Call variation and calling site preference of three sympatric Boulenophrys frogs

Abstract

Animals living in syntopy share acoustic space. Asian horned frogs are well known for their sympatric distribution, but little is known about their strategies to avoid acoustic niche competition. This study focused on three sympatric Boulenophrys frog species from southern China—B. nanlingensis, B. ombrophila, and B. shimentaina, with the former two species call in similar frequencies but breed in different seasons. First, we checked the call variation during the change of individual body size and the ambient air temperature in three species. We have found call frequencies were the most static parameters in each species that were associated with body size and contributed most to species identification. Temporal call parameters shift with temperature but are mostly influenced by low temperatures. Second, we checked the interactions between the environment and call properties. The calling site preferences of each species corresponded well with the prediction of the acoustic adaptation hypothesis (in view of higher frequency better transmission in open habitat), and species with similar call frequencies have some aspects of common calling sites. Third, we checked the species–species interaction by using playback tests with male B. nanlingensis. Results from playback experiments showed species that call at similar frequencies could hardly share the same habitat during the same season. These findings expanded the knowledge of acoustic coexistence in closely related anuran species and provided insights into the vocal behavior of Asian horned frogs.

Anurans, the most diverse group of amphibians, have evolved various acoustic signals facing complex physical environments on land, enabling them to transmit information on mating, territorial defense, alarm, or other aspects of social interactions (Gerhardt 1994; Wells 2007; Köhler et al. 2017). Such acoustic diversity is supposed to be driven by sexual selection, the environment, phylogeny history, and species–species interactions (Blair 1964; Schwartz and Wells 1983; Ryan and Rand 1993; Goutte et al. 2016, 2018). In closely related species, the divergence of acoustic signals plays an important role in premating isolation (Blair and Pettus 1954; Littlejohn 1959; Köhler et al. 2017). Therefore, in their overlapping distribution areas, they could exhibit character displacement in acoustic signals (Blair 1974).

Breeding activities of anurans restricted to wet areas (mostly water bodies), with mix-species assemblage often calls at the same time. Pioneer researchers in the 20th century have discussed acoustic niche partitioning in frogs (Fouquette 1960; Hödl 1977; Duellman and Pyles 1983; Ptacek 1992). Such questions continue to be discussed (Chek et al. 2003; Sugai et al. 2021). Acoustic niche partitioning typically includes the separation of call frequencies, call times, and position of calling individuals (Chhaya et al. 2021). While how do they distinctive calls from different species? The mechanism of frogs avoiding others’ acoustic signals in a large mix-species chorus assemblage has been deeply investigated (Vélez et al. 2012, 2017; Bee and Christensen-Dalsgaard 2016; Bee and Vélez 2018). Recently, Lee et al. (2021) have provided evidence that inflated lungs could improve the signal-to-noise ratio for communication by enhancing the spectral contrast in received vocalizations. Thus, one can avoid acoustic signals from unexpected species with different frequency ranges, and this could somewhat explain why most static characteristics of frog calls always appear at dominant frequencies (Köhler et al. 2017).

The diversification of call frequencies could be explained by different mechanisms (reviewed in Littlejohn 1977). Typically, body size is a main physical restriction because larger individuals have larger vocal apparatus, which leads to the vibration of low frequencies that resulting sounds with lower frequencies (Walkowiak 2007). This was called the principle of acoustic allometry, a common rule in vertebrates (Charlton and Reby 2016; Tonini et al. 2020; Muñoz et al. 2023). For species that emit sounds inconsistent with the allometry principle, ecological selection may play an important role. Such as frogs inhabiting streams tend to emit higher frequencies to avoid disturbances from stream noise (Goutte et al. 2018; Zhao et al. 2021). Interactions between call frequency and the environment were frequently observed, and that could be explained by acoustic signals that tend to evolve for better transmission in their environment, which was called the acoustic adaptation hypothesis (Morton 1975). However, several reviews have discovered no general consensus on this hypothesis in current research, especially in anurans (Ey and Fischer 2009; Erdtmann and Lima 2013; Hardt and Benedict 2021).

Frogs in the subfamily Megophryinae have experienced a long, diverse evolutionary history, forming 10 monophyletic clades representing different dispersal routes from the Himalayas to Southeast Asia (Mahony et al. 2017; Liu et al. 2018; Lyu et al. 2023). This has led to sympatric distribution modes in many areas, especially in the largest Boulenophrys genus (Liu et al. 2018). The similarity of Boulenophrys calls to the human ear has raised the question of how these frogs avoid acoustic niche competition in syntopy (Shi et al. 2020; Tapley et al. 2021). There have been several observations that Boulenophrys frogs with similar call frequencies would not appear at the same time, either during different seasons or separated by distribution elevations (Wang et al. 2014; Wu et al. 2020; Tapley et al. 2021).

In recent works, we have reported the sympatric distribution mode of three Boulenophrys from Mangshan in southern China, which are Boulenophrys nanlingensis, B. ombrophila, and B. shimentaina (Qian et al. 2023a, b). Three species could be found in the same habitat during each breeding season. However, B. nanlingensis and B. ombrophila share similar call frequencies exhibiting separation in calling season (Qian et al. 2023a). In order to figure out the mechanism for acoustic segregation of the three sympatric species, we focus on the following questions: (1) testing the call variation of each species to find out if there are overlapping in acoustic signals, (2) investigating the calling site preferences of each species to evaluate the interactions between call properties and the environment, and (3) testing if there are any acoustic disturbance in three species. Field recording collections and playback tests were applied in this study.

Materials and Methods

Field techniques

Call recordings and calling site data collections

Field works were conducted from July to October 2022 in Hunan Mangshan National Nature Reserve, Hunan Province, China. Call recordings were mainly made by a Zoom F3 field recorder (Japan) with a Sennheiser ME66/K6 microphone (Germany), set at 32-bit float resolution and 192 kHz sample rate. Several recordings were made by a Tascam DR05X recorder (Japan) with built-in microphones (as detailed in Supplementary Table S1). A sound level meter (Aihua AWA5636-1, Hangzhou, China) was used for measuring sound pressure levels (SPLs). While measuring SPLs, the sound level meter was put in front of the individual at ~20 cm distance, with the peak amplitude scored as the Z-weighted value (Gerhardt 1975, 1998). Ambient air temperature was measured by a digital anemometer AZ Instrument 8918 (Taiwan, China). Because the body temperature was changed fast while handling the frog, air temperature was used to represent the body temperature following Fouquette (1980). Individuals were captured after recording, the snout-vent length (SVL) was measured by a RAYENR (Ningbo, China) digital caliper (0.01 mm, to the nearest 0.1 mm), and the body mass (BM) was weighed by a Meilan MT 300g (Shenzhen, China) electronic balance (0.01 g). After measurement, individuals were photographed on three aspects: the dorsum, the venter, and the lateral side. Individual coloration patterns were used to recognize each individual and avoid analyzing calls repeatedly on the same individual. This recognition method was partly demonstrated by Qian et al. (2023a). All individuals were released in situ after photographing.

After recording, a soft ruler (1 mm, to the nearest 1 cm) or a Leica D1 (Germany) infrared rangefinder (to the nearest 1 cm) was used to measure the vertical height (VH) to the nearby bottom and distance to the nearest water body (DW) of the calling site. Depending on the substrate each frog’s feet stand on, the substrate of the calling site was categorized as (1) on the floor, including feet standing in water, on sand, and on leaf litter, and (2) elevated, including feet stand on rocks, fallen logs, and plants. Shelter types were classified as (1) open: the frog was located in an empty position; (2) burrow: the frog was in a rock cave or crevice, thus helping to broadcast the calls; and (3) hide: the frog was hidden under leaf litter, fallen logs, or plants that could not distinctly amplify the sound.

Playback tests

To test the hypothesis that similar call frequencies in closely related species could cause a breeding disturbance, we conducted field playback tests with male B. nanlingensis. We have known that local male B. nanlingensis could exhibit aggressive behavior (i.e., aggressive calls) while an intruder conspecific male calling nearby (due to field observation), similar to that reported from Xenophrys frogs (Mahony et al. 2013, 2020). Thus, we made stimulations extracted from field recordings of advertisement calls and aggressive calls of B. nanlingensis as control groups. We were curious about whether male B. nanlinensis could react against non-conspecific voices with similar call frequencies. Then we made stimulations extracted from field recordings of B. ombrophila, the sympatric species that separated calling season with B. nanlingensis and calls at a similar frequency. We also made stimulations from recordings of another species (B. lishuiensis) that did not occur in the study area but also shared call frequency that was similar to B. nanlinensis to evaluate that the reactions from male B. nanlinensis were directly based on call simulations but not other social interactions. Additionally, we have observed B. nanlingensis and B. shimentaina calling in syntopy, which means they were not disturbed by each other’s calls. This interaction was expected as they have distinct differences in call frequencies. Thus, we made additional stimulations from calls of B. shimentaina, as well as a species (B. sanmingensis) that did not occur in the study area. Both have call frequencies obviously higher than that of B. nanlingensis. For the definition of aggressive calls, we followed Toledo et al. (2015). Detailed information and spectrograms of stimulations are provided in Supplementary Table S2 and Supplementary Figure S1.

Playback experiments were conducted in the field with randomly selected calling individuals of male B. nanlingensis between 19:00 and 24:00. An iPhone SE 2022 (Apple, USA) with a built-in speaker was used to play the stimulation recordings. Based on field observations, the aggressive calls were usually emitted by local male frogs while the intruder male called within a distance smaller than 1 m. Within that distance, the SPLs of invading males to the local males approximately correspond with our speaker’s loudest volume measured at about 20 cm distance in front of the speakers. Thus, we played our stimulated recording in front of the frog of approximately 20 cm with maximum volume settings (SPLs of each stimulated recordings are detailed in Supplementary Table S2). For each frog, stimulation recordings were randomly selected. We take an approximately 1-min break after playing one type of stimulation and change to another stimulation. Due to our target of playback tests, we only count if there are any reactions (or not) from the stimulated frog. The reaction time and pre-reaction time were not counted. Once the frog exhibited the following reactions, we marked it as a RESPONSE, including (1) warning, which means the stimulated frog responded by aggressive calls and (2) seeking, which means the stimulated individual started to seek the sound source. While the frog ignored the stimulations, we marked it as a NON-RESPONSE. For the frogs that stopped calling and did not show other behaviors after stimulations, we regarded it as a failed trial since the calling behavior of the frog could be interrupted by the experimenter. The total effective tests were counted 43 times with 12 individuals.

Data analysis

Call analysis

The advertisement call of megophyinid frogs typically consisted of repeated single-note calls that could be emitted during a single inspiration (illustrated in Supplementary Figure S2). The “call-centered” terminology summarized by Köhler et al. (2017) was used, which means a single acoustic unit was defined as a “call,” and a group of calls was defined as a “call series” (Figure 1). Within a “call,” a series of well-defined energy bursts were consisted, each was defined as a “pulse.” To avoid bias between settings on a different digital recorder, all recordings used in this study were resampled to a 24-bit resolution and 44.1 kHz sample size by Adobe Audition 2022 before analysis. Raven Pro version 1.6.4 was used to perform the acoustic analysis using the following settings: Hann windows, DFT = 512 samples, and overlap = 50%. Selections on each acoustic unit were manually created to measure the acoustic parameters: call duration (CD), the duration of a call; call interval (CI), the interval between each call within a call series; dominant frequency (DF), the spectral peak of highest relative amplitude; pulse number (PN), number of pulses per call; call rate (CR), number of calls per second, calculated by counting the total number of calls (k) within a call series and dividing k-1 by the duration between the onset of the first call and the onset of the last call of the call series (modified from Bee et al. 2012); and call number (CN), number of calls per call series. The duration of a call series was not measured as it depended on the number of calls, not the interval between call series, because it was always affected by the individual’s alertness. To avoid bias caused by data unbalance, five call series of high signal-noise ratios per individual were randomly selected for analysis. In total, 175 call series from 35 individuals of B. shimentaina, 105 from 21 individuals of B. nanlingensis, and 60 from 12 individuals of B. ombrophila were analyzed, which represents 1,443, 1,464, and 2,887 calls, respectively. Notably, there were always present calls with indistinct pulses which were unable to count. Thus, the dataset of pulses was not as much as the calls analyzed. The box plots of each call parameter compared three species based on individuals’ mean value were generated by using the ggplot function in ggplot2 package v.3.4.1 (Wickham 2016) and given in Supplementary Figure S3. The spectrogram and oscillograms were generated in R program using seewave v.2.2.0 (Sueur et al. 2008) and tuneR v.1.4.2 (Ligges et al. 2022) packages with the same settings of above analysis in Raven Pro.

Figure 1.

Advertisement calls for three Boulenophrys in this study. (A–C) 40 s oscillograms showing four, one, and eight call series for B. nanlingensis, B. ombrophila, and B. shimentaina, respectively. (D) Spectrograms and corresponding oscillograms of a single call of each species.

Effect of body size and temperature on call parameters

The effect of body size and temperature on call parameters on each species was tested by using linear mixed models (LMMs). We used individual id as random effect, with the six call parameters as response variables (CD, CI, DF, PN, CR, CN), and temperature and body size as predictors. To avoid collinearity of snout-vent length and body mass, we used the index of body condition (BC) to represent the body size, which was calculated by dividing the residuals of linear regression of the cube root of body mass on snout-vent length (obtained in R program by using the lm function) by snout-vent length (Baker 1992). The models were generated in R program by using the lmer function in lme4 package v.1.1.31 (Bates et al. 2024), following the code shared by Röhr et al. (2020). In total, 12 models per species were generated. Since LMMs are generally robust to violations of model assumptions (Schielzeth et al. 2020), we visually assessed the QQ plots (by using qqnorm and qqline function in R program) of the model residuals, which indicated general satisfaction with the assumption of residual normality. We reported the marginal and conditional r squared values to understand the model fit (Nakagawa et al. 2017) by using the performance function in performance package v. 0.12.2 (Lüdecke et al. 2021).

Coefficients of variation of call parameters

Prior to the analysis of the variation of advertisement calls, we corrected all call parameters significantly correlated with temperature by adjusting their value to a common temperature (Bee et al. 2010; Kaefer and Lima 2012; Pettitt et al. 2013; Vélez and Guajardo 2021; Prasad et al. 2022). For our recording datasets, the average temperature for B. nanlingensis was 17.5°C (standard deviation [SD] = 2.9, range = 9.0–21.4 °C), and for B. ombrophila was 18.9 °C (SD = 1.1, range = 17.5–21.5 °C), and for B. shimentaina was 19.4 °C (SD = 1.2, range = 17.2–21.6 °C). We then chose the common temperature of 19.0 °C for all species since it is relatively close to the datasets for B. ombrophila and B. shimentaina, and most importantly to reduce the effect of low temperature on temporal parameters (Poyarkov et al. 2017; Qian et al. 2023a). The temperature-adjusted acoustic data were calculated following the equation from Platz and Forester (1988):

$${\displaystyle \begin{array}{l}{\displaystyle Y_{adj}=y-(b\ast T_{callingsite})+(b\ast T_{mean})}\end{array}}$$

where y is the original parameter, Y_adj is the adjusted parameter, b is the regression coefficient, T_{calling site} is the raw temperature measurement, and T_mean is the common temperature. The regression coefficient was obtained from above LMMs. Coefficients of variation (CVs) were used to analyze the within-individual variation (CV_w) and among-individual variation (CV_a) of different call parameters for each species (Bee et al. 2016), as follows:

$${\displaystyle \begin{array}{l}{\displaystyle CV=\left(\frac{SD}{X}\right)\ast 100}\end{array}}$$

where X is the average value for each call parameter. CV_w was calculated by dividing the SD of each call parameter for each frog individual by their average value. CV_a was calculated as the average and SD of call parameters of all individuals. CV_w was categorized into three types following Gerhardt (1991), including dynamic (CV_w < 5%), intermediate (5% ≤ CV_w ≤12%), and dynamic (CV_w > 12%). We report the ratio CV_a/CV_w to test whether a call parameter varied among than within individuals. Call parameters and CVs were reported based on individuals’ mean value. Additionally, we performed model II ANOVAs on each call parameter, and reported the effect size (partial η²) for each ANOVA model to provide additional estimates of which call parameters vary more among individuals than within individuals (Beecher 1989; Bee et al. 2010; Pettitt et al. 2013). In total, six models per species were generated. The Model II ANOVAs were generated in R program by using the Anova function of car package v.3.1.1 (Fox et al. 2023).

Differences in calling site preferences

To investigate the calling site differences between species, we ran one-way ANOVAs on vertical height and distance to water by using the aov function in R program. Two models were generated. We obtained the pairwise differences from Tukey’s post hoc test by using the TukeyHSD function to evaluate the result from ANOVAs. For the categorical variables (substrate type and shelter type), we use Chi-squared test by using the chisq.test function. The dataset of calling sites and call SPLs was inconsistent with the call recording analyzed, as many calling individuals were observed but not recorded well (detailed in Supplementary Table S3). We generated boxplots for visualizing differences in vertical height and distance to water for calling sites by using the ggplot function in ggplot2 package v.3.4.1 (Wickham 2016).

Playback experiment result analysis

Since we got no responses from high-frequency stimulation groups (see results), we only analyzed if there are differences between similar-frequency stimulation groups and the control groups. We gathered results from stimulations of advertisement calls and aggressive calls together and categorized by species. Then, we got three datasets, results from the control groups (B. nanlinensis), results from stimulations of calls of B. ombrophila, and results that of B. lishuiensis. We compared the differences between experiment group and the control group separately. The comparisons were performed by Fisher’s exact test in R program by using the function fisher.test since there are samples smaller than 5. We expected that there are no significant differences due to our hypothesis.

All data analyses were performed in R program v.4.2.2 (R Development Core Team 2011).

Results

Effect of body size and temperature on advertisement calls

The effect of body size and temperature on six call parameters of each species are given in Tables 1–3. The linear mixed models showed that in B. nanlingensis, call duration could be predicted by both the body size (b = −4799.919, P < 0.05) and the temperature (b = −8.731, P < 0.001), call rate could also be predicted by both the body size (b = 96.426, P < 0.05) and the temperature (b = 0.450, P < 0.05), dominant frequency was negatively correlated with the body size (b = −29202.690, P < 0.05), call interval was negatively correlated with the temperature (b = −47.499, P < 0.001), and call number was positively correlated with the temperature (b = 0.450, P < 0.05). In B. shimentaina, only temperature predicted some temporal call parameters, which are negatively correlated with call interval (b = −4.632, P < 0.05) and pulse number (b = −0.674, P < 0.05), and positively correlated with call rate (b = 0.212, P < 0.05). Not as that in the above two species, call duration shows positively correlated with temperature in B. ombrophila (b = 1.658, P < 0.05).

Table 1.

Effect of body size and temperature on the acoustic parameters of the advertisement call of Boulenophrys nanlingensis based on LMMs

Call parameter	Predictor	Estimate	χ2	P	R 2 conditional	R 2 marginal
Call duration	Body size	−4,799.919	8.114	0.004	0.917	0.252
	Temperature	−8.731	52.507	<0.001	0.909	0.638
Call interval	Body size	−16,661.001	3.032	0.082	0.854	0.108
	Temperature	−47.499	96.015	<0.001	0.838	0.681
Dominant frequency	Body size	−29,202.690	4.910	0.027	0.684	0.130
	Temperature	6.348	0.145	0.703	0.676	0.004
Pulse number	Body size	264.283	1.366	0.242	0.678	0.042
	Temperature	−0.310	1.391	0.238	0.678	0.041
Call rate	Body size	96.426	4.254	0.039	0.977	0.158
	Temperature	0.235	51.577	<0.001	0.977	0.701
Call number	Body size	277.145	2.379	0.123	0.329	0.048
	Temperature	0.450	5.305	0.021	0.327	0.094

Numbers in bold indicate significant correlations between variables (P < 0.05).

Table 2.

Effect of body size and temperature on the acoustic parameters of the advertisement call of Boulenophrys ombrophila based on LMMs

Call parameter	Predictor	Estimate	χ2	P	R 2 conditional	R 2 marginal
Call duration	Body size	−441.692	0.721	0.396	0.122	0.007
	Temperature	1.658	4.058	0.044	0.118	0.031
Call interval	Body size	1,391.644	0.125	0.723	0.076	0.001
	Temperature	0.047	<0.001	0.995	0.076	<0.001
Dominant frequency	Body size	−17,902.380	1.264	0.261	0.425	0.040
	Temperature	43.117	2.474	0.116	0.423	0.074
Pulse number	Body size	−21.986	0.014	0.906	0.383	<0.001
	Temperature	−0.093	0.076	0.783	0.383	0.002
Call rate	Body size	−6.659	0.018	0.893	0.617	0.001
	Temperature	−0.038	0.180	0.671	0.616	0.010
Call number	Body size	−1,784.174	0.861	0.354	0.368	0.034
	Temperature	2.761	0.614	0.433	0.370	0.025

Numbers in bold indicate significant correlations between variables (P < 0.05).

Table 3.

Effect of body size and temperature on the acoustic parameters of the advertisement call of Boulenophrys shimentaina based on LMMs

Call parameter	Predictor	Estimate	χ2	P	R 2 conditional	R 2 marginal
Call duration	Body size	188.803	0.227	0.633	0.242	0.002
	Temperature	−0.979	1.102	0.294	0.241	0.008
Call interval	Body size	−192.116	0.065	0.799	0.366	0.001
	Temperature	−4.632	8.932	0.003	0.360	0.074
Dominant frequency	Body size	−8,215.264	0.672	0.413	0.407	0.007
	Temperature	0.842	0.001	0.973	0.409	0.000
Pulse number	Body size	180.343	2.829	0.093	0.441	0.029
	Temperature	−0.674	7.779	0.005	0.423	0.077
Call rate	Body size	17.333	0.281	0.596	0.633	0.004
	Temperature	0.212	7.701	0.006	0.626	0.127
Call number	Body size	185.769	2.614	0.106	0.174	0.023
	Temperature	−0.089	0.102	0.750	0.184	0.001

Numbers in bold indicate significant correlations between variables (P < 0.05).

CVs of call parameters

The CVs of call parameters of each species are given in Table 4. Dominant frequencies were static in all species, call rate was static in B. nanlingensis (mean CV_w = 4.1) but intermediate in B. B. ombrophila (6.5) and shimentaina (5.8), and call duration and pulse number were intermediate in B. nanlingensis (7.3 and 8.2, respectively) but dynamic in B. ombrophila (19.9 and 16.9) and B. shimentaina (16.5 and 21.0). All parameters exhibited CV_a larger than CV_w in B. nanlingensis (CV_a/CV_w = 1.03–3.48). However, this relationship was only present in call rate of B. ombrophila (1.36) and dominant frequency and call rate of B. shimentaina (1.06 and 1.32). The result from model II ANOVA further confirmed the estimates of CV_a/CV_w. The largest value of partial η² (0.933) occurred in call rate of B. nanlingensis, which also has the largest value of CV_a/CV_w (3.48). While the smallest value of partial η² (0.056) occurred in call interval of B. ombrophila, which has the smallest value of CV_a/CV_w (0.31).

Table 4.

Descriptive statistics and variation of temperature-adjusted acoustic parameters

Species	Call parameters	Mean	SD	Range	CVa	CVw	CVa/CVw	F	P	η2
B. nanlingensis	Call duration (ms)	121.3	15.3	98.3–155.3	12.6	7.3 (3.8–13.3)	1.73	184.731 (20, 1443)	<0.001	0.719
	Call interval (ms)	228.8	61.6	101.5–348.4	26.9	26.2 (7.3–90.4)	1.03	59.466 (20, 1338)	<0.001	0.471
	Dominant frequency (kHz)	3,423.3	211.9	3,075.5–4,014.0	6.2	3.5 (0.6–9.9)	1.77	140.966 (20, 1443)	<0.001	0.661
	Pulse number	25.6	3.5	21.2–34.1	13.5	8.2 (3.8–14.1)	1.64	163.457 (20, 1406)	<0.001	0.699
	Call rate (calls/s)	2.9	0.4	2.3–3.6	14.1	4.1 (2.8–9.1)	3.48	58.230 (20, 83)	<0.001	0.933
	Call number	14.6	3.6	9.0–26.0	24.6	21.1 (5.6–40.7)	1.16	2.571 (20, 83)	< 0.001	0.383
B. ombrophila	Call duration (ms)	47.3	2.9	44.1–54.4	6.1	19.9 (10.8–23.9)	0.31	24.567 (11, 2875)	<0.001	0.086
	Call interval (ms)	246.2	25.3	204.0–292.5	10.3	33.2 (16.3–61.2)	0.31	15.271 (11, 2815)	<0.001	0.056
	Dominant frequency (kHz)	3,133.2	102.7	2,994.2–3,361.6	3.4	3.6 (1.6–7.0)	0.95	165.287 (11, 2872)	<0.001	0.388
	Pulse number	8.6	1.2	6.8–11.3	13.9	16.9 (10.5–24.7)	0.82	101.665 (11, 2462)	<0.001	0.312
	Call rate (calls/s)	3.5	0.3	3.0–4.1	8.8	6.5 (2.9–11.1)	1.36	8.234 (11, 48)	<0.001	0.654
	Call number	47.8	12.8	32.8–76.2	26.8	29.9 (7.0–63.5)	0.90	3.605 (11, 48)	<0.001	0.452
B. shimentaina	Call duration (ms)	66.4	6.4	48.7–79.0	9.6	16.5 (10.2–28.6)	0.58	16.140 (34, 1408)	<0.001	0.280
	Call interval (ms)	84.4	10.6	64.8–119.9	12.6	16.8 (9.7–42.6)	0.75	15.193 (34, 1233)	<0.001	0.295
	Dominant frequency (kHz)	4,938.4	170.2	4,604.0–5,329.1	3.4	3.3 (0.6–13.4)	1.06	28.513 (34, 1408)	<0.001	0.408
	Pulse number	9.6	1.6	6.9–13.1	17.0	21.0 (12.0–37.4)	0.81	25.270(34, 1384)	<0.001	0.383
	Call rate (calls/s)	6.8	0.5	5.6–8.0	7.6	5.8 (1.4–15.4)	1.32	7.436 (34, 140)	<0.001	0.644
	Call number	8.2	1.9	5.4–14.0	23.1	31.2 (9.3–59.8)	0.74	2.063 (34, 140)	0.002	0.334

Numbers in bold indicate significant effects (P < 0.05).

Differences in calling site preferences

Among the three species, B. nanlingensis preferred the most to call near the stream, with a mean distance to water of 10 ± 16 cm (range 0–60 cm, N = 42), and at a relatively low position of vertical height at 9 ± 20 cm (range 0–80, N = 42). However, B. nanlingensis also has the loudest call of 93.2 ± 4.4 dB (89.4–99.6 dB, N = 4). Another species that preferred lower positions was B. ombrophila, which exhibited a vertical height of 7 ± 11 cm (0–50 cm, N = 45), but the distance from water was relatively far (93 ± 103 cm, 0–416 cm, N = 35). The call SPL of B. ombrophila was measured as 81.8 ± 2.8 dB (77.9–86.5 dB, N = 12). While B. shimentaina preferred to call at high perched positions, with a mean vertical height of 89 ± 71 cm (1–300 cm, N = 82), and a mean distance to water of 44 ± 59 cm (0–313 cm, N = 80). For the substrate type, 30 of 42 calling sites were on the floor, and 12 of 42 were elevated in B. nanlingensis, and that were 27 of 50 on the floor and 23 of 50 elevated for B. ombrophila, and 77 of 86 and 9 of 86 for B. shimentaina.

Figure 2 provided boxplots of distance to water and vertical heights of three species; Table 5 shows the Tukey’s post hot test results on the two variables. Except for B. nanlingensis and B. ombrophila showing non-significant differences in vertical heights of calling sites (P = 0.981), all other comparison pairs exhibit significant differences (P < 0.05 or P < 0.001). The Chi-squared test results on differences in substrate type (P < 0.001) and shelter type (P < 0.001) show significant differences among the three species (Table 6). The typically calling site categories are shown in Figure 3.

Figure 2.

Comparison of distance to the nearest water body (A) and vertical height (B) of calling sites of three Boulenophrys frogs. Significant differences were derived from Tukey’s HSD test with ANOVA (***P < 0.001, **P < 0.01, *P < 0.05).

Table 5.

Chi-squared test on substrate type and shelter type for calling site preferences

Variables	χ2	df	P
Substrate type	19.468	2	< 0.001
Shelter type	44.434	4	< 0.001

Numbers in bold indicate significant effects (P < 0.05).

Table 6.

Tukey’s Honestly Significant Difference on quantitively calling site data

Variables	Models	P	95% Confidence interval
Distance to water	B. ombrophila vs. B. nanlingensis	<0.001	51.698, 121.353
	B. shimentaina vs. B. nanlingensis	0.015	5.403, 62.795
	B. shimentaina vs. B. ombrophila	<0.001	−83.495, −21.358
Vertical height	B. ombrophila vs. B. nanlingensis	0.981	−27.904, 23.779
	B. shimentaina vs. B. nanlingensis	<0.001	5.403, 62.795
	B. shimentaina vs. B. ombrophila	<0.001	−83.495, −21.358

Numbers in bold indicate significant effects (P < 0.05).

Figure 3.

Typical calling sites of B. nanlingensis (A and D), B. ombrophila (B and E), and B. shimentaina (C and F).

Playback test results

Figure 4 shows the results from playback tests. A total of 11 tests were performed as control groups (calls of B. nanlingensis), and 5 responses were obtained. For the stimulations from calls of B. ombrophila, we have got 7 responses from 15 tests. The stimulations from calls of B. lishuiensis resulted 4 responses from 7 tests. No responses were obtained from simulations from calls of B. shimentaina and B. sanmingensis according to 6 and 4 tests, respectively. Table 7 shows stimulation groups with non-significant differences with the control group (both with P = 1.000), which means the playback stimulations based on recordings of B. ombrophila and B. lishuiensis could cause a similar effect as B.nanlingensis self’s calls on evoke male B. nanlingensis’s aggressive behavior.

Figure 4.

Playback test results comparing male B. nanlingensis in response to different types of stimulations.

Table 7.

Comparisons of playback results from different groups of stimulations by Fisher’s Exact test

Comparisons	P	95% Confidence interval
B. nanlingensis vs. B. ombrophila	1.000	0.152, 5.862
B. nanlingensis vs. B. lishuiensis	1.000	0.061, 6.035

Discussion

Call variation in three Boulenophrys frogs

In this study, we aim to find out the mechanism of acoustic segregation in three sympatric Asian horned frogs by accessing their call variation and calling site preferences, as well as using playback tests to figure out their acoustics interactions. We find that both temperature and body size could affect the call performance in Asian horned frogs. Previous studies have reported that the body condition of frogs at an increasing temperature could exhibit a higher metabolic rate that could affect call performance (Zweifel 1968; Gillooly and Ophir 2010; Ziegler et al. 2016; Ord and Stamps 2017). Thus, the positive correlation between temperature and temporal call parameters (call duration and interval) has been frequently reported (as reviewed in Gerhardt 1994). Poyarkov et al. (2017) first hypothesized that the cool weather could elongate the call duration and interval in Asian horned frogs based on the comparison of recordings of Ophryophryne elfina in February and April (11.3–17.5 °C). In this study, all temporal parameters (call duration, call interval, and call rate) of B. nanlingensis showed a significant relationship with temperature. Such a significant relationship was only detected from call interval and call rate in B. shimentaina, and there was even a reverse direction of significant correlation in call duration in B. ombrophila, although with a very low estimation. Such conditions were previously reported from B. boettgeri (under the name Xenophrys boettgeri; Wei et al. 2019), which did not show temperature correlations in call parameters. It could not be ignored that the relationship between temperature and call length (i.e., call duration and call interval) usually depends on low temperature (Bellis 1957; Zweifel 1959). Previous cases of call length affected by temperature in Ophryophryne and B. nanlingensis were based on recordings in relatively low temperatures (Poyarkov et al. 2017; Qian et al. 2023a; this study). In a previous study, Gunderson and Leal (2015) reported that physiological constraints limited the increase of call performance after reaching a particular temperature. Thus, the sampling effort of the latter three species conducted in a relatively narrower temperature range that was quite warm (24.3 ± 2.7 °C in Wei et al. 2019) could somewhat reach the limit of increasing metabolic rate, thus not showing a correlation of call parameters with temperature. The spectral parameter (i.e., dominant frequency) was supposed to be related to body size, as the mass of vocal cords generally correlates with body size (Walkowiak 2007). This relationship was quite common not only in frogs but also in all vertebrates (Gingras et al. 2013; Tonini et al. 2020; Zhao et al. 2021; Muñoz et al. 2023). In this study, all species had a positive relationship between body size and dominant frequencies, although these relationships were not always significant. Besides, although the relationship between body size and dominant frequency was consistent within species, it was obvious that the smaller B. ombrophila could emit calls with similar frequency to B. nanlingensis, which could be a type of allometric escape of body size (Tonini et al. 2020). After controlling the effect of temperature, results from CVs showed the dominant frequency was the most static parameter in each species, corresponding to a previous study in B. boettgeri (Wei et al. 2019) and that in most anurans, as reviewed by Köhler et al. (2017). In contrast, call rate was the second most (relative) static parameter in each species, and the largest value of CV_a/CV_w was always present in call rate. The call series of Asian horned frogs were emitted by several repeated expirations (Supplementary Figure S2), which depended on the ability of continuous contraction and relaxation of the trunk muscles. This suggested call rate could be another main clue of female selection when evaluating the potential partner’s body condition.

Interactions between call properties and the environment

Different Megophryinae species, especially sympatric species, always show distinct differentiation in calling sites. For example, Pope (1931) reported that B. boettgeri (under the name Megophrys boettgeri) from Wuyi Mountains always calls at “vantage points,” but not the sympatric Boulenophrys kuatunensis (under the name Megophrys kuatunensis). Being prolonged breeders, selecting a suitable calling site could benefit from maintaining a waterbody to spawn, promote the well-being of their offspring, and better transmit their calls (Wells 1977). The physical characteristics of the environment could affect the transmission of sound, and the environment noise will reduce the signal-to-noise ratio and affect the transmission efficiency of acoustic signals, thus not suitable for the receiver to get information from acoustic signals (Ryan and Brenowitz 1985; Wells 2007; Muñoz et al. 2020). The acoustic adaptation hypothesis predicted that the call evolves for better transmission, although this was rarely proven in anurans (Bosch and De la Riva 2004). Several field observations provide evidence that small species call on elevated perches to increase their broadcast range, not only in megophryinid frogs (Shi et al. 2020) but also in the neotropical genus Syrrhophus (Reyes-Velasco et al. 2015). A classic study by Wiley and Richards (1978) pointed out that (1) calls with low frequency have better transmission ability but are easy to disturb by the stream noise and (2) the decrease of high-frequency signals is lower in a high environment with empty air. In Amazon forests, frogs from open habitats call at higher frequencies than frogs from the forest habitat (Zimmerman 1983). In this study, B. shimentaina called at elevated positions, with a mostly open type, which corresponds well with point (2). In contrast, B. nanlingensis and B. ombrophila called at low positions, with relatively lower frequencies, perhaps corresponding with point (1). However, they selected different strategies to better transmit their calls. B. nanlingensis is relatively large-sized, indicating that they have larger resonance cavities to emit louder calls. To go against the stream noise, B. nanlingensis chose to call mostly in a crevice or under rock caves, with 18 of 42 calling sites classified as “burrow.” These habitats may amplify the sound by acting as a secondary resonator, such as in some Chinese music frogs (genus Nidirana) and South American Eupsophus frogs (Penna and Solís 1999; Penna 2004; Cui et al. 2012). Although with a relatively smaller body size and lower call volume, choosing to call relatively far from the stream could reduce the effect of the stream noise as much as possible in B. ombrophila. The difference in calling site preferences of anurans enhances the ability for species coexistence, which could be a type of niche partitioning. In some cases, this spatial separation of calling sites could be regarded as premating isolation role in species with similar calls (Hödl 1977; Ptacek 1992).

Temporal acoustic niche partitioning

As discussed by different authors, the acoustic niche partitioning was common in Asian horned frogs (Wang et al. 2014; Tapley et al. 2018, 2021; Shi et al. 2020; Wu et al. 2020). This results not only in separations by calling sites but also by call frequencies or calling phenology. In the playback experiments, the male B. nanlingensis was responsible for calls of B. ombrophila with similar frequencies despite having distinct differences in most call parameters (i.e., call duration, pulse number, call number, and call rate). That could be why the two species separated their calling phenology to avoid breeding interruptions (Qian et al. 2023a). Further testing with calls of B. lishuiensis, a species that does not occur in the B. nanlingensis distribution area, also results in responses. Although our playback tests were based on random field attempts with limited sampling size, such inter-specific interactions that could affect species co-existence should not be ignored. That mechanism was quite similar to that of female túngara frogs’ response to ancestral calls (Ryan and Rand 1995), as the similarity of call frequencies and call structure (i.e., repeated single-note calls) could be a type of original calls from a common ancestor. However, frogs in chorus could shield calls from non-suitable frequencies of other species (Lee et al. 2021). Thus, the two species with higher dominant frequencies could not evoke the response of male B. nanlingensis, enhancing the ability to coexist in breeding males of B. shimentaina and B. nanlingensis. As usual, the auditory awake of anurans was sensitive to congeneric species (Zhu et al. 2017; Gupta et al. 2021; Fan et al. 2022; Wang et al. 2023). Interspecific interactions were occasionally reported; Hylodes nasus and Crossodactylus gaudichaudii with similar frequency ranges displayed interspecific territorial behaviors (Wogel et al. 2004), and aggressive interactions were detected from sympatric neotropical Hyla species (Schwartz and Wells 1984; Reichert and Gerhardt 2014). In such situations, some species exhibit temporal partitioning of calling behavior to avoid acoustic disturbance (e.g., Littlejohn and Martin 1969). Because Asian horned frogs are prolonged breeders, males call in several months during a year. Acoustic interference could happen if species with similar frequencies call at the same time. Individuals who separated their breeding activities could gain more offspring in such cases. This led species with similar call frequencies to gain more fitness if their breeding seasons were separated. In contrast, the tadpole development cycle of Asian horned frogs is long, and all were reported to cross winter (Tapley et al. 2020; Qian et al. 2023b). It could be hypothesized that the selection of breeding season restricted by tadpole development is not an emergency need. In our examples from B. nanlingensis and B. ombrophila, two species share similar call frequencies that could cause acoustic disturbance, and relatively conflict calling habitats of similar distance to water. It could be a fitness to divide the acoustic space by avoiding breeding in the same season, in case it is a type of temporal partitioning.

Funding

This work was supported by the Project for Endangered Wildlife Investigation, Supervision and Industry Regulation of the National Forestry and Grassland Bureau of China (202407-HN-001), and the Project for Endangered Wildlife Protection of Hunan Forestry Bureau of China (HNYB-2024001).

Authors’ Contributions

T.Q., D.Y., and P.L. conceived the study. T.Q., Y.S., and W.Z. participated in the field work. T.Q. performed the experiments, analyzed the data, and drafted the manuscript. P.L. and D.Y. corrected the manuscript. All authors read and approved the final manuscript.

Ethics Statement

Animal capture and experiments adhered to the Guidelines for the treatment of animals in behavioral research and teaching (ASAB/ABS, 2022). We used a noninvasive method (body patterns) for individual identification, which was illustrated in Qian et al. (2023a). Frogs were released to their original habitats after measurements and photos. Fieldwork was approved by Hunan Mangshan National Nature Reserve under China’s wildlife protection law and regulations on the management of nature reserve.

Conflict of Interest

The authors declare that they have no conflict of interest.