Hydrophone placement yields high variability in detection of Epinephelus striatus calls at a spawning site

Abstract

Passive acoustic monitoring is a cost-effective, minimally invasive technology commonly used to study behavior and population dynamics of soniferous fish species. To understand the strengths and limitations of acoustic monitoring for this purpose at fish spawning aggregations (FSA) requires an assessment of the variability in aggregation-associated sounds (AAS) as a function of time, space, and proximity for spawning fishes of interest. Here, we evaluate temporal and spatial trends in the detection of AAS by Nassau Grouper (Epinephelus striatus) using an array of six hydrophones deployed across a large Nassau Grouper FSA at Little Cayman, Cayman Islands. We collected continuous data for nine days during a winter spawning season and subsequently used an automatic classifier to extract the embedded Nassau Grouper AAS. Using these data, we analyzed variability in spatiotemporal AAS detection rates across the array with a Bayesian mixed effects model. We found high variability in the detection of AAS across the spawning site, with positive correlations among neighboring hydrophone pairs trending toward negative correlations with distances exceeding 350 m. Indeed, temporal trends in AAS rates at the spawning site were approximately inverted at the two most distant hydrophones (~600 m). Across the hydrophone network, our model predicted strong positive effects of fish proximity, spawning behavior, and crepuscular periods on detected AAS. Our findings suggest hydrophone placement can strongly influence AAS detection rates and even basic temporal patterns in AAS across the spawning season. Given both the vagaries of movement and behavior of aggregating fish at spawning sites and the limits of AAS detection using standard monitoring tools, we suggest spawning site acoustic monitoring programs deploy hydrophone arrays of sufficient size to capture the site-wide trends in AAS rates if possible; this is particularly true if researchers hope to compare/contrast AAS rates between spawning sites or across seasons for the purpose of population assessment.

Keywords: FSA; automatic classifier; fish calls; hydroacoustics; movement ecology; spawning aggregation.

FIGURE 1

Little Cayman, Cayman Islands. (a) The Cayman Islands in the Caribbean Sea relative to Cuba, Florida, United States, and Mexico. Grand Cayman, the largest island in the Caymans, is just southwest of the drawn black box. (b) Little Cayman and Cayman Brac. A black dot off Little Cayman's west end signifies the fish spawning aggregation's (FSA) location. (c) Bathymetry of the west end of Little Cayman. Blues lighten to symbolize increasing depth on the order of 10 m, with land represented in black. Depth beyond 40 m is transparent. A dark red, bold “X” marks where divers first observed spawning individuals (between ST6 and LS5), and a dark red, bold “*” marks the traditional FSA location where fish gathered during the first few days of observation (adjacent to LS3). Triangles represent hydrophones and are labeled with their station. Black triangles signify SoundTrap (ST) hydrophones and white triangles represent LS1 (LS) hydrophones.

FIGURE 2

Heat maps of (a) categorized proximities of the bulk of Nassau Grouper (Epinephelus striatus) to each hydrophone per hour (see Data analysis for category descriptions) compared to (b) effort normalized aggregation‐associated sounds (AAS) detection rates at each hydrophone per hour. (a) Categorized proximities of the bulk of Nassau Grouper to each hydrophone per hour. As dark green panels lighten, the bulk of Nassau Grouper approach the hydrophone, with the lightest green signifying within 20 m of the hydrophone. (b) Continuous effort normalized AAS detection rates at each hydrophone. Dark greens lighten to white to signify increasing rates of AAS detected. Gray regions indicate when no data were collected. Black bars positioned above both panel (a) and (b) signify days when divers observed Nassau Grouper spawning (February 12 0:00 to February 15 0:00).

FIGURE 3

Temporal trends of effort normalized Nassau Grouper (Epinephelus striatus) aggregation‐associated sounds (AAS) detection rates across all hydrophones. Because ST4 abruptly halted recording at midnight of 3 days after the first night of spawning (DAFS), black rectangles cover 3–5 DAFS for ST4 in (b) and (c). (a) Box plots of AAS detection rates for all hours of day at each hydrophone. Dark gray rectangles signify hours at night while light gray rectangles signify hours in which the sun rose (06:00) and set (18:00). (b) Box plots of AAS detection rates per hour binned by day at each hydrophone. Yellow circles outlined black symbolize the full moon which occurred −3 DAFS. (c) AAS detection rates per hour for all continuous hours of observation. Orange points indicate effort normalized AAS counts per hour of day. A green spline fits to the orange points by the loess method with a span factor of 0.1. The dark gray shade around the fitted spline represents the 89% CI. Dotted black lines separate days. Note that the first night of spawning (DAFS 0) is February 12, 2020. For all box plots: the black bar of the box is the median value; the area of the box encapsulates the middle 50% of the data (i.e., the first and third quantiles of data); the whiskers extend to the data points nearest to 1.5 × the interquartile range from the first quantile (low end) and third quantile (high end); and outliers are omitted.

FIGURE 4

Correlation pairs plot of model‐predicted correlations between each hydrophone pair after accounting for all other model effects (i.e., time of day, day of spawning period, and proximity of the mass of fish). The slope and color of the ellipses symbolize the strength of correlation (e.g., strong positive correlations are narrow ellipses sloped positively and colored dark green). Straight, positively sloped lines occupy comparisons of identical pairs (e.g., LS1‐LS1).

FIGURE 5

Linear regression of correlation coefficients against distance between hydrophone pairs in meters. Light green points symbolize the correlation coefficient of all hydrophone pairs at their distance of separation, with corresponding labels of the pair adjacent to each point. Whiskers extending from each point represent 25% and 75% posterior quantiles around the median estimated correlation. The 95% CI in the regression is shaded in light gray. A dashed black line extends from y = 0 for legibility. The “geosphere” package (Hijmans, ; version 1.5‐20) in R estimated distances between pairs of hydrophones from their coordinates.

FIGURE 6

Box plots of posterior predictions of the parameters (a) T _i , (b) D _i , and (c) FP _i (see Data analysis for variable descriptions). (a) Posterior predictions of aggregation‐associated sounds (AAS) detection rates by T _i . Dark gray panels signify nighttime hours while light gray panels represent the hours of sunrise (06:00) and sunset (18:00). (b) Posterior predictions of AAS detection rates by D _i . A yellow circle outlined black signifies the full moon on −3 days after first spawn (DAFS). (c) Posterior predictions of AAS detection rates by FP _i . Above the boxplots are cartoons roughly representing each category of fish proximity to a hydrophone. For all box plots: the black bar of the box is the median value; the area of the box encapsulates the middle 50% of the data (i.e., the first and third quantiles of data); the whiskers extend to the data points nearest to 1.5 × the interquartile range from the first quantile (low end) and third quantile (high end); and outliers are omitted. Illustration credit: Cameron J. Van Horn.