Unusual call characteristics and acoustic niche adaptation of Limnonectes larvaepartus Iskandar, Evans \ Mcguire, 2014 in its habitat (Anura: Dicroglossidae)

The advertisement call of Limnonectes larvaepartus is unique among frogs of the genus Limnonectes which have a true acoustic organ. The calls of three adult males L. larvaepartus were recorded at the foothill of Hutadelita Mountain (N 000 48’ 48.3” E 1220 23’ 00.8”; 396 m asl), Nantu Wildlife Sanctuary, Gorontalo, northern Sulawesi by using an Olympus LS-11 recorder with built-in microphones. Air temperature during recording was around 26°C. Adobe Audition 3.0 software was used to visualize and analyze the calls. The advertisement calls consist of 1-6 pulses, though only one pulse is usually produced. The wave structure of the call of L. laervaepartus is unusual because it is characterized by a short series of pulses that form a pulse train. Each pulse from the beginning of the call until the middle of the call has its own frequency spectrum, whereas the pulse train during the second half of the call becomes tightly compressed and forms one frequency spectrum. Calls of L. larvaepartus do not have a dominant frequency. Minimum frequencies decrease gradually from the beginning to the end of the pulse train; however, the maximum frequency rises gradually. Ascending frequency characters in one pulse also occur in the energy spectrum and bandwidth of frequency. The lowest minimum frequency at the end of pulse train is ~165 Hz; while the highest maximum frequency is also located at the end of the pulse train and is ~6860 Hz. The short pulses call of L. larvaepertus is similar to that of L. hasceanus, although the two species are not close relatives; in L. larvaepartus, the pulse only consists of one period, whereas in L. haschenus it consists of two periods. Considering acoustic adaptation, call of L. larvaepartus has similar frequency spectrum with call of L. modestus, we assume that they would not call at the same time and at the same localities because of their similar acoustic niches.

Based on our observations of specimens of Limnonectes housed in the Museum Zoologicum Bogoriense (MZB), Sulawesi is the island where mostly of Limnonectes spp have true advertisement calls, due to possession of vocal sacs. Most of Limnonectes species (described and undescribed) that are found on Sulawesi have vocal sacs, except for Limnonectes sp i which is included in L. grunniens group and L. microtympanum. As a comparison between Sulawesi and the large surrounding islands, there is only one species of Limnonectes on Java which has true advertisement calls, L. microdiscus [12], while L. kuhlii produces its call by passing air from its lungs through its buccal cavity because it does not have a vocal sac. In Kalimantan, there are two species that have true advertisement calls, i.e. L. finchi and L. palavanensis [13]; but another species, L. hikidae, which is in the same group with L. kuhlii does not have vocal sac [5]. In Sumatra, only L. microdiscus has true advertisement calls; whereas L. dammermani and L. kadarsani from the Lesser Sunda Islands were found to produce advertisement calls. Among described species of Limnonectes that produce sound in Sulawesi, only two species have described advertisement calls, L. arathooni by [14] and L. modestus by Kurniati & Hamidy [15].
Limnonectes larvaepartus is a very special type of frog with a unique reproductive mode ( Figure 1) involving internal fertilization and birth of tadpoles [10]; [16]. The distribution of L. larvaepartus overlaps broadly with L. modestus [10]; [11], with both species found from the northern part of Sulawesi to the western part of central Sulawesi. However, based on microhabitat type, L. modestus inhabits rocky big river with strong water flow, while L. larvaepartus has tendency to inhabit slower side streams and seep areas [10]. Males of L. modestus and L. larvaepartus produce each advertisement calls; with the call of L. modestus already described by Kurniati & Hamidy [15], and the call of L. larvaepartus not yet characterized.

Material and Methods
Calls of three adult male L. larvaepartus were recorded on the lower slope of Mount Hutadelita (N 000 48' 48,3": E 1220 23' 00,8";; 396 asl), Nantu Wildlife Sanctuary, Gorontalo, northern Sulawesi ( Figure 2). The frogs were recorded on 2 October, 2017 using an Olympus LS-11 recorder with built-in microphones. The recorder was set by using a sampling frequency of 94 kiloHertz (kHz) and a bit rate of 24 bits in WAV format. Air temperature during recording was ~26°C. The length of time for call recording for each individual was about five minutes. Adobe Audition 3.0 software was used to visualize and analyze the calls. Recorded calls were first normalized at -1 decibel (db); due to stereo recording that displayed double visualization of sound wave at the software, then the sound files were converted to 48000 Hz and 16 bits for mono visualization to ensure highest possible recording quality for frequencies up to 24 kHz (https://helpx.adobe.com/audition/using/convertingsample-types.html). Adobe Audition 3.0 software was also used to describe the oscillogram and spectrogram of the call type of L. larvaepartus. Visualization process of sound waves in the form of oscillograms, spectrograms and the dominant frequency were obtained using FFT (Fast Fourier Transformation; 1024 points) using a Hanning window. The terms used in the description of the calls follow Kohler et. al. [17], including pulse duration, call duration, inter-pulse interval, pulse rate, minimum frequency, maximum frequency and bandwidth. The minimum frequency, maximum frequency and bandwidth were sampled at three points of the pulse train frequency spectrum, i.e. at the beginning, middle, and end of the pulse.

Results and Discussion
A total of 31 advertisement calls were recorded for the three males [(Individual A; MZB Amph. 30503; SVL 42.70 mm; 3 calls); individuals B (MZB Amph. 30504; SVL 40.81 mm; 15 calls); individuals C (MZB Amph. 30505; SVL 40.62 mm; 13 calls)]. The tempo of the vocalizations that were produced by L. larvaepartus during the recordings were strongly influenced by the presence of nearby humans, and even a 5-minute recording time for one individual did not result in the male generating his calls following a regular rhythm. Therefore, call rate cannot be determined from this study.
The type of wave from the call of L. larvaepartus is a pulse ( Figure 3); with one advertisement call consisting of 1-6 pulses. A total of 31 advertisement calls from three males consisted of 10 advertisement calls with only one pulse (32%), 4 advertisement calls that included two pulses (13%), 6 advertisement calls that included three pulses (19%), 5 advertisement calls that included four pulses (16%), 4 advertisement calls that included five pulses (13%), and 2 advertisement calls composed of six pulses (6%); therefore the number of pulses in one advertisement call varies greatly.

Oscillograms
The sound wave structure of the call of L. larvaepartus ( Figure 4) is very different from the sound wave structure of the call of the other Limnonectes that have pulse type calls, including L. dabanus [18] and L. modestus [15]. One call of L. larvaepartus consists of a series of short pulses, with each short pulse formed by one wave period. The interval duration between pulses inside the pulse train was between 1-6 milliseconds (ms) at the beginning to the middle of the pulse train, followed by the short pulse becoming more compressed at the end of the pulse train without interval.
The number of pulse trains in one advertisement call varies greatly; each male can release 1-6 pulse trains in one advertisement call. When there were more pulse trains in an advertisement call, the duration was longer. The duration of an advertisement call that consists of six pulse trains can reach 450 ms, while the duration of a call that is comprised of one pulse train varies considerably; the shortest duration was ~64 ms.

Spectogram
The spectrogram of the advertisement call of L. larvaepartus ( Figure 5) is also very different from the spectrogram of the calls of other Limnonectes species. Each pulse has a minimum frequency, a maximum frequency and a frequency spectrum of its own frequency which is different from that of another pulse in a pulse train, whereby the pulse attached to the back of the pulse train forms frequency energy with the widest bandwidth. The minimum frequencies decreased from the beginning to the end of the spectrum (see Figure 5, Figure 6 and Table 1), while the maximum frequency and the frequency energy spectrum increased gradually.
Of the 31 calls obtained from three males, there was no flat minimum frequency and flat top frequency in a single pulse train. The bandwidth of one pulse train showed that from the beginning to the end of the wave the bandwidth was becoming enlarged at the end ( Figure  5 and Table 1). The average bandwidth of the three males showed different frequency values; individual A showed that the bandwidth at three frequency data retrieval points (beginning, middle and end) had the lowest frequency value compared to the other two males. However, the pulse spectrum energy of the L. larvaepartus's call did not show any dominant frequency; the peaks of the spectrum look almost parallel to each other. The spectral peak sequence will increase as the width of the frequency band increases. Table 1. Average value, Standard Deviation (SD), range and bandwidth average from three minimum frequency data of retrieval points and three maximum frequency data of retrieval points over one pulse train of three males of Limnonectes larvaepartus. Pulse train rate is quite variable from the three male individuals; the value is between 10.7/sec and 12.7/sec. Each male did not show a different pulse train rate with each other. Call rate cannot be analyzed in these recordings, since the call duration that were produced by the three males were probably influenced by the presence of the person performing the sound recording, so the interval duration among the advertisement calls varied greatly.   6. The average value of the three minimum frequencies from data retrieval points (above) and the three maximum frequencies from data retrieval points (below) for a single pulse train of individual C of Limnonectes larvaepartus.

Unusual call characteristic
The sound wave type of the advertisement call of L. larvaepartus is formed by a short pulse sequence, each short pulse producing its own frequency. The oscillogram and spectrogram bandwidth spectrum of L. larvaepartus resemble the sound wave type of the advertisement call of L. hascheanus (Figure 7). The call of L. hascheanus was described by Ohler et.al. [19], but the detailed structure of the sound wave was not clear. The difference between the short pulse of L. larvaepertus and the short pulse of L. hascheanus is in the number of periods that form a pulse; in L. larvaepartus the pulse only consists of one period, whereas in L. haschenus it consists of two periods ( Figure 8). The sound wave type of L. hascheanus and L. larvaepartus calls is unusual and very simple when compared with the call sound wave type of L. arathooni [14], L. dabanus [18], L. microdiscus [12], L. palavanensis [13], and members of the Limnonectes group lacking vocal sacs, i.e. L. fujianensis [20], L. hikidai [5], and L. kuhlii [12]. Although the calls of L. larveapartus and L. hascheanus are similar to one another, these species are not closely related [9]; [11] and this similarity therefore might reflect convergent evolution.

Acoustic niche adaptation
The distribution of L. larvaepartus and L. modestus is regionally sympatric [10], but when observed from the microhabitat scale, these two frogs tend to live in different types of habitat. Limnonectes modestus occupies larger rocky river habitats with strong water flow, while the L. larvaepartus typically inhabits narrower and shallower rocky streams and seeps with very slow water current. The spread of the microhabitat scale of these two frogs on the Adudu River was clearly allopatric [15]; L. modestus dominated the fast-flowing Adudu River habitat, while the L. larvaepartus dominated the upstream habitat of Adudu River, the distance the streams of about 100 meters from the