Methods from Kindinger and Albins (2016) "Consumptive and non-consumptive effects of an invasive marine predator on native coral-reef herbivores" doi: 10.1007/s10530-016-1268-1

To quantify NCEs of invasive lionfish on native herbivores, we observed the grazing behavior of herbivorous fishes at each of the 10 experimental reefs over 10 consecutive days in July 2011, observing paired reefs on adjacent days. Each day, we collected 20 haphazardly selected pieces of algal-covered coral rubble (0.43–0.94 m2 surface area) from a nonexperimental reef containing an extensive area of dead Acropora cervicornis coral rubble inhabited by a high density of three-spot damselfish (Stegastes planifrons). This territorial fish maintains higher standing stocks of farmed palatable seaweeds via interspecific aggression in response to intruding herbivores (Ceccarelli et al. 2001 ). 

Each piece of algal substratum was carefully placed into a plastic bag filled with seawater, photographed out of water onboard a boat, returned to its plastic bag, and transported in a cooler of seawater to a nearby experimental reef. At high-lionfish-density reefs, we randomly assigned paired substrata to two similar, but separate microhabitats (e.g., next to a coral head, on a ledge, etc.) that differed only in the presence (\ 0.25 m away) versus absence ([ 3 m away) of lionfish at the time of observation. At low-lionfish-density reefs, we placed algal substrata in paired microhabitats that were similar to those used at high-lionfish-density reefs, except lionfish were always absent during observation. All replicates were therefore placed in types of microhabitats frequented by lionfish, regardless of actual lionfish presence. Overall, we observed grazing of translocated algal-covered substrata at three levels of lionfish presence: (1) low-lionfish-density reef with lionfish absent from the observed microhabitat (n = 100); (2) high-lionfish-density reef with lionfish absent from the microhabitat (n = 50); and (3) high-lionfish-density reef with lionfish present in the microhabitat (n = 50); hereafter referred to as low -absent , high -absent , and high -present treatments, respectively. These treatments were designed to provide insight on the spatial scale at which lionfish presence affects herbivorous fish behavior by allowing simultaneous comparisons of grazing behavior between (1) low- and high-lionfish-densities at the reef-scale while controlling for lionfish presence at the within-reef scale (i.e., low-absent vs. high-absent treatments) and (2) lionfish presence-absence at the within-reef scale while controlling for lionfish density at the reef-scale (i.e., high-absent vs. high-present treatments). 

At each experimental reef, we monitored four of the translocated algal substrata—one pair in the morning (0900–1200) and one pair in the afternoon (1400–1600)—for 60 min each using automated underwater video cameras placed approximately 3 m away. Meanwhile, we observed the remaining 16 algal substrata with SCUBA (8 replicates per diver) one at a time for 20 min each, with observations divided evenly throughout the day (2 pairs in the morning and 2 pairs in the afternoon per diver). All observations were therefore performed during the day when the probability of lionfish predation is greatly reduced (Green et al. 2011 ; Cure et al. 2012 ) and all lionfish observed were inactive. We identified the species of each fish that visited these substrata, visually estimated its TL to the nearest cm, and counted the number of times it took a bite of algae. Each fish was considered to be a unique individual once it entered the diver’s field of view (approximately 2 m surrounding the focal rock), and continuing until the time it left the field of view and could no longer be visually tracked. At the end of each observation period, the algal substratum was carefully returned to its plastic bag full of fresh seawater and kept underwater until all 20 replicates had been observed. We then rephotographed each replicate onboard the boat. 

Grazing behavior observed at each replicate algal substratum was comprised of the following response variables: (1) visitation rate (number of fish/minute); (2) percent visitation rate (percent fish/minute); (3) bite rate (number of bites/minute); and (4) individual bite rate (number of bites per fish/minute). The percent visitation rate and individual bite rate allowed us to account for any potential differences in herbivorous fish densities between low- and high-lionfish-density reefs. Percent visitation rates were calculated by dividing the total number of fish observed grazing (per substratum) by the total number of herbivorous fish counted at each reef during the reef fish surveys conducted just prior (June 2011) to the grazing observations (July 2011). For all the herbivorous fish that grazed on each experimental substrate, the number of bites each fish took during individual grazing bouts was averaged to measure the individual bite rate. We also used the before and after photographs of each substrate to estimate the percent loss of algal cover from observed grazing. We quantified percent cover from photographs using the image processing program, ImageJ. 

We analyzed the response of all herbivorous fishes that grazed on the experimental substrate by fish size class (small and large , with large encompassing the response among fishes[ 10 cm TL, which remained consistent regardless of further size binning into medium and large size classes). Parrotfishes accounted for 69.2 % of the herbivorous fishes that we observed grazing. Therefore, the behavioral response (same variables as above) of this fish family was also analyzed by fish size class. The remaining fish families (surgeonfishes, angelfishes, and damselfishes) were not further divided by size class, because such extensive division of each response variable would have resulted in highly zero-inflated data. The percent loss of algae from substrata was not analyzed by fish size class nor by fish family, because individual contributions of each fish to the overall algal loss could not be distinguished. 

We fitted LMMs using a similar procedure as the one described above to account for the nested design of the fish grazing surveys when comparing grazing behavior of herbivorous fish among lionfish treatments. Random effects consisted of paired microhabitats nested within paired reefs. In addition to lionfish treatment (lowabsent, high-absent, and high-present), all full models included the initial algal percent cover (algae ) of each replicate substratumas a fixed factor in order to account for any influence this parameter could have on grazing behavior, as well as an algae 9 lionfish interaction. With the exception of the model of percent loss in algal cover, we log-transformed all rate response variables and allowed variances to differ among reefs with weighted terms to meet all assumptions of normality, homogeneity, and independence. When lionfish treatment was significant in the model based on LRTs, we performed multiple comparisons of the response at every combination of lionfish treatments using Tukey’s Honestly Significant Difference (HSD) method. All statistical analyses of both reef fish surveys and fish grazing observations were conducted using the statistical software R (R Core Team 2014 ) with the associated packages, nlme (Pinheiro et al. 2014 ) and multcomp (Hothorn et al. 2008 ).