Introduction

The acoustic niche integrates a threefold dimension: spatial (habitat used for calling; dos Santos Protázio et al. 2015), temporal (time when calling takes place; Llusia et al. 2013a), and acoustic (physical features of the call; Lima et al. 2019; Sinsch et al. 2012). In species-rich assemblages, it is expected that species’ acoustic niches be narrower and less overlapped to minimize acoustic interference, and thus facilitating mating of conspecifics by acoustic resource partitioning (Duellman and Pyles 1983; Heard et al. 2006; dos Santos Protázio et al. 2015). Moreover, species saturation in a community should generally lead to an even distribution of niche breadth (Tilman 2004). Thus, species adopt different calling strategies to avoid acoustic niche overlap during breeding period (Sinsch et al. 2012). Such strategies can involve differences in any of the three dimensions of acoustic niche, namely the calling site (Silva et al. 2008), the calling period (Duarte et al. 2019), and the acoustic signal, particularly between closely related species (Vieira et al. 2016).

Calling activity is a well-established indicator of the daily and seasonal patterns of reproduction in a variety of animal groups, such as anuran amphibians (Wells 2007). The timing, intensity and duration of calling and reproductive activities of anurans are mainly influenced by environmental factors (Llusia et al. 2013b; Heard et al. 2015; dos Santos Protázio et al. 2015; da Silva Ximenez and Tozetti 2015; Ulloa et al. 2019). Abiotic conditions, such as temperature, humidity and precipitation (Márquez 1992; Saenz et al. 2006; Llusia et al. 2013a), pond hydroperiod (Jakob et al. 2003; Cayuela et al. 2012), barometric pressure (Oseen and Wassersug 2002), and light intensity (Almeida-Gomes et al. 2007), are known to modulate phenology of anuran calling. In some species, calling activity can also vary along altitudinal and latitudinal gradients (Morrison and Hero 2003; Llusia et al. 2013b), or even due to the presence of closely related species reproducing in the same breeding site (Duellman and Pyles 1983; Márquez et al. 1993). Anurans of different climatic zones can follow different breeding strategies due to the variation in predictability of environmental factors (Llusia et al. 2013a, c). Species may vary their calling activity over the season and throughout the day (Heard et al. 2015; dos Santos Protázio et al. 2015; da Silva Ximenez and Tozetti 2015). On the daily cycle, there are daylight, crepuscular and nocturnal species (Almeida-Gomes et al. 2007; Wells 2007; Farina and James 2016; Duarte et al. 2019). At an even finer temporal scale, anurans may time the emission of their calls to avoid masking from heterospecific and conspecific callers (Gerhardt and Schwartz 1995). Therefore, a combination of social and environmental factors synergistically influences calling and reproductive activity.

Determining the factors that regulate the onset and intensity of reproduction can improve our understanding of the evolution of this key life-history trait. Moreover, since anuran amphibians are ectothermic organisms with permeable skin and largely aquatic reproduction (Duellman and Trueb 1994; Wells 2007), this comprehension is crucial for predicting how populations may deal with climate change (Chambers et al. 2013; Llusia et al. 2013a). Thus, through phenology of calling activity, it is possible to evaluate the adaptive capacity of species and their resilience, and to use this information for the management of threatened species (Chambers et al. 2013). Studies on phenological trends provide significant evidence of the effects of anthropogenic impacts on biota (Beaumont et al. 2015), and shed light on the constraints on species adaptation (Chambers et al. 2013). Although studies dealing with this question in tropical anuran assemblages have been largely conducted on a seasonal scale (e.g. Duellman and Pyles 1983; Bevier 1997; Moreira and Barreto 1997; Duarte et al. 2019), little is known about factors that may act on a daily scale (Navas 1996).

Here we studied the nightly variation in phenology (timing and duration) and intensity (relative abundance) of the calling activity of tropical anuran assemblages from Cerrado, Central Brazil, as well as the effects of weather conditions on temporal activity patterns. Specifically, we addressed the following questions: (1) Are there daily and seasonal differences in calling activity patterns among species? (2) Is there temporal overlap in the calling activity of the species and assemblages? (3) Do temperature and relative humidity influence the calling activity of the study species?

Methods

Study area

The tropical anuran assemblages were monitored in five water bodies within the municipality of Caldas Novas, state of Goiás, Brazil. Habitat features were similar in all the study sites, i.e. all permanent water bodies with presence of vegetation on the edges and connected with forested areas (see supplementary material S1). The vegetation of the region is characterized by the presence of phytophysiognomies of the Cerrado, including a range of savannas (cerrado sensu stricto, campo sujo, campo de murundus) and forest (cerradão and mata galeria), as well as areas dominated by pastures (sensu Eiten 1978). The soil type is red latosol and neosol with flat relief, and the flora is represented by herbs, with predominance of grasses, and trees varying from 3 to 5 m in height (Lima and Eterovick 2013). The shortest distance among the sampled sites was 1 km (Sites 3 and 4), and the farthest distance was 11.5 km (Sites 1 and 2). The local climate is tropical, classified as Aw in the Köppen classification system, with two well-defined seasons (dry winter and wet summer; Peel et al. 2007). The dry season, in the colder months, extends from May to September and the rainy season, in the hotter months, from October to April. Annual mean temperature is 24.8 °C, annual mean of relative humidity is 60%, and annual mean rainfall reaches 137.8 mm (from period 2007–2016; INMET 2017).

Sampling techniques

We conducted a series of acoustic surveys to count the number of calling males in each of the five study assemblages. Sites were sampled once per month, during six consecutive months, between October 2014 and March 2015, with a total of 30 sampling days for the monitoring period. Field work began 1 h before sunset (nearly 19:00 h) and finished after sunrise (nearly 07:00 h), and consisted in a survey every hour, so that a total of 14 surveys were carried out per day (14 h per day/site sampled), with a total sum of 420 sampling hours (5 sites × 6 days × 14 h). We chose to sample only between sunset and sunrise because most species that vocalize near aquatic environments in tropical and subtropical regions show nocturnal reproductive activity (e.g. Madalozzo et al. 2017; Duarte et al. 2019). During daytime visits to collect tadpoles in the five water bodies, only Pseudopaludicola mystacalis and some individuals of Leptodactylus cf. podicipinus (mainly on rainy days) were found actively calling. The time of sunset and sunrise was obtained from the Interactive National Observatory Yearbook (Observatório-Nacional 2014), following the convention adopted by the reference text Supplementary Explanatory Astronomical Ephemeris and Nautical Almanac (Urban and Seidelmann 2012). Sampling method was standardized, being always carried out by the same researcher (VG) and with the same protocol for each visit. The counts and identification of calling males were based on acoustic search by walking slowly along the complete extension of the water body, counting the number of all calling males of each species (Heyer et al. 1994). Approximately every 5 m, the researcher made a stop for about a minute to hear any calling male that was not registered, and to ensure that all males from each species were counted. The researcher took between 30 and 50 min during each acoustic survey, depending on the size of the sampled site. To count the individuals of Barycholos ternetzi, a forest species with direct development that reproduces in the leaf litter, a 25-m transect was also carried through the forested environment near the pond (mata galeria). Each transect had an average duration of 15 min and the researcher walked slowly counting all calling males of B. ternetzi. To confirm the identification of the species, a sample of the vocalizations of some individuals of each species was recorded and deposited in the sound library (FonoZoo.com) of the Museo Nacional de Ciencias Naturales (CSIC), Madrid, Spain (see supplementary material S2).

Air temperature and relative humidity at approximately 10 cm above the ground were registered using a digital thermohygrometer Incoterm (to the nearest 0.1 °C and 1% humidity). Individuals of every species were collected, euthanized with 5% lidocaine, fixed in 10% formalin solution and later stored in 70% alcohol. Voucher specimens are deposited in the Zoological Collection of the Federal University of Goiás (ZUFG) (see supplementary material S2).

Statistical analyses

Three types of analyses, based on different datasets, were performed to identify: (1) daily patterns of the calling activity of species and assemblages; (2) temporal overlap in the calling activity of species and assemblages; and (3) influence of daily periods and climatic factors on calling activity of species. First, temporal patterns of nightly calling activity at species and assemblage levels were identified using circular statistical analyses (Zar 1999) with ORIANA 4.0 software, particularly Rayleigh’s test (Kovach 2004). Two hourly response variables were applied to each level of analysis: (1) abundance and (2) presence/absence of each taxon for the species-level analysis; and (3) total abundance and (4) species richness for the assemblage-level analysis. The hour of the night was converted to angles varying from 0° to 330° (0° being midnight) and associated with each response variable. The primary parameters considered for this analysis were: (a) mean vector (µ), which corresponds to the mean period when most species were active or there was the highest activity rate of a given species; (b) circular standard deviation (SD); and (c) the r vector, which corresponds to the mean of the data concentration around the circle (day), ranging from 0 (dispersed data) to 1 (aggregated data in the same direction). A significant result in the Rayleigh’s test (P < 0.05) indicates that data were not uniformly distributed and the presence of a significant mean angle or hour of the night (Kovach 2004) in which calling activity of the studied species and assemblages reaches their peak.

Second, to assess temporal niche overlap among species within the studied assemblages, we used hourly species abundance per site. The basis of this analysis was the relative number of calling individuals for each anuran species during each hour interval. The statistical analysis was performed first for all species sampled and subsequently for species of the same genus, namely Dendropsophus, Boana and Leptodactylus. We also performed an analysis for all sampled months and another one using months divided by periods: (1) October and November, (2) December and January, and (3) February and March. Overlap was quantified as the average of all pair-wise overlap values calculated via the Czekanowski index (Feinsinger et al. 1981), following the analytical methods of Castro-Arellano et al. (2010). Null distributions of overlap values were generated using the randomization algorithm Rosario, which was designed specifically for use with interval data, in which the order of categories is important (Castro-Arellano et al. 2010). Rosario maintains the shape of the empirical activity distributions (i.e., temporal autocorrelation) for each species in the randomly generated matrices by shifting entire activity patterns a random number of intervals. For each analysis, overlap indices were calculated for 10,000 randomly generated matrices of temporal activity patterns. Significance was determined by comparing each empirical value to its associated null distribution. Analyses of niche overlap were conducted as two-tailed tests, which may detect a higher overlap in temporal niche than expected by chance alone (temporal coincidence) or a lower overlap than expected by chance alone (temporal segregation). Simulations for overlap in temporal activity were conducted with the TimeOverlap program (Castro-Arellano et al. 2010; software available for download at http://hydrodictyon.eeb.uconn.edu/people/willig/Research/activity%20pattern.html).

Finally, to examine the calling activity of anuran assemblages along the night, two response variables were considered: (1) species richness and (2) species diversity (Shannon index; Shannon and Weaver 1949). These variables were calculated hourly per site and sampling day, and the Shannon index was estimated from the hourly abundance of each species recorded. A Generalized linear mixed-effects model (hereafter GLMM) was performed for each response variable (richness and species diversity). In both models, the sampled time (14 h), air and water temperatures (°C) and relative air humidity (%) were used as fixed factors. Air and water temperature, and humidity were z-transformed. We also treated each site and day as random effects to address the potential non-independence of the observations (spatial and temporal correlation). First, to investigate the relationship between species richness and time along the night and abiotic factors (temperature and humidity), a first GLMM was set using binomial error structure and the logit link function. Such analysis was conducted assuming that all sites had equal chance to contain the same species number (19 spp.). Then we coded the response variable as a two-column matrix, the first column being the number of recorded species and the second one the number of absent species. For this model we separated the dataset of anuran assemblages in three seasons: early season (first two months, October and November); middle season (two middle months, December and January); and late season (last two months, February and March). The season period (three levels) was also used as fixed factor. Second, to verify how diversity (Shannon index) varied along the night, a second GLMM was set using Gaussian distribution and the restricted maximum likelihood function. For this model, we did not separate the anuran assemblages in three seasons.

GLMMs were fitted in R (R Core Team 2017) using the functions glmer (species richness model) and lmer (diversity Shannon index model) of the package lme4 (Bates et al. 2013). We established model inference by full-null model comparisons using a likelihood-ratio test with the R-function anova (Dobson 2002; Forstmeier and Schielzeth 2011). Null models were fitted using only the random effects and control fixed factors and then compared to the full models by applying maximum likelihood (Bolker et al. 2009). To determine the significant effect of individual predictors, P values were based on likelihood-ratio tests of the full model using the R-function drop1 (R Core Team 2017).

Before performing analyses, we conducted graphical data exploration to check for normality and homogeneity by visually inspecting probability plots (Q–Q plots) and the residuals plotted against fitted values (Bolker et al. 2009; Zuur et al. 2010). Multicollinearity was assessed with Generalized Variance Inflation Factors (GVIF; Fox and Monette 1992, Field et al. 2013) in R with the function vif (Fox and Weisberg 2011). The GVIF was performed from a standard linear model excluding the random effects and revealed absence of collinearity between fixed factors (< 2.5 in all cases). Regression diagnostics confirmed the absence of multicollinearity, autocorrelation and significant departures from normality in all final models. For all analyses, significance was set at a P value < 0.05.

Results

We monitored nightly variation in calling activity of a total of 19 anuran species in the five assemblages (see supplementary material S3). Circular statistical analysis revealed a nightly peak of total abundance (Rayleigh = 9668.30, µ = 334.69°, SD = 41.14°, r = 0.77, P < 0.01) and species richness (Rayleigh = 1008.04, µ = 347.12°, SD = 51.40°, r = 0.70, P < 0.01) in the anuran assemblages (Fig. 1), with most individuals and species calling in the time frame between one (~ 20:00 h) and 4 h (~ 23:00 h) after sunset. After this period, calling activity typically decreased following a progressive drop until dawn (~ 06:00 h; Fig. 2). Species with nocturnal activity showed activity peaks ranging from 20:00 to 01:00 h (Table 1; supplementary material S4). Only Pseudopaludicola mystacalis commonly exhibited crepuscular habits, with a rapid fall in calling activity after sunset (see panel m of the supplementary material S4). At species level, slight differences in nightly peak were found between the circular analysis using abundance and that using presence/absence data (see Table 1 and supplementary material S5). The peak of calling activity was identified to be almost 1 h later for presence/absence than for species abundance.

Fig. 1
figure 1

Rose-diagram of the circular analysis for a species richness (r = 0.70) and b total abundance (r = 0.77) of anuran species exhibiting calling activity along the night in the studied anuran assemblages (Caldas Novas, Goiás, Brazil). The arrow depicts the mean vector length (r), which indicates the concentration of species exhibiting calling activity along the hours of the night

Full size image
Fig. 2
figure 2

Average species richness along the night in the five studied anuran assemblages (municipality of Caldas Novas, State of Goiás, Brazil). The night time is indicated by the shaded background

Full size image
Table 1 Results of the circular statistical analysis of calling activity of each anuran species (in abundance) sampled in the assemblages in the municipality of Caldas Novas, Goiás, Brazil
Full size table

The temporal niche overlap of the species in the anuran assemblages was greater than expected by chance (coincident activities), considering all periods (Czekanowski Index = 0.14, P < 0.001) and each period separately (Period 1, Czekanowski Index = 0.14, P = 0.01; Period 2, Czekanowski Index = 0.17, P < 0.01; Period 3, Czekanowski Index = 0.13, P < 0.01). For congeneric species, namely Dendropsophus, Boana and Leptodactylus, no differences (segregated or coincident activities) were found (P value for Czekanowski > 0.05). Thus, no pattern of coincidence or avoidance in the calling activity throughout time was observed for these syntopic species.

As shown by the GLMMs of hourly calling activity, the two full models were different from the null models (species richness model, likelihood-ratio test: χ2 = 28.99, df = 3, P < 0.001; diversity Shannon model, likelihood-ratio test: χ2 = 14.69, df = 1, P < 0.001), revealing that richness and diversity were significantly influenced by fixed factors (Table 2). Hour had a negative significant effect on anuran richness (z = − 0.232, P < 0.001) and diversity (t = − 0.211, P < 0.001), and period influenced significantly the richness of anuran species on assemblages throughout the reproductive season (Table 2). Also, air and water temperature had a negative significant effect on calling anuran richness (z = − 0.383, P < 0.001; z = − 0.176, P < 0.001) and diversity (t = − 0.311, P < 0.001; t = − 0.162, P < 0.001), and humidity had a positive significant effect on calling anuran richness (z = 0.265, P < 0.001) and abundance (t = 0.179, P < 0.001).

Table 2 Results of the general linear mixed-effects models testing the effects of abiotic variables, period and time on richness and abundance (Shannon Diversity Index) of anuran assemblages in Cerrado, Brazil
Full size table

Discussion

Acoustic monitoring of tropical anuran assemblages revealed that calling activity was strongly influenced by weather conditions and followed a similar pattern at both species and assemblage level. Few studies have previously examined the daily variations in calling activity of anuran communities in tropical regions (e.g. Duellman and Pyles 1983; Bevier 1997; Moreira and Barreto 1997; Duarte et al. 2019). Studying calling activity along the night in several species, Bevier (1997) found that calling rates declined throughout the evening, and most species stopped calling shortly after midnight (around 5 h after sunset), in agreement with our findings. The depletion of energy reserves may contribute to this decrease (Brepson et al. 2013), since sound production in anurans is highly costly in relation to metabolic rates and oxygen consumption (Wells and Taigen 1989; Wells 2007). Moreover, in prolonged breeders (sensu Wells 1977), some calling males may also reduce or cease calling activity to save energy reserves for consecutive nights (Castellano and Gamba 2011). Nevertheless, factors other than energetic costs have also been evoked to explain this temporal activity pattern, such as nightly decrease in temperatures or the variation in the timing of females of each species arriving at the breeding site (Bevier 1997; Dias et al. 2017). Thus, many factors potentially influence male behavior, including energy reserves, weather, predation risk and probability of finding food (McCauley et al. 2000; Wells 2007; Llusia et al. 2013a).

As found by other studies in tropical regions, most anuran species exhibit predominantly nocturnal activity (Duellman and Pyles 1983; Bevier 1997; Moreira and Barreto 1997; Duarte et al. 2019). After sunset, air temperature typically decreases and air relative humidity increases, making the night period most suitable for the activity of animals with permeable skins, such as anuran amphibians. Furthermore, during the night, anurans may avoid visually oriented predators, such as diurnal bird species, and also do not compete for the acoustic space with other animal groups (Farina and James 2016). Thus, although it is expected that species adjust the period of reproductive activity to avoid competition (Kopp et al. 2010), suitable conditions within a particular nightly time window seem to favor such male aggregations at anuran assemblages (Llusia et al. 2013c; Ulloa et al. 2019), composed by large numbers of both conspecific and heterospecific calling individuals. In such noisy environments, females present evolutionary adaptations to identify conspecific males within the choruses (Vélez et al. 2013), with high density of species vocalizing at the same time.

The high overlap in temporal acoustic niche unveiled at assemblage level suggests a relaxed competition in this dimension of the acoustic space. Species likely overcome such niche overlap and avoid masking interference by presenting variation in other dimensions of the acoustic niche, such as the spatial (using different calling sites) and acoustic dimension (physical proprieties of the call or fine-adjustments of timing of calls; Duellman and Pyles 1983; Gerhardt and Schwartz 1995; dos Santos Protázio et al. 2015; Vieira et al. 2016). Besides, no temporal pattern of coincidence or avoidance in the calling activity throughout the night was found for syntopic congeneric species. Therefore, even if phylogenetic effects influence some aspects of the acoustic niche (Goicoechea et al. 2010), ecological factors may also lead to decreasing competition (Leite-Filho et al. 2015; Caldas et al. 2019).

At a specific level, studied anurans presented small variations in the peak of calling activity. During the reproductive season, species may avoid masking interference by being active at different hours of the day (Bridges and Dorcas 2000). For example, each Leptodactylus species registered showed a different nightly peak of calling activity. In addition to differences in timing, congeneric species may also show marked differences in some acoustic parameters, avoiding acoustic niche overlap, such as species of Boana that vocalize in sympatry (Boana lundii and B. raniceps—Guimarães et al. 2001; Guimarães and Bastos 2003; Boana albopunctata and B. paranaiba—Vieira et al. 2016). Moreover, aquatic environments in the Cerrado have a high environmental heterogeneity and therefore provide many available microhabitats, allowing the co-occurrence of species that have great similarity in the use of resources (De Marco et al. 2014; Gambale et al. 2014).

Anuran reproduction in the Cerrado biome is markedly seasonal, increasing in spring or early summer when rain is abundant, and decreasing throughout the rest of the season, with most species having completed their reproduction (Moreira and Barreto 1997; Oda et al. 2009; Kopp et al. 2010). Our results corroborate this pattern, with an increase of species during the second studied period (December and January) and then a decrease in the third one (February and March). Although anurans living in tropical species-rich communities may experience competition at both temporal and spatial scales, attending chorus is favored for several reasons. Choruses usually are formed in areas with suitable environmental conditions required by females and offspring, and occur during seasonal and daily periods that are favorable for courtship and reproduction (Gerhardt and Huber 2002; Wells 2007; Ulloa et al. 2019). Species that aggregate in choruses increase the probability of mate attraction, though also increase competition among males and the predation risk by acoustically oriented predators (Tuttle and Ryan 1981). Choruses are thus focal points for sexual and natural selection (Gerhardt and Huber 2002; Wells 2007).

Overall, timing and duration of anuran calling activity are highly dependent on environmental cues (Heard et al. 2015; dos Santos Protázio et al. 2015; da Silva Ximenez and Tozetti 2015; Duarte et al. 2019). As ectotherms, temperature is the main abiotic factor that determines calling activity in many anuran species (Saenz et al. 2006; Wells 2007; Lowe et al. 2016). Water temperature is crucial for species that call from or near the water (e.g. Gambale and Bastos 2014; Oseen and Wassersug 2002), while changes in air temperature are likely key for other species (Llusia et al. 2013a, c). Air relative humidity is also relevant due to the sensitivity of anurans to desiccation, especially for species that lay their eggs out of water, such as Phitecopus hypocondrialis (in leaves above water; Dias et al. 2017), or those with direct development, such as Barycholos ternetzi (eggs in leaf-litter of the forest floor; Caramaschi and Pombal-Jr 2001). Thus, both temperature and relative humidity modulate calling activity and reproduction of most anuran species, and thereby alterations in climate regimes related to climate change might adversely affect their acoustic communication and reproduction (Llusia et al. 2013b; Mainwaring et al. 2017).

Temporal patterns of species richness and diversity were influenced by similar environmental factors. Studies based on species surveys without standardization and ignoring the time of activity of the species that may potentially occur in a given area may reveal low detection probabilities of focal species, leading investigators to the false conclusion that some species are rare or absent (Heard et al. 2006; Schmidt and Pellet 2009). The use of different sampling techniques can also influence species detection (Madalozzo et al. 2017). A recent automatic recording technique (Passive Acoustic Monitoring, PAM) has been widely used to sample and measure the phenology of acoustically active animals (Madalozzo et al. 2017; Sugai and Llusia 2019). Although PAM has become a standard technique, it has some limitations, such as not being able to estimate the total number of individuals of certain species (e.g. Pellet et al. 2007), especially in tropical anuran communities, which contain species with high abundance and some of them with very similar calls. Our study showed that the use of species richness may be as efficient as diversity indices based on abundance data. Thus, presence/absence data obtained by automated digital recording equipments are expected to be effective in assessing community patterns in studies of long sampling periods (Madalozzo et al. 2017).

In this study, we verified that the circadian rhythms of calling activity in anurans of the Cerrado are highly influenced by climatic factors such as temperature and relative air humidity, and that there is no segregation pattern at assemblage or genus levels in the calling activity along the night. The lack of structure in the sharing of the time resource indicates that the competitive forces acting between species are not sufficiently strong, suggesting that the use of the acoustic temporal niche is not a limiting resource for Cerrado anurans. To better understand phenological patterns in species activity other factors such as photoperiod (Jaeger and Hailman 1981), social factors (Höbel 2017) and anthropogenic noise (Vélez et al. 2013) deserve further studies. We also show that the best hours to monitor anuran assemblages in Cerrado would be between 2 and 3 h after sunset in order to ensure maximum detection probabilities for anuran species in these communities, highlighting the influence of sampling time in species detection. Phenological studies allow a better understanding of how species would respond to future climate changes (e.g. Klaus and Lougheed 2013), in addition to improving the knowledge of the biology and ecology of each species.