Effectiveness of removals of the invasive lionfish: how many dives are needed to deplete a reef?

Study site

We sampled seven sites on Turneffe atoll (17.3638°N, 87.8581°W), Belize, Central America, (Fig. 1). Turneffe atoll is located 9–23 km offshore of the main Belizean Barrier Reef, the largest barrier reef in the Caribbean and second largest globally (Gibson & Carter, 2003). The atoll consists of a number of mangrove islands which provide important nursery habitat for a variety of fish species, including lionfish (Gibson & Carter, 2003; Mumby, 2006; Claydon, Calosso & Traiger, 2012). The perimeter of the atoll consists of a barrier reef that transitions into a drop-off (Chittaro et al., 2006). Numerous reef patches are found within the reef lagoon. The northern portion of the windward side consists of spur and groove habitat while the southern portion is composed of drop-off and wall habitats starting at a depth of ∼18 m. A more detailed description of the area can be found in (Garcia & Holtermann, 1998).

Sampling sites were located along the outer edge of the windward forereef in the central portion of the atoll, ranging in depth from 15–30 m and area from 1,148 to 1,800 m² (1,643 (±238 s.d.) m², Fig. 1, Table 1). Similar to much of the Caribbean, sampling sites generally have low coral cover (4–19%), with some evidence of a coral to algal dominated community shift (McClanahan & Muthiga, 1998; Wild, Jantzen & Kremb, 2014). Between the reef crest and wall habitats, along the windward side, the reef is dominated by numerous Orbicella colonies.

Sampling sites were haphazardly located using natural breaks, such as sand channels, as site boundaries (Fig. 1). Additional marker buoys were placed at the corners of each site to act as guides and help minimize the accidental inclusion of offsite lionfish. Site area was measured using fiberglass tape measures.

Depletion fishing

Divers on SCUBA removed lionfish from each site using pole-spears. For each dive, the number of divers, the time spent hunting for lionfish, the total number of lionfish, and the total length (TL, cm) and mass (g) of each lionfish were recorded. Fishing effort was calculated as diver-hours per dive, the sum of time spent by all divers hunting lionfish at each site. For example, three divers hunting lionfish during a 20-min dive would equal one diver-hour. Each site was depleted by repetitively diving the area, over successive days, until no lionfish were observed or caught on consecutive dives. As the areas of the sample sites varied substantially, all dive times and catches were standardized to diver-hours per 1,000 m² and catch per 1,000 m², respectively. The effect of depletion fishing on abundance was estimated by calculating catch per unit effort (CPUE) as: (1)${\displaystyle CPUE=\frac{C}{h};}$ where C is the catch per dive per 1,000 m² and h is the fishing effort expressed in diver-hours per 1,000 m². The use of diver-hours per 1,000 m² as the metric of effort assumes no change in catch efficiency with increased numbers of divers, for example if divers communicate to help find fish.

Depletion model

We used a Leslie depletion model (Leslie & Davis, 1939) to estimate the initial population size per 1,000 m² on all sites. A depletion model observes how the removal of fish affects the relative abundance of fish remaining in the population (Hilborn & Walters, 1992), assuming that repeated fishing over a small area will reduce local population size (Ogle, 2016a).

Initial population size is calculated as: (2)${\displaystyle N_0=\sum _{t=1}^T{n}_t;}$ where N₀ is the initial population size per 1,000 m², T is the number of removal dives and is indexed by t, and n_t is the catch per 1,000 m² from the t-th removal.

The model assumes that catch per unit effort in the t-th removal event is proportional to the extant population at the time of the t-th removal event: (3)${\displaystyle {CPUE}_t=q{N}_t;}$ where CPUE_t is the catch per unit effort, N_t is the population size per 1,000 m² for the t-th removal, and q is the catchability coefficient representing the fraction of the population that is removed by one unit of fishing effort.

The Leslie depletion model is in the form of a linear equation: (4)${\displaystyle {CPUE}_t=q{N}_0-q{K}_{t-1};}$ where CPUE_t is the response variable, and K_t−1 is the cumulative catch up to time t − 1, q is the slope and qN ₀ is the intercept (Leslie & Davis, 1939; Ogle, 2016a). The initial number of lionfish is thus estimated by dividing the intercept by the catchability coefficient q. Confidence intervals for estimates of qN₀ and q were derived from the regression results following the methods outlined in (Ogle, 2016a).

The model assumes populations are closed to migration over the sampling period, catchability is constant through time and among individuals, enough fish are removed to substantially reduce CPUE, and the catch removes more than 2% of the population (Ogle, 2016a). We attempted to meet these assumptions by conducting depletion dives on consecutive days, with the same observers, and whenever possible, continuing depletion until zero fish were caught or seen on two consecutive dives. While some fish can exhibit fleeing behavior in response to depletion fishing activities (Giddens et al., 2014), and this has been documented for lionfish (Côté et al., 2014), we saw no such change in lionfish behavior during the course of our dives. This is likely a reflection of high capture efficiency, preventing wounded individuals from learning to avoid diver activity. Additionally, the depletion models for all sites met the standard assumptions of normality and homoscedasticity for linear regression (Figs. S1 and S2). This approach does not assume similarity in catchability among sites, or any other site characteristic, as the initial number of lionfish at each site was estimated using only measurements from that site. Despite this, there may be a relationship between the initial density of lionfish at a site and the catchability coefficient at that site due to, for example, handling time being of relatively greater importance at higher densities. To determine if the estimated initial lionfish density affected the catchability coefficient we performed a linear regression.

Lionfish caught versus dive time

Using the estimated initial population size per 1,000 m² we modeled the percentage of lionfish caught as a function of cumulative dive time per 1,000 m² using an exponential asymptotic growth model following the formula: (5)${\displaystyle C=100\left(1-e^{-\alpha qt}\right)}$ where α is the marginal reduction in CPUE, t is the cumulative number of diver-hours per 1,000 m², q is the site-specific catchability coefficient, and C is cumulative percentage of lionfish caught at time t. Models were built both including and not including the catchability coefficient to test the hypothesis that catchability interacts with dive-time to affect the cumulative percentage of lionfish caught. Including catchability coefficient as a multiplier of time acts as a site-specific scalar of effort in order to standardize effort across sites with disparate degrees of removal efficacy. To account for potential variation among sites (beyond what is captured by the catchability coefficient) in coral cover or habitat complexity, for example, this function was modeled to include site as a random factor influencing α. The best model was chosen based on minimization of the bias corrected Akaike Information Criteria (AICc) (Akaike, 1973; Arnold, 2010), which maximizes model fit while avoiding overfitting by incorporating the log-likelihood of the model, while penalizing for the number of parameters and the small sample size. The best-fit model was used to predict the time required to remove 50% and 90% of the estimated total population of lionfish per 1,000 m² based on the mean catchability coefficient.

All of the above analyses were performed with the statistical package R, version 3.0.2 (R Development Core Team, 2013), depletion models were calculated with the FSA package (Ogle, 2016b), exponential asymptotic growth models were calculated using the nlme package (Pinheiro et al., 2013), and graphing was done using the package ggplot2 (Wickham, 2009). All relevant data, analysis, and code are provided in the supplemental material (Data S1 and Supplemental Information 1).

This research was conducted under permission 000033-14 from the Belize Fisheries Department. Animal use for this project received approval by the Animal Care & Use Committee, Texas A&M University—Corpus Christi, protocol number #05-14.

Depletion fishing

Across all seven sites, divers captured 209 lionfish, ranging in size from 12–41 cm TL (mean = 25.9 (±5.9 s.d.) cm). The total number of fish captured at each site ranged from 17 to 43 lionfish (Table 1). In four of seven sites, we were unable to achieve our depletion criterion (i.e., no lionfish observed on two consecutive dives; Table 1).

Depletion model

Leslie depletion models independently estimated the initial number of lionfish present as 11 to 29 individuals per 1,000 m² per site (mean = 19.5 (±6.5 s.d.) per 1,000 m⁻², Table 2, Fig. 2). The catchability coefficient was found to range from 0.43 to 0.91 (mean = 0.694 (±0.170 s.d.), Table 2, Fig. 2) with higher values indicating lionfish were easier to capture at a particular site. The models’ r² values ranged from 0.68–0.98 and were all significant, suggesting a good fit of the data to the model (Table 2). There was no effect of the initial estimated density of lionfish on the catchability coefficient (r² = 0.13, p = 0.45; Fig. S3A).

Lionfish caught versus dive time

The best-fit model incorporated among-site variability as a random factor affecting α and the interaction between the catchability coefficient and dive-time (Eq. 6; Table 3; Fig. 3). The model predicted, with a 95% CI, that one hour of fishing effort per 1,000 m², scaled using the mean catchability coefficient (0.69), results in a 58%–66% reduction in the lionfish population per 1,000 m². Four diver-hours per 1,000 m², scaled by the mean catchability coefficient, achieves 97%–99% reductions of the initial population per 1,000 m² (Fig. 3A). The percent reduction of lionfish for a specified amount of effort is inversely related to the catchability coefficient (Fig. 3B). As such when one hour of fishing effort per 1,000 m² is scaled based on the minimum observed catchability coefficient (0.43) there is a 41%–49% reduction in lionfish; when scaled with the maximum catchability coefficient (0.91) there is a 68%–76% reduction in lionfish (Fig. 3B). Similarly, a 50% decrease of the initial population size was achieved after between 1.16–1.17 (minimum observed catchability) to 0.54–0.55 (maximum observed catchability) diver-hours per 1,000 m²depending on the catchability coefficient (0.71–0.72 diver-hours per 1,000 m² based on the mean catchability coefficient). A 90% reduction required between 3.83–3.88 (minimum observed catchability) to 1.80–1.83 (maximum observed catchability) diver-hours per 1,000 m² depending on the catchability coefficient (2.37–2.41 diver-hours per 1,000 m² based on the mean catchability coefficient; Fig. 3B).