530 



Fishery Bulletin 105(4) 



and to quantify fishing effort in a large square than 

 in a typical narrow rectangle used for belt transects. 

 Within each permanent quadrat, fish were captured 

 by squirting an anesthetic (10% solution of Quinaldine 

 sulfate in seawater) into the water where the fish were 

 hiding and by casting a surround net (for smaller and 

 more sedentary fishes) around corals heads or small 

 patch reefs. For each species, an attempt was made to 

 tag the same number of individuals at each MPA and 

 reference site. Because target species density was gen- 

 erally lower at the fished sites (see "Results" section), 

 fish were captured and tagged first at the fished sites by 

 exhaustively fishing each 20x20 m quadrat. The same 

 numbers of fish were then tagged in quadrats within 

 the MPAs. This procedure ensured that tagging effort 

 was equal across all sites, although fishing effort was 

 often lower in the MPAs because sufficient fish could 

 be captured in a shorter time. For recaptures of tagged 

 fish, all permanent quadrats were fished exhaustively 

 and all tagged individuals were recorded. 



Tagging took place biweekly from May through July 

 2003 and from January through March 2004. Recap- 

 tured fish were collected weekly from May through 

 August 2003 and January through April 2004 in=32 

 total recapture attempts per site), allowing 1 week to 

 6 months between tagging and recapture. Captured 

 fish were identified, measured, and tagged with vis- 

 ible implant elastomer (VIE) tags (Northwest Marine 

 Technologies, Inc., Shaw Island, WA), and immediately 

 released at the site of capture. The VIE tag was im- 

 planted under a fish's skin and thus would not become 

 entangled, scraped off, or fouled with algae. Tag loss 

 can lead to underestimates of recapture rates if a fish 

 is recaptured after losing its tag. Past studies with 

 several reef fish families (Labridae, Scaridae, Acanthu- 

 ridae, and Serranidae) showed high (>90%) retention of 

 elastomer implants, particularly for individuals greater 

 than 150 mm standard length (Tupper, 2007). The ef- 

 fective life of the VIE tag in most reef fish is about 6 

 months, after which the tissue surrounding the tag 

 generally has overgrown and obscured the tag (Tupper, 

 2007). The use of surround nets allowed for capture 

 of resighted tagged fish. This approach enabled much 

 higher recapture rates than those in more conventional 

 studies where external tags and nonselective gears 

 (such as traps) are solely used. 



Analysis of data 



To calculate spillover (S) for a given species, the num- 

 bers and biomass of tagged fish emigrating or immigrat- 

 ing across an MPA boundary were estimated. Spillover 

 was calculated as the number (or biomass) of emigrants 

 minus the number (or biomass) of immigrants. Percent 

 spillover was calculated as the proportion of tagged 

 fish (numbers and biomass) exported to adjacent fished 

 areas minus the proportion of tagged biomass imported 

 to the MPA: 



S = {B/Bp - BJBff) X 100, 



where S = percent spillover; 



B^. = biomass emigrating from the preserve; 



Bp = biomass remaining in the preserve; 



B, = biomass immigrating into the preserve; 



and 

 Sp = biomass remaining in the reference site. 



A positive value would indicate net spillover; a negative 

 value would indicate net influx of biomass to the MPA. 

 Thus, a positive value indicates that the MPA is a source 

 of biomass for adjacent fished areas, where a negative 

 value indicates that the MPA is a biomass sink and 

 perhaps better suited to conserving biomass of a given 

 species than to enhancing local fisheries. 



Before analysis, all raw data were tested for normal- 

 ity by using the Shapiro-Wilk W test and for homogene- 

 ity of variance by using Levene's test (Sokal and Rohlf, 

 1995). Because raw density and raw spillover data did 

 not initially meet these assumptions, they were square- 

 root transformed. Percent spillover data were arc-sin 

 transformed. All transformed data met the assumptions 

 of parametric analysis of variance (ANOVA). Variation 

 in mean fish biomass between locations and between 

 MPAs and fish sites was determined with 2-way ANO- 

 VA. For this analysis, each MPA was paired with its 

 adjacent fished site and this grouping resulted in three 

 pairs: North (Tumon and Tanguisson), central (Piti and 

 Asan Bay), and south (Achang and Cocos Lagoon). The 

 ANOVA design was crossed, with location (north vs. 

 central vs. south) as one factor and protection status 

 (MPA vs. fished site) as a second factor. Variation in 

 mean spillover between locations and between species 

 was also analyzed bby using 2-way ANOVA. Tukey's 

 honestly significant difference (HSD) was used as a post 

 hoc comparison test to determine pairwise differences 

 in mean biomass and mean spillover in MPAs and ref- 

 erence sites. Linear regression was used to explore the 

 relationship between density (expressed as biomass) of 

 fish within the MPAs and the spillover rate from the 

 MPAs. 



Results 



Biomass estimates 



Mean biomass of the three herbivorous species was 

 higher in MPAs than in the fished sites (Fig. 2). Mean 

 biomass of convict surgeonfish did not differ between 

 locations (i.e., between south, central, and north, 2-way 

 ANOVA, F=0.79, P=0.46) but was significantly higher 

 in MPAs than in fished sites at all locations (F=13.47, 

 P<0.001; Tukey's HSD, P<0.05 for all paired compari- 

 sons). There was no significant interaction between 

 location and protective status. Mean biomass of orang- 

 espine unicornfish also did not differ between locations 

 (2-way ANOVA, F =0.90, P=0.42) but was significantly 

 higher in MPAs than in fished sites (F=12.02, P<0.0001). 

 There was a significant interaction (F=9.4, P<0.01) 

 between location and protective status because biomass 



