Biology of the Colombian Horned Lark and preliminary observational study
The study was done in La Copa reservoir, a montane ecosystem, located at an elevation of 2660 m in the central Andean mountains at the municipality of Toca in the Boyacá state (Fig. 1). We selected this area because it harbors the largest populations of the Neotropical or Colombian Horned Lark and has reports of nesting (Valencia and Armenteras 2004; Botía-Becerra and Echeverry-Galvis 2010). A previous study by our research group showed that at the beginning of the reproductive season, both males and females presented an intense vocal activity and there were some fights between adult males (Arias-Sosa et al. 2021). Later the larks were observed mainly in couples and both parents were involved in the care of the nest and offspring, indicating that this bird is at least seasonally monogamous (Arias-Sosa et al. 2021).
In North America, the Horned Larks tend to define territories during the reproductive season that are defended by the males (Beason and Franks 1974; Cannings and Threlfall 1981; DeGraaf and Yamasaki 2001). Nevertheless, in our study area, there were fights between adult males but they didn’t define territories during the mating season; because the area was small and the population density was high, the larks were hindered to have independent well-defined territories (Arias-Sosa et al. 2021).
As we aimed to make playback experiments, we implemented a preliminary test with some recordings of males and females obtained during field observations. In this test, we realized that both males and females responded to the playback similarly, emitting responding songs and getting closer to the speaker. We distinguished males and females based on the distinctive facial marks of the adult individuals. About 60% of responses were from females and 40% from males. The majority of playback experiments only take into account male response, based on the assumption that the female response (mate choice) would follow the same pattern, but this may not be true in all cases (Matessi et al. 2000; Dingle et al. 2010; Pegan et al. 2015; de Oliveira Gordinho et al. 2016). Considering the female response is particularly important in the tropical region where it is more common that female birds present a high song activity and response to playback stimuli (Slater and Mann 2004; Illes and Yunes-Jimenez 2009), we therefore decided to include both sexes in the experiments.
Also, during the playback preliminary test, we realized that multiple individuals responded during the same experiment (as the Horned Larks in the zone presented a gregarious behavior). It was difficult to differentiate the individual response due to the small size of the birds and because they usually move fast, which makes challenging to distinguish which individual of the group was emitting the response singing. Thus, we decided to implement the playback with all the birds in a radius of 25 m per experiment rather than individual analysis.
During the preliminary experiments, we also observed the time it takes for the birds to return to their normal behavior after the acoustic stimuli. We observed that after 1 min the birds returned to look for food on the ground and move away from the speaker. Thus, we established that 2 min of silence between experiments would be adequate to avoid interference between them.
As we mentioned previously during the preliminary studies, we could recognize the sex of several birds using binoculars at a close to medium distance (2‒10 m) and by following a single individual. However, when we tried to measure the playback response at a far distance (30 m, required for the playback experiments), it was difficult to accurately recognize the sex of each bird. Because the identification of the sexual dimorphism in this species requires great attention to particular marks on the face and chest, we thus could not perform a separate sex analysis.
Acoustical traits analyses
Recording collection
The songs of the Neotropical Horned Lark were obtained by field recording in La Copa Reservoir from November 2017 to May 2018. The recordings were obtained from natural vocalizations using a unidirectional microphone (Takstar SGC 568) and were stored in a digital device (Tascam DR 07 MK II MDR12). The same device was used for all recordings, and files were converted to WAV 16-bit, 44.1 kHz. To avoid pseudoreplication we took the recordings in line transect across the 6 points studied (Fig. 1). We took recordings of one singing lark, and then we moved on to take a record of another individual. Overall, we got 40 recordings from E. a. peregrina.
The most recent work in the taxonomy of the Horned Larks proposed it split into four different species or clades, three of them distributed in the western Palearctic and one distributed in the North Palearctic and the Nearctic (Ghorbani et al. 2019). This last species of the Horned Lark was named E. alpestris (sensu stricto) and is subdivided into two subclades: the Nearctic and the North Palearctic subclade. We chose to compare the songs of E. a. peregrina with the Nearctic subclade because it is the geographically closest one and the one with more recordings available. Also because the previous genetics studies indicated a single colonization process of the New-world and a common evolutionary history of the New-world larks (although they only considered some North American populations) (Ghorbani et al. 2019).
We seached for recordings of the Nearctic E. alpestris in the Macaulay Library (Cornell Lab of Ornithology 2018; https://www.macaulaylibrary.org/) and Xeno-canto (Xeno-canto Foundation 2018; https://www.xeno-canto.org/). We had into account only recordings longer than 30 s. These files were obtained in MP3 format and converted to WAV 16-bit, 44.1 kHz. We retrieved 88 recordings.
Song analysis
All recordings were digitized and analyzed using AvisoftSAS Lab Lite software (Sound Analysis and Synthesis Laboratory Version 5.2.12.01 December 2017) (Turčoková et al. 2010; Hamao et al. 2013). From an initial analysis of the recordings, we found two types of vocalizations of the Horned Lark, a short call composed of two or three notes with a length of 0.5 s, and another with a mean of 10 notes and a length of about 2 s. This last one was considered the principal song and was the only one used in the next analysis. We selected this song because it presented more notes and traits that allow better analysis. We observed that the short call is used to communicate between close individuals, especially between the parents with their offspring.
To eliminate external interferences like wind, songs of other birds, and human noise we applied the infinite impulse response (IIR) and finite impulse response (FIR) filters to each recording. These are common and effective mathematical algorithms used in digital acoustic management and in birds’ bioacoustics studies to reduce noise and avoid artifacts (Leonardo 2004; Mockford and Marshall 2009; Rimell et al. 2015). We selected the adequate cut-off frequencies for each song based on the sonogram as it varied between recordings (Leonardo 2004; Mockford and Marshall 2009).
From the original 40 Colombian and 88 Nearctic recordings retrieved, we excluded 20 and 60 recordings respectively because they only contained short calls, had incomplete songs, presented too much background noise, or had a bad sound quality. Thus, we continued the analysis with only 20 Colombian and 28 Nearctic recordings (Additional file 1: Table S1).
For each recording, we extrapolated the songs and excluded those that presented significant noise based on the clarity of the spectrogram. From this analysis, we got 71 songs (obtained from the 20 recordings) of the Colombian lark and 78 songs (obtained from the 28 recordings) of the Nearctic larks. We analyzed these songs using the software AvisoftSAS (with a – 5 to – 10 dB threshold) to get the number of notes, the number of notes per second, song length, the time between notes (average), the highest frequency, and the lowest frequency. We selected these song features because they have been described as parameters of divergence between passerines (Dingle et al. 2008, 2010).
Data analysis
All statistical analyses were performed using the free software R (3.6 version) and the package R Commander (2.5–3 version), with a p < 0.05 as significant (R Core Team 2020). We obtained several songs from the same recording (and potentially the same individual) and we observed interesting variation between them. Thus, we made two separate analyses: (1) using all the obtained songs to get an analysis that included the individual variation and (2) using one randomly selected song per recording to get an analysis not affected by individual pseudoreplication.
As the variables did not present normal distribution (evaluated by Shapiro–Wilk and Kolmogorov–Smirnov tests), the differences in acoustic features were analyzed by the nonparametric Mann–Whitney U test (Turčoková et al. 2010). Later we implemented a principal component analysis (PCA), using Past software (Hammer et al. 2001) to combine the acoustic variables into a single one named “song divergence”. It corresponded to the first axis of the PCA (PC1) that presented an eigenvalue > 1 and was evaluated by the ANOVA test as it was normally distributed. The PCA was done because there was a probable correlation between the variables based on Bartlett’s Test of Sphericity (BTS) and Kaiser-Meyer Olkin test (KMO) in both analyses: using all songs (BTS p < 001, KMO = 0.48) and one song per recording (BTS p < 001, KMO = 0.46).
Finally, we plot the PC1 vs PC2 using a scatterplot to see the grouping of the Nearctic and Neotropical songs. As the Nearctic songs come from different locations in North America, we show the locations in the scatterplot classified in: Central-West (CW), North-East (NE), North-West (NW), South-West (SW), and South USA-Mexico (SM). This classification was based on the subdivision of the Nearctic Horned Larks proposed by Ghorbani et al. (2019) (except South USA-Mexico that was not included in this previous study).
Playback experiment
General description of the experiments
The playback experiments evaluated the degree of recognition and response strength by individuals of E. a. peregrina exposed to the conspecific and Neotropical songs stimuli. The experiments were carried out in the northern part of La Copa Reservoir, where our research group registered the highest densities of E. a. peregrina and several nests (Arias-Sosa et al. 2021).
In each playback experiment, we tested the response to three stimuli: (1) songs of the conspecific populations (Neotropical), (2) songs of the allopatric populations (Nearctic), and (3) response to a different species (Rufous-collared Sparrow Zonotrichia capensis) as a control group. Z. capensis is a common bird of the Colombian Andes that cohabitates naturally in sympatry with the lark and present similar size and behavior (feeding on seeds and insects in grasslands). The songs from the Rufous-collared Sparrow were obtained from Xeno-canto recordings from Boyacá, Colombia. In the conspecific (Neotropical) stimuli, we used songs from other localities (the farthest away in the southern region of the reservoir) to avoid the familiarity of the larks with the local dialect.
Preparation of the experiments
We implemented a total of 36 playback experiments. Each experiment consisted of the broadcasting of a recording, composed of 2 min mix of songs of each one of the three stimuli (Neotropical, Nearctic, and control) separated by 2 min of silence between each other. We prepared each recording with a different mix of songs to avoid potential bias (Pegan et al. 2015). To prepare each recording we had a set of 38 songs of each of the three stimuli. We randomly selected 10 songs per stimuli, added a random pause between songs, and randomly repeated them until complete the 2 min stimuli. All recordings were normalized in the range of 30–80 dB.
Field conditions of the experiments
We broadcast the recordings using an iPod Touch 4G 8 GB player through amplification with an EXTRA BASS™ XB32 portable speaker at 40–80 dB at 1 m, placed on the ground (because the Horned Larks stay most of the time foraging on the floor). We started the experiments only if the close birds were looking for food and not singing by themselves (indicators of a non-stimulated state), to reduce the risk of getting an intrinsic response (not related to the playback stimuli) (de Oliveira Gordinho et al. 2016; Slender et al. 2018).
To reduce the effect of pseudoreplication the experiments were performed in different locations and with one-hour intervals to avoid interference between each other. To avoid potential bias the conspecific (local population) songs were always reproduced after the allopatric and control songs (Sagario and Cueto 2014). This because the conspecific song could overstimulate the response of the birds in the subsequent stimulus (Sagario and Cueto 2014). In the case of Nearctic and control treatment, we alternated its order. The response to the conspecific songs is unlikely to be affected by the reproduction order because as we observe in the preliminary observational study 2 min of silence is enough for the birds to return to their normal (unstimulated) behavior. During all the playback experiments we confirm that the target birds returned to their feeding behavior and get away from the speaker within the 2 min of silence.
Data collection
Once we started the experiment, we registered the behavioral responses at approximately 30 m from the speaker using binoculars. We followed the response during the 2 min of each stimulus and the subsequent 1 min of silence. As the Horned Larks in our study area presented a gregarious behavior that makes it difficult to take individual data, we retrieved the information of all responding larks per experiment. As we reproduce the 3 archives in a consecutive way the same number of individuals were exposed to each stimulus. In some of the experiments, we could observe that both sexes sang in response and approached the speaker; however, we could not differentiate the sex of most of the individuals and we were not able of making a sex-specific analysis. We evaluated the following response variables:
-
1.
The approach response: the number of birds that approached the speaker. This variable was subdivided into: approach within a 20-m radius and an approach within a 10-m radius. A higher number of larks approaching indicates a stronger response.
-
2.
The response vocalization: the number of songs emitted by the larks during the experiments on a radius of 25 m. A higher number of response songs indicates a stronger response.
-
3.
The latency to approach: the time (in seconds) that it took for the first birds to approached the speaker within a 5-m radius. When no bird approached, we reported the measure as 180 s (the 3 min of observation). A lower time is associated with a stronger response.
These variables were selected because they have been described as consistent indicators of behavioral response to playback in passerine birds (Matessi et al. 2001; Jankowski et al. 2010; de Oliveira Gordinho et al. 2016; Freeman et al. 2016; Slender et al. 2018).
The research was done under a research project made in collaboration with the “Corporación Autónoma Regional de Boyacá” the governmental entity that regulates environmental activities in the state. None individual was captured, manipulated, or suffered any damage.
Data analysis
The differences in the playback variables were evaluated by the Kruskal–Wallis test with a Bonferroni post-hoc test. Because there is a probable correlation between the variables (BTS p < 001, KMO = 0.7), we used principal component analysis (PCA), using Past software to combinate the playback variables into a single one named “response intensity”. This variable corresponded to the first axis of the PCA (PC1) with an eigenvalue > 1, and was evaluated by the ANOVA test with a Bonferroni post-hoc test as it was normally distributed. Also, we used a PC1 vs PC2 scatterplot to observe the difference in the response to the three stimuli.