Difference in acoustic responses to urbanisation in two African passerines

Modification of ecosystems as a consequence of urbanisation alters natural habitat structures and soundscapes, creating constraints for vocal communication in animals. Birds are able to adjust their vocalisation to the prevailing acoustic features of their habitat. As such, their sounds have been shown to reflect the level of anthropogenic disturbances across landscapes. While the effect of a single anthropogenic disturbance like ambient noise on birds’ vocal communication is widely investigated, the combined effects of various disturbances remain less explored. We tested single and combined effects of anthropogenic noise and urban physical structures on the vocalisations of two African passerines, the Common Bulbul Pycnonotus barbatus and Northern Grey-headed Sparrow Passer griseus. We predicted that (i) both species would increase the minimum frequency of their vocalisation to avoid masking by ambient noise, (ii) both species would decrease their maximum frequency to evade distortion by physical structures, (iii) the two urban components would have a significant combined effect on the vocalisations, and (iv) the change in minimum or maximum frequency will lead to a corresponding change in the vocalisation bandwidth. For the Common Bulbul, the minimum frequency increased significantly as the ambient noise level increased, but the maximum frequency did not change in response to urban physical structures. For the Northern Grey-headed Sparrow, the minimum frequency did not show a response to ambient noise, but the maximum frequency decreased significantly as physical structures and ambient noise increased. We did not find any evidence for a combined effect of urban physical structures and ambient noise on the vocal adjustment of either species. A change in either the minimum or maximum frequency resulted in a corresponding change in the vocalisation bandwidth of each species. Our findings highlight how the same vocalisation traits of different songbird species can be affected differently by novel selective pressures in acoustic communication that arise in urban environments.


Différence dans les réponses acoustiques à l'urbanisation chez deux passereaux africains
La modification des écosystèmes due à l'urbanisation altère les structures de l'habitat naturel et l'environnement sonore, créant des contraintes pour la communication vocale chez les animaux.Les oiseaux sont capables d'ajuster leur vocalisation aux caractéristiques acoustiques dominantes de leur habitat.Il a donc été démontré que leurs sons reflètent le niveau de perturbations anthropiques dans les paysages.Alors que l'effet d'une seule perturbation anthropique comme le bruit ambiant sur la communication vocale des oiseaux est largement étudié, les effets combinés de diverses perturbations restent moins explorés.Nous avons testé les effets uniques et combinés du bruit anthropogénique et des structures physiques urbaines sur la vocalisation de deux passereaux africains, le Bulbul commun Pycnonotus barbatus et le Moineau à tête grise Passer griseus.Nous avons prédit que les deux espèces: (i) augmenteraient la fréquence minimale de leurs vocalisations pour éviter le masquage par le bruit ambiant, (ii) diminueraient leur fréquence maximale pour éviter la distorsion par les structures physiques, (iii) les deux composantes urbaines auraient un effet combiné significatif sur les vocalisations, et (iv) le changement de la fréquence minimale ou maximale conduirait à un changement correspondant de la largeur de bande des vocalisations.Pour le Bulbul commun, la fréquence minimale a augmenté de manière significative lorsque le niveau de bruit ambiant a augmenté, mais la fréquence maximale n'a pas réagi aux structures physiques.Pour le Moineau à tête grise, la fréquence minimale n'a pas répondu au bruit ambiant, mais la fréquence maximale a diminué de manière significative lorsque les structures physiques et le bruit ambiant ont augmenté.Nous n'avons trouvé aucune preuve d'un effet combiné des structures physiques et du bruit ambiant sur l'ajustement vocal des deux espèces.Le changement de la fréquence minimale ou maximale a entraîné un rétrécissement correspondant de la largeur de bande des vocalisations des deux espèces.Nos résultats mettent en évidence comment les mêmes traits de vocalisation de différentes espèces d'oiseaux chanteurs peuvent être affectés différemment par de nouvelles pressions sélectives dans la communication acoustique qui surviennent dans les environnements urbains.. Keywords: anthropogenic noise, bioacoustics, Passer griseus, Pycnonotus barbatus, songbirds, urban physical structures, vocal adjustment, vocalisation frequency Supplementary material: available at https://doi.org/10.2989/00306525.2024.2310683In addition to climate change, land-use changes (e.g.urban development) are considered one of the most significant threats to wildlife (Isaksson 2018).Infrastructure and noise pollution generated by humans exacerbate the negative ecological effects of habitat change by interfering with sound transmission and by changing the soundscape of the affected areas (Isaksson 2018).These changes interfere directly with sound transmission (and, correspondingly, the signal reception), which negatively affects animal species for which sound plays a central role in communication (Slabbekoorn et al. 2007(Slabbekoorn et al. , 2017;;Dowling et al. 2012;Schroeder et al. 2012).Therefore, the nature of the environment becomes an important subject of discussion in animal communication.Many species including birds utilise vocalisation as a channel of communication, and their vocalisations serve important functions for survival and reproductive strategies, such as mate attraction, territorial defence and species recognition (Winkler 2001;Osinubi et al. 2012).Urban environments are evolutionarily new to birds (Slabbekoorn and Peet 2003;Slabbekoorn et al. 2007Slabbekoorn et al. , 2017;;Halfwerk et al. 2011).Adaptations required to overcome the novel constraints may result in trade-offs between signal transmission and important traits necessary for survival (Halfwerk et al. 2011).Dowling et al. (2012) state that two salient features of the urban environment that birds have to cope with are high-density physical structures and increasing levels of background noise.
Noise pollution is defined as undesired or excessive sound that can have negative effects on the environmental quality or health of living organisms (Goines and Hagler 2007;Nathanson and Berg 2023).It masks sound signals transmitted by the sender, thereby reducing the receiver's ability to hear the signal's original content (Leventhall 2004;Dowling et al. 2012;Isaksson 2018).Two characteristics of noise that play a role in masking are the frequency and the noise level (i.e.degree of loudness).Most anthropogenic noise is low frequency in nature, at <2 000 Hz (Slabbekoorn and Peet 2003), although it can reach 4 000 Hz (Wood and Yezerinac 2006;Nemeth et al. 2013), thereby masking signals by the sender, especially those occurring within this frequency spectrum.Noises are loudest at frequencies of <2 000 Hz (Wood and Yezerinac 2006), and therefore birds that naturally vocalise at much lower frequencies (≤2 000 Hz) are at greater risk of acoustic interference or masking by low-frequency background noise when compared with species that vocalise at higher frequencies (e.g.>4 000 Hz: Oden et al. 2020).Numerous studies have reported that birds sing at higher minimum frequencies in response to urban noise to maximise the functionality of their signals (Dowling et al. 2012;Nemeth et al. 2013;Narango and Rodewald 2015;Ernstes and Quinn 2016;Zollinger et al. 2017;To et al. 2021).In simple terms, birds sing at a higher pitch in a noisy environment.As such, bird sounds have been shown to reflect the level of anthropogenic noise across landscape gradients (Slabbekoorn and Smith 2002;Slabbekoorn and Peet 2003).
Unlike noise that mostly affects low-frequency signals, the effects of physical structures-a dominant feature of the urban environment-have been shown to be stronger on higher-frequency and broader band signals, creating many overlapping echoes (reverberation) that decay randomly (scattering), with unpredictable arrival times that ultimately mask, cancel or distort structural features of the signals (Boncoraglio and Saino 2007;Slabbekoorn et al. 2007;Dowling et al. 2012;Phillips et al. 2020).The acoustic adaptation hypothesis (a theoretical framework for predicting the effects of habitat structure on acoustic signal structure and calling behaviour) predicts that a low-frequency rather than a high-frequency signal carries best in the presence of physical structures because it is diffracted around barrier edges and continues to propagate (Morton 1975;Richards and Wiley 1980;Wiley 1991;Rossing and Fletcher 2004).Therefore, in habitats with many physical structures, birds decrease their song frequency to avoid distortion of their signals.For example, the maximum frequency and frequency bandwidth have been shown to decrease with an increasing concentration of physical structures (Dowling et al. 2012;Phillips et al. 2020).The maximum frequency has also been shown to increase in response to ambient noise, in which case a shift in the entire song frequency in response to noise is suggested (Zhan et al. 2021).This may also be linked to a trade-off whereby the songbird responds more to urban noise than to the density of physical structures.
A rural-urban landscape that varies in the level of physical structures and anthropogenic noise can be seen as a natural laboratory for studying the impact of urbanisation on bird communication (Narango and Rodewald 2015).To date, most studies on this subject have focused on ambient noise.These range from early work by Slabbekoorn and Peet (2003) that found that Great Tits Parus major in Europe sang at higher minimum frequencies in noisy territories compared with in quiet ones, to recent findings by Derryberry et al. (2020) that White-crowned Sparrows Zonotrichia leucophrys sang at lower minimum frequency during the relatively quiet COVID-19 lockdown period than during a parallel period in previous years when urban noise levels were higher.
Little is known about the singular effect of other anthropogenic disturbances or their combination on bird vocalisations.Also, not much is known about how Afrotropical bird species respond to these factors.Here, we tested single and combined effects of anthropogenic noise and urban physical structures on the vocalisations of two common African bird species, the Common Bulbul Pycnonotus barbatus and the Northern Grey-headed Sparrow Passer griseus.From a conceptual overview of the effect of ambient noise and urban physical structures on bird vocalisations, based on the studies cited above, we predicted that: (i) both species would increase the minimum frequency of their vocalisations in a bid to avoid masking by ambient noise; (ii) both species would decrease their maximum frequency to evade distortion by physical structures; (iii) the two urban components would have a significant combined effect on minimum and maximum frequencies of the vocalisations of each species (for example, the minimum frequency might increase in the presence of ambient noise, yet might have less potential to do so when physical urban structures are also present), and (iv) the frequency Introduction bandwidth of the vocalisations would decrease owing to predicted changes in minimum and maximum frequencies.

Study area
This study was carried out in Jos, Plateau State, Nigeria (09°55′00″ N, 08°53′25″ E), from October to December 2020.Jos is a metropolitan city and the largest human settlement on the Jos Plateau (total population of ~900 000 based on the 2006 census: Federal Republic of Nigeria [2012]); it also has pronounced differences in land-use types along a rural to urban gradient.
Human population size is a relatively good indicator of city-level impacts on birds (Isaksson 2018), therefore we classified a settlement as 'rural' if it was characterised by farmlands and small communities, scattered buildings and the availability of open land; 'suburban' if it was composed mainly of residential areas with some commercial activities, new communities in the suburbs of established urban areas as a result of urban expansion, or just areas outside the major urban centres; and 'urban' if it had a high population density, high density of human-created structures, little open space, and many road networks along which most of the commercial activities were carried out.A total of 15 sites were selected-5 sites each (at least 1 km apart) representing rural, suburban and urban settlements (Figure 1).

Study species
The songbird species were chosen based on the following criteria: (i) they are oscine passerines; (ii) they are synanthropic (found both within and outside human settlements); (iii) they are highly vocal species; and (iv) they could be observed in relatively high numbers across all three habitat types (rural, suburban and urban settlements).
The Common Bulbul Pycnonotus barbatus (family Pycnonotidae) is an African endemic resident that is considered the most widespread and abundant bird on the African continent.Common Bulbuls are habitat generalists occupying wooded or bushy habitats, except dense forest.They are also associated with human settlements, being common in villages and urban centres (Keith et al. 1992).
The Northern Grey-headed Sparrow Passer griseus (family Passeridae) is a common Passer species in Nigeria (Okore and Amadi 2016).Its habitats range from plains, open woodland, coastal scrub, and the edges of marshes and forest clearings (but not dense forests), and it is widespread in villages and urban areas (Fry and Keith 2004).

Bird vocalisation recording
We identified vocalising birds and recorded them along paths (streets and minor paths within settlements) between the hours of 06:30 am and 11:00 am, the most active time of day for singing (Wood and Yezerinac 2006;Villegas et al. 2018).All bird vocalisations were recorded with an Olympus DM720 digital recorder attached to an Olympus unidirectional ME31 Compact Gun Microphone, digitised at a sample rate of 44.1 kHz at 16-bit amplitude resolution, and saved in WAV format.
Recording points were marked using a GPS device.To reduce the risk of double recording individuals, we visited each site only once, and established each of the 5 sites in each location at least 1 km apart.During the survey, there were rarely two or more birds calling simultaneously from the same point.However, if two individuals were close enough, the unidirectional microphone was pointed towards one individual at a time and the song recorded before turning and pointing it towards another individual while the recordist slowly moved nearer to that individual.If at any point we were confused about which individual was calling, we did not record them.The distance between vocalising birds and the recordist was difficult to maintain because the birds easily flew away when approached.We began the recording when a vocalising bird was close enough for its vocalisation to be picked up by the digital recorder.

Determination of density of physical structures
Physical structures include buildings, electric poles and impervious surfaces like tarred roads, concrete floors and interlock pavers (Dowling et al. 2012;Isaksson 2018).We determined the percentage of physical structures within a 200-m radius around the central coordinate of each point where each bird was recorded.We obtained satellite imagery (Landsat 8) of the study area for the year 2020 from the Earth Explorer website (earthexplorer.usgs.gov).Using ArcGIS 10.8, we carried out supervised image classification after creating training samples and selecting the different land-use types from the composite layer of bands 1-7 (for this study all selections were grouped into 'built-up' and 'others').We extracted the shape file of the study area from the map of Nigeria and imported into ArcGIS 10.8.We also converted coordinates into a shape file and created a 200-m radius buffer around each coordinate.Finally, we calculated the density of physical structures within each buffer using the 'calculate geometry' function and expressed it as a percentage.Most of the physical structures were buildings (observation from ground truthing).

Determination of ambient noise
On an hourly basis, we recorded ambient noise at a height of 1.5 m at each site (Slabbekoorn and Smith 2002) using a calibrated Extech 407730 digital sound-level meter for 3 min from the four cardinal directions (N, E, S, W).The noise meter was set to 'C weighting' and 'slow' mode during recording (this mode captures most of the low-frequency background noise and automatically averages the noise levels in case of abrupt high levels).After 3 min, we pressed the 'MAX' button and the maximum noise level for that direction was displayed on the screen.We took the average of the maximum noise levels from the four cardinal directions for each point, and then the mean from five recording stations (taken at hourly intervals) was taken as the ambient noise level of that site.

Sound analysis
We used Raven Pro 1.6 software (Cornell Lab of Ornithology, Ithaca, New York) to analyse bird vocalisations.Spectrogram settings in the Raven program were at Fast Fourier Transform (FFT) 512 and temporal overlap of 75%.The minimum frequency (lower frequency bound of the selection or the lower frequency limit), maximum frequency (upper frequency bound), bandwidth (difference between minimum and maximum frequency) and note duration (difference between 'Begin Time' and 'End Time' of a selection) were determined using visual inspection and manual selection of each sound element (Charif et al. 2010; Barišić et al. 2018).

Statistical analyses
We imported spectrogram parameters of vocalisations measured in Raven Pro 1.6 into Microsoft Excel 2016, organised the data and saved it as a CSV file.We used the R 4.2.2 (R Core Team 2022) for all analyses.To establish variation in ambient noise and physical structures across the settlement gradient (rural, suburban and urban), we ran a one-way analysis of variance.Settlement type was used as an explanatory variable, and ambient noise and physical structure were used as individual response variables.
To determine the relationship between vocalisation frequencies and ambient noise and physical structure we used linear mixed-effect models (LMMs) with 'lmer' function in R.Here we set ambient noise and physical structure as fixed factors and sampling site as a random factor.We also used a LMM to test the combined effect of ambient noise and physical structure on vocalisation frequencies.
Ambient noise and physical structure were strongly correlated (r = 0.59).We determined the effect of their collinearity on the model estimates by calculating the variance inflation factor (VIF) for the model.The VIF of the model (containing the two variables) was high with a value of >1 000.Therefore, to prevent collinearity problems, separate models were run for the two explanatory variables.We tested combined effect of ambient noise and physical structure on minimum and maximum vocalisation frequencies of each species, by grouping the predictor variables into two levels (low and high) using K-means clustering (Dowling et al. 2012).We then ran an interaction model having one grouped predictor and another ungrouped but different predictor, as follows: [Frequency = noise + grouped physical structures + noise*grouped physical structures + (1 | site), and Frequency = physical structures + grouped noise + physical structures*grouped noise + (1 | site)].
The reason for using K-means clustering was to avoid the ambiguity that may result when the variables are grouped by mere settlement classification.For example, some rural sites had a high density of physical structures comparable to suburban and urban sites when the different settlements were clustered.Likewise, some urban and suburban sites had lower than expected levels for these variables.K-means clustering was able to group the values into categories irrespective of settlement classification by pooling similar values from all settlements around a given mean and creating non-overlapping levels.With this we are able to tell the effect of a particular level of ambient noise or physical structure irrespective of location.In addition, Zhan et al. (2021) suggested that the simple classification of study sites into rural and urban may not fully describe the variation in the physical habitat between sites.

Results
In total, vocalisations of 75 Northern Grey-headed Sparrows (hereafter 'sparrows') and 45 Common Bulbuls (hereafter 'bulbuls') were recorded at 15 sites (i.e. 5 sites in each settlement type: rural, suburban and urban) with a mean of 6.53 ± 2.85 and 3.33 ± 1.75 per site for the respective species.The maximum distance between the recordist and birds was 69.9 m, and the maximum recording duration was 3 min.Mean values of the minimum and maximum frequencies were 1 781 and 3 504 Hz for bulbuls, and 1 819 and 6 236 Hz for sparrows (Table 1; Figure 2).

Variation of ambient noise and physical structure across the rural-urban settlement gradient
Ambient noise level increased significantly along the rural-urban gradient (ANOVA: F 2 = 109.650,R 2 = 0.65, p < 0.001).After removing outliers, it ranged from a mean minimum of 57.0 dB in the rural area to a mean maximum of 62.4 dB in the urban area (Figure 3), with overall mean (±SD) levels of 58.65 ± 0.89 in rural area, 59.31 ± 1.02 in suburban area, and 61.52 ± 0.99 in urban area.Pairwise comparisons with post hoc tests (Tukey HSD) showed that the difference between any pair of settlements was significant at p < 0.001, except for suburban-rural (p = 0.02) where there was some overlap.Low-frequency noise such as sounds from vehicles and motorcycles (acceleration and horns), engines, standby generators, panel beating and construction works were the chief components of the ambient soundscape.Physical structure, like ambient noise level, increased along the rural-urban gradient and differed significantly among the settlement types (ANOVA: F 2 = 87.191,R 2 = 0.60, p < 0.001).Its percentage ranged from 0.49% in the rural area to 99.15% in the urban area (Figure 3), with mean values of 23.25 ± 15.14% in rural area, 45.35 ± 19.97% in suburban area, and 71.20 ± 17.98% in urban area.All pairwise comparisons with a post hoc test revealed that settlement types were significantly different at p < 0.001.The majority of physical structures were buildings with galvanised roofing sheets, electric poles, and pavements such as roads, concrete floors and interlock pavers.

Frequency bandwidths
Ambient noise did not affect the frequency bandwidth of either species, but physical structure did (Table 2).

Interaction/combined effect of ambient noise and physical structure on vocalisation frequency
None of the interaction models including the grouped variables (ambient noise and physical structure) from K-means analysis was significant for either species (Supplementary Table S1).The minimum frequency of the song of bulbuls increased with ambient noise at all levels of physical structure (Figure 8), whereas the maximum frequency of the song of sparrows decreased with physical structure at all levels of ambient noise (Figure 9).

Baseline levels of ambient noise and physical structure
In this study, anthropogenic physical structures and ambient noise increased significantly along the rural-urban gradient.This is not unexpected since what defines the different settlement types includes the number of people, the density of physical structures and human activities that generate noise, which, by definition, occur at increasing levels along the settlement gradient (Warren et al. 2006;Babalola 2012).Other studies have found a similar pattern along a rural to urban gradient (Slabbekoorn et al. 2007;Dowling et al. 2012;Giraudeau et al. 2014;Job et al. 2016;Zhan et al. 2021).

Birds respond to ambient noise and physical structures in different ways
The main aim of this study was to determine how ambient noise and the density of physical structures affect acoustic communication of two Afrotropical songbird species, separately and taken together, in the face of fast-paced urban development associated with human population increases.This scenario is typical in Nigeria, the most populous country in Africa, where our study was conducted.Urbanisation in the form of human incursion and alteration of natural habitats presents an environment that is evolutionarily novel (Slabbekoorn et al. 2007;Halfwerk et al. 2011).Africa for the most part has large areas of near-natural environment compared with Europe, North America and parts of Asia, but has recorded very high rates of urbanisation in recent decades (United Nations 2018).The effect on acoustic communication in birds, which is important for breeding and survival in many species, may be of particular concern because of the rapid transformation of natural habitats into built-up areas that present highly reflective surfaces and a noisy soundscape.
Ambient noise and physical structure had significant effects on the frequencies of the vocalisations of our two study models.However, their effects were different for each species.While the minimum frequency of the vocalisations of the Common Bulbul increased significantly with increased ambient noise (as expected), no change in this trait was observed in the Northern Grey-headed Sparrow.However, the maximum frequency of vocalisation by the sparrows decreased significantly with both increased ambient noise and the density of physical structures, but the maximum frequency in bulbuls was not correlated with those variables (i.e. the peak frequency did not change in this study; see Supplementary Material).As such, predictions from the acoustic adaptation hypothesis were confirmed but in a different manner in the two species.
The mean minimum frequency of the Northern Grey-headed Sparrow was just 37.92 Hz higher than that of the Common Bulbul.This would suggest that such a difference is not large enough to cause the bulbuls to respond to increased ambient noise whereas the sparrows were not affected.However, Dowling et al. (2012), in a comparative study of six bird species, found that species that tend to vocalise at lower minimum frequencies were more affected by noise.For example, the mean minimum frequency of vocalisations by the Gray Catbird Dumetella carolinensis was 73 Hz lower than that of the Northern Cardinal Cardinalis cardinalis, a sufficient difference for the Gray Catbird song to be more affected by ambient noise.Therefore, it may not be impossible that the small difference in the mean minimum frequency of our species may have influenced the differing responses to ambient noise.
Nonetheless, other studies found that where the minimum frequencies of bird vocalisations were not affected by ambient noise corresponded to species where the minimum frequency of their songs is relatively high and above the peak of ambient noise (Dowling et al. 2012;Zhan et al. 2021), which was not the case for the Northern Grey-headed Sparrow.The minimum frequency of the vocalisations of the sparrow were in the range most affected by ambient noise (<2 000 Hz).It may be possible, therefore, that adaptation of the vocalisations of the Northern Grey-headed Sparrow to the urban environment represents a trade-off between an increase in minimum frequency because of ambient noise and a decrease in maximum frequency and bandwidth, traits that are susceptible to degradation by physical structures.This suggests that the high maximum frequency and wide broadband of the sparrow vocalisations (relative to the bulbul songs) are more affected by interference from physical structures than by ambient noise.High-frequency and broadband sounds propagate well in an urban environment only if produced at lower frequencies (Kight and Swaddle 2015;Billings 2018;Phillips et al. 2020).Our field observations support this hypothesis, as the sparrows often called from electric wires, poles and other structures that were mostly located along noisy roads.The species also calls frequently from rooftops and open vegetation, which may be a way of evading the effects of reverberant physical structures, especially buildings, and a means to maintain an active space for long-range song propagation (Polak 2014).Similarly, urban structure had an overall greater influence than ambient noise on the vocalisations of the Chipping Sparrow Spizella passerina (Job et al. 2016).
In the Northern Grey-headed Sparrow, the maximum frequency also showed a significant relationship with ambient noise.This is contrary to the expectation that ambient noise affects only the minimum frequency of songbird vocalisations.Given that the same ambient noise did not affect the minimum frequency in this species, the decrease in the maximum frequency owing to ambient noise may be an indirect response to the overwhelming effect of physical structures.Although the two predictors were put in separate models because they were statistically collinear, in nature this may be unavoidable since ambient noise and physical structures are present at similar levels in urban areas (Slabbekoorn et al. 2007;Dowling et al. 2012;Isaksson 2018).Another possible reason for the significant negative relationship between the maximum frequency of the vocalisations and ambient noise in the sparrows is that at high ambient noise levels the sparrows may be vocalising at a higher amplitude to be heard over background noise (especially considering that the minimum frequency of their vocalisations falls within the low frequency of ambient noise).However, 'higher and louder' vocalisations may be too noisy to transmit intelligible signals, which will drive a decrease in the maximum frequency.This kind of response is not in consonance with the Lombard effect, which postulates that increased amplitude leads to higher frequency because of shared production mechanics or, simply, because an elevated frequency is a byproduct of increased amplitude (Nemeth and Brumm 2010;Brumm and Zollinger 2011;Zollinger et al. 2012).Although our findings are contrary to this effect, they are in agreement with other studies that reported a decrease in maximum frequency and bandwidth with increasing ambient noise levels (Hanna et al. 2011;Kight and Swaddle 2015;To et al. 2021).Evidence that the amplitude of ambient noise decreases as its frequency increases (Kight and Swaddle 2015;Phillips et al. 2020) further supports our claim that the decrease in the maximum frequency of the vocalisations of the Northern Grey-headed Sparrow may be because it vocalised at a higher amplitude as a response to increasing levels of ambient noise.Unfortunately, we did not measure the amplitude of the sparrow vocalisations to be able to confirm this claim.Still, the present results add to an increasing body of evidence suggesting that an increase in the amplitude of bird vocalisations may not always lead to higher frequencies nor may frequency shifts be just a byproduct of the Lombard effect but also a functional adjustment to noise (Cardoso and Atwell 2011).
The Common Bulbul increased the minimum frequency of its vocalisations in response to greater ambient noise but did not decrease its maximum frequency in response to denser physical structures.Background noise is generally low-frequency noise at <4 000 Hz and has the ability to mask other sounds in that frequency (Slabbekoorn and Peet 2003;Wood and Yezerinac 2006).As such, the entire or a large extent of the vocalisation frequency of the Common Bulbul (1 781-3 504 Hz) lies in the low-frequency spectrum and risks being masked by ambient noise.This may be why the minimum frequency increased, as well as why the density of physical structures did not affect its maximum frequency.Similar responses from different bird species have been reported (Boncoraglio and Saino 2007;Dowling et al. 2012;Nemeth et al. 2013;Kight and Swaddle 2015;Narango and Rodewald 2016).It has also been shown that birds are able to return to the typical minimum frequencies of their songs when ambient noise decreases (Derryberry et al. 2020).Given the low maximum frequency (3 504 Hz) and narrow bandwidth (1 723 Hz) of the vocalisations of bulbuls, one would think that it would be possible for the species to increase its maximum frequency as a result of the upward shift of the entire vocalisation in response to ambient noise, as reported elsewhere (Dowling et al. 2012;Phillips et al. 2020;Zhan et al. 2021).We did not find any evidence for that, probably because the increase in minimum frequency was not strong enough to cause a corresponding increase in the maximum frequency, or because the maximum frequency is located already in the optimum for sound transmission.
We did not find evidence to suggest a combined effect of ambient noise and physical structure on the vocalisation of the study species.Minimum frequency of the vocalisations of the bulbul continued to increase in response to ambient noise at all levels of physical structure density, whereas maximum frequency of the vocalisations of the sparrows decreased in response to the density of physical structures at all levels of ambient noise.These results refuted our hypothesis that the birds would adjust their vocalisation frequencies in the presence of a constraint, but may be less able to do so when a second constraint is present.
Contrary to the acoustic adaptation hypothesis, physical structure had no effect on the maximum frequency of bulbul vocalisations, and ambient noise had no effect on the minimum frequency of sparrow vocalisations.This suggests that the same vocalisation traits of different songbird species can be affected differently by novel selective pressures in acoustic communication that arise in urban environments.
Considering the observed changes in vocalisation frequencies (bulbul: minimum frequency increases, maximum remains the same; sparrow: minimum frequency remains the same, maximum decreases), the frequency bandwidth of their songs narrowed in both species, but the entire call frequency did not shift as seen in Zhan et al. (2021), where both minimum and maximum frequencies increased in response to ambient noise.
In summary, the acoustic traits of each species' vocalisations determined their response to changes in the acoustic properties of their habitat.Our findings established that urban physical structures do not represent a serious constraint to vocal communication in the Common Bulbul because of its low-frequency vocalisations that can 'bend' around those structures and continue to propagate; but they do so in the Northern Grey-headed Sparrow because its high-frequency vocalisations are more affected by physical obstacles.On the other hand, noise did not seem to be a problem to the sparrows' vocal communication but it did to the bulbuls.Successful evasion of one constraint and the ability to navigate the other as we have seen here may present good communication opportunities for them.This may further contribute to the success of these species in urban environments, despite the disturbances they face.
Further research should focus on the behavioural strategies these species adopt to successfully communicate in urban environments.Our study suggests that these species are able to choose the path of least resistance by evading the disturbance that will more easily allow them to successfully communicate.For example, we would expect that the sparrow will successfully communicate in noisy but open areas, whereas the bulbul will do the same in closed but quieter areas.To achieve this in noisy environments, the bulbul should choose high perches to escape masking of the signal, whereas the sparrows should perch away from physical structures to avoid distortion of the signal.This may provide a clearer understanding of how birds adjust their vocalisations and what behavioural adaptations they employ to thrive in an urban environment.

Figure 2 :
Figure 2: Spectrograms of vocalisations of (a) the Northern Grey-headed Sparrow Passer griseus and (b) the Common Bulbul Pycnonotus barbatus at rural sites in Jos, Nigeria

Figure 5 :Figure 7 :
Figure 5: Relationship between presence of physical structures in the environment and the minimum frequency of vocalisations of the Common Bulbul Pycnonotus barbatus (n = 45, p = 0.1042) and Northern Grey-headed Sparrow Passer griseus (n = 75, p = 0.6292)

Table 1 :
Frequency parameters of song of study species.Values are the mean (standard deviation)

Table 2 :
Summary of linear mixed-effect models showing the relationship between frequency bandwidth of the two study species with ambient noise and percentage physical structure.Bold font denotes significance at the 0.05 level