Eleutherodactylus frogs show frequency but no temporal partitioning: implications for the acoustic niche hypothesis

Temporal partitioning

The pattern of calling activity was estimated using the Amphibian Calling Index (ACI) (Nelson & Graves, 2004). The pattern during the night for six of the eight species was significantly different at some part of the night (Table 2). Five species had their peak of activity between sunset and midnight: E. coqui, E. hedricki, E. portoricensis, E. richmondi, and E. wightmanae, measured as a higher proportion of samples with ACI values of 2 or 3 (Fig. 2). The calling activity of these five species was highest between 1900 and 2100 h and declined steadily after midnight. In the case of E. gryllus, the species had a short peak of activity between sunset to about 2000 h. Two species, E. locustus and E. unicolor, showed no significant difference in their activity during the night.

Three species had a small peak of activity during the last hours of the night (Fig. 2). The species E. gryllus, E. portoricensis, and E. wightmanae increased their activity two hours before sunrise from the declined activity level of the hours after midnight. These peaks were smaller than the main peak before midnight.

To illustrate some of these patterns, an example series of audio files recorded every hour, between sunset and sunrise, is available as Dataset S1.

Acoustic frequency partitioning

From the whole dataset, the 64 recordings made at 2000 h were analyzed to determine acoustic partitioning. In all sites with more than one species, most species exhibited frequency partitioning (Fig. 3), where the frequency range of their signals did not overlap. Between E. coqui and E. portoricensis there was a large overlap in one of the notes. Their first note, “co”, did not overlap. Their second note, “qui”, showed an overlap that averaged 63.9% (54.5–85.9%) of the frequency range of the note of E. coqui and 57.8% (45.2–80.3%) of the frequency range of the note of E. portoricensis (n = 6).

Two other cases showed some overlap to a lesser degree. The calls of between E. wightmanae and E. richmondi at the Carite State Forest (Site 14 in Table 1) overlapped 42.0% (10.3–83.0%) and 11.5% (4.2–21.0%) respectively (n = 5). At the El Yunque National Forest, the call of E. hedricki and the “qui” note of E. portoricensis showed some overlap at the Tradewinds Trail sites (Sites 6 and 7 in Table 1). The overlap in the call of E. hedricki was 9.7% (0–17.9%) and for the “qui” note of E. portoricensis was 7.0% (0–10.9%).

Temporal partitioning

The results from this study indicated that the level of activity of six of eight highland Eleutherodactylus species studied was not uniform during the night. Most species called from sunset to midnight, with a peak around 2000 h. These results suggest that this anuran community does not have temporal partitioning of their calling activity during the night. Seasonal differences will need to be studied using long-term datasets.

A previous study found some temporal partitioning in several species at El Yunque National Forest (Drewry & Rand, 1983). In particular, two species in that study, E. portoricensis and E. richmondi, were calling later that the other species in the night. However, the data used in that study were collected using different methods and from field notes, not from a systematic survey during the night. Another study using ADRS in a palustrine herbaceous wetland in Puerto Rico also found temporal clustering in four Eleutherodactylus, among them the generalist E. coqui. The four species at this site exhibited a peak of calling activity also at 2000 h (Ríos-López & Villanueva-Rivera, 2013), which is the first period of complete darkness. The sunset in Puerto Rico happens between 1800 and 1900 h (Ríos-López & Villanueva-Rivera, 2013).

It was expected that species should limit their calling activity to a period when its benefits (attracting females) are outweighed by its costs (energy expenditure, reduced foraging, and predation risk). Reproductive success of E. coqui is determined only by calling effort (Townsend & Stewart, 1994). In several studies that compare the energetic cost of calling, the metabolic rate increases up to ten times, making it a very energetically expensive activity (Gerhardt, 1994; Wells, 2001). In E. coqui, males reduce the number of prey they consume while calling (Woolbright & Stewart, 1987) and the energy requirements of calling stop their growth (Woolbright, 1989). Predation on Eleutherodactylus by owls (Megascops nudipes) has been reported (Zelick & Narins, 1982), so it is possible that these predators may sometimes use the call of the males to hunt them.

Acoustic surveys are a standard method for anurans (Zimmerman, 1994; Rödel & Ernst, 2004; Dorcas et al., 2009), but communities with temporal clustering present some problems. Very loud species, like E. coqui, can mask other species present at the sites (Villanueva-Rivera, 2007). Audio recordings can be a better method for monitoring these species, in particular when combined with automated identification (Aide et al., 2013). Results from this study suggest that acoustical monitoring of Puerto Rican Eleutherodactylus species should take place from sunset to midnight, when most of the species are highly active, with recorders to reduce false negatives due to masking. Furthermore, surveys conducted after midnight should be avoided as low calling activity levels could be due to the time and not a local extinction or a declining population. Special attention should be given to cases like E. gryllus, with a very short peak of calling activity limited to the first two hours of the night.

Acoustic frequency partitioning

The acoustic community of Eleutherodactylus species exhibited partitioning in the acoustic frequency of their sound signals. The calls of E. wightmanae and E. richmondi showed some overlap, however the calls are very different, which could reduce the pressure for frequency displacement. The call of E. wightmanae is a repetition series of a note while the call of E. richmondi is a short click with a very broad range in frequency (Rivero, 1998). In the other case of overlap, the species have spatial partitioning. Eleutherodactylus portoricensis is found in the understory up to 3 m of the ground, while E. hedricki only calls from holes in old branches near the canopy (Stewart & Woolbright, 1996).

The large overlap in the second note, “qui”, of the call of Eleutherodactylus coqui and E. portoricensis found in this study deserves further study, in particular because the species do not show temporal or spatial partitioning. Both species partition the frequency space of their “co” note. In invasive E. coqui populations in Hawaii, the “co” has very little inter-individual variation, while the “qui” note seems to be more variable (Benevides & Mautz, 2013). This high variability and overlap between sympatric E. coqui and E. portoricensis could indicate that this is not an important signal for distance communication since evolution has not separated this signal as the others. The two-note call of E. coqui has been studied in some detail. The first note, “co”, seems to be important to maintain distance between calling males and to establish their territory, while the second one, “qui”, is used to attract females (Narins & Capranica, 1976; Narins & Capranica, 1978; Zelick & Narins, 1982).

A study that tested the effect of two levels of density of sympatric E. portoricensis on the dominant frequency of the notes of E. coqui found that the “co” did not change, while the acoustic frequency of the “qui” was lower in sites with high densities of E. portoricensis (Luther et al., 2012). E. coqui might be trying to avoid masking of the noise or the higher densities of E. portoricensis could be triggering a suppression of the call in individuals with higher overlap in the frequency range (Zelick & Narins, 1982). The mechanism that is driving this effect of lower frequency could be studied by determining whether the females select males with lower frequency due to masking or if males that call at higher frequencies have less reproductive success because they are suppressed by the heterospecific call. In Hyla cinerea, the presence of H. gratiosa was related to displacement in female preference and in the advertisement call of the males (Höbel & Gerhardt, 2003). However, a study with sympatric populations of two Pseudacris species showed that the character displacement was variable among sites (Lemmon, 2009).

Several studies seem to indicate that acoustic frequency partitioning in anuran communities is not common. In a review of 11 assemblages, only 3 showed acoustic partitioning (Chek, Bogart & Lougheed, 2003). However, there was no data on temporal or spatial partitioning in most assemblages and some included data from a large geographical region. Other factors can make it harder to study, including separation of the acoustic frequency due to factors other than competition for the acoustic resource (Gerhardt & Schwartz, 1995).

A null model of a 7-species community in a pond in Costa Rica did not find differences in the frequency partitioning with random assemblages (Guyer & Donnelly, 2005). However, this study documented 28 combinations of up to 6 species during 47 sampling nights. This large variability of species may not be enough selective pressure to induce displacement of the acoustic frequencies (Pfennig & Pfennig, 2009).

In a community of five anurans in the Andes of Colombia, three of them Eleutherodactylus, the four species that were nocturnal had their peak of calling activity between one and two hours after sunset (Lüddecke et al., 2000). The species partitioned both the calling sites used and the acoustic frequency range (Lüddecke et al., 2000). In a Thailand assemblage of 11 species in 3 families, a study found partitioning in acoustic frequency, timing and space (Garcia-Rutledge & Narins, 2001).

In a community of 13 species in Brazil studied in permanent and temporary ponds and swamps, species that did not partition in space, partitioned their acoustic signals (Santos & Rossa-Feres, 2007). In turn, species that had similar calls partitioned their use of space (Santos & Rossa-Feres, 2007).