Age, growth, and recruitment patterns of juvenile ladyfish (Elops sp) from the east coast of Florida (USA)

Study area

Field-collections were made at numerous locations throughout the Indian River Lagoon (IRL) and Volusia County (VC (Tomoko River Basin, Ponce de Leon Inlet, and Mosquito Lagoon complex)). Field sampling was conducted by Florida Fish and Wildlife Conservation Commission (FWC), Fisheries Independent Monitoring (FIM), personnel at 21 (seines [8], trawls [11], and gillnet [2]) pre-determined stations (i.e., fixed stations [FS]) in the IRL (Fig. 1) and 29 (seines [14] and trawl [15]) FS in VC (Fig. 2); FS were stratified by geographical location, habitat, and depth (McMichael et al., 1995). Further details on site descriptions are provided by Levesque (2013).

Gear and sampling methodology

Field sampling at FS was conducted once a month during daylight (i.e., the period between one hour after sunrise and one hour before sunset). Three haul repetitions were made at each station with a center-bag seine (21.3 m long by 1.8 m high; center bag constructed of 3.2 mm #35 knotless nylon Delta mesh). Based on the profile of the beach (i.e., bank slope) and water depth, one of three deployment methods (beach, boat, or offshore) were used to deploy the center-bag seine (i.e., seine) at each station (McMichael et al., 1995). The first deployment technique was the beach method. A beach deployment method was used when the water depth was shallow and the bank had either a gradual slope or no slope. The beach deployment method consisted of the seine being pulled parallel to shore by two biologists for a total distance of 9.1 m; a 15.5 m line stretched between each seine pole was used to assure the net was being pulled the same inner-pole distance for every haul. The second deployment technique was the boat deployment method. A boat deployment method was used when the water was either to deep (water depth 0.7–1.2 m) or the bank was too steep to use a beach deployment. The boat deployment method consisted of deploying the seine from the stern in a semi-circular pattern along the bank. Once the seine was fully deployed, two biologists would pull the seine toward shore. The third and final deployment method was the offshore deployment method. An offshore deployment was used when there was either no available beach or it was too shallow to reach the beach bank by boat. The offshore deployment followed the same procedures as the beach deployment with one minor difference; at the end of the 9.1 m distance, two biologists worked the seine using a stationary pivot pole to ensure the catch did not escape (McMichael et al., 1995). Given the seine dimensions and the distance traveled (9.1 m) along the beach, the total area sampled with the beach seine was 100 m².

Data

The FWC used two experimental field sampling approaches in the 1990s to survey fish throughout Florida (McMichael et al., 1995): monthly FS and year-round stratified random sampling (SRS). The FWC conducted fisheries monitoring using a variety of sampling gears, such as center-bag seines, otter trawls, gillnets, blocknets, and dropnets. For these analyses, data was restricted to monthly FS collections of ladyfish collected with a center-bag seine because fewer juvenile ladyfish were collected with the SRS approach. Therefore, pooling the datasets (SRS and FS) could have bias the analyses by under- or over-estimating size-at-age. Also, most ladyfish collected by the SRS approach were larger and older than the pre-selected maximum cut-off length of 100 mm SL. Following Levesque (2014), a maximum cut-off length of 100 mm SL was chosen because ladyfish larger than 100 mm SL could avoid some field sampling gear (i.e., small-mesh center-bag seines). After every net haul, fish were sorted, enumerated, and measured to the nearest 1 mm standard length (SL); a total of 20 individuals of every species were measured. It should be noted these data were collected prior to McBride et al. (2010) describing Elops smithi; Elops saurus and Elops smithi are both found on the east coast of Florida. Unfortunately, the FWC team was unaware at the time of the study that there were potentially two different species of ladyfish that could be found within the study area. As such, my findings only refer to the genus Elops.

Statistical analysis

Data were evaluated for normality and homoscedacity (variance (equivalently standard deviation) are equal) using Kolmogorov–Smirnov (Zar, 1999) and Bartlett (Bartlett, 1937a; Bartlett, 1937b) tests, respectively. If the data passed the normality tests, then parametric procedures were followed; otherwise, the data were log-transformed [log(X + 1)] to meet the underlying assumptions of normality (Zar, 1999). Non-parametric procedures were applied if the data could not meet the assumptions of normality after transformation. A post-hoc multiple comparison test was used to perform pairwise comparisons in the presence of significance at the 95% confidence level for either the Analysis of Variance (ANOVA) or Kruskal–Wallis non-parametric multi-sample tests. All analyses were conducted using Microsoft Excel^® and Statgraphics Centurion XVI^® Version 16.1. Statistical significance was defined as p < 0.05.

To estimate growth, monthly field collections of cohort lengths were categorized into 5 mm SL size classes, graphed, and evaluated. Descriptive statistics (e.g., mean, standard deviation, variance, standard error) were derived and cohorts identified using modal progression analysis (MPA); MPA consisted of plotting the mean SL and the collection date (Petersen, 1892). Before evaluating cohort modal progressions, a one-way ANOVA test was used to distinguish whether there was a significant difference in length among months, years and locations. Annual ladyfish growth was estimated by regression analyses of the monthly geometric mean SL on capture date. Growth was described by linear (SL = slope [age] + y-intercept) and nonlinear regression. The coefficient of determination value was used to choose the most parsimonious (i.e., the model that best fit the data) growth model. Exponential growth regression was described with the following equation: ${\displaystyle \mathrm{SL}=L_o{\mathrm{e}}^{Gt}}$ where, SL = standard length (mm); G = instantaneous growth coefficient (per month); L_o = initial SL (mm) size at first capture; t = the time (per month) for the average individual in the length-class to achieve the indicated size.

The relative instantaneous growth coefficient (G) was estimated by calculating the average time individuals in a year-class attained a certain length (Deegan, 1990). The instantaneous growth coefficient was used to represent the average growth of the population during the time period (Ricker, 1975). The absolute daily growth rate was estimated by the following equation: ${\displaystyle G=\Delta l\left(l_2-l_1\right)/\Delta t\left(t_2-t_1\right)}$ where, l₂ = SL (mm) at the end of a unit of time; l₁ = initial SL (mm) at time 0; t₂ = at the end of a unit of time (days); t₁ = initial time 0 (days).

Analysis of Covariance (ANCOVA) was used to determine if the slopes of the regression lines were significantly different (homogeneity of slopes assumption); significance criteria (homogeneity of y-intercepts and coincidental slopes and intercepts of the regression lines) was achieved when the parallelism of slopes assumption was met. If annual growth rates were equal, then the data were pooled. Following Ricker (1975), it was assumed: (1) the population sampled had a normal distribution; (2) the size classes (captured) were not influenced by gear or sampling methods; (3) mortality was the only natural population influence; and (4) the population was resident to the sampling location (i.e., lack of immigration or emigration). In general, I assumed a steady-state and a closed-population. These basic assumptions are often applied to derive various life-history estimates, including mortality and recruitment (Ziegler, Welsford & Constable, 2011). Based on Levesque (2014) and McBride et al. (2001), these population assumptions seemed reasonable because the data was limited to seine gear, and most of the sampling stations were located in ideal ladyfish habitat (Eldred & Lyons, 1966; Gilmore et al., 1981; McBride et al., 2001; Florida Fish and Wildlife Conservation Commission, 2006).

Growth and growth rates were evaluated to ensure estimates were realistic and biologically accurate given ladyfish have a metamorphic development (i.e., leptocephalus). Since ladyfish early development consists of the body shrinking before it transitions into the juvenile stage, estimating growth is somewhat challenging compared to most fish, especially if attempting to back-calculate size and age. If the derived size was unrealistic both in terms of recruitment and projected age-1 length, then size was corrected (y-intercept) to compensate for the unrealistic smaller predicted recruitment size and larger projected age-1 length. Using linear regression, the y-intercept of the exponential regression formula was corrected (standardized) to 21 mm SL to better reflect natural growth. The 21 mm SL was selected because it is generally the length ladyfish have before transitioning from the leptocephalus to the juvenile stage (Alikunhi & Rao, 1951; Gehringer, 1959). It is also the minimum size usually collected with a 3.2 mm #35 knotless Delta mesh beach seine. It should be noted that this mesh size seine can potentially capture smaller individuals, but 20 mm SL is a conservative size.

Length-frequency

A total of 767 juvenile ladyfish ranging from 1 to 99 mm SL ($\bar{x}=48.8\mathrm{mm}$, S.D. ± 26.3 mm) were collected in the IRL during 1991 through 1995. Annual monthly length-frequency distributions were confounded because a few small individuals were collected throughout the year; monthly length-frequency data generally demonstrated a cyclical pattern (Figs. 3 and 4). The smallest ladyfish ($\bar{x}=12.5\mathrm{mm}$ SL, S.D. ± 13.4 mm, n = 2) were collected in September and the largest ($\bar{x}=65.3\mathrm{mm}$ SL, S.D. ± 28.2 mm, n = 174) in May [F(11, 753) = 31.87, P < 0.001]. Post-hoc analysis showed no significant difference in length between August and May, or among the other months. Two separate one-way ANOVAs showed length during April [F(2, 97) = 0.15, P = 0.86] and June [F(3, 50) = 2.35, P < 0.08] was not significantly different among years; however, mean length in May was significantly different among sampling years [F(2, 187) = 21.44, P < 0.001]. The smallest ladyfish ($\bar{x}=42.7\mathrm{mm}$ SL, S.D. ± 16.73 mm, n = 44) captured in May occurred in 1993 and the largest ($\bar{x}=64.6\mathrm{mm}$ SL, S.D. ± 14.01 mm, n = 98) in 1995.

One hundred and sixty-nine juvenile ladyfish ranging from 2 to 99 mm SL ($\bar{x}=34.3\mathrm{mm}$ SL, S.D. ± 16.92 mm) were collected in VC waters during 1993 through 1995. Annual monthly length-frequency distribution demonstrated that growth generally occurred from late-winter and spring to summer (Figs. 5 and 6). The smallest ladyfish ($\bar{x}=19.3\mathrm{mm}$ SL, S.D. ± 19.61 mm, n = 4) were collected in February and the largest ($\bar{x}=70.8\mathrm{mm}$ SL, S.D. ± 34.24 mm, n = 4) in August [F(8, 160) = 6.04, P < 0.001]. Post-hoc analysis showed no significant difference in length among September, October, March, January, April, May, June, and August. In addition, no significant difference in length was found among February, September, October, March, January, April, and May. Three separate one-way ANOVAs showed length in April [F(2, 114) = 0.65, P = 0.52], May [F(2, 4) = 2.27, P = 0.22], and June [F(2, 10) = 1.88, P = 0.20] was not significantly different among years.

Length-frequency progressions

Ladyfish growth in the IRL was unable to be estimated by the progression of monthly cohort sizes as recruitment of smaller individuals occurred throughout the year. Therefore, for comparison purposes, and to eliminate recruitment bias (i.e., influx of small individuals), growth evaluations in the IRL were limited to catches occurring from April to June. This corresponded to the period when recruitment was not only consistent, but monthly mean size generally increased from one month to the next. The monthly instantaneous growth coefficient ranged from −0.0677 in 1995 to 0.094 in 1991. Absolute growth ranged from 0.55 in 1992 to 0.63 mm day⁻¹ in 1993 and 1994. On average, the absolute growth rate was 36.3 mm in 60 days or 0.605 mm day⁻¹. Cohort-specific daily growth rates, elevations, and coincidentals (slopes and intercepts of the regression lines) were similar among sampling years [F(1, 2) = 0.0035, P = 0.3146]; [F(1, 3) = 1.545, P = 0.2702]; [F(2, 2) = 0.5177, P = 0.3121], respectively. Cohort-specific growth rates ranged from 1.807 in 1993 to 1.811 mm day⁻¹ in 1994 ($\bar{x}=$ 1.811 mm day⁻¹, S.D. ± 0.003 mm day⁻¹). The overall growth was best (i.e., goodness of fit) described by an exponential regression having the formula: SL =9.5030^{0.3226 (age)}; r² = 0.8474 (Figs. 7 and 9). If the exponential trajectory rate was maintained over 365 days, ladyfish would attain a standard length of 457.5 mm corresponding to an estimated growth rate of 1.25 mm day⁻¹ (Tables 1 and 2). The corrected exponential growth equation yielded a size-at-age 1 of 156.0 mm SL, which corresponded to an estimated growth rate of 0.4356 mm day⁻¹ (Tables 1 and 2).

Estimating ladyfish growth from VC collections was also problematic because recruitment of small individuals occurred throughout the year and the estimated growth rate varied among sampling years. Therefore, to compensate for the recruitment of small individuals in VC, growth evaluations were limited to catches occurring from March to August. The monthly instantaneous growth coefficient ranged from −0.3061 in 1995 to 0.3324 in 1994. Absolute growth ranged from 0.3833 in 1993 to 0.5833 mm day⁻¹ in 1994. On average, the absolute growth rate was 28 mm in 150 days or 0.1866 mm day⁻¹. Cohort-specific daily growth rates were significantly different among sampling years [F(2, 15) = 3.6921, P = 0.0497]; however, the elevations and coincidentals were similar [F(2, 17) = 0.4349, P = 0.3927]; [F(4, 15) = 2.1324, P = 0.1402], respectively. Cohort-specific growth rates ranged from 1.741 in 1994 to 1.933 mm day⁻¹ in 1993 ($\bar{x}=1.837\mathrm{mm}{\mathrm{day}}^{-1}$, S.D. ± 0.14). Mean ladyfish growth was best (i.e., goodness of fit) described by a linear regression having the formula: SL = 5.4429 (age [days]) + 11.1; r² = 0.8711. However, natural growth was explained better by the exponential regression formula: SL =16.846^{0.1545 (age)}; r² = 0.8659 (Figs. 8 and 9). If the exponential trajectory rate was maintained over 365 days, ladyfish would attain a standard length of 107.6 mm corresponding to an estimated growth rate of 0.2951 mm day⁻¹ (Tables 1 and 2). The corrected exponential growth equation yielded a size-at-age 1 of 80 mm SL corresponding to an estimated growth rate of 0.2361 mm day⁻¹ (Tables 1 and 2).