratio smaller than 25 dB [based on humans and 40 bird species; (17)] can negatively impact hearing ability.

Mitigating sound can take many forms, at the source (e.g., reducing the type or characteristics of a produced sound), during transmission (e.g., how it is reflected or absorbed by the environment), or by the receiver [e.g., how the animal perceives it; (78)]. Sound mitigations such as sound-proofing will briefly be noted here. Different sound-proof barriers will either reflect or absorb sound, so if a barrier is required for research purposes this ability must be tested experimentally before use. For example, comparing the noise-reducing qualities of plastic, wood, and foam barriers for zoo enclosures can be achieved by simply measuring the reduction in decibels from one side of the barrier to the other (78). Planting more trees in an enclosure may reduce sound levels but ironically may lead to animals finding it harder to switch to visual signals (modality switching, Section 2.3.1). Sound conditioning refers to masking unpleasant acoustic stimuli (noise) with another sound, such as music, white noise, or a naturalistic recording (121).

In contrast to blocking or masking aversive sounds, a small branch of research has considered adding beneficial sound into the environment, which we interpret as a form of environmental enrichment. Acoustic enrichment in zoos has had very mixed methods and results, calling for more concise statements of methods to allow replication (8). Across captive environments, it is typical to find animal behavior is compared before, during, and after the addition of pre-recorded music [e.g., chickens: (122); Western lowland gorillas: (123): Psittacines: (124)]. As one example, farmed chickens have been exposed to pre-recorded machinery sound and music composed by Mozart (122). These conditions were played for different durations of time and at different dB levels, which raises interesting debate over standardizing sound conditions vs. making “naturally” sporadic and variable. The sound metrics of different music types have not been rigorously studied in the context of acoustic enrichment [although see two studies that reported beats per minute in studies on domestic cats and dogs, (125), (126)], so there is scope to apply a range of indices in this field (Section 4.2). An interesting sideline has been to provide sound stimuli on-demand to animals; chimpanzees and orangutans (Pongo spp.) have been given control over sound production either by pressing control units or by moving physical objects in the enclosure (127, 128). Sound has also been used more as a prompt to perform more naturalistic behaviors, with little consideration for the information held in the sound per se. Generic bird sound has been used as a cue for an African leopard (Panthera pardus) to explore her enclosure, but the sound itself had little real relevance to the leopard, who had only learned the connection between the sound and a food reward (129).

The most common α soundscape indices measure the biodiversity of the soundscape and were inspired by traditional biodiversity indices used in ecology (136, 137). The acoustic complexity index [ACI; (138)] is a commonly used index comparing the difference in amplitude from one time interval to the next within a narrow frequency band. Therefore, the data required to calculate an ACI can be extracted from a spectrogram divided into temporal and frequency bins [Section 2.4, (138)]. High ACI values are obtained from soundscapes with high biophony (e.g., bird and insect calls), or geophony (e.g., storms), thus it is hard to make a clear distinction between these two sound categories. Similarly, the acoustic diversity index [ADI; (139)] measures evenness across frequency bands the required data can be extracted from a spectrogram. A soundscape containing a high range of frequencies will yield a high ADI value [but so will a completely silent recording, reinforcing the importance of listening to recordings in addition to any computer analysis, (137)]. ACI has been used to reliably estimate the number of indri lemurs (Indri indri) participating in a chorus, which is beneficial because inspecting spectrograms was only reliable for two detecting three or fewer singers (140).

Another common application of α indices is to summarize the degree of naturally produced (biophonic) or manmade (anthrophonic) sound in a soundscape. The normalized difference soundscape index [NDSI; (131)] is used to measure the relative ratio of biophonic to anthrophonic sound, working on the assumption that these categories of sound fall within particular frequency bands. Note that geophonic sound is merged with biophonic sound because the index works to distinguish between natural vs. manmade sound. Biophonic sound tends to fall within 2–11 kHz whereas anthrophonic sound tends to fall within 1–2 kHz frequencies (131). To calculate an NDSI, an readings are taken from a power spectral density graph (Section 2.4) for anthropogenic and biophonic frequency ranges, and a ratio is calculated. A higher NDSI value indicates less anthrophony in the soundscape, but the NDSI is by no means a flawless method. The main issue is that the frequency cutoffs are artificial; animals can produce sound below 2 kHz which would be wrongly classified as anthrophony, and the sound of wind and rain can also register as low-frequency sound. Assessing the “naturalness” of animal environments has also been achieved without using soundscape indices, but requires more subjective comparisons and reliance on decibel levels. For example, amplitudes and frequencies of zebrafish soundscapes in the laboratory and five wild habitats have been compared (141). Another study found that ambient noise level in an aquarium pool was 15–25 Db higher than the wild habitat this pool was intended to simulate, due to the life support system (142).

β acoustic indices are less common than α acoustic indices and are used to compare how similar or different soundscapes are across space or time. A dissimilarity index estimates the (dis)similarity in the composition of two recordings (143), and examples include the Spectral Dissimilarity (Df), Temporal Dissimilarity (Dt), and Acoustic Dissimilarity Index (D). Soundscape dis/similarity is difficult to deduce, given that several confounds such as time of day and distance between the microphone and the sound source could wrongly be interpreted as soundscape differences, and there is no universally agreed metric to estimate sound similarity or difference (136). Spectral dissimilarity (Df) for example is derived by comparing the power spectrums (Section 2.4) of different recordings, calculating differences in average Fourier-transformed data for each frequency bin. Given that current β acoustic indices have been criticized for their simplicity and are not as straightforward to calculate or interpret as α indices (136), we issue caution with their use.