Introduction

The structure, form and volume of coral reefs are shaped by the interaction between reef growth and reef destruction (Hutchings 1986). Reef growth is achieved mainly by the carbonate production of framework building (scleractinian) corals and encrusting coralline algae, while reef destruction is driven by the physical or biological erosion (bioerosion) of the carbonate structure. Physical erosion is often described as coral breakage. Bioerosion is due to the chemical dissolution or the mechanical abrasion of the substrate by organisms. Physical erosion is episodic (Grigg 1995), while bioerosion is a constant process which over the long term contributes significantly more than physical erosion to the destructive processes of reefs (Harney and Fletcher 2003). In the long term, bioerosion facilitates the formation of characteristic reef structures such as boulder tracts and eroded reef flats (Hutchings 1986). Bioerosion plays a significant role in the production of reefal sediment that produces cays and lagoonal sediment habitats (Neumann 1966; Warme 1976; Hutchings 1986; Perry 1999).

Bioeroders can be divided into two groups: internal bioeroders (borers) and external bioeroders (grazers) (Glynn 1997). Grazers, in their removal of algae, are often the dominant agents of bioerosion and can account for up to 70% of the total bioerosion (Chazottes et al. 2002; Tribollet et al. 2002; Tribollet and Golubic 2005; Pari et al. 2002). Parrotfishes (Scaridae) have been identified not only as a dominant group of grazers affecting coral–algae competition (e.g., Mumby et al. 2006), but also as the dominant group of bioeroders and sand producers on coral reefs, particularly in the Pacific (e.g., Gygi 1975; Ogden 1977; Kiene 1989; Bellwood 1995a, b; Bellwood et al. 2004). The ecological effects of grazing and bioerosion on the reef ecosystem are species dependent. Recent species-specific studies have shown that different species of grazers have different effects on the rate and trajectory of algal community succession (Hixon and Brostoff 1996; Ceccarelli et al. 2005). Similarly, recent studies that have directly quantified the bioerosion rates of different species of parrotfishes have indicated that bioerosion rates are not uniform across species (Bellwood 1995a; Bruggemann et al. 1996; Bellwood et al. 2003; Fox and Bellwood 2007).

Parrotfish species can be divided into scrapers and excavators and are distinguished by their jaw morphologies and feeding behavior (Bellwood and Choat 1990). Scrapers have smaller jaw muscles and more complex jaw articulations. This enables scrapers to take larger bites, but only crop the epilithic algae, leaving the substrate relatively intact. Large scrapers may also be able to remove underlying substrate. Excavators have larger jaw muscles (but simple jaw articulations) allowing for short, powerful bites that penetrate into the substratum and excavate carbonate material (Bellwood and Choat 1990). Studies in the Caribbean have shown that the daily erosion rates of excavators can be 3–6 times higher than the removal rates of scrapers of equivalent size (Bruggemann et al. 1996). These results suggest that bioerosion by parrotfishes will be dominated by excavating species (Bellwood and Choat 1990). No equivalent study comparing species with different feeding modes has been conducted in the Pacific, where the diversity of parrotfishes of both feeding modes is much greater.

The aims of the present study were (1) to quantify and compare the magnitude of bioerosion of a scraper species and an excavator species of parrotfish in Hawaii and (2) to quantify the factors affecting bioerosion rates within a species. We then postulate how harvesting of parrotfishes would impact their ecological role as bioeroders in the reef ecosystem.

Materials and methods

Study area

All observations were conducted on the reef crest, fore reef and reef shelf at Hanauma Bay, Oahu, Hawaii (21.26°N, 157.69°W) as defined by the CCMA (Center for Coastal Monitoring and Assessment) benthic habitat maps for Hawaii. Due to the narrowness of the reef crest, data from the fore reef and reef crest were combined and are hereafter referred to only as the fore reef. The depth of the fore reef ranges from 0.5 to 5 m. The reef shelf consists of a spur and groove system, with spurs starting at depths of 5 m and progressing seaward to a maximum depth of 10 m. Hanauma Bay is a 41 ha Marine Protected Area and fishing has not been allowed since 1967.

Species studied

Hanauma Bay contains five common species of parrotfishes; Calotomus carolinus (browser), Chlorurus sordidis and C. perspicillatus (excavators), and Scarus psittacus and S. rubroviolaceus (scrapers) present in all size classes (Streelman et al. 2002; Bellwood and Choat 1990). This study focused on the two largest species, S. rubroviolaceus (scraper feeding mode) and C. perspicillatus, (excavator feeding mode), to quantify the effects of size and feeding modes on bioerosion rates.

Individuals were sized by visually estimating their fork length (FL) in centimeters. All size estimations were made by a single observer. Individuals were categorized into 5 cm size classes, starting at 25 cm. Individuals less than 25 cm FL of both species were observed to rarely produce bite scars (hence causing little bioerosion), so observations were confined to individuals between 25 and 55 cm FL. There were few individuals of either species larger than 55 cm FL.

The color phase of an individual, whether an initial phase or terminal phase individual, was also recorded. Typical of most parrotfish species, these two parrotfish species have two different color phases as adults (Randall 2007). For both species, the first mature phase is drab brown and is called the initial phase. Initial phase individuals, depending on species, can consist of both females and males (diandric) or, less commonly, only females (monandric). Initial phase individuals of S. rubroviolaceus are diandric (Hawaii Cooperative Fisheries Research Unit 2008), while the sex composition of initial phase C. perspicillatus is still unknown. The second phase of both species in this study is a brightly colored blue-green and is called the terminal phase. The terminal phase is always male.

Visual census

Visual censuses of the distribution and abundance of S. rubroviolaceus and C. perspicillatus were conducted from February to October 2006 by observers snorkeling over the fore reef and the reef shelf. Given that the same individuals were observed in the same locations in Hanauma Bay throughout the year, seasonal effects on population size were assumed to be negligible (pers. obs.). Transects were 10 m wide and of variable length. Transects on the fore reef spanned the entire length of the reef from one side of the bay to the other, while the length of transects on the reef shelf were determined by the length of the individual spurs within the spur and groove system. A GPS was towed and waypoints were recorded along the transect, and the total length of each transect was calculated from the added distances of one waypoint to another. Thirteen transects were conducted on the fore reef (transect lengths: 27–132 m, transect width 10 m; depth <5 m). Fourteen transects were conducted on the reef shelf (transect lengths: 57–168 m, transect width 10 m; depth 5–10 m). The numbers and sizes of S. rubroviolaceus and C. perspicillatus were recorded on each transect.

Feeding rates

To estimate feeding patterns, behavioral observations were conducted from September to October 2002 (average water temperature: 27.0°C) and from February through April 2003 (average water temperature: 25.0°C), which are the times when water temperatures are most divergent. Water temperatures were chosen instead of day length to represent seasons because fish feeding rates have been shown to be influenced by water temperatures (Smith 2008; Ferreira et al. 1998; Horn and Gibson 1990). Because the warming of the water temperature lags behind day length, day length during the warm water months (11 h) was shorter than the cold water months (12 h).

To record the time of the first and last bites (and thereby determine the length of the feeding day), observations were conducted at sunrise and sunset. Snorkellers followed individual fish and recorded the time when the first or last bite was taken. First bite was observed by following individuals that had emerged from their sleeping holes and recording the time the first bite was taken. The last bite at sunset was defined as the bite taken with no other bites observed in the following 10 min or if the individual went into a sleeping hole. Parrotfish feed continuously throughout the day; at no other time than at sunset do parrotfish stop feeding for periods longer than 5 min.

Bite frequency observations were recorded during 5-min sessions between 0700 h and 1100 h and between 1300 h and 1700 h. During each session, an effort was made to select different individuals by targeting individuals of different sizes and by not revisiting areas where observations had already been taken. Observations were conducted while snorkeling, and fish were not approached within 2 m, unless the fish swam toward the observer. Hanauma Bay is a no-take area and frequented by tourists so the parrotfish were ‘diver neutral’ and no avoidance behavior was seen.

The number of bites taken in a 5-min period was recorded, along with the species, phase (initial or terminal phase) and size of the fish. If visual contact with the fish was lost before the end of 5 min, the data were discarded.

Daily feeding rates

In order to have sufficient sample sizes to calculate feeding rates for different sizes, the 5 cm size classes were pooled into 10 cm size classes (25–34, 35–44, 45–54 cm). Feeding rates per individual (y) were adjusted to bites min−1 for each 10 cm size class and plotted against time (x). A quadratic regression

$$ \begin{aligned} & y = b_{0} + b_{1} x + b_{2} x^{2} + e\end{aligned}$$

where b 0, b 1 and b 2 are fixed constants and e is the error term was fitted to the data for each size class.

Feeding rate trends

The quadratic equations were centered (i.e., zeroed) around noon (1200 h). An examination of the first-order constant in the quadratic equation determines whether the feeding rates increase (positive constant) or decrease (negative constant) from early morning to early afternoon.

Number of bites taken per day

Using feeding rates for the start (x 1) and end (x 2) of the feeding day, the area under the quadratic equation is an estimate of the number of bites taken per day (I) which was obtained by integrating the quadratic regression:

$$ \begin{aligned} I & = \left[ {\left( {x_{2}^{3} - x_{1}^{3} } \right)/3\left] {b_{2} + } \right[\left( {x_{2}^{2} - x_{1}^{2} } \right)/2} \right]b_{1} + \left( {x_{2} - x_{1} } \right)b_{0} \\ & = {\text{A}}b_{0} + {\text{B}}b_{1} + {\text{C}}b_{2} \\ \end{aligned} $$

A, B and C are fixed known constants

To calculate the standard error for the number of bites per day, the variance (var(I)) is estimated as

$$ \text{var} (I) = m \cdot \text{var} (b) \cdot m' $$

where b is the vector of estimated regression coefficients [b 0 b 1 b 2], var(b) is the estimated parameter covariance matrix, m is the row vector [A B C] and m′ is the transpose of m;

therefore,

SE(I) = var(I)0.5

(Taylor personal communicatiion).

Standard errors for the number of bites per day were calculated for all size classes for both species and for both seasons. Comparisons between species and seasons were carried out by calculating the z statistic and the P values obtained. The Bonferroni correction was applied to the P values when there were multiple pairwise comparisons.

Bite descriptions

Bite volumes, proportion of bites creating scars

Working in pairs, two SCUBA divers followed individual fish until the first visible bite was seen. The size (5 cm size class) and phase of the fish were recorded. One SCUBA diver measured the maximum length and breadth of each bite scar with vernier calipers. The depth was measured at the middle of the bite scar using the depth probe on the vernier calipers (Bellwood 1995a). The volume was assumed to be approximated by a rectangular solid (see Bite size validation below). Most bites consisted of two bite scars that did not meet (one for each jaw). Therefore, both scars were measured and the volumes added to give an estimate of the total substrate removed by each bite. The substrate type consumed was assumed to be the same as the substrates immediately surrounding the scars (turf algae, crustose algae, turf and crustose algae, others (including live coral)). The contour of the substrate (flat or convex) and the water depth at which the scar was made were also recorded. The second SCUBA diver followed the fish constantly and recorded the fraction of bites taken by the particular individual that actually resulted in bite scars (7–12 total bites for each observation). All observations were taken in less than 10 m of water.

Bite size validation

The accuracy of the bite volume estimates was checked by comparing the estimated volumes based on caliper measurements with direct measures of volume of the same bite scars. Pieces of substratum with bite scars (n = 9) were collected and each bite scar was filled with dental wax. The wax was then removed and weighed and the volume estimated based on the density of the wax (Bellwood 1995a). Five to eight molds were taken per bite scar and the resulting volumes averaged. The volumes calculated from measurements made with calipers (mean vol = 21.4 mm3 ± 4.3 SE) were not significantly different from the volume measured by wax impressions (21.9 mm3 ± 3.4 SE) (paired Student’s t-test: t = 1.92, df = 8, P = 0.89). Therefore, the bite volume estimates obtained by caliper measurements conducted in the field were supported by the wax impression technique.

Determining factors affecting bite volume and proportion of bites causing scars

The general linear model (GLM) procedure (Minitab 2007) was used to determine the factors that significantly affect bite volume. The volumes of bite scars as well as fish sizes were log transformed (bite volume offset by +1 to avoid the singularity at zero) to normalize the data (Table 1). Fish size was the midpoint of each 5 cm size class. Both main factors and interaction terms were included in the model (Table 1). Stepwise backward elimination was performed (factors were removed if P > 0.15) where interaction terms were removed before removing main effects. If there was an interaction term remaining with P < 0.15, no main effects were removed. Main variables that were significant in the model and had more than two levels were subjected to Tukey’s pairwise comparisons to determine which levels were significantly different.

Table 1 General linear models (GLM) factors for bite volume and proportion of bites leaving scars
Full size table

GLM procedures and stepwise backward linear regression were similarly conducted to determine factors affecting the proportion of bites causing scars (Table 1).

The equations derived from the GLM procedures were then used to determine the size of fish for which bite volumes decreased to zero and for which the proportion of bites leaving scars became zero. This gives a theoretical threshold size where parrotfishes start functioning as bioeroders.

Substrate preferences

Percent cover of the different benthic substrates (turf algae, crustose algae, live coral and others) on the fore reef and reef shelf were determined using 1-m2 quadrats placed on the benthos. The substrate types at 25 equally spaced points within the quadrat were recorded. Quadrat locations on the fore reef and reef shelf were determined by randomly generated waypoints using the Animal Movements extension in ArcVeiw 3.2 GIS program (Hooge and Eichenlaub 1997). Percent cover was combined for both the fore reef and reef shelf, adjusted for their respective areas (the reef shelf area is much larger than the fore reef area and simple average of the cover in the two areas would not be accurate). Substrate preferences comparing available substrate with substrate type chosen (see section on Bite volumes, proportion of bites creating scars) were calculated using the selectivity index (Manly et al. 1993). A threshold of 1.5 was used to indicate positive selectivity and a value of 0.5 or less for negative selectivity.

Substratum density

Substratum samples from feeding sites were collected for both C. perspicillatus and S. rubroviolaceus. Fish were followed until a feeding bite scar was produced and then the substrate containing the bite scar was chipped off. These samples were air dried and scrubbed to remove the epilithic algae. The samples were then weighed. Each sample was later sealed with a thin layer of Vaseline® (Bellwood 1995a) and its density calculated from its volumetric displacement in water.

Bioerosion rates of different sized individuals

Fish were grouped into three size classes, 25–34, 35–44, 45–54 cm. Mean annual bioerosion rates were calculated for an individual in each of the above size classes with the following formula:

Mean bioerosion rate (m3 ind−1 year−1) = mean bite volume (m3) × mean proportion of bites leaving scars × mean bites per day × 365

Overall error terms were calculated using the Goodman’s estimator following Travis (1982) where:

$$ {\text{SE}}_{{\left( {x,y,z} \right)}}^{2} = \left( {x,y} \right)^{2} \cdot{\text{SE}}_{z}^{2} + \left( {x,z} \right)^{2} \cdot{\text{SE}}_{y}^{2} + \left( {y,z} \right)^{2} \cdot{\text{SE}}_{x}^{2} + \left( x \right)^{2} \cdot{\text{SE}}_{y}^{2} \cdot{\text{SE}}_{z}^{2} + \left( y \right)^{2} \cdot{\text{SE}}_{x}^{2} \cdot{\text{SE}}_{z}^{2} + \left( z \right)^{2} \cdot{\text{SE}}_{x}^{2} \cdot{\text{SE}}_{y}^{2} + {\text{SE}}_{x}^{2} \cdot{\text{SE}}_{y}^{2} \cdot{\text{SE}}_{z}^{2} $$

where x is the mean bite volume, y is the mean proportion of bites leaving scars and z is the mean bites per day

Bioerosion values (in kg) were subsequently calculated by multiplying the volume eroded by the estimated average substratum density. Z tests were used to test whether these bioerosion rates were significantly different between size classes and species. The Bonferroni correction was applied to the P values due to multiple pairwise comparisons. Because log-transformed bite volumes and proportion of bites causing scars all had significant linear relationships with log-transformed fish size, both fish size and the bioerosion rates calculated were also log transformed and fitted with a linear regression. This log-linear relationship was assumed as a best description of the data collected as there were too few fish size classes (three points total) to conduct meaningful statistics to test the relationship. No r 2 or P values are reported for this regression. The regression was extrapolated back to determine the bioerosion rates of the 15–24 cm size class for each species and to estimate the threshold size for bioerosion. Threshold size was determined as the size at which bioerosion rates were calculated to be zero. Log transformation, rather than assuming a direct linear relationship between size and bioerosion rates, also gives a more conservative estimate of the threshold size. The threshold size calculated from the linear regression of bioerosion rates against size was then compared with the threshold size calculated for bite volume and the proportion of bites leaving scars.

For determining how bioerosion rates varied with biomass, size classes for each species were converted to biomass using length-weight equations derived from S. rubroviolaceus and C. perspicillatus individuals collected on the island of Oahu (Hawaii Cooperative Fisheries Research Unit 2008).

Bioerosion rates over the fore reef and reef shelf

Bioerosion rates per unit area were calculated by combining fish densities and size classes in each area, as determined from the visual census surveys, with the estimated bioerosion rate of each size class. Fish in the 15–24 cm size class were assigned bioerosion rates extrapolated from larger size classes. From the calculations of threshold size for bioerosion in the present study, fish in the 5–14 cm size class for both species were assumed to not cause any bioerosion. A general linear model (GLM) was run to compare the bioerosion rates by the two species on the fore reef and reef shelf.

Results

Bioerosion rates

Bioerosion rates for both species increased with increasing size (Table 2). There were no inter-species differences in bioerosion rates within size classes (Table 3). Within species, each larger size class of S. rubroviolaceus eroded significantly more than the size class below it (Table 3). For C. perspicillatus, there were no differences in the bioerosion rates of the 35–44 and 45–54 cm size classes. The two larger size classes eroded significantly more than the 25–34 cm size class (Table 3).

Table 2 Summary table of the variables used for calculating bioerosion rates per individual, varying with size class (±SE)
Full size table
Table 3 Z values of bioerosion rates between size classes and between species
Full size table

The calculated bioerosion threshold size (i.e., bioerosion rate = 0 kg) was 9 cm for S. rubroviolaceus and 7 cm for C. perspicillatus. Since several analyses converged in estimating bioerosion threshold sizes ranging from 9 to 13 cm (see Size of grazing scars and Proportion of bites leaving scars below) and because these values fall within the upper range of the 5–14 cm size class used in the visual surveys, 15 cm was used as the minimum size threshold for the onset of bioerosion. Consequently, for subsequent calculations of ecological impact, the bioerosion rate for the 5–14 cm size class was assumed to be zero for both species. The bioerosion rates of the 15–24 cm size class for both species were subsequently calculated from the log-linear regressions fitted to the bioerosion rates and were 14 and 21 kg ind year−1 for S. rubroviolaceus and C. perspicillatus, respectively.

When the measured bioerosion values from this study were used to determine the biomass required from each size class for each species to bioerode an equivalent amount of substrate, larger size classes of S. rubroviolaceus in a population bioerode disproportionately more than smaller size classes. The removal of the 45–54 cm size class of S. rubroviolaceus, will require a disproportionately larger biomass of smaller individuals to compensate for the loss of bioeroding activity of large fish. For C. perspicillatus as long as the total biomass of a population of C. perspicillatus (greater than 14 cm) remains the same, overall bioerosion rates will be maintained (Fig. 1).

Fig. 1
figure 1

The biomass required by each size class to bioerode 10,000 kg of substrate. Error bars are standard error

Full size image

Ecological impacts of parrotfish bioerosion

Scarus rubroviolaceus was more abundant than C. perspicillatus for all size classes at both reef zones (Fig. 2).S. rubroviolaceus abundance peaked at 45–54 cm on the fore reef. Because of these factors, S. rubroviolaceus dominated the bioerosion impact of both species, accounting for 96% of the bioerosion on the fore reef and 78% on the reef shelf (Fig. 2).

Fig. 2
figure 2

Densities of S. rubroviolaceus and C. perspicillatus on the fore reef and reef shelf at Hanauma Bay. Calculated bioerosion rates (±SE, kg m−2 year−1) for each species and for each reef zone are in parentheses

Full size image

Turf and crustose algae were consumed 98% of the time for the two studied species. Live coral constituted a very small fraction of the diet (1–2%) in both these species in Hanauma Bay (Fig. 3). The selection indices showed that turf algae was grazed in the proportion available, whereas crustose algae was quite strongly selected for (Manly index for S. rubroviolaceus = 2.4 and C. perspicillatus = 2.1 with a threshold of 1.5 for positive selectivity), and live coral and other substrates were largely avoided (Manly index for S. rubroviolaceus = 0.04 and C. perspicillatus = 0.08 with a threshold of 0.5 or less for negative selectivity). Convex surfaces were grazed more often than flat surfaces (62 vs. 38% for S. rubroviolaceus and 59 vs. 41% for C. perspicillatus).

Fig. 3
figure 3

Feeding preferences of S. rubroviolaceus (n = 75) and C. perspicillatus (n = 82) compared with availability in the environment (n = 71). Percent abundances of food type available in the environment are normalized for the size of the area of the fore reef and reef shelf

Full size image

Feeding rates

The commencement and cessation of the feeding day was abrupt for both species. The first-order constant in the quadratic equation used to describe feeding rates was consistently positive for both species and all size classes, indicating that feeding rates increased from early morning to early afternoon. Thereafter, feeding rates remained constant or declined slightly (see Fig. 4 for examples).

Fig. 4
figure 4

Feeding rates during cooler months for S. rubroviolaceus. A quadratic equation centered around noon (1200 h) was fitted for each size class. The equations are as follows: 25–34 cm: y = −112.68 x 2 + 28.353 x + 17.561 r 2 = 0.397; 35–44 cm: y = −26.254 x 2 + 23.646 x + 13.447 r 2 = 0.3422; 45–54 cm: y = −96.857 x 2 + 21.135 x + 12.156 r 2 = 0.4568

Full size image

S. rubroviolaceus showed no seasonal change in total bites taken per day except for the 25–34 cm size class (z = 2.5, P < 0.006) which made more bites in cooler than warmer months (Table 4). C. perspicillatus feeding rates showed seasonal effects for both the 35–44 cm (z = 2.05, P < 0.02) and 45–54 cm (z = 2.04, P < 0.02) size class, also with more bites taken in cooler than warmer months (Table 4). To compare average bite rates between size classes of the same species and between species, bite rates for warmer and cooler months were pooled (Table 2). For S. rubroviolaceus, feeding rates for each size class were significantly different, with larger size classes taking fewer bites per day than smaller size classes (Table 3). For C. perspicillatus, bites rates were similar for all size classes (Table 3). Comparing species, the same size class took similar number of bites per day except in the largest size class (44–54 cm) where C. perspicillatus took 20% more bites per day than S. rubroviolaceus (Table 3).

Table 4 Seasonal feeding rates (bites taken per day: mean values ±SE) of S. rubroviolaceus and C. perspicillatus
Full size table

Bite descriptions

Size of grazing scars

Bite volume does not differ between species, color phases or substrate shape, but is significantly affected by the size of fish, food type and water depth (Table 1, regression equation: log(bite vol + 1) = −3.5 + 3.04 log(size) + food type constant + 0.04 (depth) where food type constant for turf algae = −0.03; crustose algae = −0.22; turf and crustose algae = 0.05; others = 0.47; R 2 = 0.37, P < 0.0001). Bites taken on surfaces covered with turf algae are larger in volume than those on crustose algae (Tukey’s pairwise comparisons of turf vs. crustose algae: t = −2.658, P = 0.04). Bite volume also increases with increasing water depth.

The threshold size at which parrotfish start bioeroding was determined by assessing at which size the regression model predicted that bite volume would be zero. The threshold size estimated for bioerosion for both parrotfish species was 11 cm on turf algae and 12 cm for bioerosion on crustose algae at water depth 10 m.

Proportion of bites leaving scars

Of the factors analyzed, fish size was the only significant factor influencing the proportion of bites leaving scars on a substrate. There were no differences between species or color phases (Table 1, Proportion of bites leaving scars = −1.12 + 1.13 log(size), R 2 = 12.2, P < 0.0001). Once again, using the resulting regression model, the threshold size for bioerosion (i.e., proportion of bites leaving scars = 0) is 10 cm.

Substratum density

Analyses of substrata with bite scars made by C. perspicillatus (n = 5) and S. rubroviolaceus (n = 4) showed no significant difference in density (t-test: t = 1.1, df = 7, P = 0.3). Pooling these samples resulted in an average density of 1.76 g cm−3 ± 0.06 SE. This value was used for all calculations.

Discussion

This study provides the first estimates of bioerosion by large parrotfishes in Hawaii. The results indicate that fish size is the paramount determinant of bioerosion rates for S. rubroviolaceus and C. perspicillatus. Previous estimates of bioerosion rates for parrotfish from other regions vary widely, depending on the species and sizes measured. Scarus rubroviolaceus and C. perspicillatus in this study fall in the middle of the spectrum of studied species and can be classified as “moderate” bioeroders. Bioerosion rates from literature range from as little as 23.6 kg ind−1 year−1 for Chlorurus sordidus (15–20 cm SL) to 5,690 kg ind−1 year−1 for a full grown (120 cm TL) Bolbometopon muricatum (Table 1 Supplemental Material). Contrary to results from the Caribbean and predictions of Bellwood and Choat (1990), the bioerosion rate of S. rubroviolaceus (scraper) was equal to that of the excavator C. perspicillatus. In fact, there were no measurable differences between these species in almost all factors examined for all size classes. This result also contrasts with that of Bruggemann et al. (1996) who found that in the Caribbean, three to sixfold more bioerosion was caused by the excavator S. viride than by the scraper S. vetula of similar size. One explanation is that despite their weaker jaw morphology (which consists of lighter jaw bones and smaller muscle mass than excavators), scrapers may be capable of increased excavation when they achieve larger sizes (Bellwood and Choat 1990). The current study is the first to compare scrapers and excavators in the large size range of 45–54 cm.

The density of the substratum may also have a role to play in explaining the difference in the current results from those from the Caribbean. The substratum densities reported by Bruggemann et al. (1996) and Bruggemann (1995) at equivalent water depths to the current study ranged from 1.77 to 2.01 g cm−3, which are higher than the 1.76 ± 0.06 g cm−3 (range 1.46–2.08 g cm−3) measured at Hanauma Bay. The densities reported by Bellwood (1995a) and Bellwood and Choat (1990) are even higher (2.44 g cm−3) On lower density substrates, scrapers may be able to bioerode as much, if not more, than excavators that are limited by their smaller gape size (due to simpler and less mobile jaw articulations than scrapers) (Bellwood and Choat 1990). In Hawaii at least, the assessment of the ecological impact of bioerosion by grazers will have to include both scrapers and excavators.

Parrotfishes are the most mobile of all reef bioeroders. They have two impacts on the reef; they bioerode the existing substrate as well as reduce the gross carbonate production by consuming carbonate producers. On most coral reefs, scleractinian corals and coralline algae are the major contributors to carbonate production (Stearn and Scoffin 1977; Hubbard et al. 1990; Harney and Fletcher 2003). Feeding observations in the present study showed that live coral was a very small part of the diet (less than 1–2%). Crustose algae (which in this study includes coralline algae and encrusting non-coralline algae such as Peysonellia) was a preferred food type and a significant proportion (37–43%) of the diet. These feeding preferences are consistent with other studies (Bellwood and Choat 1990; Bellwood 1995a; Bruggemann et al. 1996), although there are exceptions where significant amounts of live coral are consumed (e.g., Chlorurus gibbus (Red Sea), Bolbometopon muricatum (Indo-Pacific), Sparisoma viridae (Caribbean); Bellwood and Choat 1990; Bruckner and Bruckner 1998; Bruckner et al. 2000; Alwany et al. 2009). Even though consumption of live coral in Hanauma Bay was small, the large amount of crustose algae consumed indicates that parrotfishes can significantly reduce the gross carbonate production of Hawaiian reefs.

Parrotfish grazing plays a large role in the carbonate dynamics of Hawaii reefs. The gross carbonate production rates of the Hanauma Bay reef are estimated as 1.84 kg m−2 year−1 for the fore reef and 4.76 kg m−2 year−1 for the reef slope (Ong, unpubl. data). The current data indicate that bioerosion by the populations of these two parrotfish species in Hanauma Bay results in removal of 60% of the gross carbonate production of the fore reef and 19% of the production of the deeper reef shelf (Fig. 2). Thus, the ecological consequences of parrotfish bioerosion in Hanauma Bay, particularly in shallow areas, are significant.

In Hawaii and in many areas in the Pacific, large parrotfishes are frequently targeted in fisheries. Conventional fisheries management aims to preserve biomass, but not necessarily the historical size structure of populations (Jennings and Lock 1996; Law 2000). Given the results of this study, this management approach raises several issues of concern for coral reef ecosystems. The bioerosion rate of an individual fish is primarily determined by body size and approaches zero for individuals of size class 5–14 cm FL. These values are very similar to threshold sizes found by Bruggemann et al. (1996) of 10 cm for Scarus vetula and Sparisoma viride in the Caribbean. If larger individuals are removed by harvesting, any increase in numbers of individuals at or below the bioeroding threshold size of 5–14 cm will not compensate for the loss of bioerosion function even if the overall biomass of the species remains the same.

If only individuals above 15 cm are taken into consideration, changes in stock size structure of C. perspicillatus and S. rubroviolaceus will have different effects. For C. perspicillatus, bioerosion rates increase proportionately with biomass and thus the bioerosion impact of a population can be maintained as long as the biomass remains the same, regardless of the size of the individuals. For S. rubroviolaceus, the bioerosion rate of an individual increases faster than the biomass of the fish (Fig. 1) and the loss of larger individuals will significantly reduce the bioerosion contributions of this species even if the overall biomass remains the same. A similar observation was made by Lokrantz et al. (2008) in Zanzibar, where the surface area cleared by three species of parrotfish while feeding increased disproportionately with size. The ecological impact of the harvesting of S. rubroviolaceus is of concern because S. rubroviolaceus dominates the commercial catch in Hawaii (Hawaii Cooperative Fisheries Research Unit 2008). Fish species (such as S. rubroviolaceus) that have disproportionately greater ecological impact with increasing size may benefit from a harvest slot limit whereby the largest fish as well as fish smaller than size at first reproduction are protected.

While some effects of parrotfish grazing on algal–coral competition and its importance in conferring reef resiliency have been quantified (e.g., Mumby 2006; Mumby et al. 2006; Fox and Bellwood 2007), the effects of bioerosion and its subsequent sediment production have not. Bioerosion is often viewed with negative connotations, especially in the light of inhibiting reef accretion. While the immediate effects of bioerosion include the reduction of reef structure, the attendant sediment production is crucial for the sediment fauna which can constitute a significant proportion of the reef ecosystem (Hutchings 1986; Grigg 1998). In the long term, sediment is also incorporated into the internal reef framework. These sedimentary facies have been shown to be important in maintaining reef structure on the Great Barrier Reef (Marshall and Davies 1982; Davies and Hopley 1983; Perry 1999).

Brock (1979) showed in Hawaii that grazing by parrotfishes at virgin population densities increased coral and coralline algae recruitment. This may be because parrotfish grazing exposes bare substrate for coral settlement during months of coral reproduction (Steneck 1988). Results of the present study (conducted in an area where parrotfish populations probably reflect virgin conditions) indicated that approximately 4% (400 cm2 per m2) of the reef is exposed at any time (Ong, unpubl. data). This amount of bare substrate would be significantly reduced if the larger size classes of parrotfish were removed and this in turn might jeopardize the continued recruitment of reef building corals.

This study was limited to Hanauma Bay. It will be important to study other protected areas to determine whether Hanauma Bay’s assemblages of parrotfishes and size distributions are typical and whether feeding preferences (algae versus live coral) are the same in other areas and if preferences change with availability. Comparisons of substrate condition (e.g., coral and algae cover and diversity) between fished and non-fished areas would enable determination of the magnitude of ecological effects caused by the removal of larger individuals. Experimental studies on the importance of bare substrate availability for successful coral recruitment will also be necessary to determine to what extent parrotfish grazing can affect coral recruitment. Exposure of bare substrate may also enable a greater diversity of organisms to settle on the reef.