The rainbow parrotfish Scarus guacamaia is a prominent herbivore in the coastal waters of southeastern Florida whose life history is strongly linked to a dependence on both mangrove and coral reef habitats. Rainbow parrotfish also serve in maintaining the health of coral reefs by keeping algal populations in check. Using NOAA fisheries data from the Mangrove Visual Census and the Reef Visual Census, this study focused on observations of this species in Biscayne Bay and the Upper Florida Bay in order to quantify occupancy and to examine the different factors that affect the presence and absence, and the ontogenetic shifts present in this species between juvenile and adult stages. Logistic regression was used to predict abundance and occurrence using the environmental variables of temperature, dissolved oxygen, salinity, average depth, and distance from channel openings. Presence and absence were also measured against mangrove cover, bottom substrate type, and shoreline development. It was found that salinity, average depth, and distance from channel openings were significant in predicting the occurrence of this species, while temperature and dissolved oxygen were not. Conservation efforts for this species, listed as vulnerable under the IUCN, need to be given greater consideration as the health of this and other parrotfish may be useful in determining the management breadth and priorities on coral reef ecosystems across the Caribbean Sea.

Key words: rainbow parrotfish, mangroves, logistic regression, conservation, land-use planning.


In completing this thesis research, I would foremost like to thank my advisor, David W. Kerstetter, Ph.D., and committee members John F. Walter III, Ph.D. and Richard E. Spieler, Ph.D., whose input and guidance has been critical in moving forward through this project. I would like to thank David L. Jones, Ph.D. for his assistance on equations and statistics. For their assistance in various aspects of ArcGIS, I would like to thank Brian K. Walker, Ph.D. and Kristian Taylor. Notably, I would like to thank James A. Bohnsack, Ph.D. and Joseph E. Serafy, Ph.D. and their work, without whom, this research could not have taken place. I would like to thank my lab mates, especially Bryan Armstrong, Shannon Bayse, Amy Heemsoth, Cheryl Cross, and Kerri Bolow for all their feedback, inquiries, assistance and advice throughout the entire research process. Finally, I would like to thank my family and all my friends for their tireless support and unfailing encouragement in the completion of my thesis work.


Life History of the Rainbow Parrotfish

Rainbow parrotfish Scarus guacamaia is the largest herbivorous fish in the Atlantic Ocean and Caribbean Sea and is found in both mangrove and coral reef habitats (Mumby 2006). The rainbow parrotfish is a large, heavy-bodied, and laterally compressed fish, compared with other species of reef fish. It has a fusiform body shape with dull orange fins possessing streaks of green extending into the dorsal and anal fins; median fin margins are blue in color with the dental plates appearing a blue-green. In this species there appears to be no obvious color differentiation based on sex (Cervigón 1994). Rainbow parrotfish are behaviorally cautious in nature, and are generally observed in isolation, though they can be found in schools of up to thirty individuals (Dunlop and Pawlik 1998).

It has a daily home range of about 1000 m3 (Smith 1997), and occupies varying depths from the surface to 25 m. It depends on corals for shelter and space to inhabit (Cole et al. 2008) and seeks shelter under ledges at night or when threatened. The species has been shown to use the angle of the sun as an aid in returning to these shelters (Smith 1997). Rainbow parrotfish are herbivorous fish that, like most members of the Scaridae family, feed mainly by scraping macro-algae from coral structure (Bellwood et al. 2004). However, it has also been observed to feed directly on coral (Rotjan and Lewis 2006) and gut content analyses have revealed spicules from feeding on sponges (Dunlop and Pawlik 1998).

Rainbow parrotfish life history characteristics are reasonably well known. It is a protogynous hermaphrodite, meaning individuals in this species undergo a sex change between their initial phase, where they are generally female and terminal phase, where they are male. Terminal phase male rainbow parrotfish defend a territory and a harem of females, and when the male dies, the most dominant female will become the dominant male, with her ovaries becoming functional male testes (Streelman et al. 2002). Like other species in this family, peak spawning occurs primarily in warmer summer seasons from May to August, but can occur year-round, and there is an active period of recruitment into the population occurring around February in this region (Haus et al. 2000). Spawning is found to take place generally around dusk, and may correlate to either the lunar cycle or the high tide, as this is an optimal time for egg dispersal. The initial phase is composed of females while the terminal phase is composed of sexually mature males. Rainbow parrotfish aggregate into territories that contain a group of females and the dominant male, which pair-spawns almost exclusively within this group (Munoz and Motta 2000).

The rainbow parrotfish is a relatively large reef fish, compared to most species of reef fishes in the Caribbean, and can achieve a maximum length of 120 cm (TL). The estimated K value of 0.293 equates to a minimum population doubling time of approximately four and a half to fourteen years (Robins and Ray 1986; Randall 1962). Observations of rainbow parrotfish have been made in waters with temperatures ranging from 12-36 °C, salinities ranging from 23.74 to 39.1 ‰ (parts per thousand), and dissolved oxygen concentrations ranging from 2.4 to 14.07 ‰ (Serafy et al. 2003). The species' wide range of tolerances to these factors is most likely an adaptation to the wide range of its known habitats. These habitats range from estuaries to offshore areas, both of which are subject to large pulses of freshwater and storm events. The varied thermal and oxic conditions cannot be exploited by less tolerant species and may be beneficial in providing refuge from predators, foraging grounds, or potential nursery areas (Rummer et al. 2009).

The diet of rainbow parrotfish has been shown to be variable across life stages and habitats. In the Dunlop and Pawlik (1998) study, sponge spicules were found in higher masses in the individuals collected from the mangrove sites as compared to those from coral reefs, suggesting there are shifts in diet preference based on the food sources available. A secondary food source is coral, as rainbow parrotfish has been classified as a facultative corallivore based on direct observations, meaning coral can be either a majority of their diet or only a minor component. These fish impose more permanent and chronic pressures on scleractinian corals (those that generate a hard skeleton such as Montastrea and Porites species) meaning there is repeat scraping activity on these corals, and the damage caused is longer lasting. However, chronic predation may play a factor in regulating distribution, abundance, and fitness of certain prey corals (Cole et al. 2008). Though not fully known, this corallivory may be part of an ontogenetic diet shift, meaning coral is only an important food source for part of their lives, accounting for less than five percent of their bites (Cole et al. 2008). Along with this diet selectivity comes the ability to cause significant damage to corals by biting off growing tips or large portions of skeletal material, which means they are capable of having a disproportionately large impact on the physical structure of Caribbean reefs (Cole et al. 2008). It has also been observed that grazing reduced the density of zooxanthellae and increased the severity of a bleaching event in Belize (Cole et al. 2008). Rainbow parrotfish use a feeding method of scraping or grinding algae from the coral or other rocky substrate, and sometimes inadvertently ingests coral animals as well. The hard coral substrate is broken down through its digestive system, and the excretion of this limestone material is one of the main sources in the creation of the sand surrounding coral reefs in the Caribbean.

Parrotfishes are known to become progressively more important to coral reef ecosystems upon reaching a certain key size around 15-20 cm, at which point they become 'functionally mature' (Lokrantz et al. 2008) and their actions provide a significant impact on the coral reef. This impact increases exponentially as there is a non-linear relationship between body size and scraping function. Calculations have suggested that up to 75 individuals with a size of 15 cm are required to functionally compensate for the loss of a single 35 cm individual, and a 50% decrease in body size can result in a 90% loss of function provided to the ecosystem (Lokrantz et al. 2008). In addition, the level of grazing impact in mangrove systems is also a power function of body length. A conservative estimate places the home range of S. guacamaia at 1600 m3 (Mumby and Hastings 2008), which is larger than that of many other scarids. Rainbow parrotfish also represents approximately 14% of the total grazing intensity measured for mangrove depauperate systems (Mumby and Hastings 2008).

The majority of the rainbow parrotfish diet consists mostly of short epilithic turf algae, cropped algae, red coralline algae, and filamentous algae (Mumby and Hastings 2008), and they feed heavily upon Halimeda opuntia, a green calcareous alga. Juvenile scarid abundance has also been shown to be positively related to the percent cover of Dictyota spp. algae at site level in the Florida Keys (Kuffner et al. 2009). Similar parrotfish species have been observed consuming whole pieces of the thallus rather than grazing on the attached epiphytes, and taking more bites from H. opuntia and fewer bites from coral than would be expected from the percent cover of different microhabitats (Munoz and Motta 2000). While not quantitatively known for rainbow parrotfish, a mean home range for similar parrotfish species, redband parrotfish and redtail parrotfish, in the Florida Keys was observed to be 4371.5 +/- 5869.5 m2 (Munoz and Motta 2000); the standard error was found to be high due to a low number (n = 7) of study sites. Due to overlap in microhabitat and foraging areas in these home ranges, interspecific aggression between parrotfish species takes place when one species attempts to use defended resources to the detriment of the defending species. This aggression involves vigorous chasing over comparatively large distances, as well as biting. Engaging in resource defense behavior was found to be advantageous as the benefits gained outweighed the cost (Munoz and Motta 2000). Aggression has also been observed to be greater when encountering other parrotfish species as opposed to non-parrotfish species and rainbow parrotfish were instigated into these aggressive encounters most often by redband parrotfish Sparisoma aurofrenatum (Munoz and Motta 2000).

Scarus guacamaia is most closely related phylogenetically to midnight parrotfish Scarus coelestinus and striped parrotfish Scarus iseri, with Scarus clades having root nodes at between 2 and 3 million years ago, thus implying that most Scarus species are products of recent speciation. This speciation likely occurred around the time of the complete closure of the Isthmus of Panama at approximately 3.1-3.5 million years ago (Smith et al. 2008). The pantropical distribution and the relatively recent ages of the divergence of the four main clades of Scarus imply that fluctuations in sea level and patterns of differential cooling of the oceans during the Pliocene and Pleistocene may be the driving forces behind the rapid radiation in this genus, which is today largely restricted to the complex reefs built by hard corals (Smith et al. 2008). Alternatively, processes of ecological speciation and divergence due to sexual selection remain a possible explanation for the rapid radiation of parrotfishes, which all have pelagic larval phases and highly similar morphology (Smith et al. 2008). The protogynous mating system of parrotfishes, where species aggregate and have male-dominated haremic systems organized by color recognition, has also been proposed as a possible driving force for speciation via sexual selection mechanisms (Smith et al. 2008). The phylogeny of parrotfish suggests a gradual shift from browsers living in seagrasses to excavators inhabiting rock and/or coral reefs to scrapers found exclusively in association with coral, with Sparisoma being considered the transitional genus (Streelman et al. 2002). It can be assumed that the Scarus genus has always had a habitat association with coral reefs as the Scarus genus is the third radiation off of the Sparisoma lineage (Streelman et al. 2002).

Of the parrotfishes, S. guacamaia is the only species that possesses an obligate and functional dependence on the mangrove habitats (Nagelkerken 2007; Mumby 2006). This dependency has been shown quantitatively in the Mumby et al. (2004) study in which the species suffered local extinctions that corresponded with the removal of mangrove stands, and the extent of mangrove coverage in a region is one of the dominant factors in structuring reef communities. Mangrove connectivity enhances the biomass of rainbow parrotfish on neighboring coral reefs, because grazing influences the cover of macroalgae on reefs and high levels of parrotfish grazing has been shown to lead to a twofold increase in recruitment of Porites and Agaricia corals in the Bahamas (Mumby and Hastings 2008). Biomass of rainbow parrotfish has been shown to more than double when coral reefs were located adjacent to rich mangrove resources, defined as mangrove stands with 70 km or greater of fringing red mangrove Rhizophora mangle located in a region of 200 km2, equating to coverage of 35% (Mumby 2006). Juveniles of this species, those less than 30 cm total length (TL), are observed almost exclusively in mangrove habitats, while all individuals observed on the coral reef were greater than 25 cm TL (Dorenbosch 2006). Average sizes of 10.1 cm and 14.6 cm TL have been recorded in mangroves and seagrass beds, respectively (Nagelkerken et al. 2000). The species of juvenile reef fishes that utilize mangroves and seagrass beds do so because of the high food availability, the presence of shade and shelter that the mangroves provide, and a reduced risk of predation due to the plant and root configurations. There is also a lessened chance of interaction with predator species as well as low predator abundance and efficiency (Verweij et al. 2006). Shallow water habitats such as mangroves and seagrasses, are believed to contain less piscivores than the reef (Verweij et al. 2006) possibly because the energetic costs of chasing the smaller fish in these habitats outweigh the gains of catching one of the prey fish. The turbidity of the water can also negatively affect predator efficiency due to scattering and reduction of light by suspended particles (Verweij et al. 2006). There is significant interannual variability in species composition that may be expected in mangrove fish communities, but spatial factors have been found to contribute more to differences in fish community structure than seasonality (Robertson and Duke 1990).

Verweij et al. (2006) tested the effects of plant structure, shade, and food upon rainbow parrotfish foraging behavior using artificial seagrass leaves and artificial mangrove roots. Rainbow parrotfish showed the same trends as those of pooled herbivores, showing highly significant Poisson regression results for the tested variables of structure, food, structure*food, and location of the experimental unit. In this study, 72 individuals were observed ranging in size from 7.5-15.0 cm. The behavior observed was broken down into 2.8% of individuals resting (spaced evenly throughout the water column), 91.7% foraging, and 5.6% swimming. Eighty-four percent of the rainbow parrotfish observed foraging in the study were found in the artificial mangrove roots, with six percent foraging on artificial seagrass leaves. It was determined that the presence of higher surface area on the root structure provided more substrate for algae, which allowed for diurnal feeding (feeding that occurs in the daytime) on the fouling algae and epiphytes in mangroves and seagrass beds. Rainbow parrotfish observed in this study were also found to be preferential to experimental units with the highest structural complexity. Caribbean region mangroves and seagrass beds function as foraging habitats, but are not used continuously as shelter during the daytime (Verweij et al. 2006). The value of these habitats is diminished with decreased water clarity from turbidity originating from terrestrial run-off, leading to population declines in this and other species (Freeman et al. 2008). Seagrass minimum light requirements differ between species and systems. Halodule and Syringodium seagrass species often require more than 24-37% surface light intensity (Freeman et al. 2008). These seagrass species consistently require minimum light levels that are an order of magnitude higher than the requirements of terrestrial plants or other photosynthetic marine organisms. Reduced subsurface light intensity has caused seagrass declines and the subsequent re-suspension of unstabilized sediments has impeded recovery of these seagrass systems, increasing the pressure placed on species such as the rainbow parrotfish that depend on them (Freeman et al. 2008).

However, presence of preferential habitat is not the only contributing factor determining abundance. It is possible that habitat configuration has an influence on the connectivity between mangroves, seagrasses, and coral reefs and this configuration in terms of providing pathways and connections to the reef affects fish assemblage composition, fish density and size, and species richness (Dorenbosch et al. 2007). Local recruitment patterns can also play a major role. In a study off Curaçao, juvenile densities on the reef were comparable to those in seagrass beds, suggesting that this species can also use the coral reef as a nursery (Dorenbosch et al. 2004). Dorenbosch et al. (2007) concluded that for rainbow parrotfish, migration among these habitats most likely takes place along the coastline. The presence of seagrass-mangrove bays along the coasts of these islands strongly influences the distribution pattern of this species on the coral reef (Dorenbosch et al. 2004). The absence of seagrass beds and mangroves was shown to lead to reduced density of those species that utilize seagrass-mangrove bays in juvenile stages (Dorenbosch et al. 2004). For island sites, this migration was observed to occur on the sheltered or leeward shores, where most adult individuals were observed on coral reefs between 0 and 10 km from mangroves. However, no significant linear relationship was present between mean total density of adult rainbow parrotfish on these reefs and the distance to the nearest stands of mangroves (Dorenbosch et al. 2006). There was also reduced density or complete absence of juvenile rainbow parrotfish on the coral reefs that were farther than nine kilometers from the mangrove and seagrass habitats used by fish of juvenile ages.

The density of these species is additionally regulated on local scales by variable habitat structural complexity and the available vegetation. Herbivory, measured by rates of grazing, was found to be highest at the maximum habitat complexity site (Unsworth et al. 2007). This suggests that the increased shelter and food abundance provided by denser seagrass beds may have increased fish abundance resulting in these higher levels of herbivory (Unsworth et al. 2007). Herbivory was found to increase away from patchy seagrass areas whilst increasing distance from a reef reduced the rate of herbivory due to a reduction in fish migration. Observed high levels of herbivory, however, may only be a short-term effect of irregular grazing by shoals of juvenile and sub-adult scarids (Unsworth et al. 2007).

Rainbow parrotfish migrate across habitats in accordance with its life history stage, and will grow as large as possible before moving on to the next habitat (Mumby et al. 2004). Utilization of intermediate nursery habitats has been hypothesized to increase survivorship of small fish (Mumby et al. 2004). The intermediate nursery stages between mangroves, seagrass beds, and patch reefs serve the function of alleviating predatory bottlenecks in early demersal ontogeny (Mumby et al. 2004). A predatory bottleneck occurs when pressure from predation prevents a large percentage of a population from reproducing. The presence of seagrass beds has also been linked to significantly higher densities of rainbow parrotfish on coral reefs (Dorenbosch et al. 2006) while other studies (e.g., Gonzalez-Salas et al. 2008) have found differing results with respect to these nursery habitats. Noting high abundance of juveniles and adult members of S. guacamaia in coral reef habitats and a total absence in mangrove stands, it appears that mangroves in certain regions do not function as obligate habitats and that seagrass and coral rubble become the primary alternative for nursery, growth, and reproduction (Gonzalez-Salas et al. 2008). It is possible that with removal of mangrove forests the rainbow parrotfish are adapting to utilize other habitats that offer similar survival benefits. The reduced benefits of these marginal habitats may not provide rainbow parrotfish with the resources necessary to survive across their entire life history, allowing only temporary survival through one life stage or another (Rummer et al. 2009). This selective use, which is defined as use of a particular habitat patch disproportionately relative to its availability, can be exhibited either seasonally or spatially, and proximity rather than suitability has been found as the dominant pattern of habitat use (Faunce and Serafy 2008). Mangrove shorelines across broad spatial scales are not equivalent in their value as fish habitats due to the inherent patchiness within the ecosystem. A measure of total habitat area may therefore overestimate the amount of functional habitat utilized by these fishes. In addition, species richness and total number of fishes collected adjacent to mangrove shorelines has been shown to decline with increasing inland distance from creek mouths and oceanic inlets, with water depth greatly related to fish use (Faunce and Serafy 2008).

Rainbow parrotfish are valuable members of the communities with which they are associated. The grazing activities of these parrotfish are beneficial in preventing algal overgrowth and enhance coral reef resilience to algal blooms and other competitor species (Hughes et al. 2007). The species also facilitates settlement and survival of corals by scraping and bioeroding the hard dead coral substratum and are crucial for the regeneration and maintenance of coral reefs (Lokrantz et al. 2008). Rainbow parrotfish and other scarid species participate in not only the uptake of carbon into the food chain in their direct consumption of seagrass, but also indirectly contribute to the detrital food chain with the removal of decaying seagrass material, which potentially results in the widespread dispersal of seagrass material into surface waters. Detached seagrass may also be cast onto the shore where it decays and may re-enter the system as detritus (Unsworth et al. 2007). Rainbow parrotfish may be equally important in influencing seagrass export from the system by the high rates of material discarded during consumption. The unattached plant matter, estimated to be as high as 11% of seagrass growth, becomes subsequently removed from the system by weather and currents (Unsworth et al. 2007). This figure is in addition to the amount consumed in grazing which causes the loss of at least 16% of the seagrass growth each day (Unsworth et al. 2007).

In spite of their ecological role and importance, S. guacamaia populations are thought to be in decline and to have been fished to ecological extinction in Brazil, as well as other areas of the Caribbean (Floeter 2006). Rainbow parrotfish has been listed as vulnerable on the IUCN Red List. This designation means the species is facing a high risk of extinction in the wild based on one or more of the following five criteria: reduction of population size, shrinking geographic range, a population with fewer than 10,000 mature individuals, restricted population extent, or quantitative analysis showing the probability of extinction in the wild is at least 10% within 100 years (the full explanation of which are detailed in the 2004 IUCN criteria; version 2.3, Roberts 1996). Given this information and the ecosystem role of the species developing a model that details occurrence provides a means to assess the health and function of this parrotfish in this region. In addition, one may apply the methods not only throughout the range of this species, but it may be possible to apply this model to other parrotfish species and similar families across the Florida Reef Tract and the Caribbean Sea.

Characteristics of the Biscayne Bay and Florida Reef Tract Region

The Biscayne Bay region receives high numbers of larvae from offshore spawning adults and functions as a source point for juveniles and adults to migrate to the reef tract (Wang et al. 2003). The region also contains some of the most pristine habitat within the Florida Keys (Ishman 1997). The coastal shelf of the Florida Keys is characterized by shallow and highly variable topography, where currents are influenced by tides, wind, and the very energetic offshore Florida current system (Haus et al. 2000). The eddies and meanders of the Florida Current make it possible for upwelling and larval transport to occur across the shelf, and the scale of these perturbations can vary from slow moving mesoscale gyres to faster moving, sub-mesoscale eddies (Haus et al. 2000). Velocities of these eddies can range from 0.53 m/s to 0.80 m/s along the inshore edge of the Florida Current (Haus et al. 2000) and the variability of those velocities can have an impact on dispersal and the resulting end locations of larvae (Haus et al. 2000).

Patch reefs in this region occupy a significant portion of the water column, which leads to variability in the water depth. These protrusions have the potential to change the strength and direction of the tidal flow in the bay. The northern Florida Keys contain over 4,000 patch reefs, composed generally of cemented reef (47.3 +/- 2.2% cover) and pavement (20.1 +/- 2.1%), with varying amounts of rubble, boulders and sand (Kuffner et al. 2009). The benthic community observed on these patch reefs is largely dominated by macrophytes, encrusting invertebrates, and "suitable settlement substratum" found beneath a substantial canopy of gorgonian ("soft") corals (Kuffner et al. 2009). Macroalgae occupies a large portion of space on the reefs, especially Dictyota spp. (15.4 +/- 0.8% cover) and Halimeda tuna (11.7 +/- 0.6% cover). Live scleractinian corals account for only 5.8+/- 0.6% of the benthos (Kuffner et al. 2009).

The tides are generally weak, with a semidiurnal height range of approximately 0.5 m (Haus et al. 2000). As measured in Caesar Creek, tidal velocity can exceed 25 cm/s, while current measurements within the inlets have shown peak tidal velocities in excess of 50 cm/s (Haus et al. 2000). These channels - commonly referred to as the "ABC Channels" because of their names: Angelfish Creek, Broad Creek, and Caesar Creek - form the main outlet from the southern end of Biscayne Bay onto the Florida reef tract. The ABC Channels convey large oscillating tidal flows and wind driven flows between the bay and the ocean, and transport through these corridors predominantly shows a semi-diurnal cycle with amplitudes of 500 m3/s, 300 m3/s, and 250 m3/s respectively (Wang et al. 2003). Based on observations, there is a net outflow at Angelfish and Caesar Creek, but an inconsistent inflow in Broad Creek (Wang et al. 2003). With the tidal flows and the input of freshwater, the residence times of the water varies widely from several months in the more enclosed Barnes Sound and circulation-restricted Card Sound (Ishman 1997), to about a month in the western parts of South Biscayne Bay, and nearly zero in the vicinity of the ocean inlets (Wang et al. 2003).

The area encompassing Biscayne Bay south to Card Sound and Barnes Sound forms a barrier island lagoon system that exhibits estuarine characteristics near points of freshwater inflow during the wet and early dry season (Wang et al. 2003). This lagoon system leads to broad salinity regimes that are highly variable throughout the year, and vary greatly across relatively small areas of only several kilometers due to high freshwater input through canals (as opposed to groundwater), and limited tidal flushing. Salinity variations in Biscayne Bay primarily result from canal discharges through gated control structures, as well as smaller freshwater exchanges in the Bay driven by overland runoff, rainfall, and evaporation (Wang et al. 2003) and upwelling from groundwater (Ishman 1997). The greatest salinity fluctuations occur near canal mouths in Barnes Sound and along the western margin of Biscayne Bay. The smallest fluctuation ranges were observed near ocean inlets (Wang et al. 2003), where the vertical variations of salinity in the water column ranged from less than 0.2 ‰ to a maximum salinity change of 0.8 ‰ from top to bottom in the vicinity of the inlet mouth (Haus et al. 2000). In the Pelican Bank region of Biscayne Bay (see Figure 10), good circulation results in regular flushing and average salinities range from 33 to 35 ‰ (Ishman 1997).

Water flow characteristics in this region are also determined by a network of drainage canals used for agricultural and industrial purposes. These canals also function to control flooding, which has greatly altered the distribution of freshwater within the watershed, as well as the quantity, quality, and timing of freshwater discharges to Biscayne Bay (Wang et al. 2003). This has led to greater pulses with larger peak discharges in the wet season and less freshwater reaching Biscayne Bay in the dry season due to reduced terrestrial storage and lowered groundwater levels (Wang et al. 2003). Increased runoff not only affects salinity conditions in coastal waters, but also can be a mechanism for increased nutrient loading (Rudnick et al. 2006). There exists a coastal ridge, bisecting the Bay, which acts as a groundwater divide, with water west of the ridge flowing toward Florida Bay. The outputs of freshwater from the canals have punctured massive holes through the ridge, changing the direction and characteristics of the flow, and the qualities of the watershed (Wang et al. 2003).

This region also is characterized by large coverage of submerged aquatic vegetation such as seagrasses, and wide availability of phytoplankton, microalgal and macroalgal species. Florida Bay is approximately 2000 km2 in total surface area, with 95% bottom coverage of seagrasses, characterized by sparse, patchy beds of Thalassia testudinum interspersed with locally abundant Halodule wrightii (Fourqurean and Robblee 1999). However, in the spring of 1991, Florida Bay exhibited a shift from a system characterized by clear water to one of extensive and persistent turbidity and phytoplankton blooms, which limits the ability of the seagrass to grow and function properly by reducing penetration of light in the water column (Fourqurean and Robblee 1999). This seagrass die-off was not accompanied or preceded by noticeable decreases in water clarity or increases in colonization by epiphytes, however. There were many hypothesized causes for this die-off which include hypoxia and sulfide toxicity, the loss of the estuarine nature of the system, overdevelopment of the seagrass beds, chronic hypersalinity, in-filling of the bay, and abnormally warm late summer and fall temperatures (Fourqurean and Robblee 1999).

The problem the seagrass die-off presents for rainbow parrotfish and other species of fishes is that fish abundance and diversity is highly correlated with seagrass abundance and species composition in Florida Bay (Fourqurean and Robblee 1999). Die-off events of these seagrasses often result in a lack of suitable habitats, changes in trophic position for various species, and alteration of food webs.

Seasonal algal blooms also present problems for S. guacamaia in the south Florida ecosystems, altering the physical parameters of the ecosystem outside of the utilizable range. Algal blooms are most commonly composed of cyanobacteria (formerly blue-green algae) such as those of the genera Synechocystis and Synechococcus, the most recent of which was likely initiated by an increase in total phosphorus (Rudnick et al. 2006). Southern Biscayne Bay typically has low phytoplankton biomass and low productivity, in large part because of low phosphorus availability; concentrations of total phosphorous normally range from 0.006 mg/L to very rare high values of 0.02 mg/L, yet during the 2005 bloom event, total phosphorous reached levels of 0.1 mg/L (Rudnick et al. 2006). Concentrations of chlorophyll-a, which are an indicator of phytoplankton biomass, typically have a value of 0.4 mcg/L and rarely exceed 2 mcg/L, but when the blooms occurred, values reached 8 mcg/L for chlorophyll-a (Rudnick et al. 2006). Factors causing this bloom event are thought to be linked to road construction along U.S. Route 1 between Florida City and Key Largo (Rudnick et al. 2006). This project involved an 18-mile section of U.S. Route 1 including the construction of an elevated bridge over Jewfish Creek. Wetland mitigation techniques and four pipe culverts south of the C-111 canal bridge were also installed in an attempt to restore partial freshwater flows. The bloom event was also thought to be triggered in part by the multiple hurricane impacts in the 2005 season (Rudnick et al. 2006). These factors were exacerbated by water management operations, which divert water through canals as opposed to groundwater, and other anthropogenic means, such as development that also disrupts natural groundwater flow (Rudnick et al. 2006). The proximity of the Biscayne Bay region to the metropolitan areas of South Florida presents not only a great stressor, but also a unique perspective in this field of study. Can threatened species, such as rainbow parrotfish, exist and thrive under such duress, and can management be effective enough to negate the influence of such development.

Previous Studies and Statement of Problem

Prior work has detailed the importance of the rainbow parrotfish to the coral reef communities of the Caribbean Sea. However, these studies generally only examined rainbow parrotfish as part of a group of species with similar functional traits, as opposed to a detailed study with a focus solely on this species. Other studies have taken place either on sites in the Windward Islands or in the Bahamas, and while these reefs may be similarly physically structured, being mainly fringing reefs, they are not exactly the same as the barrier coral reefs of the Florida Reef Tract. These islands are more isolated, subject to less anthropogenic interactions, and governed by the Caribbean current as opposed to the Gulf Stream. Two studies - the Mangrove Visual Census and Reef Visual Census - have taken place along the Florida Reef Tract, and this work expands on the conclusions from their data.

This research has taken point count data of the rainbow parrotfish from Biscayne Bay and Upper Florida Bay and developed a model predicting the occurrence of this species in other similar areas and regions where such data collection has not taken place. The Biscayne Bay and Upper Florida Bay locations provide good examples of shoreline mangrove forests, islands and barrier reefs, and how they are affected by anthropogenic means in a highly developed region. As is the case with many species of reef fishes there is an ontogenetic shift in habitat utilization with S. guacamaia, and this ontogenetic difference in habitat use may result from predation rate, or juveniles selecting lower predation risk habitats such as seagrass beds (Nakamura and Tsuchiya 2008).

While there is significant information about the habitat preferences of the rainbow parrotfish, there is not published information that substantiates the importance of a specific location to the species, or how dependent this fish is on the abiotic characteristics (e.g., bottom substrate) of a given site. The development of a model for these habitat characteristics has provided numerical evidence for the probability of occurrence of this fish in a particular habitat and a ranked importance of the habitat type. The rainbow parrotfish has a functional importance and dependence within these mangrove and reef habitats of Biscayne Bay and Upper Florida Bay, suggesting that it is a better study subject than other species with less functional impact. The research gives graphical evidence to support and expand the known habitat utilization and life history information for rainbow parrotfish, information that can be applied to similar species found in similar habitats to allow for more accurate and detailed management and conservation decisions for this and other unstudied species of reef fish in the future.


Hypothesis 1: Mangrove habitat predicts the presence of rainbow parrotfish.

It has been well established that there is a functional dependency of juvenile rainbow parrotfish on mangrove habitat (Mumby et al. 2004). However, absence of mangroves does not automatically equate with absence of the species (Gonzalez-Salas et al. 2008). The extent of this importance of the mangrove habitat needs to be quantified, particularly in areas that are subject to high human development. Characteristics such as percent vegetative cover per square kilometer (calculated area from GIS map coverage), and pressures such as human population density per square kilometer were used in assessing the suitability of the habitat to this fish. Regression models will be used to quantify habitat locations and they were ranked for suitability. Finally, they were ranked separately for abundance of rainbow parrotfish.

Hypothesis 2: Thickness of the mangrove forests determines rainbow parrotfish presence/absence.

Mangroves provide this species of parrotfish with structure that in turn provides shelter, shade, and sources of food. However, simply the presence of mangroves may not be enough to support occupancy by rainbow parrotfish. The distance that mangrove habitat extends from the shore was measured and correlated to parrotfish presence/absence. Based upon this relationship, one can make suggestions about the relative importance of mangrove coverage (density) to presence of this fish.

Hypothesis 3: Type of bottom substrate plays a role in abundance and occurrence of rainbow parrotfish.

Bottom substrate affects the plant biota present, as well as relating to the presence of available shelter. Seagrasses grow where the bottom type tends to be sandy, whereas encrusting algae need a hard or coralline substrate on which to attach. Rainbow parrotfish are generally herbivorous and tend towards eating encrusting algae as noted by scars on similar textured corals and sponges (Dunlop and Pawlik 1998). The predominant bottom substrate type was correlated with the abundance and occupancy results between the bottom substrate and the presence and viability of parrotfish to determine if there is a correlation.

Hypothesis 4: Presence of preferred diet species determines location

While there is some plasticity in the diet preferences of this species of parrotfish (Dunlop and Pawlik 1998, Cole et al. 2008), there appears to be a hierarchy in the nutrient benefit provided by the different sources of prey. Comparison of the optimal locations for diet species with the predicted locations of highest occurrence of parrotfish, provided by the model, was measured to show if there is a correlation between diet location and abundance of rainbow parrotfish.

Hypothesis 5: Dissolved oxygen (DO) concentration is the most important abiotic factor determining presence or absence of parrotfish in a particular location.

Rainbow parrotfish was observed across a range of temperatures that varied between 12°C and 36°C, and salinity, which varied between 0 ‰ and 42 ‰, while the range for DO varied between 0 ‰ and 17 ‰. One could then assume that a change in DO might have a much greater consequence than would a change in one of the other factors. This change was simulated by altering the DO numbers in the model and determining how it affects the occurrence numbers.

Materials and Methods

To evaluate these hypotheses, a predictive model was developed to explain the occurrence of S. guacamaia in a given area, and investigate the relationship between juvenile and mature stage individuals and habitat utilization in greater detail. In accomplishing this, two NOAA datasets were examined for rainbow parrotfish in the Biscayne Bay and the Upper Florida Bay region.

The datasets used in developing this model are the Mangrove Visual Census and the Reef Visual Census. Data primarily come from the Mangrove Visual Census (MVC) conducted over the years 1998-2007, as detailed in Serafy et al. (2003). This ongoing study examined the fish assemblages along two types of mangrove-lined shoreline using a visual 'belt-transect' census method over consecutive seasons in Biscayne Bay and the Upper Florida Bay. This method entails snorkeling 30 m long transects parallel to the shore and recording identity, number, and size-structure (minimum, mean and maximum total length) of fishes observed. The width of each belt-transect was 2 m, giving an area of 60 m2 per transect. Visual surveys were conducted between 09:00 and 17:00, thereby minimizing visual identification problems associated with low light conditions. Censuses were conducted during consecutive wet and dry seasons (i.e., July to September and January to March, respectively) and transect locations were chosen at random each season. Measurements of water quality and depth were obtained for each fish census. Water temperature, salinity, and dissolved oxygen were measured using a Hydrolab® multi-probe instrument. Depth was measured along each transect at intervals of 15 m using a 2 m long polyvinyl chloride pole marked at 2 cm intervals (Serafy et al. 2003). This work examined both ontogenetic shifts for various fishes from mangroves to reefs and the trophic roles of varying shoreline habitats (Serafy et al. 2003).

Secondary sets of data were used from the Reef Visual Census (RVC), which was begun in 1979 and was conducted through 2005. This study occurred on the reef tract parallel to the Florida Keys and involved sampling fish community structure in virtual cylinders with a radius of 7.5 m around randomly selected, stationary points (see Bohnsack and Bannerot 1986). Divers in this study began each sample site by facing in one direction and listing all fish species within the field of view. When no new species were noted, new sectors of the cylinder were scanned by rotating in one direction for the duration of the five-minute period slated for each site. After the initial five minutes, data were then collected on the abundance and minimum, mean, and maximum lengths for each species sighted. An "all-purpose tool" (APT; a ruler connected perpendicularly to the end of a meter stick) was utilized as a reference device to reduce potential magnification errors in fish size estimates. Species with few individuals were counted and their size estimated immediately. Highly mobile species, such as sharks and carangids, which are unlikely to remain in the area, were tabulated when first observed and then ignored (Bohnsack and Bannerot 1986).

The Mangrove Visual Census measured a wide range of physical parameters, but is limited temporally in that it was only begun in 1999. The Reef Visual Census covered a much longer timeframe, but also recorded fewer physical data in the observations. These datasets are beneficial since they both employ similar methods and occurred concurrently (for at least part of the studies). In addition, both studies were performed in the same region using low impact assessment methods. The drawbacks in using these censuses are that there are limited data for the sand flats, seagrass beds, and the channels that link these two habitats. The sampling locations are both a benefit and a drawback because it provides coverage across a large range of area, but the sites selected were random, and a more accurate picture might have been obtained in limiting the size of the sampling area. Human error is also involved, leading to measurements and observations that will not be absolutely accurate in situ. The use of different observers over the course of the surveys will present slight problems, too, due to the inherent differences between each observer's individual technique and abilities. The fact that the MVC used snorkeling methods and RVC used SCUBA is not a limitation as the methods reflected the differences in depth of the survey sites. In using both of these studies, it is possible to analyze and compare traits such as ontogenetic shifts, migration, and habitat utilization that would not be possible with analyses using only one dataset.

The rainbow parrotfish was chosen as the study species because of its role as the largest herbivorous fish in the Caribbean region and because its population numbers are in decline, with an IUCN listing as vulnerable. The rainbow parrotfish was also selected because it had broad observation coverage across the two datasets, and there has been limited work done with this species using data from the Florida Keys and Biscayne Bay region, unlike species such as sailor's choice Haemulon parra and gray snapper Lutjanus griseus. These are two species recorded in the MVC and RVC datasets that have been studied in greater detail with more research having been done in the region on grunts and snappers.

The data sets used for this work consisted of 1,812 sites in the Mangrove Visual Census (MVC), of which 1,798 were used, and 13,443 sites in the Reef Visual Census (RVC), of which 764 points were used. Sites from the MVC were omitted if there was one or more physical parameter for which there was no data and sites from the RVC were omitted if there was not rainbow parrotfish observed at that site.

Logistic regression was chosen because this type of analysis has predictive power when using many variables, and tests the range between two given outcomes; "1" or "0" in this case, representing presence or absence respectively. This technique has proved valuable in other studies (e.g., Sampson and Al-Jufaily 1999, Fransen et al. 2006, Bi et al. 2007) using variables such as stream gradient, temperature, and sediment type, and a 91% accuracy was found in determining occupancy in the Fransen et al. (2006) study. While these studies examined sole (Sampson and Al-Jufaily 1999), salmon (Bi et al. 2007), and multiple species (Fransen et al. 2006) as opposed to rainbow parrotfish or other reef fish species, they are still valuable examples of habitat modeling using physical and environmental factors for management purposes.

The model was set up initially in Microsoft Excel (Microsoft Corporation, Version 12.0.6214.1000) and included the following parameters: temperature, dissolved oxygen, salinity range/regime, average depth, distance from corridors, and mangrove density. Habitat type, mangrove cover, substrate type, and shoreline development were researched and used for comparison, but not incorporated into the model. The parameters, such as mangrove cover and substrate type, not included in the original data sets were derived from available GIS map coverage data from agencies such as the Florida Fish and Wildlife Conservation Commission and the United States Geological Survey. The value for mangrove cover/density is defined as the distance of continuous mangrove cover extending from the shoreline. These data were incorporated into a custom ArcGIS shapefile for analyses. Using the programs R (The R Foundation for Statistical Computing, Version 2.6.0) and SAS (SAS Institute, Version 9.2), this set-up was evaluated by a logistic multiple regression. Logistic regression was chosen because this analysis is used to predict the probability that an event of interest will occur as a linear function of one or more continuous or dichotomous independent variables (Karp 2001). The preferred approach of habitat association modeling is to use regression modeling to relate the variations in large-scale distributions to key habitat variables (Freckleton et al. 2006). In logistic regression, maximum likelihood estimators are approximately normally distributed, with little or no bias (Ragavan 2008), but there is often the need for at least two types of model using the data from one study: one for description/interpretation and another for prediction (Shtatland et al. 2001). The corresponding informational outputs and regression values - the numerical values that resulted from the model's initial setup parameters - were then be ranked according to habitat, abundance, and size, and then plotted spatially on a GIS map using ArcGIS (ESRI, Version 9.2).

Using the GPS-based positions included in the respective data sets, the locations of the rainbow parrotfish individuals were plotted on a map using ArcGIS. Initially, all observed individuals (Figure 1) were plotted, and then these individuals were broken down into groups by size of 0-100 mm, 100-200 mm, 200-300 mm, and greater than 300 mm (Figures 3A, 3B, 3C, 3D) based on ontogeny and similar work from the MVC. Each of these maps were further separated by number of individuals at each site in groups corresponding to 1-5 individuals, 6-15 individuals, or 16-80 individuals. These groups were represented on the maps with points of increasing size; the breakdown of these numbers is shown in Table 2.

The size distribution values for the rainbow parrotfish individuals are depicted in Figure 4. With this size information, the sites were mapped again, broken down into the locations of observed juvenile and mature individuals, based on total length, for the MVC (Figure 9A) and RVC (Figure 9B).

Next, a logistic regression analysis was performed using the variables in the MVC of temperature, dissolved oxygen, salinity, average depth, and the distance from island channels. Distance from the island channels was not included in the initial measurements of the data set but rather was calculated by taking the GPS location in the center of each of the eight channels (Figure 2) and measuring the distance to each site using the equation (D.L. Jones, personal communication):

Distance = R*(acos((sin(lat1/r)*sin(lat2/r))+(cos(lat1/r)*cos(lat2/r)*cos((lon2-lon1)/r)))),

where R = 6370 (Earth's radius in kilometers) and r = (360/(2*p)).

The distance from each point to the closest channel opening was the value used. The RVC data were omitted because the environmental variables other than depth were not recorded, and depth was not recorded at any study site where a rainbow parrotfish was not observed. To set up the regression analysis, the five aforementioned values were measured against the presence (1) or absence (0) of Scarus guacamaia individuals at each site. A multiple logistic regression was performed using all the variables and separate logistic regression analyses were then performed for each of the five variables individually against presence or absence.

Finally, mangrove coverage was measured from Virginia Key to Tavernier Creek, an approximately 100 km straight-line distance, using spatial analysis and measure tools in ArcGIS. The perimeter (km) and area (km2) were recorded of mangrove stands present on the Keys, the mainland shore, and islands in Upper Florida Bay, Blackwater Sound, and Card Sound from the derived mangrove coverage shapefile. Mangrove patches that were not adjacent to a water outlet were omitted.


In the MVC dataset rainbow parrotfish were observed at 107 of the 1,798 sites, consisting of a total of 533 individuals. In the RVC dataset rainbow parrotfish were observed at 764 of the 13,443 sites, consisting of a total of 1,499 individuals. Of these individuals, 57 were of mature size in the MVC and 688 were of mature size in the RVC (refer to Table 1). This gives naïve occupancy, a measure of total observed individuals divided by the total number of locations, of 0.3052 for the MVC, and naïve occupancy of 0.1115 for the RVC.

The total regression was found to be significant with a Pr > Chi-square value of <0.0001 and salinity, average depth, and channel distance were found to be individually significant with Pr > Chi-square values of 0.0014, <0.0001, and 0.0001, respectively. Temperature and dissolved oxygen were found to not be significant with Pr > Chi-square values of 0.8359 and 0.7855, respectively. These results coincide with the results of the ROC (receiver operating characteristic) curve, a measure of sensitivity versus specificity, better termed as a representation of the relationship between false positive and false negative rates for every possible value. An ideal test has a value of 1.0 indicating 100% sensitivity and specificity, meaning there is a low false positive rate and there is a high false negative rate. When the ROC curve follows a diagonal path from the lower left hand portion of the graph to the upper right hand portion this means that every increase in the false positive rate is equaled by a corresponding reduction in the false negative rate, and the area under the curve is closer to 0.5 indicating 50% sensitivity and 50% specificity. For the total regression, the area under the ROC curve was a value 0.921, for temperature it was a value of 0.544, for dissolved oxygen it was a value of 0.599, for salinity it was a value of 0.774, for average depth it was a value of 0.821, and for distance from channel openings, it was a value of 0.886.

The results of the logistic regression indicate that using the five variables together is the best predictor of presence or absence; salinity, depth and distance from channel openings are adequate predictors of presence or absence, while temperature and dissolved oxygen are not individually significant in predicting presence or absence. The inflection point of the logistic regression curve for salinity occurred at 36.4 ‰ with a predicted occurrence value of 0.100 (Figure 5). Average depth had an inflection point of 103.5 cm on the logistic regression curve with a predicted occurrence value of 0.150 (Figure 6). The inflection point of the logistic regression curve for the distance from the closest channel opening was 3.5 km, with a predicted occurrence value of 0.11633 (Figure 7). Out of the 1,798 sites in the MVC, 414 were above the inflection point for salinity values, 148 were above the inflection point for average depth values, and 316 were above the inflection point for channel distance values. Table 3 shows the complete breakdown of the predicted regression values for all of the five variables, as well as the total for the three significant values.

The total predicted occurrence values are mapped out for the Biscayne Bay region, seen in Figure 8A, as well as for temperature (Figure 8B), dissolved oxygen (Figure 8C), salinity (Figure 8D), average depth (Figure 8E) and distance from channel openings (Figure 8F).

Mangrove coverage from Virginia Key to Tavernier Creek was found to occupy an area of 124.1 km2, with a perimeter on these patches of 1292.1 km. This coverage counted 309 unique stand-alone parcels across 13 separate locales in the region.


This study has shown that in using the combination of the five aforementioned variables - temperature, dissolved oxygen, salinity, average depth, and distance from channel openings - it is possible to significantly predict the occurrence of rainbow parrotfish individuals. Furthermore, the variables of salinity, average depth, and distance from channel opening are each individually significant in predicting the occurrence of rainbow parrotfish, while temperature and dissolved oxygen are not.

Recalling the hypotheses, it was surmised first that mangrove presence is essential to rainbow parrotfish, and second that density of mangrove stands was more important than presence alone. Stronger presence versus absence data reflected the importance of mangroves, and regression data predicted that the higher density of mangroves on the keys as compared to the mainland shore led to higher occupancy of rainbow parrotfish; however, more work needs to be done to quantitatively justify this assessment. The third and fourth hypotheses also correspond with each other, in that benthic algae being the preferred diet species are generally found on hard bottom substrate as opposed to sandy bottoms or seagrass beds. This corresponds to the data, as presence was higher in areas of hard bottom substrate. The fifth hypothesis that dissolved oxygen, based on its narrow range, would be the most important variable was not proven, as dissolved oxygen was not found to be significant, while salinity, average depth, and distance from channel openings were.

Temperature turned out to be not significant as a predictive variable for presence of rainbow parrotfish. Temperature can vary greatly across very small distances and the variations can be highly localized due to factors such as sunlight intensity or groundwater seeps. The ability of this species for movement across daily ranges of up to several thousand meters means adjusting for slight changes in temperature of only a couple degrees Celsius to more suitable conditions would not be a limitation.

Dissolved oxygen was initially hypothesized to be a more important factor than salinity, because the measurements taken of this variable were across a smaller range than variables such as temperature and salinity, which turned out not to be the case. This could be because pockets of high or low dissolved oxygen values that vary due to such factors as vegetative cover or algal bloom events are dispersed across wider areas. The rainbow parrotfish inhabits areas that are suitable and avoids those regions without adequate amounts of oxygen, and the variations within the suitable habitat may not differ enough to determine specific locations of observed individuals.

Salinity would present more of a problem of avoidance for rainbow parrotfish should conditions become unfavorable. The problem occurs as salinity gradients are more gradual, but fluctuations have the potential to spike drastically and persist for greater time periods over larger areas. The inflection point for salinity occurred at 36.4 ‰, with a predicted regression value of 0.100, meaning that there is a lower than 10% chance of occurrence at any site where salinity is below that value. With these values, occurrence is much more likely to occur in waters than have salinities in the range of 35-40 ‰, corresponding with the clear, warm, natural oligotrophic waters of the Florida reef tract. The clear, shallow waters of coastal estuaries from Biscayne Bay to Barnes Sound possess natural salinities near saltwater strength, providing the optimal salinity conditions as nursery grounds and transition habitats.

Average depth is a significant variable due to several reasons. The inflection point for this regression analysis was 103.5 cm, meaning a water depth of about a meter is the critical value for predicting occurrence. The rainbow parrotfish diet consists of benthic turf algae with incidental corallivory (Cole et al. 2008) and it has also been shown (Verweij et al. 2006) that plant structure is preferred by rainbow parrotfish. In depths of water less than a meter, there is less available space for swimming, foraging and shelter, and conditions at that depth may be suboptimal for epiphytic plant and algal growth. Halimeda species have shown optimal photosynthesis at depths of one meter (Häder et al. 2008) and Caulerpa species have shown optimal photosynthesis at depths of five meters (Häder et al. 1997). The Häder et al. (1997) study also showed that exposure to solar radiation at the surface caused a drastic decline in the effective photosynthetic quantum yield, and to a smaller extent nonreversible photodamage. Much of the coral substrate of the Florida reef tract on which rainbow parrotfish inhabits is greater in depth than a meter, generally depths of five to greater than ten meters (Duffy 2007), so it becomes a question of necessity that to find suitable habitat, inhabiting depths of greater than a meter is required. The deeper waters are also less subject to changes in sediment loads, which can affect bay water clarity leading to changes in the benthic algae cover, as well as seagrass and drifting macroalgal communities, which comprise most of the submerged aquatic vegetation.

Distance from channel opening is significant due to the separate juvenile and mature life history stages that take place in the mangrove communities, and on the coral reef, respectively because to there needs to be a pathway for this transition to take place, and the closer an individual is to these channels, reduces the costs due to predation, energy expended, and search for food. Rainbow parrotfish is known to have a large home range with an ability to swim over great distances, but the less time spent transitioning between suitable habitat allows for allotting less energy on avoidance, swimming, and escape, and more on foraging and reproduction.

The maps of predicted regression values for the Biscayne Bay region illustrate these points. Mangrove-lined shorelines of the leeward keys in Biscayne Bay and Upper Card sound have the highest predicted regression values. Here, the mangrove thickness and coverage is the greatest and substrate coverage is hardbottom, as well as a large proportion being part of Biscayne National Park. The next tier of regression values could be found on the selected windward sites outside of Biscayne National Park, and in the areas of Card Sound and Upper Florida Bay, where mangrove coverage is high near where the Overseas Highway enters onto the Keys.

The lowest predicted values are found on the mainland shorelines. In this area, there is low mangrove coverage, and low density of mangroves back from the shoreline. The bottom coverage is also composed of seagrass, rather than hard substrate, further limiting available diet resources for rainbow parrotfish. In addition, predators on juveniles such as barracuda, sharks, snappers and grunts are higher in number in this area (MVC data), and the anthropogenic influence is higher due to the continued development in Miami-Dade County that continues to encroach farther south. The impact of continued population growth cannot be emphasized enough. A study has found that fecal cortisol and corticosterone levels will increase in rainbow parrotfish in response to hypoxia or restraint stress, responses that are exacerbated under the influence of anthropogenic development (Turner et al. 2003). In a study by Turner et al. (2003), juvenile rainbow parrotfish ranging in size from 25-35 cm TL were observed off of St. John in the U.S. Virgin Islands. The averages for cortisol and corticosterone were significantly greater at two sites, Harbor Point and Gallows Point, compared to the average of any other sites. These two sites are the ones situated closest to the towns of Coral Bay and Cruz Bay, showing that even small communities can have a large impact on coral reef ecosystems.

Returning to the importance of bottom substrate, sediment cores from Manatee Bay show a history of deposition of mangrove peat material followed by progressively increasing amounts of marine sands and mud (Ishman 1997). Similarly, Card Sound and Pelican Bank show patterns of peat and high levels of vegetation accumulation early in their history followed by increases in the deposition of marine and or carbonate clastics (Ishman 1997). The normal bottom coverage has been disrupted because removal of natural flora in the area has impeded biophysical cycling regimes, resulting in sub-optimal habitat for juvenile reef fishes. Often activities such as feeding, reproduction, and resting occur in different habitat types, and the home range consists of two areas joined by a narrow movement path (Kramer and Chapman 1999). With reduced viability of hardbottom substrate between mangrove and reef habitats there is limited effectiveness of the transitions between life stages.

The mainland edge consists of freshwater wetlands, which contain fringing, riverine, and basin mangrove communities, separated from marine habitats in certain places by a mud embankment that impedes freshwater runoff. Natural freshwater input from rainfall and runoff is now disrupted as water flows through artificial drainage canals as opposed to natural channels and circulation. This has led to a variable freshwater flow in many near shore areas of Biscayne and Florida Bays, such that downstream salinity regimes are suboptimal for freshwater, brackish, and marine organisms (Ishman 1997).

Rainbow parrotfish are pelagic spawners, leaving larval settlement subject to flow regimes and currents. Larval settlement could then be impaired in areas that are meant to function as nursery grounds by the compounding problems of lack of suitable habitat and increased distance from adult coral reef habitats. Greater numbers of S. guacamaia individuals are found on the northern edge of the Florida reef tract where suitable mangrove stands are present as opposed to farther south, where development has left much of the windward and leeward shores absent of mangrove stands.

Mangroves have historically been one of the prominent ecosystems of coastal South Florida however these vital habitats are under threat. Of the approximately 30,000 km2 that make up the Everglades, only 1500 km2 is occupied by mangroves (Olmsted and Loope 1984). While that measurement takes into account a much larger area, the area from Virginia Key to Tavernier Creek only possesses approximately 125 km2 of mangroves, indicating that there has been a sizable decline of these ecosystems. Freshwater marshes have been found to experience a 21.95% loss in total area in the Everglades since 1927 due to draining for agriculture and other uses (Foster and Smith 2001). The loss of these marshes throws off the balance between this ecosystem and that of the mangroves, adding more support that there is a much greater degree of interconnectedness that needs to be examined in future studies. With the decline of these ecosystems, species such as rainbow parrotfish that depend upon them will also suffer reductions in population. This population loss will not entirely be able to be halted by managing only the coral reefs, which make up only a part of their life history.

Rainbow parrotfish are listed as vulnerable under IUCN criteria, a designation given not only due to a reduction in suitable habitat, but also reduced size of individuals. Reduced individual size leaves fewer numbers of those that are functionally mature and have a significant impact on their ecosystem. Rainbow parrotfish and other similar species need to be looked at in greater focus, as to how their loss will impact the Caribbean region. In most parts of the Caribbean, parrotfishes are a major component of reef and subsistence fisheries, especially where their slower-growing predators have long been depleted (Hughes et al. 2007). In a study by Hawkins and Roberts (2003), average sizes of all parrotfish species tended to decrease with increasing fishing pressure and smaller species constituted an increasing proportion of the total parrotfish assemblage as fishing pressure increased. This population level fishing may eliminate behavioral traits transferred through generations by larger, older fish (Hawkins and Roberts 2003). Fishing may also lower recruitment by preventing fish from living long enough to survive through periods when conditions are poor for offspring survival (Hawkins and Roberts 2003). Also of concern for rainbow parrotfish are ontogenetic sex changes, which leave sex ratios out of balance within communities. This problem is exacerbated by the reduction in numbers of large terminal phase males by fishing or other stressors.

Rainbow parrotfish have several crucial roles in the dynamics of tropical reefs: they graze fleshy seaweeds that compete with juvenile and adult corals for space; they erode dead coral skeletons and generate reef sediments, and they are an important trophic link between their natural predators and algal primary producers (Hughes et al. 2007). Parrotfish density will vary spatially and temporally in response to local rates of recruitment and mortality (Hughes et al. 2007). Meanwhile, the suitability of a habitat varies spatially, so a specific threshold of habitat quality must be achieved to allow populations to persist (Freckleton et al. 2006).

Implementation of an adequate marine reserve in this region could help to ensure the viability of not only this species, but also the biodiversity present across these habitats. The Florida Keys are protected by a National Marine Sanctuary, multiple state parks, and National Wildlife Refuges farther south than the scope of this study. However, within these boundaries, very little area is designated as ecological reserve, and only small, individual reefs are designated as sanctuary preservation areas. Reserves can only protect fish species if movement of individuals remains within a localized home range, contained fully within the reserve boundaries, during at least part of their life cycle (Kramer and Chapman 1999). Rainbow parrotfish, being highly mobile with ranges of several thousand meters (Munoz and Motta 2000), and a life cycle that requires varied, disparate habitats, do not fit within these constraints. Home range size has been shown to be a power of body size (Kramer and Chapman 1999), so without an adequately sized preserve area, rainbow parrotfish may face pressure or even inability to reach maximum or functional size. In addition, on fringing coral reefs, where habitat zonation parallels depth contours, the home ranges of larger species are considerably longer in the dimension parallel to the depth contours than in the dimension perpendicular to it (Kramer and Chapman 1999). This illustrates the need for not only adequately sized reserves, but also properly shaped ones as well. Relocation of the home range is one possibility, which involves selection for and movement to a novel habitat patch. This can occur as a response to changes in a despotic distribution, a distribution with one dominant male and a harem of submissive females, either by competition with other parrotfish species or with sex change within a harem. In this case, however, the costs of relocation will increase with the investment required to learn the characteristics of the home range, which may be larger for species such as S. guacamaia that use the complex substrate for feeding and refuge (Kramer and Chapman 1999).

Solutions proposed in other marine regions are "no-take" areas and networks of Marine Protected Areas or MPAs. A "no-take" reserve is a protected area within which extractive fishing activities are not permitted (Van Lavieren 2009). This regulation can also apply to other activities such as pollution, construction, research, boating and diving (Van Lavieren 2009) and relate to one or several species. These MPA networks can function to sustain resident populations both by local replenishment and through larval dispersal from other reserves (Planes et al. 2009). Protecting spawning areas of reef fishes spills over and replenishes neighboring areas of coral reef and related habitats, and larval subsidies from a single reserve may contribute to the resilience of subpopulations at other reserves within a network of MPAs (Planes et al. 2009). This method has proved to be effective for protecting grouper spawning aggregations, but may prove difficult with rainbow parrotfish, and a broader coverage net would most likely be necessary. In Mumby et al. (2006), it was found that biomass of parrotfish was reduced by 30-60% on adjacent reefs, compared with biomass in the no-take area. In another study the elevated biomass of parrotfish in a no-take area resulted in an estimated grazing intensity that was six times higher and the cover of seaweed within the park was five times lower (Hughes et al. 2007).

Proper stewardship of marine parks by local communities can enhance grazing and help to prevent regime shifts from coral- to algal-dominated systems. Many marine parks have been established by central governments or foreign non-governmental organizations (e.g., the Nature Conservancy), but they remain ineffective because they lack local support or adequate management (Hughes et al. 2007). In the Florida Keys, estimates range that 33-75% of local income is due to diving and tourism, with an additional 5-8% coming from commercial fishing (National Marine Sanctuaries website), and to restrict use of these offshore areas to "no-take", or even limiting areas used for recreational purposes, such as catch-and-release, would put a great strain on the residents of these communities. However, marine parks and protected areas have shown to be effective in preserving species such as rainbow parrotfish and total biodiversity in other regions while not impinging on the local livelihoods. In the Mumby et al. (2006) study, populations had not leveled off even after the marine park had been in existence for over twenty years, implying that results may not be seen immediately, even if the course of action is highly restrictive. With populations of rainbow parrotfish vulnerable and in decline, and the mangroves and coral reefs where they inhabit threatened by anthropogenic means, marine protected areas need to be developed that ensure future biodiversity and sustainable resident populations.

If reserve design is not changed to meet these goals would mean losing not only the aims of preservation, but also the livelihood that depends on these waters and species. A concerted effort needs to be undertaken to create self-sufficient marine preserves that incorporate mangrove and reef habitat as well as connecting corridors, based first on solid science, and one that incorporates the needs of all the stakeholders. The necessity of taking a broader, ecosystem based goal in conservation, that accounts for inputs and stressors, cannot be stressed enough, especially in the much more open, vulnerable marine environments.