Temporal and spatial mapping of red grouper ... - Wiley Online Library

Journal of Fish Biology (2014) 85, 1470–1488 doi:10.1111/jfb.12500, available online at wileyonlinelibrary.com

Temporal and spatial mapping of red grouper Epinephelus morio sound production C. C. Wall*†‡, P. Simard*, M. Lindemuth§, C. Lembke§, D. F. Naar*, C. Hu*, B. B. Barnes*, F. E. Muller-Karger* and D. A. Mann*|| *University of South Florida, College of Marine Science, 140, 7th Avenue South, St Petersburg, FL 33701, U.S.A. and §University of South Florida, Center for Ocean Technology, 140, 7th Avenue South, St Petersburg, FL 33701, U.S.A. (Received 24 February 2014, Accepted 9 July 2014) The goals of this project were to determine the daily, seasonal and spatial patterns of red grouper Epinephelus morio sound production on the West Florida Shelf (WFS) using passive acoustics. An 11 month time series of acoustic data from fixed recorders deployed at a known E. morio aggregation site showed that E. morio produce sounds throughout the day and during all months of the year. Increased calling (number of files containing E. morio sound) was correlated to sunrise and sunset, and peaked in late summer (July and August) and early winter (November and December). Due to the ubiquitous production of sound, large-scale spatial mapping across the WFS of E. morio sound production was feasible using recordings from shorter duration-fixed location recorders and autonomous underwater vehicles (AUVs). Epinephelus morio were primarily recorded in waters 15–93 m deep, with increased sound production detected in hard bottom areas and within the Steamboat Lumps Marine Protected Area (Steamboat Lumps). AUV tracks through Steamboat Lumps, an offshore marine reserve where E. morio hole excavations have been previously mapped, showed that hydrophone-integrated AUVs could accurately map the location of soniferous fish over spatial scales of 2100 t of commercial landings and >130 000 recreationally caught E. morio were reported for Florida’s west coast (SEDAR, 2009). Epinephelus morio are ecological engineers that excavate holes and expose structure that serve as habitat for themselves, other reef fishes and invertebrates (Jones et al., 1994; Coleman & Williams, 2002; Coleman et al., 2010). Sustaining E. morio populations is clearly an important management goal (Jones et al., 1994; †Author to whom correspondence should be addressed. Tel.: +1 303 497 4459; email: [email protected] ‡Present address: University of Colorado at Boulder, 216 UCB, Boulder, CO 80309, U.S.A. ||Present address: Loggerhead Instruments, Inc., 6576 Palmer Park Circle, Sarasota, FL 34238, U.S.A.

1470 © 2014 The Fisheries Society of the British Isles

E P I N E P H E L U S M O R I O S O U N D P RO D U C T I O N

1471

Wright & Jones, 2006). Ensuring sustainability, however, first requires improving the understanding of the life-history characteristics and distribution of these fish. Similar to all grouper species, E. morio are slow-growing, late-maturing, relatively stationary and long-lived. Epinephelus morio are protogynous hermaphrodites that mature first as females and change into males later in life (Moe, 1969). Collins et al. (2002) estimated that 50% sex transition for E. morio populations in the eastern Gulf of Mexico occurred at 13 years (or 800–900 mm total length, LT ); however, this change can vary considerably from 9 to 15 years (Moe, 1969; Jory & Iversen, 1989; Heemstra & Randall, 1993; Musick, 1999; Coleman et al., 2000; Sadovy, 2001). Immature females (or juveniles) are found year-round inshore (7–27 m depth), whereas juveniles, mature females, transitionals and mature males are found offshore (30–90 m depth) (Brulé et al., 1999). This is consistent with the observed increased size of E. morio with distance from the shore (Burns, 2009), adult site fidelity (Coleman et al., 2010) and offshore (c.70 m depth) spawning (Coleman et al., 1996). Epinephelus morio form small polygamous spawning groups dispersed widely throughout the West Florida Shelf (WFS), Gulf of Mexico, from late winter to early spring, peaking from March to May (Jory & Iversen, 1989; Coleman et al., 1996; Brulé et al., 1999; Koenig et al., 2000; Collins et al., 2002). These small groups, compared with large aggregations common to many grouper species, and batch fecundity (>20 spawning occurrence for one female during the spawning season) (Sadovy, 2001; Collins et al., 2002) enable E. morio to be relatively resilient to fishing pressure (Coleman et al., 1996). Yet, populations have experienced a truncated age structure (Coleman & Koenig, 2010). To date, most analyses of spawning populations derive from invasive, point-source, gonadosomatic index (I G ) examinations (Moe, 1969; Jory & Iversen, 1989; Coleman et al., 1996; Johnson et al., 1998; Brulé et al., 1999; Collins et al., 2002), leaving knowledge of the shelf-wide spatial range of spawning habitat largely undefined. Passive acoustic technology can be used to monitor soniferous fishes effectively over a wide range of habitat, depths and time periods (Mann & Lobel, 1995; Lobel, 2002; Luczkovich et al., 2008a; Van Parijs et al., 2009; Lobel et al., 2010; Locascio and Mann, 2011). Passive acoustic monitoring (PAM) systems can record large amounts of acoustic data and their application to soniferous fish (100–2000 Hz) studies has been successfully demonstrated using moored devices (Locascio & Mann, 2008; Nelson et al., 2011) and autonomous vehicles (Wall et al., 2012a). Underwater audio and video recordings have shown that E. morio produce a unique sound during courtship and territorial behaviour (Nelson et al., 2011) (Fig. 1). Videos 4 and 5 within the Supporting Information of Nelson et al. (2011) may be of particular interest as they demonstrate a correlation between sound production and courtship (www.int-res.com/articles/suppl/b012p097_supp). The sounds recorded had a peak frequency of 180 Hz. The highest sound pressure level (SPL) measured from recorders deployed next to E. morio excavations was 142 dB re 1 μPa RMS (root mean square), which provides a surrogate estimate for source level. Short duration deployments showed that E. morio produced sounds throughout the day and night with peaks near sunrise and sunset (Nelson et al., 2011). The goals of this project were to determine daily and seasonal patterns of E. morio sound production and map sound production on the WFS for this species. Passive acoustic recordings from across the WFS were collected using fixed location recorders and Slocum gliders fitted with hydrophones to explore the feasibility of this novel

© 2014 The Fisheries Society of the British Isles, Journal of Fish Biology 2014, 85, 1470–1488

1472 (a)

C . C . WA L L E T A L.

(b) 1000

3

800

Frequency (Hz)

Pressure (Pa)

2 1 0 –1 –2 –3

IP

G

P

600 400 200

1

2

3

4

5

6

7

1

2

3

4

5

6

7

Time (s)

Fig. 1. (a) Waveform and (b) spectrogram of Epinephelus morio sound. This example shows a call with four introductory pulses (IPs), a grunt (G) and a pulse train (P). The spectrogram was created using a 2048 point Hanning-windowed fast Fourier transform with 50% overlap.

method for mapping habitat and range over large scales. The daily patterns of sound production covering an entire year across a large portion of the WFS were examined and compared with the diel patterns reported by Nelson et al. (2011) determined from only two stationary locations over a 7 day period.

MATERIALS AND METHODS S T U DY A R E A The WFS extends over 200 miles off the west coast of Florida, U.S.A. and features a wide, gently sloping shelf. The inner WFS consists of a nearly flat, drowned and partially dissolved lithified carbonate (karst) platform covered by a thin layer of sediment (Hine, 1997; Brooks et al., 2003). Sediment types overlying the bedrock are highly varied and range from organic-rich mud, muddy sand, shelly sand, mixed siliciclastic and carbonate and fine quartz sand (Edwards et al., 2003; Robbins et al., 2008). To mitigate fishing pressure on E. morio aggregations during spawning, in June 2000, two marine reserves covering 687 km2 (200 square nautical miles) were established on the shelf break (50–120 m deep) of the north-eastern Gulf of Mexico–Madison Swanson (29∘ 06′ –29∘ 17′ N; 085∘ 38′ –085∘ 50′ W) and Steamboat Lumps (28∘ 03′ –28∘ 14′ N; 084∘ 37′ –084∘ 48′ W) Marine Reserves (Coleman et al., 2004a). Similar to much of the WFS, the Steamboat Lumps Marine Reserve (Steamboat Lumps) lacks geologic relief. Holes excavated by male E. morio 5–25 m in diameter, however, create structure and uncover carbonate nodules in the otherwise flat, sandy bottom of Steamboat Lumps (Scanlon et al., 2005; Coleman et al., 2010; Wall et al., 2011). The Florida Middle Grounds is a 1193 km2 area east of Steamboat Lumps, c. 200 km north-west of Tampa Bay. This area consists of two north–northwesterly parallel ridges separated by a valley. It is home to stony corals that provide extensive hard bottom and structure for numerous species of reef fishes, algae, sponges, molluscs, crustaceans and echinoderms (Coleman et al., 2004b; Waddell & Clarke, 2008). Artificial reefs and wrecks are common between the Florida Middle Ground and the coast, on the mostly flat sediment bottom. D ATA C O L L E C T I O N All acoustic data were recorded using digital spectrogram recorders (DSG; Loggerhead Instruments; www.loggerheadintruments.com). The DSG is a low-power acoustic recorder controlled by script files stored on a secure digital (SD) memory card (16 or 32 GB) and an on-board



1473

(a)

(b)

(c)

Fig. 2. Recorder designs deployed for the study. (a) Trawl-resistant housing containing the digital spectrogram recorders (DSG) ( ) with exposed hydrophone and cement base ( ) deployed in June and July 2008. (b) Mid-water column housing with protective PVC arms ( ) and PVC tubes encasing the DSG ( ) with exposed hydrophone employed for the June 2009 deployment. (c) Hydrophone-integrated Slocum glider.

real-time clock. The DSG clock maintains its accuracy with temperature-compensated drift. The DSG file system is a data file structure that stores embedded time stamps with the raw data, allowing each file to remain in synchrony with other glider or mooring data. Hydrophone [HTI-96-MIN, sensitivity −186 dBV (June and July 2008) or −170 dBV (June 2009 and Glider), ±3 dB from 2–37 kHz, High Tech, Inc.; www.hightechincusa.com] signals were digitized with 16 bit resolution by the DSG recorders.

Stationary recorders Moored acoustic recorders were deployed on the WFS for 1 month to 1 year between 2008 and 2010. Several designs were implemented as this project progressed from short-term, nearshore pilot studies to a year-long, shelf-wide deployment (Dudzinski et al., 2009). Initial deployments conducted for 1 month in June 2008 and 2–5 months in July 2009 (n = 5 and 18, respectively) employed bottom-mounted, trawl-resistant casings in which the hydrophone was exposed through a fibreglass flat-top pyramid-shaped cover; the cover was connected to a 1 m2 cement base with stainless steel cables [Fig. 2(a)]. A PVC tube located inside the casing provided a water-tight housing for the DSG and battery packs; the DSG was connected to the hydrophone through a bulkhead connector. During these deployments, the DSG recorded sound for 10 s every hour at a sample rate of 50 kHz. A larger deployment aimed at recording sound for 1 year at 63 sites on the WFS occurred between June 2009 and May 2010. The recorders were deployed in a grid 20 km apart from the coast (10 m depth) to c. 150 km offshore (100 m depth). The greater depth precluded the continued use of bottom-mounted recorders. Therefore, a 10 m subsurface design was implemented in which the PVC tubes that housed the hydrophone, DSG and battery packs were hose-clamped to polypropylene line 10 m (or shallower) below the water surface. The polypropylene line extended from the water surface where it was connected to surface and subsurface buoys


1474


down to the seafloor where it was connected to the bottom mooring constructed from two cement-filled cinderblocks joined with chain. In an attempt to protect against impact from boats and shrimp trawls, PVC cages consisting of four semicircular, buoyant arms surrounded 1–2 PVC tubes [Fig. 2(b)]. The entire PVC material was covered with anti-fouling paint. Additional recorders were attached directly to the polypropylene line near the mooring blocks for sites located in deep water (>30 m, n = 42) to ensure the recording of demersal E. morio. DSGs recorded sound for 6–8 s every hour at a rate of 36⋅4 or 50 kHz. Sample rate and frequency varied slightly among sites in an attempt to optimize recording longevity and storage capacity of the SD card. Seven additional recorders were deployed in Steamboat Lumps (71–73 m depth) and six at nearshore sites to target specifically E. morio (RG) (15–40 m depth) between April 2009 and May 2010. PVC tubes housing the hydrophone, DSG and battery packs were hose-clamped to polypropylene line within 1 m from the mooring blocks. Surface and subsurface floats attached to the polypropylene kept the line and recorder upright in the water column. All Steamboat Lumps recorders were deployed within 0⋅2 km of each other and, due to the close proximity, are considered one site in the analyses resulting in data from all recorders being binned together. These recorders sampled for 10 s every 6 min at 20 kHz.

Hydrophone-integrated gliders A hydrophone was integrated into the aft cowling of four Slocum electric underwater gliders (Teledyne Webb Research; www.webbresearch.com) for passive recording of sound with an on-board DSG board while concurrently collecting a suite of environmental and optical variables (Wall et al., 2012a) [Fig. 2(c)]. Slocum gliders are buoyancy-driven autonomous underwater vehicles (AUVs) 1⋅8 m in length and are shaped like a winged torpedo (Webb et al., 2001; Schofield et al., 2007). They can traverse over 600 km using a single set of alkaline batteries and contain sensors tailored towards scientific applications (Schofield et al., 2007). Glider depth in the water column and the bottom depth were measured using an on-board pressure sensor and an altimeter. The glider’s DSG recorded sound for 25 s every 5 min at a sample rate of 70 kHz. The gliders move c. 75–85 m between acoustic recordings (Wall et al., 2012a). The DSG clock is synchronized to the clock on-board the glider’s computer and thus the environmental sensors. A total of 15 glider deployments of 1–4 weeks in length were conducted between April 2009 and April 2011 covering a range of depths (up to 984 m) depending on the deployment path. These paths were opportunistic, exploratory and, in part, dictated by collaborating scientists. Gliders were run and maintained at the University of South Florida, Center for Ocean Technology. D ATA C ATA L O G U E Stationary recorders All acoustic data and associated metadata were catalogued in a custom MySQL database, and the files were stored on a 192TB Sun Fire X4400 server called Ocean Observing Metadata Archive (OOMA) using a MATLAB (Mathworks; www.mathworks.co.uk) interface. OOMA is stored at the University of South Florida, College of Marine Science. Metadata included hydrophone sensitivity, latitude and longitude co-ordinates of the recorder site, water depth, sample rate, file size, coordinated universal time (UTC) time stamp of the recording, file path and file name on the server, and UTC time stamp of the recorder deployment and recovery times. Hydrophone-integrated gliders Acoustic data collected by the gliders were also catalogued in the database and stored on the server. Metadata collected throughout the deployment included glider depth in the water column, bottom depth, UTC time stamp, roll, pitch and heading. Latitude and longitude positions were collected using GPS satellites when the glider was at the surface. The position of the glider when



1475

not at the surface was estimated from the surface latitude and longitude co-ordinates using linear interpolation and a 10 point moving average.

D ATA A N A LY S I S Steamboat Lumps time series An automatic detection algorithm was developed to identify E. morio calls within the time domain of sound files. The number of introductory pulses and presence of a pulse train vary among E. morio calls (Nelson et al., 2011), so the algorithm targeted the statistical variables of one introductory pulse and the grunt, which are always present. Each acoustic file was first resampled at a lower rate (one tenth of the original sample rate with low-pass filtering) to enable faster processing. The RMS values of three bands, each 130 Hz wide, were used to measure proxies of E. morio sound level (band 1; 50–180 Hz), sound level of other fish calls (band 2; 270–400 Hz) (Lobel, 1992; Coleman et al., 2010; Lobel et al., 2010) and the level of ambient and anthropogenic noise (band 3; 869–999 Hz). Steamboat Lumps is heavily populated with E. morio and, therefore, this species dominates the sound production in band 1. This circumstance, however, would not necessarily hold in other areas. Band 2 represents calls from other fishes, such as Pomacentrids, Scarids and Haemulidae (Coleman et al., 2010). Ambient noise covers a broad range of frequencies (Urick, 1983), but sounds produced by fishes are less likely to be present in the band 3 frequency range. For consistency, the 130 Hz bandwidth selected for band 1 was applied to bands 2 and 3. For automated detection within band 1, data were normalized by the maximum signal level, rectified and enveloped with a 75 ms window. High-amplitude, narrow-band pulses that result from hydrophone interference were removed through automatic detection to reduce noise and increase the ability of the algorithm to detect accurately the more subtle introductory pulse and grunt. Once noise-related pulses were removed, data were normalized again and signal values above a threshold, representing potential grunts, were identified. The threshold value of 0⋅12 was determined from previous analyses of files containing E. morio to optimize the detection of calls with high signal to noise and ignore those with low amplitude. For each potential grunt, the duration, time in the file, peak frequency and 3 dB bandwidth were calculated. The duration of any prior grunts, representative of introductory pulses (introductory pulse intervals), was also identified. The peak frequency and 3 dB bandwidth were calculated from a fast Fourier transform (FFT) of the signal with a 5 Hz resolution. A MySQL table was created to catalogue each file that contained grunts. Catalogued information included file attributes (file time stamp and location), grunt attributes (duration, time in file, peak frequency, 3 dB bandwidth and introductory pulse intervals) and RMS levels (bands 1–3). These data were then subjected to a suite of variable restrictions to select for E. morio and minimize false detections due to noise. Variable restriction values were determined from E. morio call statistics calculated by Nelson et al. (2011) and from a training library recorded during this study (1306 files containing E. morio calls; 1476 files containing noise). Grunt duration was set between 0⋅35 and 0⋅71 s, peak frequency was set between 78 and 194 Hz, 3 dB bandwidth was set to ≤140 Hz, introductory pulse intervals were set to 0⋅49 and 0⋅79 s and RMS band values were set to ≤0⋅34, 5⋅8–9⋅6 and 5⋅8–6⋅1 for bands 1, 2 and 3, respectively. Files with detections that met these restriction criteria were then manually verified. When run on the training library, the detection algorithm accurately detected E. morio 44% of the time (true positives), falsely detected E. morio 4% of the time (false positives) and missed detecting E. morio 30% of the time (false negatives). Although the level of true positives is low, a conservative detector was deliberately sought to reduce significantly the number of false positives. The algorithm showed positive results when applied to the Steamboat Lumps dataset. Manual verification of files with detections that met the restriction criteria showed that 92⋅5% contained E. morio sounds (7⋅5% false detection rate). The false-negative rate is unknown. As only 30% of E. morio calls were missed by the detector, the conservative variables used in this approach should neither significantly affect the results nor impede the validity of the algorithm. Furthermore, it is assumed that false negatives occur at random (both diurnal and seasonally), and that the overall distribution of calls detected will not change.


1476


The detection algorithm was applied to the Steamboat Lumps acoustic files stored on OOMA using MATLAB scripts run on a computer cluster. Analysis of the restriction criteria was run using MATLAB programmes directly on OOMA. Visual verification of files that met the restriction criteria was analysed using MATLAB on a local machine. Daily counts of files containing E. morio sounds were extracted from the Steamboat Lumps analysis. An FFT was applied to the time series of daily counts to determine whether there were cyclical peaks in calling. These peaks were then compared with the cyclical pattern of moon phases to determine if E. morio calling was influenced by lunar phase.

Mapping of E. morio sound production Although the detection algorithm showed promise for sites offshore and highly populated with E. morio, inshore sites that were more susceptible to boat traffic, equipment noise (e.g. movement of the PVC tube) and less likely to have E. morio proved to be very difficult for the detection algorithm to detect E. morio with high accuracy and avoid false positives. Therefore, files collected at the remaining stationary sites were analysed manually by visually inspecting spectrograms to identify the acoustic presence of E. morio. Spectrograms were created using 2048 point Hanning-windowed FFTs with 50% overlap. Due to the large number of acoustic files recorded throughout the deployments, only files recorded between 1600 and 2200 hours (local time) were reviewed. Epinephelus morio call 24 h per day with peaks at dawn and dusk ensuring that this timeframe will reveal their sound production if present (Nelson et al., 2011). The RG recorders recorded more frequently (10 s for every 6 min) than the other stationary recorders (6–10 s every hour). To reduce sampling bias, one file per hour of the RG recorders was analysed. Acoustic files recorded during the glider missions contained electrical and mechanical noises, which similarly prevented the successful application of the detection algorithm. Therefore, all files were analysed manually. Detections were binned into 1 h intervals over the glider track. Temporal binning also resulted in spatial binning as the glider moved continuously. Therefore, the interpolated co-ordinates of the file closest to the 30 min mark were used to display the spatial position for that 1 h bin. As the glider moves at c. 0⋅9–1⋅8 km h−1 (c. 0⋅5–1⋅0 knots), binning over a 1 h period did not appreciably affect the spatial representation of these data. Hourly bins and data display were completed using MATLAB (mathworks; www.mathworks.com) and ArcGIS (Environmental Systems Research Institute; www.esri.com). Environmental data Temperature was recorded once every 12 min using HOBO data loggers (www.onstcomp.com) at Steamboat Lumps and analysed using HOBOware Pro software and MATLAB. Sea-surface temperature (SST) data were derived from infra-red data collected by NASA’s moderate resolution imaging spectroradiometers (MODIS) onboard satellites Aqua and Terra. Chlorophyll a concentrations (Chl) were calculated from visible data collected by ORBIMAGE’s sea-viewing wide field-of-view sensor (SeaWiFS) using standard SeaDAS processing. SST anomaly data were calculated as the difference between a location’s weekly mean SST and weekly mean climatology based on SST data from 2001 to 2010. All satellite data processing was performed using IDL (Research Systems, Inc.; www.exelisvis.com). SST, Chl and SST anomaly were correlated to sound production data using Pearson’s product–moment coefficient in MATLAB. High-resolution bathymetry data were collected in Steamboat Lumps in 2009 using a Kongsberg EM3000 (300 kHz) multibeam sonar (Wall et al., 2011). These data show holes excavated by male E. morio and indicate areas of potential spawning habitat. The locations of E. morio sound production in Steamboat Lumps identified here with the passive acoustic data were compared with the locations of holes identified in the 2009 active acoustic data. Daily and seasonal patterns in Steamboat Lumps E. morio sound production were compared with sunrise and sunset, and lunar cycle data obtained from the UNSO database from June 2009 to May 2010 (USNO, 2012a, b). The presence of boat noise and other soniferous fish species within the acoustic files collected from both stationary recorders and gliders was noted to understand the effect, if any, of these sounds on the detection of E. morio sound production.



1477

RESULTS D ATA C O L L E C T I O N Stationary recorders

Nineteen of the 23 recorders deployed in 2008 were recovered (five from June; 14 from July; 83% recovery rate). A total of 16 482 files were recorded from these stationary sites. Of the 63 recorders deployed in 2009, 29 were recovered (46% recovery rate) resulting in an additional 121 524 acoustic files. Four recorders deployed in Steamboat Lumps in 2009 (162 877 files) and four recorders deployed at the RG sites in 2009 (101 862 files) were successfully recovered (57 and 67% recovery rate). After removing files recorded before and after the deployment period, the total acoustic library collected during this study consisted of 377 728 files. The location of recovered and unrecovered stationary recorders is illustrated in Fig. 3. Details of recording duration for each recorder are outlined by Wall et al. (2013). Some recorders stopped recording before deployment or ended prematurely. On retrieval, four recorders showed battery corrosion and failure due to a leak in the PVC tube. When battery leaks were not evident, software errors are suspected to be the cause. Hydrophone-integrated gliders

A total of 15 glider missions were conducted between April 2009 and April 2011 (Fig. 3). During this time, missions were run during all months except May, August, November and December, and data were collected during all hours of the day. A total of 25 760 files were recorded over the various glider tracks. All gliders deployed were successfully retrieved; however, acoustic recordings stopped before recovery due to filled storage space on the SD card for some of the longer missions. Details of recording duration for each glider mission are outlined by Wall et al. (2013). D ATA A N A LY S I S Steamboat Lumps time series

In Steamboat Lumps, E. morio produced sound throughout the day and night, and during all months sound was recorded (Fig. 4). Monthly calling increased progressively from spring to summer (March to August), decreased in the autumn and peaked again in early winter (November to December). Diurnal peaks in calling coincide with the seasonal shift in sunrise and sunset times. Peaks in E. morio sound production (day 12) did not correspond to any lunar cycle (7, 15 or 29 days). Weekly binned E. morio detection counts from the stationary sites were significantly correlated to SST, SST anomaly and Chl (Pearson’s product–moment correlation coefficient, r = 0⋅66, 0⋅61, −0⋅45, d.f. = 7480, P < 0⋅05, respectively). Daily binned E. morio detection counts from Steamboat Lumps were significantly correlated to the water temperature collected near the seafloor (r = 0⋅62 d.f. = 321, P < 0⋅05). The positive correlation of weekly detection counts to SST and SST anomaly, and the negative correlation to Chl are illustrated in Fig. 5. Bottom temperature shows a similar but more muted pattern of seasonal temperature changes compared with SST. Mapping of E. morio sound production

The spatial area over which E. morio sounds were identified from manual analysis is illustrated in Fig. 6. The symbol size is proportional to the percentage of files


1478


(a)

84° W

83° W

3

2

Tampa Bay

1

28° N

28° N

5

4 6

8

9 9 14 14

0m

0m

20

10

13 12

17 17 16

15

20

19

27° N 0m

5

(b)

B2

B3

0m

0m

84° W

83° W

84° W

83° W

B4

B5

B6

B13

28° N

B7 B8 B7b B15

0 m 1

20

3

4

B9

m

27° N

Tampa Bay

B17

28° N

B33 B40

RG3

B49

0m

20

0m

B42 RG1

B50

B51

B44 B52

B58

10

B53

B61

B62

RG2

27° N 50

84° W

(c)

86° W

85° W

m

40

m

30

m

20

0 m 1

m 27° N

83° W 84° W

83° W

82° W

Tampa Bay

28° N

28° N

27° N

27° N 0 00

1

86° W

85° W

m 2

m 00

m 00

1

84° W

m 40 m 30 m 20 m 0 m 50 1

83° W

82° W

Fig. 3. Map of the stationary acoustic recorders. (a) Positions of recovered recorders deployed in June ( ) and July 2008 ( ), and unrecovered recorders deployed in July 2008 ( ). Station labels for the June 2008 deployment are in grey. (b) Positions of recovered ( ) and unrecovered recorders ( ) deployed in June 2009, and recorders deployed to target E. morio inshore (RG) ( ) and in Steamboat Lumps ( ). B5 and B5b, B6 and B6b, B8 and B8b, B9 and B9b, and B33 and B33b are located at the same site and only one station label is shown. (c) Interpolated tracks for hydrophone-integrated glider missions between 2008 and 2011. indicates the area highlighted in the inset ( ). indicates area of Steamboat Lumps ( , mission 16; , mission 25; , mission 31; , mission 37; , mission 39; , mission 40; , mission 43; , mission 44; , mission 46; , mission 47; , mission 49; , mission 50; , mission 51; , mission 52; , mission 53).


1479


(a) (b)

0·06

Calls per unit effort

0·05 0·04 0·03 0·02 0·01 December

November

October

August

September

July

10 12 14 16 18 20 22 Hour

June

8

May

6

April

4

January

2

March

ND 0

February

0

Month 2 4 6 8 10 12 14 16

60 50 40

No data

Hour

(c)

30 20

18 20 22 December

November

October

August

September

July

June

May

April

March

January

February

10 0

Month

Fig. 4. Epinephelus morio calls identified by the detection algorithm in the Steamboat Lumps time series. Calls per unit effort were calculated based on the number of files that contained E. morio sound production normalized by the number of files analysed binned (a) hourly and (b) monthly. indicate mean sunrise (0705 hours) and sunset (1940 hours) throughout the year. (c) Matrix of the number of E. morio calls per hour for each month. The colour bar indicates more calls per hour for each month. indicate sunrise (top) and sunset (bottom) times. All times are local (EST, eastern standard time). No acoustic data (ND) were recorded in October.

that E. morio were present in stationary recorders compared with the number of files analysed. The highest percentages of files containing E. morio are found between the 30 and 50 m isobaths (0–10 m: 0⋅32, n = 1224; 11–20 m: 1⋅0, n = 10 159; 21–30 m: 0⋅7, n = 10 908; 31–40 m: 15⋅9, n = 12 579; 41–50 m: 4⋅0, n = 12 096; 51–60 m: 1⋅0, n = 686; 61–70 m: no data; 71–80 m: 0⋅89, n = 1677). Hourly bins of E. morio sounds recorded along the interpolated track for each glider mission are also shown. The bin values represent the number of files that contained E. morio sound within that hour. As files are recorded every 5 min, the maximum number possible is 12. Directly west of Tampa Bay, all recordings identify that most E. morio calls are detected between the 30 and 50 m isobaths. In the northern portion


1480 40

300

30

200

20

100

10 23/04/2009

(b)

12/06/2009

01/08/2009

20/09/2009

09/11/2009

29/12/2009

17/02/2010

08/04/2010

0 300

4

SST anomaly (°C)

2 200 0 100 –2

–4 23/04/2009

Chlorophyll a (mg m–3)

(c)

12/06/2009

01/08/2009

20/09/2009

09/11/2009

29/12/2009

17/02/2010

08/04/2010

0

10

300

1

200

0·1

100

0·01 23/04/2009

12/06/2009

01/08/2009

20/09/2009

09/11/2009

29/12/2009

17/02/2010

08/04/2010

Weekly detection count

Water temperature (°C)

(a)


0

Fig. 5. Time series of sound production and environmental data from Steamboat Lumps. Weekly Epinephelus morio detection counts ( ) are shown with associated (a) sea surface temperature (SST) ( ) and in situ bottom temperature ( ), (b) SST anomaly and (c) chlorophyll a.

of the study area, E. morio are consistently observed from 30 to 93 m water depths. No E. morio were detected in depths greater than 93 m. The percentage of glider files containing E. morio by 10 m isobath occurred as follows: 0–10 m: no data; 11–20 m: no data; 21–30 m: 1⋅7, n = 1929; 31–40 m: 16⋅8, n = 7223; 41–50 m: 10⋅3, n = 4864; 51–60 m: 7⋅0, n = 2207; 61–70 m: 7⋅2, n = 1505; 71–80 m: 19⋅0,


1481


85° W

84° W

83° W

Tampa Bay

28° N

28° N

27° N

27° N 100

85° W

m

84° W

5

0m

0m

4

30

m

20

0m

m 1

83° W

Fig. 6. Epinephelus morio call detection rates from glider data (detections by hour) and manually analysed stationary data (file percentage). For glider data, symbol colour is proportional to the number of files per hour that contain E. morio sound production ( , 1–2; , 3–4; , 5–6; , 7–9; , 10–12). For stationary data, symbol size is proportional to the per cent of files E. morio sound production was detected, out of the total number of files analysed per site ( , 0⋅11–0⋅65; , 0⋅65–1⋅33; , 1⋅33–4⋅92; , 4⋅92–9⋅23; , 9⋅23–16⋅17). Stationary recorder locations in which no E. morio calls were identified are indicated ( ). Locations of the Steamboat Lumps recorders ( ), the boundaries of Steamboat Lumps ( ) and the lower boundary of the Florida Middle Grounds ( ) are also shown.

n = 2556; 81–90 m: 5⋅1, n = 1082; 91–100 m: 0⋅5, n = 403; 101–984 m: 0, n = 3006. Boat noise was recorded in 2⋅6% of the files from stationary recorders (1647/63 993) and 1⋅6% of the files from glider deployments (403/25 129). Sounds made by other fish species were observed in 14⋅9% of the files recorded by stationary recorders (9580/63 933) and 37⋅2% of the files recorded during glider deployments (9356/25 129). Calls from other fish species co-occurred in sound files containing E. morio in 7⋅4% of the stationary (554/7481 files) and 26⋅2% of the glider recordings (678/2581 files). The track for mission 53 deployed during 29 March to 15 April 2011 was specifically selected to overlap the area within Steamboat Lumps where stationary recorders had been deployed. Holes excavated by male E. morio in Steamboat Lumps have been identified from multibeam data collected in 2009 (Wall et al., 2011). The location of these holes, the position of stationary Steamboat Lumps recorders deployed in 2009 and the tracks and E. morio detection counts by hour identified from mission 53 glider data are illustrated in Fig. 7. High (10–12) E. morio hourly counts were observed within 200 m (mean ± s.d. = 81 ± 66 m, n = 15) of higher hole density areas. The number of E. morio files h−1 decreased in areas away from the densely located holes, with the exception of one area located near the right-centre of the image. The location of the stationary recorders deployed in Steamboat Lumps is within the densely populated area of holes.


1482


1 km

0·5 km

Fig. 7. Epinephelus morio detection counts identified in mission 53, binned by hour, overlaid on multibeam data collected in Steamboat Lumps in 2009. Location of the mission 53 interpolated track ( ), stationary recorders deployed in Steamboat Lumps ( ) and the multibeam data delineating the E. morio holes that were presented in Wall et al. (2011) are also shown. To aid the reader, an area densely populated with holes is outlined on the multibeam data ( ). within the inset shows where multibeam data were collected ( , 1–2; , 3–4; , 5–6; , 7–9; , 10–12).

DISCUSSION T I M I N G O F S O U N D P R O D U C T I O N A N D S PAW N I N G

Epinephelus morio sound production was observed throughout the day and night, and during all months in which data were recorded. Peaks in calling corresponded to sunrise and sunset, a pattern that remained as the seasons changed and daylight hours varied. Sound production peaked in late summer (July and August) and early winter (November and December). The hourly pattern in sound production observed in the Steamboat Lumps time series reflects the hourly calling peaks of E. morio recorded over a shorter deployment period in the same area by Nelson et al. (2011). Despite a low positive detection rate expected from the detection algorithm applied to the Steamboat Lumps dataset, this analysis still accurately captured the diel pattern of E. morio sound production over an 11 month time period. Gonadosomatic index (I G ) data collected from E. morio in the eastern Gulf of Mexico indicate that spawning occurs in the late winter to late spring (March to May) with a peak in April (Moe, 1969; Coleman et al., 1996; Collins et al., 2002). This is contrary to the peaks in yearly sound production identified in this study. A 10 year dataset (1991–2001) of female I G show variation within spawning peaks (36–82% of



1483

active females observed between March and May), and active females were observed, although infrequently, throughout the year (Collins et al., 2002). Collins et al. (2002) hypothesized that larger and older fish may spawn longer than smaller and younger fish due to the observation of a large, old E. morio with a large histological stage 4 ovary in August. It may be possible that spawning patterns in Steamboat Lumps may vary from previous I G work, namely if this offshore area contains a higher percentage of older fish. Spawning patterns, however, would have to be shifted 3 months later in Steamboat Lumps (from March to May to June to August) if sound production was completely resultant from spawning behaviour. Although linkages between fish sound production and spawning have been identified for numerous species (Mok & Gilmore, 1983; Lobel, 2002; Lowerre-Barbieri et al., 2008; Luczkovich et al., 2008a, b; Walters et al., 2009), the lack of synchrony between yearly sound production peaks observed in this study and known spawning peaks strongly suggests that E. morio do not produce sound solely during courtship. An alternative possibility is that sounds produced outside of the expected spawning period are associated with territorial behaviour (Mann & Lobel, 1995; Nelson et al., 2011). Species that maintain territories all year, such as male E. morio, vocalize outside of the breeding season (Lobel et al., 2010). In addition, increased sound production is expected from movement of new fish into an area of older fish with established territories (Lobel et al., 2010). Therefore, offshore movement of younger E. morio (Burns, 2009) could result in an increased occurrence of conspecific interaction outside of spawning season. Unfortunately, seasonal patterns cannot be discerned from the coarse resolution of these tagging data, which serves as the only work to date that documents migration patterns of E. morio (Burns, 2009). Further research that incorporates long-term video monitoring in addition to acoustic recordings is needed to better understand seasonal peaks in E. morio sound production and associated behaviour. B I O G E O G R A P H Y O F S O U N D P RO D U C T I O N

The spatial range over which E. morio sounds were detected is quite extensive, from c. 15 to 93 m water depth. The majority of E. morio sound production occurred between 30 and 50 m depth. Along the northern boundary of the study area, E. morio sounds were detected predominately between 40 and 93 m depth. Preferential habitat in the Florida Middle Grounds (hard bottom) and Steamboat Lumps (excavated sediment) is suspected to be the cause for the slight shift in depth range in this region. The lack of sound production in depths beyond 100 m does not imply that E. morio are absent as they are expected to inhabit depths up to 400 m (Moe, 1969). It does suggest, however, that it is unlikely. The acoustic presence of E. morio in shallow water depths (