https://orcid.org/0000-0002-7289-7115
Abstract
Given the common physical overlapping between coastal developments and important marine mammal habitats, there is a need to identify potentially important foraging grounds for dolphins when informing marine spatial planning and management of underwater noise. Hydrophones were deployed at four locations either side of the mainland China–Hong Kong Special Administrative Region border to monitor the presence of soniferous fishes; a key prey item for Indo-Pacific humpback dolphins. Five distinct chorus-types were identified; each showing spatiotemporal variability. Each chorus-type was assumed to represent a separate species. Chorus-type diversity also differed between sites, with SP4 and SP5 types only being detected within Hong Kong waters where bottom trawling is illegal. Chorus-type SP1 was only detected at the recording sites in mainland Chinese waters. Call rates and chorus duration were highest during the spring and summer months. Given these dolphins show a predator-prey relationship, these data provide new information on the local fish communities at a much finer-scale than fish landing records and a baseline of fish activity in an environment that is challenging to explore. Overlaid with acoustic detections of foraging dolphins, these data form a basis for identifying potentially important foraging habitats that should be afforded the highest priority for protection.
Introduction
The biological impact from underwater noise on ecologically significantly regions is an internationally recognized issue (Williams et al., 2015). This is particularly relevant where large human populations overlap important marine mammal habitats, such as key foraging areas or nursery habitat. The Pearl River Estuary (PRE) in China is a good example of this whereby nine major cities, encompassing three separate reporting and management jurisdictions, share a single population of Indo-Pacific humpback dolphins (Sousa chinensis). Due to the economic importance of the PRE, the scale of coastal developments, and thus the increasing noise levels, within the estuary has been extraordinary and the potential noise impacts on the resident dolphin population is a significant conservation concern.
The Indo-Pacific humpback dolphin ranges from somewhere east of India to the Indo-Malay Archipelago, north along the East Asian coasts and south to northern Australian waters. The boundaries of the genus are poorly understood and it was only recently that Australian humpback dolphins were designated as a separate species. Populations are fragmented and usually associated with major estuaries. Little or no information exists regarding the overall number of this population; however, some areas have been better studied than others, e.g. the PRE and the Eastern Taiwan Straits. The Indo-Pacific humpback dolphin is regarded as Near Threatened by the International Union for Conservation of Nature. Under Chinese legislation, this species is considered a Grade 1 National Key Protected Species and is afforded the highest protection (Xu et al., 2015). Within the PRE, it is believed that the resident population may be over 2000 individuals (Jefferson and Smith, 2016), and it is decreasing at a rate of 2.46% per annum (Huang et al., 2012). Considering the mounting evidence for the adverse effects of noise on marine mammals (see Weilgart, 2007; Williams et al., 2015), it is possible that increasing levels of underwater noise may be a contributing factor (SCU, 2012).
Although the population of dolphins that reside within the PRE is better studied than most, relatively little is understood of their estuary wide range and the distribution of prey within this range. Recent research has shown detection rates of Indo-Pacific humpback dolphins correlating with fish call rates near the Sanjiao islands in the PRE, with no such correlation to the ambient noise levels from proximate vessel traffic (Pine et al., 2016). These data suggest that dolphins are often exposed to acoustic stress in order to forage in areas where prey and anthropogenic noise sources overlap (Pine et al., 2016). These issues have been recognized in mainland China and Hong Kong, with local governments requiring impact assessments, including noise mitigation strategies, as part of the environmental permitting process. Notwithstanding however, many developments (such as the construction of the Hong Kong-Zhuhai-Macau Bridge, the Third Runway development at the Hong Kong International Airport and the Guishan offshore wind farm), all overlap with important marine mammal habitats; a scenario that is sometimes discovered after construction has commenced. Understanding where these important foraging habitats may be before or during the environmental impact assessment phase of future developments is thus fundamental for the conservation of these dolphins within the PRE.
Similar to marine mammals, the sensitivity of fish to underwater sound has also been well documented and dolphins eavesdrop on their potential prey during foraging (Gannon et al., 2005). Like marine mammals, fish use underwater sounds to sense their environment, as well as coordinate certain behaviours such as reproduction or territorial defence (Hawkins, 1986; Rountree et al 2006). They typically produce low-frequency sounds (<1 kHz), including grunts, croaks, clicks, knocks, and snaps, through stridulation (the rubbing of skeletal components), drumming (rapid contraction of the sonic muscles on or near the swim bladder) or hydrodynamics (quickly altering swimming direction and speed) (Ladich and Fine, 2006; Fine and Parmentier, 2015). These mechanisms for producing sounds often depend on the situation. For example, stridulation often occurs during feeding when the teeth or pharyngeal teeth and jaws are engaged, during territorial displays or as a fright response (Hawkins, 1986). Often the swim bladder amplifies stridulation-generated sounds (Hawkins, 1986). However, drumming may be used to produce mating calls, with the most well known sounds classified as croaks and hums; a common acoustic feature of members of the Sciaenidae family, as well as toadfishes (of the family Batrachoididae) (Mann, 1998; Mann et al., 2006). Consequently, a lot of vocal behaviour in fishes is associated with spawning and thus shows seasonal variability. Furthermore, the size and condition of the sonic muscles also show seasonality that is related to the species’ reproduction period (Kasumyan, 2008; Lagardére and Mariani, 2006). For example, seasonal periodicity in the sonic organs have been described in the weakfish, Cynoscion regalis, haddock, Melamogrammus eaglepinus (Templeman and Hodder, 1958, Hawkins et al., 1967, Connaughton and Taylor, 1994, 1996) and the plainfin midshipman, Porichthys notatus (Kasumyan, 2008).
Members of the Sciaenidae family, also known as the croakers and drummers, are the largest known family of soniferous fishes (Ramcharitar et al., 2006). Based on local fisheries landing records and dietary studies of the Indo-Pacific humpback dolphins within the PRE (Barros et al., 2004; Fish and Mowbray, 1970; Banner, 1972; Whitehead and Blaxter, 1989; Ren et al., 2007) there are at least three sciaenid species within the PRE: the lionhead (Collichthys lucida), the Belanger’s croaker (Johnius belangerii) and the big-snout croaker (Johnius macrorhynus). However, anchovies (Thryssa sp.) have also been found within the stomach contents of stranded Indo-Pacific humpback dolphins within Hong Kong (Barros et al., 2004) and could be soniferous (Whitehead and Blaxter, 1989). Sciaenid fishes typically produce a series of rapid pulses commonly described as croaking knocking, clucking, or purring (Ramcharitar et al., 2006).
Relying on a diet almost exclusively of fish, Indo-Pacific humpback dolphins within the PRE feed primarily on these soniferous croakers (Barros et al., 2004) and dolphin detection rates have been linked to soniferous fish activity in certain areas of the PRE (Pine et al., 2017). However, very little is understood on the distribution and seasonality of these ecologically important fishes. Notwithstanding, there are some data that provide insight into potential seasonal and lunar interactions in soniferous fishes in the western waters of Hong Kong. For example, passive acoustic monitoring on dolphins north of Lantau Island indirectly revealed higher fish calling activity during the spring months (Munger et al., 2016). Although incomplete, other data from within the PRE indirectly suggest seasonality in fish calling rates, with higher call rates during the summer months compared with late autumn (Pine et al., 2017). However, while these data do suggest seasonality, no comprehensive effort that directly investigates the spatiotemporal distribution of vocal fishes in the area has been done. Given these vocal fishes are an important prey item for the Indo-Pacific humpback dolphins within mainland PRE and Hong Kong waters (Barros et al., 2004), and that positive correlations between short-term vocal activity in fish and dolphin presence have been demonstrated (Pine et al., 2017), understanding the spatiotemporal distributions of these fish would assist in the identification of important foraging habitats for dolphins. The current study is the first to establish a baseline for fish detections across several sites in both mainland Chinese and Hong Kong waters of the PRE.
Material and methods
Study sites
Four passive acoustic listening stations were set up at four locations within the PRE (Figure 1). Each listening station were maintained between August 2015 and September 2016 and where specifically located off Sanjiao Island (referred to as the Sanjiao site (N 22.118 E 113.716), Qi’An island (Qi’An site; N 22.420 E 113.679), Lamma Island (Lamma site; N 22.189 E 114.116) and Lung Kwu Chau (A5 site; N 22.373 E 113.898). The Sanjiao and Qi'’n sites were located within mainland Chinese waters, while the Lamma and A5 sites were within Hong Kong SAR waters. Due to the high fishing pressure in mainland Chinese waters, acoustic loggers at the Sanjiao and Qi’An sites were deployed at least 5 m from a disused radio tower and offshore lighthouse, respectively. By doing so, the recorders were safe from bottom trawlers whilst being distant enough for minimal noise contamination from waves against the pillar (determined during a series of pilot studies in 2015). Since bottom trawling does not occur within the Hong Kong SAR, the acoustic loggers could be deployed in isolation with fewer restrictions.
Study site in the PRE along the southern coast of China. Solid black circles represent each recording site while *represents the locations of each weather station.
Listening stations and acoustic loggers
The listening stations consisted of the acoustic logger (a SoundTrap 300 HF acoustic recorder from Ocean Instruments New Zealand) secured to an iron rod set in a 30 cm deep, 50 kg concrete base that sat directly on the seafloor. When attached, the hydrophone component of the SoundTrap recorder was 1 m off the seabed. Each armed acoustic logger was set to operate on a duty cycle (2-min recording every 15-min at a sampling rate of 288 kHz and logged temperature every 60 s) to economize on battery life and available data card storage whilst improving data handling efficiency (Pine et al., 2017). Electronic calibration checks were completed at the start of each recording and checked at the start and middle of each recording day for each season.
Environmental conditions during recording
At higher wind speeds, noise from the oscillating air bubbles entrained at the surface from wind can peak around 500 Hz (Erbe et al., 2015) and therefore has the potential to mask fish calls. Therefore, to quickly and easily identify periods when this may be the case, wind speed, direction and rainfall were continuously monitored from the closest three weather stations (Fan Lau Shan, Hong Kong International Airport and Taipa in Macau) when the listening stations were active.
Data analysis
Recordings were uploaded into Adobe Audition 3.0 software for automated batch processing and editing before the customized fish detectors in Raven Pro 1.5 and Matlab were run and daily spectrograms were created for each deployment day. All statistical tests were carried out using SPSS 17.0 and SigmaPlot 11.0 and a p-value ≤ 0.05 was considered significant.
Recordings were analysed for each deployment day over the study period and visually inspected after a 24 h period power spectral density spectrogram was generated in Matlab (using a 1 s Hanning window, 50% overlap and 10 s time-averaging). Recordings that did contain fish choruses were set aside and those PSD spectrograms were later correlated with the fish detector results. Weather station data were used to identify periods of high wind speeds (>10 km h−1) and rainfall and were excluded from the fish detector data after confirming any acoustic contamination.
In order to understand the spatiotemporal distributions of the key fish species producing evening and night choruses within the PRE, as well as the temporal variation between seasons, the duration (mins) of each chorus (of each dominate call-type) recorded from all four listening stations was calculated. Choruses consisting of a single call-type were referred to as chorus-types. Call-types were first identified based on the received spectral and temporal characteristics in Raven Pro 1.5: the centre and peak frequencies (Hz), 90% bandwidth (Hz), 90% duration (s), overall call duration (s) and the burst rate per call. A total of 150 samples of each call-type were then randomly selected for statistical testing, whereby specific call-types were confirmed by statistically significant differences between either the spectral and/or temporal characteristics using a one-way ANOVA, after confirming the required assumptions. Where appropriate, Holm-Sidak pairwise comparisons were used to distinguish which call-types significantly differed from others within each acoustic characteristic and referred to as chorus-types SP1–SP5. This statistical testing application was done due to the inability to identify the species of fish each call type was coming from. An important assumption; however, is that each fish species within the PRE produce a chorus composed of a single dominant call-type and that each different chorus-type represents the presence of a different species. This assumption was considered appropriate within the current study due to the observation that choruses were of specific spectral bandwidths lasting several hours and because of the consistent temporal partitioning of chorus-types throughout a 24-h period.
Following the identification of each chorus-type, the durations of each chorus from each recording site was calculated and plotted. Call rates for the most prevalent chorus-type was also calculated using the same band-limited energy detectors and plotted. To contextualize the spatiotemporal distributions around the PRE, the proportion of time that each chorus-type was detected for each season were calculated by totalling the duration of all choruses recorded and plotted to form a spatiotemporal map of fish activity.
Results
A total of 103 104 recordings, equating to 3436.8 recording hours, were successfully obtained.
Five distinct call-types making up evening and/or night choruses were identified across all recording sites (Figure 2). Key acoustic parameters of each call-type that made up a chorus are provided in Table 1. Peak and centre frequencies across all call-types were below 1 kHz and the 90% bandwidths were below 1800 Hz. Call-types SP2 and SP4 shared similar peak (Holm-Sidak test, p = 0.961) and centre (Holm-Sidak test, p = 0.099) frequencies but were distinguished from one another based on the differing 90% bandwidth, and temporal characteristics (90% duration, total duration and pulse rate). Similarly, call-types SP1 and SP3 where of similar peak frequencies (Holm-Sidak test, p = 0.416), however differed in centre frequency (Holm-Sidak test, p < 0.001), 90% bandwidth (Holm-Sidak test, p < 0.001), 90% duration (Holm-Sidak test, p < 0.001), call-burst duration and pulse rate (Table 1).
Summary of the acoustic parameters used to confirm the presence of a new call-type chorus (mean ± SD).
| Parameter | Chorus-type | ||||
|---|---|---|---|---|---|
| Sp1 | Sp2 | Sp3 | Sp4 | Sp5 | |
| Peak frequency (Hz) | 863.3 ± 111.8a | 668.0 ± 173.3b | 853.2 ± 291a | 665.9 ± 165.3b | 144.1 ± 14.8c |
| Centre frequency (Hz) | 875.3 ± 84.3a | 719.1 ± 159.4b | 948.5 ± 148.3c | 674.2 ± 96.6b | 144.1 ± 8.8d |
| 90% bandwidth (Hz) | 1129.8 ± 160a | 437.9 ± 186.6b | 1723.7 ± 174.7c | 572.8 ± 57.7d | 107.4 ± 30.2e |
| Burst-duration (ms) | 161 ± 31.8a | 136 ± 52a | 342 ± 51b | 230 ± 37.1c | 1011 ± 225.3d |
| 90% duration (ms) | 140 ± 49.1a | 95.5 ± 49b | 300 ± 49.5c | 206 ± 23.9d | 860 ± 199.3e |
| Pulse rate per call | 9-11 | 13-16 | 36-42 | 12-13 | 11-14 |
| Parameter | Chorus-type | ||||
|---|---|---|---|---|---|
| Sp1 | Sp2 | Sp3 | Sp4 | Sp5 | |
| Peak frequency (Hz) | 863.3 ± 111.8a | 668.0 ± 173.3b | 853.2 ± 291a | 665.9 ± 165.3b | 144.1 ± 14.8c |
| Centre frequency (Hz) | 875.3 ± 84.3a | 719.1 ± 159.4b | 948.5 ± 148.3c | 674.2 ± 96.6b | 144.1 ± 8.8d |
| 90% bandwidth (Hz) | 1129.8 ± 160a | 437.9 ± 186.6b | 1723.7 ± 174.7c | 572.8 ± 57.7d | 107.4 ± 30.2e |
| Burst-duration (ms) | 161 ± 31.8a | 136 ± 52a | 342 ± 51b | 230 ± 37.1c | 1011 ± 225.3d |
| 90% duration (ms) | 140 ± 49.1a | 95.5 ± 49b | 300 ± 49.5c | 206 ± 23.9d | 860 ± 199.3e |
| Pulse rate per call | 9-11 | 13-16 | 36-42 | 12-13 | 11-14 |
Note these are received levels averaged over 150 samples. Superscripts identify which call-types statistically differ within each parameter (Holm-Sidak test, p < 0.05).
Summary of the acoustic parameters used to confirm the presence of a new call-type chorus (mean ± SD).
| Parameter | Chorus-type | ||||
|---|---|---|---|---|---|
| Sp1 | Sp2 | Sp3 | Sp4 | Sp5 | |
| Peak frequency (Hz) | 863.3 ± 111.8a | 668.0 ± 173.3b | 853.2 ± 291a | 665.9 ± 165.3b | 144.1 ± 14.8c |
| Centre frequency (Hz) | 875.3 ± 84.3a | 719.1 ± 159.4b | 948.5 ± 148.3c | 674.2 ± 96.6b | 144.1 ± 8.8d |
| 90% bandwidth (Hz) | 1129.8 ± 160a | 437.9 ± 186.6b | 1723.7 ± 174.7c | 572.8 ± 57.7d | 107.4 ± 30.2e |
| Burst-duration (ms) | 161 ± 31.8a | 136 ± 52a | 342 ± 51b | 230 ± 37.1c | 1011 ± 225.3d |
| 90% duration (ms) | 140 ± 49.1a | 95.5 ± 49b | 300 ± 49.5c | 206 ± 23.9d | 860 ± 199.3e |
| Pulse rate per call | 9-11 | 13-16 | 36-42 | 12-13 | 11-14 |
| Parameter | Chorus-type | ||||
|---|---|---|---|---|---|
| Sp1 | Sp2 | Sp3 | Sp4 | Sp5 | |
| Peak frequency (Hz) | 863.3 ± 111.8a | 668.0 ± 173.3b | 853.2 ± 291a | 665.9 ± 165.3b | 144.1 ± 14.8c |
| Centre frequency (Hz) | 875.3 ± 84.3a | 719.1 ± 159.4b | 948.5 ± 148.3c | 674.2 ± 96.6b | 144.1 ± 8.8d |
| 90% bandwidth (Hz) | 1129.8 ± 160a | 437.9 ± 186.6b | 1723.7 ± 174.7c | 572.8 ± 57.7d | 107.4 ± 30.2e |
| Burst-duration (ms) | 161 ± 31.8a | 136 ± 52a | 342 ± 51b | 230 ± 37.1c | 1011 ± 225.3d |
| 90% duration (ms) | 140 ± 49.1a | 95.5 ± 49b | 300 ± 49.5c | 206 ± 23.9d | 860 ± 199.3e |
| Pulse rate per call | 9-11 | 13-16 | 36-42 | 12-13 | 11-14 |
Note these are received levels averaged over 150 samples. Superscripts identify which call-types statistically differ within each parameter (Holm-Sidak test, p < 0.05).
Representative amplitude and spectrograms of the dominate call-type making up each of the five chorus-types. Spectrograms generated using a 2048 sample Hann window, 50% overlap, 1024 Hop Size and 202 Hz 3 dB filter bandwidth.
The greatest diversity of chorus-types was detected at the Sanjiao and A5 recording sites, with three distinct chorus-types (SP1–SP3 at the Sanjiao site an SP2–SP4 at the A5 site), followed by Lamma (SP2 and SP5) and Qi’An (SP1 only). Interestingly, chorus-type SP1 was only detected within the Sanjiao and Qi’An recording sites, while chorus-type SP4 was only detected at A5 and SP5 at the Lamma recording site (Figure 3). The duration of each chorus-type varied over the year, with the longest durations occurring between spring and summer months (Figure 4). Over a 24-h period, chorus-type SP1 was the longest-lasting, with a maximum duration of 660 mins (during April at the Sanjiao recording site), while the shortest-lasting chorus-type was SP4, with a maximum duration of 150 min (during May, July, and August at the A5 site). All chorus-types followed the general trend of heightened activity (represented by longer chorus durations) between spring and summer months that gradually decreased through autumn to minimal activity over the winter.
Map showing the spatial variation in chorus-types at each of the recording sites. Percentages were calculated based on the total duration of each chorus after combining the durations of all chorus-types for each recording site. Basemap is the World Water Bodies map, dated 12 May 2017, downloaded from the ArcGIS website.
Chorus duration (minutes) of each chorus-type per each night between August 2015 and September 2016 at each recording site (SJD, Sanjiao site; QAD, Qi’An site; LAM, Lamma site; and A5, A5 site). The black segments respresent the periods when no data were collected.
When all chorus-type durations (referred to as chorus-mins) were totalled over the survey period, chorus-type SP1 made up 64% of the total chorus-mins at the Sanjiao recording site (31 596 mins over a total of 49 584 chorus-mins), SP2 accounted for 19% (total mins of 9648 over the survey period) and SP3 made up the remaining 17% (8340 min) (Figure 3). Broken down into seasons, chorus-type SP2 was the more common-occurring chorus-type during the summer, making up 42% of choruses at the Sanjiao recording site (Figure 5) while chorus-types SP1 and SP3 contributed 32 and 26% of the total chorus duration, respectively. Over at the A5 recording site, chorus-type SP2 made up the second largest proportion across all sites of 85% (14 700 min of a total 17 520 min), followed by 85% of total chorus-mins at the Lamma recording site (Figure 4). During the autumn, chorus-type SP2 decreased in activity within both the Sanjiao and A5 recording sites, making 15 and 61% within either site, respectively (Figure 5), while chorus-type SP1 was the more common-occurring chorus-type at Sanjiao. At the Lamma and A5 recording sites; however, chorus-type SP2 was the most common-occurring during the autumn months, making up 53% of all the total chorus-mins, while SP5 contributed to 47% at the Lamma recording site. Winter showed the lowest diversity of chorus-types among sites, with only SP1 being detected at the Sanjiao site and SP2 at Lamma, before rising again during the spring months (Figure 5).
Map showing the spatiotemporal variation in chorus-types at each of the recording sites between seasons. Percentages were calculated based the total duration of each chorus after combining the durations of all chorus-types over each season for each recording site. Basemap is the World Water Bodies map, dated 12 May 2017, downloaded from the ArcGIS website.
As a further indication of the temporal variability in fish activity within the proposed area, call rates during each SP1 chorus showed the highest rates (approximately 222.7 ± 12.4 calls per 2-min recording) during April and lowest rates during the winter months December and January (∼37.3 ± 20.9 and 45.6 ± 8.2 calls per 2-min recording, respectively) (Figure 6). Although call rates throughout the summer, autumn and winter months showed a unimodal distribution between 17:00 and 02:00 h, a bimodal distribution formed during the spring months, starting in March and by April the call rates showed two clear peaks of ∼127.1 ± 10.1 and 222.7 ± 12.4 calls per 2-min recording at around 16:00 and 22:00 h, respectively.
Mean call rates (number of calls per 2 min recording) between 12:00 and 06:00 h (the following day), averaged over each month at the Sanjiao recording site.
Discussion
Identifying potentially important foraging habitat for threatened dolphin populations is a fundamental step towards informing management and formulating mitigating strategies to minimize anthropogenic noise impacts. This was the principal rationale for the current study which provides preliminary information on the spatiotemporal distribution of soniferous fishes; a key prey item for the Indo-Pacific humpback dolphins within the PRE (Barros et al., 2004) and to which dolphin presence is linked (Pine et al., 2017). Although a complete and comprehensive survey is not possible using passive acoustics alone, by “eavesdropping” on the soundscape an understanding of the seasonal and temporal distributions of vocal fishes can be mapped. This study therefore aimed to investigate the fish choruses detected during each season to better understand the spatiotemporal variability of a key factor influencing the current habitat use of dolphins within the PRE. The data presented in this paper provides the foundation for the eventual process of relating the spatiotemporal variation in fish activity with dolphin movements.
Analyses reveal the presence of at least five different chorus-types within the PRE, each showing seasonal patterns in activity. For the purposes of investigating the distribution of fishes, the duration (mins) of each chorus-type was analysed. An important consideration within this study is that each chorus-type was assumed to be representative of a different species and was based on the temporal and spectral dynamics of a chorus’s most dominant call-type. However, fish sometimes produce different sounds during a same chorus. For example, the sea catfish (Arius maculatus), as well as the croakers Johnius sp. which occur within the PRE, often produce strumming and stridulation sounds which vary both temporally (i.e. duration and pulse-rates) and spectrally (Mok et al., 2011). However, by focussing on each chorus as a whole and characterizing it by the most dominate call-type, the probability of the same species being counted as separate chorus-types is controlled for.
The consistent bandwidth and temporal partitioning between chorus-types is further evidence that each chorus-type is likely of a separate species. Certain chorus-types were also recorded at only certain recording sites. For example, chorus-type SP2 peaked during early evening, followed by chorus-type SP1 between ∼20:00 and 23:00 h, then chorus-type SP3 with peak activity occurring between midnight and 02:00 h. Resource partitioning is a fundamental ecological concept for reducing competition between co-occurring species (Schoener, 1974; Hastings and Sirovic, 2015). Fish have appeared to evolve bandwidth and/or temporal partitioning to likely prevent species from confusing conspecific calls from interspecific ones (Luczkovich and Sprague, 2011; Wilkins et al., 2013; Parsons et al., 2016; Staaterman et al., 2013; Hastings and Sirovic, 2015), while other families that compete for acoustic bandwidth may show temporal partitioning (Luczkovich and Sprague, 2011; Hastings and Sirovic, 2015; Ruppé et al., 2015; Parsons et al., 2016). This was observed within the current study with each chorus-type being of varying bandwidths and rarely showed temporal overlap. By avoiding temporal overlap between chorus-types of similar peak-frequencies and burst-rates, such as chorus type SP2 and SP4, each fish species appears to be displaying a degree of resource partitioning, as seen in other regions (Hastings and Sirovic, 2015). Further to the temporal partitioning in chorus-types, the current study has also outlined seasonal patterns in choruses. Previous studies characterizing the diversity and variation in fish choruses have shown diurnal, lunar and seasonal patterns among choruses; thereby emphasizing the use of PAM for temporal and spatial predictive modelling (Parsons et al., 2016).
At least three sciaenid species are known to exist within the PRE (the lionhead, Belanger’s croakers, and the big-snout croaker) and are most likely the source of at least three of the chorus-types. It was not possible to identify the species of fish from which the received signals were emitted, due to the limited information on species-specific vocalisations, as well as due to the nature of the data collection. However, the dominate call-types within each chorus were largely commensurate with the acoustic characteristics of strumming and drumming. Although no published investigations on the sound production of fishes within the PRE were found, the predominate call-type making up the SP1 chorus-type showed overlapping frequency and temporal characteristics with the Belanger’s croaker (J. belangerii) (Pilleri et al., 1982). For example, calls from the Belagner’s croaker range between 500 and 1250 Hz (dominant frequencies), with bursts lasting ∼140–160 ms with 4–14 pulses per call (Ramcharitar et al., 2006). However, the Belanger’s croaker has also been known to produce calls lasting between 140 and 260 ms (Pilleri et al., 1982). Therefore, while unable to confirm this, it may be that SP1 chorus-type is predominately that of the Belanger’s croaker. Similarly, chorus-type SP3 shows a degree of overlap with calls from the big-snout croaker, J. macrorhynus (Lin et al., 2007).
Notwithstanding the challenges in species identification, there was considerable spatiotemporal variation among chorus-types. During the dry season (cooler sea surface temperatures), there was a considerable decrease in chorus-type diversity and activity. For example, a single chorus-type was recorded over only 11 nights during the dry season and the average call rates were significantly lower compared with the wet season. This dramatically contrasted to the wet season where all five chorus-types were detected, durations remained for hours at a time over most nights and call rates peaked. Croakers (Johnius sp.) and other potential prey species of the dolphins are known to move into the shallower waters during the wet season, while moving to deeper waters during the dry seasons (Chen and Liu, 1982; Munger et al., 2016). Given the previous finding that dolphin occurrence correlates with fish activity between August and November (the wet season), dolphin detections are expected to reach the lowest rates of the year over the dry season. Although not directly investigated to date, observations around north Lantau Island, near the A5 recording site within this study, have shown fewer acoustic detections of dolphins during early spring, despite higher fish activity (Munger et al., 2016). Although this could be due to lower acoustic activity in the dolphins, evidence suggests that dolphin densities are lower coming into the wet season off the northern shores of Lantau Island (Jefferson, 2000; Hung, 2008) with the dolphins’ possibly feeding on other fish species further offshore (Munger et al., 2016).
Although data from the A5 recording site were first collected during mid-autumn, only chorus-type SP2 was present until June; after which chorus-types SP3 and SP4 were then detected. Based on information as recent as 2014, this finding may be unusual as fish activity in the area is thought to be most intense during the early spring (Munger et al., 2016). With the development of the Hong Kong International Airport and increased dredging within the PRE, this finding is particularly interesting in that it provides potential insight into new methods for assessing possible impacts from developments on fish communities. Although these data do not support this as a conclusion, it does warrant further research whereby quantified comparisons between acoustic detections of fishes and dolphins, as well as anthropogenic sources, are tested. Such comparisons would provide critical information on the extent of possible ecological changes within the PRE over a very small time-frame between 2014 and 2016 due to coastal development projects.
Unique to the Lamma recording site was chorus-type SP5. Occurring mostly during the spring and autumn months, chorus-type SP5 was of considerably lower peak frequencies, shorter bandwidths and longer durations compared with the other four chorus-types. It may be possible that this call-type is also from Johnius sp., however the times during which chorus-type SP5 were detected appeared independent from chorus-type SP2 and neither chorus-types SP1, SP3, or SP4 were detected at the Lamma site; thereby suggesting a different species. Although this cannot be confirmed and care should be taken when assuming a chorus-type does in fact represent a new species, this preliminary finding may suggest chorus-type SP5 to be relatively restricted in its distribution. If chorus-type SP5 is confirmed to be produced from a different fish species, further research into the distributions of this chorus-type would be very interesting given that it was only detected at the one site and over a relatively short period of time.
Acknowledgements
We would like to thank all members of the Conservation Biology of Aquatic Animals research group at the Institute of Hydrobiology, Chinese Academy of Sciences. In particular, I thank Zhaolong Cheng, Dr Lijun Dong, Dr Liang Fang, and Changqun Zhang for their valued assistance in the field. I also sincerely thank Director Xichun Gu, Vice-director Huantong Wan and technician Xi Chen from the PRE Chinese White Dolphin National Nature Reserve for their assistance in the field. We would also like to thank Paul Hodgson at The Oceanway Corporation Limited and Lai Hong Yu, Lui Kwok Wah, and Lam Wing Sum of SMRU Hong Kong for deploying and servicing SoundTraps in Hong Kong waters. Also, a big thank you to Dr Christine Erbe at Curtin University for her assistance with the MATLAB coding. This research was supported by the Administration of Ocean and Fisheries of Guangdong Province, the Administration of Guangdong Zhujiang Estuary Chinese White Dolphin Nature Reserve and partly supported by China Ministry of Agriculture and the National Science Foundation of China.
References
