Vocalization Behavior of Chinese Bahaba (Bahaba taipingensis) during the Reproduction Season

Round 1

Reviewer 1 Report

General considerations:

The manuscript describes and characterizes 2 types of sounds made by Chinese bahaba during a reproduction period. Also, the temporal distribution of these sounds is presented along the different hours of the day.

In general, the referee thinks that there are a few days of study to be able to present sounds. It would be convenient to do another recording without reproduction or spawning. The frequency and duration characteristics of the LD (long druming)  and SD (short drumming) are different from when they are in reproduction and whether not. Would circadian behaviors also change?  The authors should consider if it is necessary to talk about direct sounds and the echoes in the water tank, some comments are about this aspect.   Specific comments of the paper:
What is the dynamic range of the device's analog-to-digital converter? Does it capture signals between -1 and 1 volt? Figures 2 and 4 (you can see echoes in them) do not contribute much, in my opinion, they are not necessary, since the same information is shown much better in figures 3 and 5.
In the case of SD sounds, have the echoes generated by the tank been taken into account?  They are low frequency impulsive sounds that spread easily through the water tank and discerning between the echoes becomes not easy. Is this taken into account when performing the analysis? nothing is said in the text.   Show real oscillograms SD and LD from acquiring signals in figures 3a and 5a. Figures 3b and 5b should include the parameters by which the fft has been calculated.    Assuming that the echoes have not been taken into account, given the large difference in ICI between the SDs that form an SD train, it should be considered not to treat them as an SP train. The number of SPs (NP) that form a train varies excessively, from 2 to 47. If there are 207 SD pulses within 96 trains. Most of the NP in the trains are formed by 2 SD. What is NP mode, median and mean? Perhaps the statistics have not been dealt with conveniently. In this case, eliminate NP and PP rows in table 2.   The analysis of daily (circadian) patterns is performed without taking into account that they are trains of SD. The duration of the LPs goes from 13 cycles to 30, is this the case or does the echoes make the LPs last longer? The 13 cycles of an 80Hz wave is 0.162.5 s and 30 cycles is 0.375 s.    The significance level in Figure 6 is 0.05 but in the discussion is 0.01. Which is correct?   Highlighted in red. Which is the duration of the sound when there is no purpose of breeding? Please, indicate the proper bibliography and comment.   Highlighted in yellow. English changes required   The referee warmly invites the authors to elaborate more on these points.

Author Response

====================   To Reviewer 1   ====================

March 14, 2023

 

To reviewer # 1

Thank you for your time and efforts spent on our manuscript. Your professional comments are helpful to improve our manuscript. We have substantially revised the manuscript based on your comments. Both the changes and those removed from the original manuscript were marked using “Track Changes” function in revised manuscript in .doc text. Our responses to your comments are listed as follows:

 

Comment 1: In general, the referee thinks that there are a few days of study to be able to present sounds. It would be convenient to do another recording without reproduction or spawning. The frequency and duration characteristics of the LD (long drumming) and SD (short drumming) are different from when they are in reproduction and whether not. Would circadian behaviors also change? The authors should consider if it is necessary to talk about direct sounds and the echoes in the water tank, some comments are about this aspect.

Response 1: These are great comments and when we realize the importance of making comparisons between recordings without reproduction or spawning and those during the reproduction in current paper, the experiment time window was gone. From our current dataset, the only conclusion that can be made is: there are at least two types of calls that the Chinese bahaba produced during the reproduction season. This paper provides the first known dataset of calls for this species under aquaculture conditions and as suggested, we are making plans to and sending applications to conduct more systematic recordings to probe into its sound production during different processes as well as those in the field. We’d believe that circadian behaviors will change during pre-spawning, spawning and after-spawing, which will be elucidated in our future work.

Great comments on the direct sounds/echoes, which indeed was found in our recordings. A great proportion of our recordings (about 54.8 %) had strong echoes/reverberations (e.g. Figure 1), which were actually inevitable in tank acoustics and these signals were rejected in our analysis. In some cases, strong reverberations were noted that overlapped the principal part of the calls (Figure 1). At these cases, it is often not possible to extract the principal part of the calls. We did not use a call if the principal part of call could not be extracted. In these cases, we manually selected high quality calls (207 of 496 For SD and 33 of 35 for LD) produced by the Chinese bahaba from approximate 264 hours of recordings. We referred to previous papers dealing with the echo/reverberation, and followed strict criteria to select clean signals to be used in subsequent analysis (Song et al., 2023a). Thus, the signals with an obvious overlap in waveform were not selected to conduct subsequent acoustic analysis and statistics.

Figure 1. Call produced by Chinese bahaba with an obvious overlap in waveform.

 

Comment 2: What is the dynamic range of the device's analog-to-digital converter? Does it capture signals between -1 and 1 volt?

Response 2: Thanks for your insightful comments. In our manuscript, SoundTrap ST300 HF recorder (Ocean Instruments Ltd, Auckland, New Zealand) was used to record calls produced by the Chinese bahaba. This recorder has been widely used in the study of underwater soundscape and bioacoustics (Dimoff et al., 2021; Song et al., 2023b). The ADC of SoundTrap was 16-bit. And the dynamic range was 96.3 dB. This recorder could capture signals between -1 and 1 volt.

 

Comment 3: Figures 2 and 4 (you can see echoes in them) do not contribute much, in my opinion, they are not necessary, since the same information is shown much better in figures 3 and 5.

Response 3: Figures 2 and 4 provides the temporal patterns of the two types of the calls as well as the IPI (inter-pulse-interval) information. Figure 3 and Figure 5 were produced to present the signals individually. These figures in our opinion give different information and we have merged the figures (2 and 3), (4 and 5) into single figures, attached below FYI.

 

Figure 2. Spectrogram (a) and oscillogram (b) of representative single drum trains emitted by Chinese bahaba. (c) Oscillogram for representative individual single drum and (d) Power spectra for the total calling signals.

 

Figure 3. Spectrogram (a) and oscillogram (b) of representative fast drum-trains in Chinese bahaba. (c) Oscillogram for representative single fast drum-trains and (d) Power spectra for the total calling signals.

 

Comment 4: In the case of SD sounds, have the echoes generated by the tank been taken into account? They are low frequency impulsive sounds that spread easily through the water tank and discerning between the echoes becomes not easy. Is this taken into account when performing the analysis? nothing is said in the text.

Response 4: See more details in Response 1 above.

We used three criteria to select quality calling signals produced by Chinese bahaba for further analysis. First, the individual calls with a signal to noise ratio (SNR) above 26.3 dB was extracted from the original acoustic files. Second, the waveform of every individual calls was check visually. Third, the calls with strong reverberations were also rejected for subsequent acoustic analysis. In some cases, strong reverberations were noted that overlapped the principal part of the calls (Figure 4). At these cases, it is often not possible to extract the principal part of the calls. We did not use a call if the principal part of call could not be extracted. In these cases, we manually selected high quality calls (207 of 496 For SD and 33 of 35 for LD) produced by the animals from approximate 264 hours of recordings. We referred to previous papers dealing with the echo/reverberation, and followed strict criteria to select clean signals to be used in subsequent analysis (Song et al., 2023a).

Figure 4. Call produced by Chinese bahaba with an obvious overlap in waveform.

 

Comment 5: Show real oscillograms SD and LD from acquiring signals in figures 3a and 5a. Figures 3b and 5b should include the parameters by which the fft has been calculated.

Response 5: As suggested, we have revised the real oscillograms of SD and LD instead of Figure 3a and 5a in original manuscript. And the related description accordingly was changed as “Power spectra were visualized using a 2048-point fast Fourier transform (FFT).” in revised manuscript. The real oscillograms of SD and LD were also attached here.

 

Figure 5. Oscillogram for representative individual single drum and fast drum-train

 

Comment 6: Assuming that the echoes have not been taken into account, given the large difference in ICI between the SDs that form an SD train, it should be considered not to treat them as an SP train.

Response 6: As suggested, we have deleted the related description of SP trains in the revised manuscript.

 

Comment 7: The number of SPs (NP) that form a train varies excessively, from 2 to 47. If there are 207 SD pulses within 96 trains. Most of the NP in the trains are formed by 2 SD. What is NP mode, median and mean? Perhaps the statistics have not been dealt with conveniently.

Response 7: Thanks for your helpful comments. Correct, most of the NP in the trains are formed by 2 SD. Here, we provide the distribution of NP from a train (Figure 6). To avoid misunderstanding, we deleted the description of “The number of pulses in the train ranged from 2 to 47 individual pulses with a mean PP of 114.4 ± 65.6 ms (Table 2).”.

 

Figure 6. Distribution of Np from a train

 

Comment 8: In this case, eliminate NP and PP rows in table 2.

Response 8: As suggested, we have deleted the NP and PP rows in the revised manuscript. And the related content of NP and PP has also been deleted.

 

Comment 9: The duration of the LPs goes from 13 cycles to 30, is this the case or does the echoes make the LPs last longer? The 13 cycles of an 80Hz wave is 0.162.5 s and 30 cycles is 0.375 s.

Response 9: See more details in Response 4 above.

We used three criteria to select quality signals of LP for further analysis. After the screening, the analyzed signals of LPs were clear without strong reverberations. Thus, the echoes were no significant influence on the DUR of signals.

 

Comment 10: The significance level in Figure 6 is 0.05 but in the discussion is 0.01. Which is correct?

Response 10: We have corrected “0.01” to “0.05” in the revised manuscript.

 

Comment 11: Which is the duration of the sound when there is no purpose of breeding?

Response 11: Referring to the previous study (Wei et al., 2021), the single drums (SD in the original manuscript) were considered as an acoustic signal without the purpose of breeding. These signals have a duration of 61.8 ± 24.0 ms.

 

Thanks again for your time and consideration. The comments are really helpful to improve the original manuscript which has been revised accordingly.

 

Sincerely yours,

Hongquan Li, Zhongchang Song, Jiangang Hui, Yingnan Su, Weijie Fu, Shufei Zhang, Lin Yan, Kuoqiu Yan, Honghui Huang, and Yu Zhang.

 

Reference

Dimoff, S. A., Halliday, W. D., Pine, M. K., Tietjen, K. L., Juanes, F. and Baum, J. K. (2021). The utility of different acoustic indicators to describe biological sounds of a coral reef soundscape. Ecol. Indic. 124, 107435.

Song, Z., Mooney, T. A., Quakenbush, L., Hobbs, R., Gaglione, E., Goertz, C. and Castellote, M. (2023a). Variability of Echolocation Clicks in Beluga Whales (Delphinapterus leucas) Within Shallow Waters. Aquat. Mamm. 49, 62–72.

Song, Z., Ou, W., Su, Y., Li, H., Fan, W., Sun, S., Wang, T., Xu, X. and Zhang, Y. (2023b). Sounds of snapping shrimp (Alpheidae) as important input to the soundscape in the southeast China coastal sea. Front. Mar. Sci. 10, 1029003.

Wei, C., Huang, J., Dai, M., Zhang, S., Huang, H. and Zhang, Y. (2021). Acoustic characteristics of synthesized signals of Chinese bahaba ( Bahaba taipingensis ). JASA Express Lett. 1, 011201.

Author Response File: Author Response.pdf

Reviewer 2 Report

The paper presents the study of the acoustic behavior of the Chinese bahaba for the first time. Although previously it has been carried out on other members of the Sciaenidae species, it is important to verify the relationship between spawning and sonic behavior, especially since it is in a critical state of extinction as done in this work.

Comments for author File: Comments.pdf

Author Response

====================   To Reviewer 2   ====================

March 14, 2023

 

To reviewer #2

Thank you again for your time and efforts spent on our manuscript. As suggested, all the figures have been changed to a better quality in the revised manuscript.

 

Sincerely yours,

Hongquan Li, Zhongchang Song, Jiangang Hui, Yingnan Su, Weijie Fu, Shufei Zhang, Lin Yan, Kuoqiu Yan, Honghui Huang, and Yu Zhang.

Author Response File: Author Response.docx

Reviewer 3 Report

The proposed manuscript shows, using recordings made with a SoundTrap that there are two distinct sound types produced by the Chinese bahaba, and that the reproductive behavior of the animals has interaction with the prevalence of these sound types. The presence of spectra and zoomed-in oscillograms and reasonably clear specifications of recording equipment, makes this a useful contribution to researchers who wish to search for this species in sound recordings.

There is absolutely no doubt that “long drumming” is produced by making a rapid succession of the clicks making up the "short drumming". You can model it by making trains of single drums and see how it compares with the actual recordings. There may be some non-linear effects, unexplained by simple convolution. I suggest changing the nomenclature to “single drums” and “fast drum-trains”. 

There are several places where the language needs seeing to. Have a native speaker read it.

“… with a high signal-to-noise ratio (SNR)”. State the criterion numerically

“… files and conducted to analyze with” unusual use of the word “conducted”

“Time between two points at which the time-integral energies of acoustic signals are 2.5% and 97.5%, respectively” How about noise? How closely were they cropped?

Figure 2 and Figure 3. The x-axes of the spectrograms are not aligned with the oscillograms. Spectrograms in Matlab do not start in 0. Use linkaxes to fix. These low-frequenct sounds call for logatithmic y-axis- This is a little bit cumbersome in Matlab, but would improve the information from the spectrogram a lot. Aren’t there two animals in Fig2?

This sentence: “The sound production of Sciaenid family has associated highly relation with courtship and spawning behavior”, needs rewriting.

Author Response

====================   To Reviewer 3   ====================

March 14, 2023

 

To reviewer #3

Thank you very much for your time spent on polishing the language and figure legends etc. Your professional comments are helpful to improve our manuscript, which has been revised substantially and the changes were marked using “Track Changes” function in revised manuscript in .doc text. Our responses to your comments are as follows:

 

Major comments

Comment 1: There is absolutely no doubt that “long drumming” is produced by making a rapid succession of the clicks making up the "short drumming". You can model it by making trains of single drums and see how it compares with the actual recordings. There may be some non-linear effects, unexplained by simple convolution. I suggest changing the nomenclature to “single drums” and “fast drum-trains.

Response 1: Thanks for your professional comments. As suggested, we have changed all the descriptions of “short-drumming” and “long-drumming” as “single drums” and “fast drum-trains”, respectively in the revised manuscript.

 

Minor comments

Comment 1: “… with a high signal-to-noise ratio (SNR)”. State the criterion numerically.

Response 1: This is a helpful comment. The SNR was calculated as:

 

Where P_s and P_n were the effective power of signal and noise, respectively.

We measured the SNR of the individual calls, subsequently. And the lowest value of SNR was 26.3 dB. As suggested, this sentence was revised as “A total of 240 individual calls with a signal-to-noise ratio (SNR) above 26.3 dB were extracted from the original acoustic files.”.

 

Comment 2: “… files and conducted to analyze with” unusual use of the word “conducted”

Response 2: As suggested, this sentence was revised as “The waveform of every individual calls was check visually. And the calls with strong reverberations were also rejected for subsequent acoustic analysis. The remained signals conduct the acoustic analysis by a customized routine based on MATLAB.”

 

Comment 3: “Time between two points at which the time-integral energies of acoustic signals are 2.5% and 97.5%, respectively” How about noise? How closely were they cropped?

Response 3: Figure 7 showed how we measured the signal duration (DUR) in an individual call. In our manuscript, the DUR was calculated as “time between two points at which the time-integral energies of acoustic signals are 2.5% and 97.5%, respectively”. Thus, the energy of noise in individual pulses was too low to influence the measurement of DUR.

 

Figure 7. Interpretation of DUR in example of fast drum-trains.

 

Comment 4: Figure 2 and Figure 3. The x-axes of the spectrograms are not aligned with the oscillograms. Spectrograms in Matlab do not start in 0. Use linkaxes to fix. These low-frequency sounds call for logarithmic y-axis- This is a little bit cumbersome in Matlab, but would improve the information from the spectrogram a lot.

Response 4: Thanks for your helpful comment. As suggested, we have aligned the start point in the y-axes in Figure 2(b) and the x-axes in Figure 3(b), respectively. A similar revision was also conducted in Figure 4(a) and Figure 5(b).

 

Comment 5: Aren’t there two animals in Fig2?

Response 5: Actually, a total of 8 sexually mature Chinese bahaba (5 males, 3 females) were maintained in the tank. So, it is hard to verify how many animals produce sound in Figure 2. To avoid the misunderstanding, we have changed the Figure 2 in revised manuscript, which was also attach here.

 

Figure 8. Acoustic recording of Chinese bahaba vocalization.

 

Comment 6: This sentence: “The sound production of Sciaenid family has associated highly relation with courtship and spawning behavior”, needs rewriting.

Response 6: As suggested, we have rewritten this sentence as “The relationship between sound production and spawning behavior has been found in other species in Sciaenid family, suggesting the acoustic monitoring could be used to monitor reproduction.”.

 

Thanks again for the time and consideration and the comments are really helpful to improve the manuscript. We have substantially revised the manuscript.

 

Sincerely yours,

Hongquan Li, Zhongchang Song, Jiangang Hui, Yingnan Su, Weijie Fu, Shufei Zhang, Lin Yan, Kuoqiu Yan, Honghui Huang, and Yu Zhang.

Author Response File: Author Response.pdf